Interpreting Archaeological Topography: 3D Data, Visualisation and Observation 1842175165, 9781842175163

Table of contents :
Cover
Foreword
List of Contributors
Acknowledgements
1. Interpreting archaeological topography: lasers, 3D data, observation, visualisation, and application
2 An overview of airborne and terrestrial laser scanning in archaeology
3 Airborne laser scanning and archaeological interpretation – bringing back the people
4 Cultivating the ‘wilderness’ – how lidar can improve archaeological landscape understanding
5 I Walked, I Saw, I Surveyed, but what did I see?...and what did I survey?
6 Reading aerial images
7 Messy landscapes: lidar and the practices of landscaping
8 Visualizations of lidar derived relief models
9 Worth a thousand words – Photogrammetry for archaeological 3D surveying
10 From lidar to LSCM: micro-topographies of archaeological finds
11 Using lidar data – drawing on 10 year’s experience at English Heritage
12 Lidar and World Heritage Sites in Ireland: Why was such a rich data source gathered,how is it being utilised, and what lessons have been learned?
13 The role of lidar intensity data in interpreting environmental and cultural archaeological landscapes
14 The changing picture of archaeological landscapes: lidar prospection over very large areas as part ofa cultural heritage strategy
15 Lidar in Mediterranean agricultural landscapes: reassessing land use in the Mauguio
16 Using lidar as part of a multi-sensor approach to archaeological survey and interpretation
17 Remotely acquired, not remotely sensed: using lidar as a field survey tool
18 Lidar survey in the Brú na Bóinne World Heritage Site
19 Immersive visualisation of survey and laser scanning: the case for using computer game engines
20 The practice of collaboration
Appendix: Key technical terms

Citation preview

Interpreting archaeological topography airborne laser scanning, 3D data and ground observation

edited by Rachel S. Opitz and David C. Cowley

Oxbow Books Oxford and Oakville

Published by Oxbow Books, Oxford, UK © Oxbow Books and the individual authors, 2013 ISBN 978-1-84217-516-3

This book is available direct from: Oxbow Books, Oxford, UK (Phone: 01865-241249; Fax: 01865-794449) and The David Brown Book Company PO Box 511, Oakville, CT 06779, USA (Phone: 860-945-9329; Fax: 860-945-9468) or from our website www.oxbowbooks.com

Front cover image: Single-direction hillshade of lidar coverage from the Hill of Ward, near Athboy, Co. Meath, Ireland. A large (130 m diameter) quadrivallate ringfort lies on the summit of the hill, while a number of other features, including outer banks, possible cairns, field boundaries and a deserted settlement have been recorded for the first time through analysis of the lidar data. Stephen Davis, UCD School of Archaeology. Data from the Ordnance Survey Ireland Back cover image (top): Colour ramp visualisation of lidar coverage from the Hill of Ward, extending to the east and west of the area shown on the front cover. Stephen Davis, UCD School of Archaeology. Data obtained from Ordnance Survey Ireland Back cover images (bottom): prehistoric clearance from the Slovenian Karst (left), terraced hillsides in Mediterranean Slovenia (centre), ridge and furrow fields Dravinja valley, Slovenia. Dimitrij Mlekuž

A CIP record for this book is available from the British Library

Printed and bound at Gomer Press, Llandysul, Wales

With the support of the Culture 2007–2013 Programme of the European Union (CU7-MULT7 Agreement Number 2010-1486 / 001-001). This work programme has been funded with support from the European Commission. This publication reflects the views of the authors, and the Commission cannot be held responsible for any use, which may be made of the information contained therein.

Foreword Aerial archaeological approaches are important in the study of sites, landscapes and regions, and are one of the major sources for discovering and understanding our past. The approach is well established, with a pedigree dating back over 60 years, though its application across Europe is not uniform. Increasingly new approaches and techniques are adding to the power of the aerial perspective, and one of the most exciting is the application of airborne laser scanning (ALS or lidar/LiDAR) to archaeology. Providing a means of recording and understanding topography from site to landscape scale and across a wide range of land use types from open agricultural land to forests, ALS is widely used in some parts of Europe, while in others its potential for archaeological research and Cultural Heritage Management has yet to be discovered. This variability in uptake lies at the core of the aims of the ArchaeoLandscapes Europe network, which seeks to promote the application of appropriate remote sensing techniques. The ArchaeoLandscapes Europe network was established to support the development of aerial approaches and remote sensing techniques across Europe, to encourage the exchange of expertise and skills, to foster cooperation between archaeological institutions, and to enhance public awareness. ArchaeoLandscapes Europe (ArcLand; http:// www.archaeolandscapes.eu) is supported by the EU under the framework of the Culture 2007–2013 programme (CU7-MULT7 Agreement Number 2010-1486/001-001). To date 58 partner institutions from 30 countries are working together to create a self-supporting network of expertise and to disseminate the methods and techniques of modern archaeological surveying to archaeological research institutions, cultural heritage management and to the general public. The aims of ArchaeoLandscapes will be achieved through eight key actions: 1. Creating an ultimately self-supporting ArchaeoLandscapes Network. 2. Using traditional and innovative methods to publicize the value of remote sensing techniques and landscape studies amongst all those who deal with cultural landscapes and heritage sites. 3. Promoting the pan-European exchange of people, skills and understanding through meetings, workshops, exchange

visits, placements and opportunities for specialist training and employment. 4. Enhancing the teaching of remote sensing and landscape studies through courses for students and teachers, and through a European Masters degree in Remote Sensing & Heritage Management. 5. Securing the better exploitation of existing air-photo archives across Europe and publicizing their potential for heritage interpretation and landscape conservation. 6. Providing support for aerial survey, remote sensing and landscape exploration in countries relatively new to their use. 7. Further exploring the uses of laser, satellite and other forms of remote sensing and web-based geographical systems in archaeological and landscape research, conservation and public education. 8. Providing technical guidance and advice on best practice in aerial survey, remote sensing and landscape studies, with a particular emphasis conservation and heritage management.

As part of our work for Actions 2 and 7 we aim to publish a number of high-quality articles and books representing the current state-of-the-art of remote sensing and other surveying techniques. It is therefore a pleasure to be able to support this book, which has been made possible by the work of Rachel Opitz and Dave Cowley, both partners of the ArcLand network. It includes contributions from some of the leading experts in the fields of (topographical) landscape interpretation and treatment of ALS data, and draws on long-established expertise as well as more recent approaches. We are sure that this volume will become a standard text for all interested in the technological aspects of topographic research, landscape understanding and remote sensing in archaeology. I am pleased to thank Rachel and Dave for drawing this volume together, and the contributors for their invaluable work. Frankfurt, 19 April 2012 Axel G. Posluschny Project Manager ArchaeoLandscapes Europe Roman-Germanic Commission of the German Archaeol­ogical Institute [email protected]

Contents Foreword List of Contributors Acknowledgements

iv vii xi

1 ������������������������������������������������������������������������������������������������������� Interpreting archaeological topography: lasers, 3D data, observation, visualisation and applications��� 1 Rachel S. Opitz and David C. Cowley 2 An overview of airborne and terrestrial laser scanning in archaeology Rachel S. Opitz

13

Section 1: Towards understanding landscapes – lidar in context 3 Airborne laser scanning and archaeological interpretation – bringing back the people Michael Doneus and Thomas Kühteiber

32

4 Cultivating the ‘wilderness’ – how lidar can improve archaeological landscape understanding Ole Risbøl

51

5

63

I Walked, I Saw, I Surveyed, but what did I see?...and what did I survey? Stratford Halliday

6 Reading aerial images Rog Palmer

76

7

88

Messy landscapes: lidar and the practices of landscaping Dimitrij Mlekuž

Section 2: Working with lidar and 3D data 8 Visualizations of lidar derived relief models Žiga Kokalj, Klemen Zakšek and Krištof Oštir

100

9

115

Worth a thousand words – Photogrammetry for archaeological 3D surveying Fabio Remondino

10 From lidar to LSCM: micro-topographies of archaeological finds Adrian A. Evans, Mhairi L. Maxwell and Gemma L. Cruickshanks

123

11 Using lidar data – drawing on 10 year’s experience at English Heritage Simon Crutchley

136

vi 12 Lidar and World Heritage Sites in Ireland: Why was such a rich data source gathered, how is it being utilised, and what lessons have been learned? Anthony Corns and Robert Shaw 13 The role of lidar intensity data in interpreting environmental and cultural archaeological landscapes Keith Challis and Andy J. Howard

Contents 146 161

Section 3: Making meaningful landscapes with lidar and being part of something bigger 14 The changing picture of archaeological landscapes: lidar prospection over very large areas as part of a cultural heritage strategy Ralf Hesse

171

15 Lidar in Mediterranean agricultural landscapes: reassessing land use in the Mauguio Nicolas Poirier, Rachel Opitz, Laure Nuninger, and Krištof Ostir

184

16 Using lidar as part of a multi-sensor approach to archaeological survey and interpretation Rebecca Bennett, Kate Welham, Ross A. Hill and Andrew Ford

197

17 Remotely acquired, not remotely sensed: using lidar as a field survey tool Stewart Ainsworth, Al Oswald and Dave Went

206

18 Lidar survey in the Brú na Bóinne World Heritage Site Stephen Davis, Conor Brady, William Megarry and Kevin Barton

223

19 Immersive visualisation of survey and laser scanning: the case for using computer game engines Keith Challis and Mark Kincey

238

20 The practice of collaboration Anthony Beck

252

Appendix: Key technical terms Rachel Opitz

266

List of Contributors The Editors Rachel Opitz Center for Advanced Spatial Technologies 304 JBHT University of Arkansas Fayetteville AR 72701 USA [email protected] Dave Cowley RCAHMS 16 Bernard terrace Edinburgh EH8 9NX UK [email protected] Contributors Stewart Ainsworth Department of History and Archaeology University of Chester Parkgate Road Chester CH1 4BJ UK [email protected] (formerly English Heritage) Kevin Barton Landscape and Geophysical Services Convent Road Claremorris County Mayo Ireland [email protected] Anthony Beck School of Computing University of Leeds LS2 9JT UK [email protected]

Rebecca Bennett School of Applied Sciences Bournemouth University Talbot Campus, Fern Barrow Poole BH12 5BB UK [email protected] Conor Brady Department of Humanities Dundalk Institute of Technology Dublin Road Dundalk Co. Louth Ireland [email protected] Keith Challis IBM Vista, Institute of Archaeology and Antiquity University of Birmingham Edgbaston Birmingham B15 2TT UK [email protected] Anthony Corns The Discovery Programme 63 Merrion Square Dublin, D2 Ireland [email protected] Gemma Cruickshanks Scottish History and Archaeology Department National Museum of Scotland Chambers Street Edinburgh EH1 1JF [email protected]

viii

List of Contributors Simon Crutchley Remote Sensing Heritage Protection Department English Heritage The Engine House Fire Fly Avenue Swindon SN2 2EH UK [email protected] Stephen Davis UCD School of Archaeology Newman Building Belfield Dublin 4 Ireland [email protected] Michael Doneus Department for Prehistoric and Medieval Archaeology University of Vienna/ LBI for Archaeological Prospection and Virtual Archaeology University of Vienna Franz-Kleingasse 1 A-1190 Vienna Austria [email protected] Adrian Evans Archaeological Sciences School of Life Sciences University of Bradford UK [email protected] Andrew Ford School of Applied Sciences Bournemouth University Talbot Campus, Fern Barrow Poole BH12 5BB UK Strat Halliday Rath Manach Romanno Bridge West Linton EH46 7BY UK [email protected]

Ralf Hesse State Office for Cultural Heritage Baden-Württemberg Berliner Strasse 12 73728 Esslingen am Neckar Germany [email protected] Ross Hill School of Applied Sciences Bournemouth University Talbot Campus, Fern Barrow Poole BH12 5BB UK Andy Howard Institute of Archaeology and Antiquity University of Birmingham Edgbaston Birmingham B15 2TT UK [email protected] Mark Kincey Department of Geography University of Durham South Road Durham DH1 3LE UK [email protected] Žiga Kokalj Institute of Anthropological and Spatial Studies/ ModeLTER Scientific Research Centre of the Slovenian Academy of Sciences and Arts/Centre of Excellence for Space Sciences and Technologies Novi trg 2, SI - 1000 Ljubljana Slovenia [email protected] Thomas Kühteiber Institute for Medieval and Early Modern Material Culture Austrian Academy of Sciences Körnermarkt 13 A-3500 Krems Austria [email protected] Mhairi Maxwell Archaeological Sciences School of Life Sciences University of Bradford UK [email protected]

List of Contributors William Megarry UCD School of Archaeology Newman Building Belfield Dublin 4 Ireland [email protected] Dimitrij Mlekuž Department of Archaeology University of Ljubljana Askerceva 2 P.O. 580 SI-1001 Ljubljana Slovenia [email protected] Laure Nuninger L CNRS – UMR 6249 Chrono-Environnement Université de Franche-Comté France Krištof Oštir Institute of Anthropological and Spatial Studies/ ModeLTER Scientific Research Centre of the Slovenian Academy of Sciences and Arts/Centre of Excellence for Space Sciences and Technologies Novi trg 2, SI - 1000 Ljubljana Slovenia [email protected] Al Oswald Department of Archaeology University of York The King’s Manor York YO1 7EP UK [email protected] (formerly English Heritage) Rog Palmer Air Photo Services 21 Gunhild Way Cambridge CB1 8QZ UK [email protected]

ix Ole Risbøl NIKU – Norwegian Institute for Cultural Heritage Research Storgata 2 P.O. Box 736 Sentrum 0105 Oslo Norway [email protected] Nicolas Poirier N CNRS – USR 3124 MSHE C. N. Ledoux Besançon France Fabio Remondino 3D Optical Metrology Unit Bruno Kessler Foundation (FBK) Trento Italy [email protected] Robert Shaw The Discovery Programme 63 Merrion Square Dublin, D2 Ireland [email protected] Kate Welham School of Applied Sciences Bournemouth University Talbot Campus, Fern Barrow Poole BH12 5BB UK Dave Went English Heritage 37 Tanner Row York YO1 6WP UK [email protected] Klemen Zakšek Institute of Geophysics University of Hamburg/Centre of Excellence for Space Sciences and Technologies Bundesstrasse 55 20146 Hamburg Germany [email protected]

Acknowledgements The origins of this volume lie in several conferences, a workshop and informal contacts over several years. Thus, it will be no surprise that it owes a debt to a large number of people who have contributed in a whole variety of ways. First and foremost our thanks to the contributors who have produced stimulating papers within a tight timescale, inevitably fitted in around other pressing work. Some of the contributors have also helped us in undertaking peer review, principally Anthony Beck, Keith Challis, Anthony Corns, Simon Crutchley, Strat Halliday, Ralf Hesse, Žiga Kokalj, Rog Palmer and Fabio Remondino. For assistance in peer review and other ways we are grateful to David Edwards, Damian Evans, Sorin Hermon, Pete Horne, Fraser Hunter, David Kennedy, Fred Limp and Billy MacRae. Our thanks also to Julie Blackmore, Julie Gardiner, Lizzie Holiday and Val Lamb at Oxbow for their support. The Training and Research in the Archaeological Interpretation of Lidar Workshops held at the European Research Centre at Mont Beuvray, organized with generous support from the BQR PRES MuTER and in cooperation with the MSHE Ledoux and the Bibracte Établissement Public de Coopération Culturelle, brought many of the volume’s contributors together and highlighted the growing importance of the use of lidar within the archaeological community. AARG conferences have provided a forum

for discussions and ours thanks to the various organising committees and delegates for their contributions. The central role of AARG is reflected in the identification of this volume as an Occasional Publication of the Aerial Archaeology Research Group. Fundamentally, this volume draws on informal networks of professionals from various backgrounds, who bring a variety of experience, perspectives and specialisms to bear. We hope this provides a thought-provoking mix that will inspire reflection on practice, both in the specific applications of laser scanning, but also in the wider arena of landscape archaeology. The ArchaeoLandscapes Europe (ArcLand) network is a more formal manifestation of these connections across Europe and we are especially grateful for the financial support of this European Union funded multi-partner project in defraying publication costs (see Foreword); we are especially grateful to Axel Posluschny (ArcLand Project Manager) for his enthusiasm for the volume. Central to the aims of ArcLand is promoting the application of appropriate remote sensing techniques to archaeological research and Cultural Heritage management. We hope this collection of papers will support this aim by providing a multiplicity of views and examples of best practice in a range of contexts that will encourage applications that are best fitted to the particular environment or archaeological research issue. Rachel Opitz. Fayetteville, Arkansas, USA Dave Cowley. Lamington, Lanarkshire, UK

May 2012

1 Interpreting archaeological topography: lasers, 3D data, observation, visualisation and applications Rachel S. Opitz and David C. Cowley The central issues of this volume are introduced, describing laser scanning and 3D data for archaeology. Archaeological interpretations of topography and the skills sets required are discussed in the context of rapidly changing approaches and the need to integrate field experience and computer aided analysis. An outline for understanding and working with ALS ranges across issues of scale, certainty of interpretation, processing, visualisation and integration, concluding with thoughts on the impact of regional research traditions and prospects for the future. Keywords: 3D data, ALS, experience, archaeological topography, scale, integration, research traditions

Preamble Airborne Laser Scanning (ALS), or lidar (terms used interchangeably throughout this book), has been described as one of the most important innovations in data collection and interpretation for archaeology (Bewley et al. 2005), and it is the principle focus of the volume. However, the themes, approaches and methods discussed in this volume are broadly applicable to laser scanning, close-range photogrammetry and other 3D data collection methods in use in archaeology, and these too are included. Laser scanning is a technology which accurately and repeatedly measures distance, based on a precise measurement of time, and combines these measurements into a collection of coordinates. These coordinates are normally stored as a point cloud, from which information on the morphology of the object being scanned may be derived. Photogrammetric approaches also produce 3D point clouds describing the shape of an object based on the triangulation of matched points from multiple images, and many of the applications and processes overlap with those discussed in this volume in the context of laser

scanning. Objects recorded with these techniques can range from a small artefact recorded by a triangulation scanner in a laboratory, to vast extents of landscape totalling many 1000s of km2 recorded from an airborne platform. The scope of what may be recorded and studied through these techniques, from objects to surface structures to buried archaeology, traversing spatial scales, and the range of questions which may be approached, has led to their adoption across the archaeological community. These specialist techniques are now an integral part of many academic, popular and heritage management projects. In particular, the last decade has seen an exponential growth in the use and awareness of ALS by archaeologists and cultural resource managers. The ‘magic’ of a new technology with the ground recorded in detail, the ability to ‘see through’ trees and the powerful images produced, all promised a brave new world. And so it is – a world of possibilities and challenges, both in ensuring appropriate, archaeologically reliable applications that inform us about the past, but also in developing practices that integrate the strengths of new possibilities in manipulation



Rachel S. Opitz and David C. Cowley

Figure 1.1: A relief shaded digital surface model (DSM) generated from an extract of high resolution Flimap 400 lidar survey flown by helicopter at the deserted medieval settlement of Newtown Jerpoint, Co. Kilkenny, Ireland. Remarkably fine detail is present in the model, the result of a ground sampling distance for the survey of 15 cm. Centre left, tyre marks from vehicles can be seen converging at a break in the field boundary; returns from power lines can be seen running diagonally across the model; and towards the bottom right sheep can be seen grazing in the field. This image records features with minimal surface expression, for example rendering field surfaces in remarkable detail. However, these ephemeral landscape features are mixed in with textures that are a product of data collection and processing such as the slight ‘grain’ that runs from top to bottom across some areas. Interpreting such an image benefits from an ability to manipulate lighting (at the very least), an understanding of how it was created, and experience built on field observation. © The Discovery Programme/Heritage Council

and interrogation of vast digital datasets with so-called ‘traditional’ skills of archaeological observation and interpretation (Figure 1.1). This theme is at the core of this volume, by drawing on 10 years of archaeological engagement with ALS to explore the technical and interpretational challenges of this data, and to address the integration of approaches drawn from direct observation, data manipulation and visualisation, and to move beyond the purely technical or observational to engagements with past lives. The use of archaeological topography

in the title of this volume identifies a focus on the topographic expressions of the past in the present, whether the earthworks (humps and bumps) of past settlements and field systems at landscape scale or the micro-topography of tool marks on an artefact. And it is this manifestation of past activities and processes in 3D, whether at micro- or macro-scale, that creates an exciting challenge in combining approaches from a wide variety of archaeological practice. To evoke an obvious generalisation, these approaches range from the muddy-booted fieldworker engrossed

1  Interpreting archaeological topography: lasers, 3D data, observation, visualisation and applications in the topography of a hillside to the computer geek sitting in a darkened room surrounded by humming hardware and writing complex software to create a virtual environment! Of course these are polar extremes, but they do highlight the importance of combining ‘field-craft’ and observation with the powerful algorithms and visualisation techniques that dense and/or extensive 3D data demand if we are to do anything more than scratch the surface. This volume draws together expert papers from across a broad spectrum of engagement with archaeological topography as expressions of developing practice in a rapidly evolving field. This introductory essay provides some background and introduces the main themes of the volume. Principally these are the growing archaeological applications of 3D data, for which laser scanning is now a major source, and how these relate to the interpretation of topography. ALS is not straightforward ‘data’ and its incorporation into routine practice demands a level of understanding and critical thinking, which range across scale of analysis, certainty of interpretation, the roles of processing and visualisation, integration and varying uptake and regional traditions. The essay concludes with some comments on prospects for the future and a brief description of the contents and origins of this volume. The emphasis throughout is on the underlying principles, rather than technical descriptions or issues. For an overview of the technical aspects of laser scanning see Chapter 2 (Opitz) and attention is also drawn to the glossary of key technical terms on p. 266.

3D data in archaeology Digital 3D data is now entirely embedded in archaeological recording, interpretation and visualisation, within a wide variety of projects and at a variety of scales. ALS has found applications in mapping and prospection surveys in woodland, scrub and open ground, and may provide the only means of survey in difficult to access areas. While basic mapping may draw on relatively low-resolution, usually second-hand, data, the potential of high point densities to provide incredibly detailed recording of small landscapes and individual sites has been demonstrated. ALS has shown its value in providing landscape context, drawing on geomorphological mapping of palaeo-features

and landscape characterization, and informing conservation studies focused on the impacts of erosion or modern land use. On the site scale, terrestrial laser scanners (TLS) are used to collect bespoke data for specific archaeological projects. Such projects may include recording a site or monument before excavation or conservation work takes place, detailed documentation during excavation of complex or easily damaged features, or scanning of highly eroded or abraded surfaces to highlight subtle features. Like ALS, this type of recording has immediate primary applications in erosion and stability monitoring, particularly of monuments exposed to weather, pollution and tourists. With even smaller scale scanners used in laboratories the same processes of documentation and applications apply in equal measure to small objects, and are increasingly part of the study of the manufacture and use of artefacts. In all cases the integration of 3D data into archaeological practices promotes the use of ever more sophisticated modelling and visualisations, from the creation of virtual replicas for display in a physical or digital museum or dissemination over the internet, to virtual reality and immersive visualization projects. Throughout, while the primary aim of these products may be to communicate and engage with a wide audience, these approaches also have a vital role for the investigating archaeologist in supporting interpretation where the visualization and measurement of very small scale and subtle features is essential (e.g. tool marks or rock art), and to under-pin spatial analyses such as viewsheds and least cost-paths, and inclusion in interactive virtual reality models. Universally, it is the use of 3D data as an articulation of archaeological topography that lies at the heart of the processes.

Archaeological topography Topographic survey, in the first instance the interpretation and survey of archaeological earthworks, is a long-established tradition with antiquarian origins (e.g. Bowden 1999). Early cartographers experimented with depictions of slopes, developing through military surveys depictive techniques like hachures and shading that were used to denote natural and anthropogenic earthworks. This approach has a rich history as an archaeological technique for





Rachel S. Opitz and David C. Cowley documenting whole swathes of landscape, as a means to think through, understand, record and communicate sites like Iron Age earthworks, deserted medieval villages, and, on a grand scale, ancient Rome. Topographic surveys at a range of scales create coherence and aid in understanding the features surveyed, providing a plan view, but also attempting to convey something of the topography of the site/landscape. This latter aspect, of conveying topography, has varied in the degree to which conventions and depiction have successfully allowed the viewer to interpret slope, but is of course inherent in contours and other expressions of height differences such as shaded relief models. For archaeological survey the collection of 3D data was, less than 20 years ago, a time-consuming process, and while for map-making the innovation of photogrammetric pairs of aerial images was a major step-forward for quickly capturing the form of large areas, its uptake by archaeologists has been uneven. The last ten years have witnessed an increase in pace and intensity in archaeological engagement with 3D data, ranging from data collection by total station and DGPS through the increasing use of terrestrial laser scanning and close range photogrammetry. In tandem, huge improvements in 3D modelling software are rapidly changing how topographic documentation is undertaken on excavations. While these developments encompass the full range of archaeological recording, from close detail in an excavation trench, or within a building, to extensive landscapes even at country-scale, these advances in 3D recording primarily impacted practice on the site to object scale and the links with archaeological topography in the traditional landscape sense were not emphasized. This is the mix into which ALS has been added, greatly advancing the ability to collect and work with 3D models of large areas. The popularity of ALS for studying forested areas, floodplains and rural areas in general has renewed interest in the topic of topographic survey, and further spurred integration with digital technologies and applications. The potential of these synergies demands critical examination of working practices, especially in the area of generating archaeologically meaningful and stimulating interpretations of topography and in revitalizing the use of topographic data in landscape projects. To illustrate our views on the intersection of 3D data and the practices of recording, visual

depiction and interpretation a brief commentary on the use of hachures and shading in traditional approaches is instructive. These conventions have been used to depict slopes and as a means of recording and communicating the results of an analytical engagement with earthworks – What are the humps and bumps? How do they interrelate? How do they express structures from the past? and so on. Here, the processes of archaeological interpretation and depiction are intertwined. Thus, while the results of such analytical site survey should be metrically accurate, the depiction is a product of the interpretative engagement of the surveyor with the earthworks, and the translation of their interpretation into a drawing. This is a process that is heavily dependent on experience, a sound knowledge-base and a reflexive, self critical approach. The site of Braidwood in southern Scotland is a good example, where the subtle, incredibly complex earthworks are a product of a sequence of construction of timber round houses and palisaded and earthwork enclosures (Gannon 1999). The survey drawing (Figure 1.2) is a result of about three days in the field and an intense engagement with the humps and bumps of the site, what they might mean and how to translate that into a meaningful plan – undertaken by two highly experienced fieldworkers with many years of experience between them. This has produced a plan which expresses their observations and can be ‘read’ – a plan in which the observation and interpretation of the earthwork remains are explicit and completely intermeshed (Figure 1.3). The plan of Braidwood makes an important point – that the interpretation of archaeological earthworks (or natural topography for that matter) is a skill built on experience and knowledge, where intuition and subjective judgements are very much to the fore. This may seem, at first glance, old-fashioned and irrelevant to the new reality of digital survey data, where height data has often replaced depictive survey and digital drawing packages have taken the place of the draughtsman’s pencil. However, while ALS is providing digital surface models at a scale and level of detail that would have been unimaginable 20 years ago, it also presents considerable methodological and interpretative challenges that relate to the nature of the dataset and how it is processed, manipulated and used to generate archaeological information. Many of these require new approaches rooted in

1  Interpreting archaeological topography: lasers, 3D data, observation, visualisation and applications

 Figure 1.2: A masterful example of an earthwork survey at Braidwood in southern Scotland produced by Angela Gannon and Strat Halliday. The complex palimpsest of ephemeral earthworks has been examined in detail building an understanding that has been translated into an analytical drawing. This process is a complex interplay of fine-tuned observation, experience and drawing skills. © Angela Gannon, reproduced by kind permission

Figure 1.3: The slight earthworks on Gibbs Hill in southern Scotland are deciphered by a team and translated into an interpretative analytical drawing similar to Figure 1.2.The people are standing between two of the shallow trenches which once held timber palisades. © Strat Halliday, reproduced by kind permission



Rachel S. Opitz and David C. Cowley processing algorithms and visualisation software and new skills in digital data manipulation to go along with them, but others take us back to the skill-set that produced the plan of Braidwood or that have been honed examining aerial photographs. We would argue that in engaging with digital 3D data the skills of reading the topography and the employment of experience and knowledge are still very much at the fore.

Understanding ALS As archaeological use of ALS has developed it has become increasingly clear that it is not an ‘objective’ dataset that can be used uncritically. Indeed terrain models, like any models, are constructs and often riddled with unspoken assumptions. Firstly, the primary data collection and processing parameters have a major impact on output, while the ability to ‘see’ is heavily dependent on software for manipulation and visualisation, and data artefacts may be a trap for the unwary. These factors are a complex mix of objective parameters (e.g. point density) and subjective judgements (‘this visualisation looks better than that’) that are inextricable from the pervasive issue of archaeological interpretation. So, for all the new technology and software, the basic issues of how archaeological information are created remain central. While fundamentally we see the use of ALS and other 3D digital data sources as a continuation of existing practices, the specifics of how we work with these technologies throw up some important differences of approach which impact on established survey practices and workflows. Firstly, the emphasis of much ALS work in archaeology has been desk-based, with limited engagement in concurrent or subsequent ground observation. A purely desk-based approach carries certain dangers – principally that there may be no or limited feedback between ground observation and ALS interpretation. Thus knowledge of the site types that may be expected in an area and artefacts created by modern landuse, for example, may not inform the desk-based work (processing – manipulation – visualisation – interpretation) as it should. Lack of this basic type of knowledge of a landscape is the main factor in the misinterpretation of aerial photographs (e.g. Wilson 2000) and the same will certainly be true for ALS. A related point concerns the interplay of manipulation,

visualisation and interpretation, and at a basic level the role of archaeologist and ALS specialist – two roles which, in current practice, usually do not overlap much. Many ALS data are used by archaeologists who have little understanding of the processes by which it has been generated (e.g. a hillshade model), while ALS data may be processed with little or no consideration of archaeological imperatives (inevitable if ALS data is ‘second-hand’). Such divisions are not desirable and best practice projects have developed a synergy of these different skill-sets where the ability to manipulate data interacts with knowledge of the archaeological landscape. Addressing scale and certainty One of the major challenges for archaeological uses of ALS is how to work at an extensive scale, with potentially huge and complex datasets (this also applies to multi/hyper-spectral data (Beck 2011) and to datasets collected over smaller physical areas but with very high spatial or temporal detail). This problem is particularly relevant to areas without good archaeological databases that can support effective heritage management, where lidar may be a key (or the sole) source of new archaeological information and much depends on its interpretation. Archaeologists working in a research context may choose to study relatively small areas for which the lidar and other archaeological and supporting information can be inspected in detail. However, for cultural resource managers using lidar to help set planning priorities ahead of development across large areas, particularly where the overall archaeological record is poor or variable in coverage and quality, a strong dependence on lidar is problematic as a close inspection of all areas of the dataset is impractical and only studying parts of the dataset will not support prioritisation and protection. To address the challenges of fairly assessing large areas given limited resources, and to enable Baden-Württemberg (35,751 km2) to be mapped in six years, Ralf Hesse has developed as many automated data processing stages as possible (Hesse this volume). Thus local relief models (LRM, Hesse 2010) were developed as a technique to extract small-scale ‘local relief ’ features from the digital elevation model (i.e. archaeological earthworks usually have low relief relative to the rest of the landscape), and if data with these characteristics can be rapidly extracted there will be a massive timesaving over visual inspection.

1  Interpreting archaeological topography: lasers, 3D data, observation, visualisation and applications This approach may ring alarm-bells with some archaeologists, but it cuts straight to the heart of the problems of dealing with complex, dense or extensive datasets, whether they are aerial/satellite photographs, hyper/multi-spectral data or ALS. If archaeologists and heritage managers are going to make effective use of these data for extensive survey to underpin effective management, then they must engage with techniques that shortcut exclusively ‘manual’ inspection, in the manner of, for example, traditional aerial photograph interpretation – and that means auto-extraction, or more accurately, semi-automated extraction of features (Cowley 2012). No-one really advocates ‘fully automatic’ extraction, and in practice workers mean semi-automatic or supervised feature extraction (i.e. De Laet et al. 2007, 2008, 2009; Trier and Pilø 2012). This bears directly on the type of information that may be expected and the degree of certainty that will attach to interpretations. In the vast area covered in Baden-Württemberg (above and Ch 14) the approach will generate many potential archaeological features, whose character and even certainty of being ‘real’ are provisional. Within a development control context, where an archaeological flag will identify a location that requires inspection if it is threatened, this is a valuable enhancement of the dataset for cultural resource management. Alternatively, high-point densities, integration with other data sets and ground observation will generate deep information with a high degree of interpretative integrity. Variable scales of reliability or certainty are inevitable, but so long as fitness for purpose is identified from the outset, weaknesses in the survey product should be minimised. An acceptance that some interpretations made relying on semi-automatic techniques will be proved incorrect is a necessary evolution of management practice. In essence, there is a trade-off between a decrease in the confidence of interpretations and an increase in the area studied. Accepting this trade-off is absolutely central to increasing our ability to address management needs and research questions at the regional and supra-regional scale. Processing and visualisation Processing and visualization have real and important impacts on interpretations. Anyone who has begun to ‘play’ with different visual­ izations and terrain model processing techniques quickly realizes that features of potential interest

can dramatically change depending on the settings and algorithms used. There are near endless possibilities for the creation of new models and visualizations of these models, and a distinct danger of continually tweaking parameters in the hope of ‘improving’ the appearance. The many possibilities for manipulating digital models forces consideration of two points. First, there is the question of how much information can be retrieved, which is somewhat like asking how long a piece of string is, but is fundamentally about how to assess cost/benefit. Crutchley (this volume) makes an admittedly entirely subjective observation that if one model gives 90% of the nominal ‘total’ information, then the decision not to chase the other 10% may be taken on cost-effectiveness, depending on the requirements and constraints of a project. At the heart of this assertion is a pragmatic approach to avoiding the dangers of loosing sight of survey objectives in an endless round of data processing and manipulation. It is however, an area that requires structured assessment, perhaps to create benchmarks of cost/benefit that allow users to decide where on a sliding scale they wish to curtail manipulation of data, because a certain approach gives them enough to be fit for purpose. Second, as noted by Kokalj et al. and Beck (this volume) providing detailed information on how a model and visualization was created is essential for others to understand and evaluate the end product and interpretation. Is a feature really present or is it simply a digital artefact or trick of the virtual light? As noted above, close collaboration and sharing of knowledge between the ALS specialist and the archaeological interpreter (as two people or one person performing two distinct tasks) will enhance the links between manipulation of digital models and the interpretations based on them. Documentation of the process is vital, and linking directly to the first point, including the basis on which ‘when to stop’ is determined – i.e. when ‘enough’ information and an appropriate level of confidence in that information have been achieved. A clear definition of fitness for purpose and survey design are central to this decision-making process. Context and integration When integrating a new(-ish) data source into archaeological practice there are two modes of





Rachel S. Opitz and David C. Cowley engagement. On one level there is informal integration wherein the data is used in an increasing number of projects in innovative ways to engage with a growing variety of questions, coupled with a growing understanding of how the new data source interacts with existing techniques. On another level we have formal integration where recognized standards and archives are adjusted to accommodate the new data source and best practices are developed and recognized. ALS, TLS and 3D recording from photogrammetry in archaeology have made important strides in both modes. While early ALS and 3D recording publications often focused on the technology itself, showcasing benefits and pitfalls (e.g. see references throughout the volume), the recent trend is for publications of case studies in which 3D recording is a well-integrated part of a collection of data supporting research into specific archaeological questions (Crutchley, Poirier et al., Bennett et al., Davis et al. this volume). The variety of projects using 3D data has increased, and now ranges from close studies of an object (Evans et al. this volume) or single site to broad regional surveys (Hesse this volume) and from those collecting fundamental basic data for regional surveys (Risbøl this volume) to those using ALS to study the links between manuring patterns and field systems (Poirier et al. this volume). Formal guidelines for ALS and 3D recording have appeared in the last five years, including the ADS Guides to Good Practice and English Heritage Guides to 3D Laser Scanning, which provide important documentation of current practices, advice and standards for these data. These guides and standards are most widely applied in the Anglophone community, primarily the UK where they have been developed, but similar initiatives elsewhere are developing. Advances in the adoption of ALS, TLS and photogrammetry by archaeologists are closely tied to future developments as well as past ones. The need to efficiently manage large 3D datasets, to more closely tie together technical and interpretational skills, and to link together methodological and regional or local archaeologically driven advances will push forward the agenda in areas like the use of webbased platforms, data sharing and metadata standards, and research in topics like the archaeology of forested environments.

Uptake and regional traditions Differential uptake of ALS reflects many things, from the history of aerial approaches in a region to restrictions on flying, from the prevalence of non-archaeological agencies collecting and using lidar to the character of archaeological sites/ monuments/landscapes and local or regional research agendas. The content of this volume reveals the dominance, both in the number of projects and length of engagement, of Northern and Continental Europe in archaeological lidar. More recently a few Mediterranean European countries and a limited number of North and Central American (but run by North American research consortiums and Universities) projects have emerged. The first Asian archaeological lidar project is planned for 2012 in Cambodia under the aegis of a consortium of international researchers and heritage managers (below). As far as the authors know, currently there are no archaeological lidar projects based in South America, Africa, or the Middle East. While in places such as Cyprus or much of the Middle East the absence of ALS projects is clearly due to restrictions on airspace and the practical difficulties of finding data collectors who operate locally, the tradition of the use of aerial imagery, especially amongst cultural resource managers, is a major factor in the adoption of ALS. Such data for large areas is often collected and held by a government environmental or land management agency, and other government groups including those responsible for heritage management are typically the first to be made aware of and gain access to the data. This can be seen, for example, in the UK and Netherlands, where data collected by Environmental Agencies for flood risk mapping and monitoring sparked the widespread adoption of ALS by the archaeological community. Whether state-funded archaeologists seek access will depend to some extent on whether they believe that aerial approaches are inherently useful, and the degree to which primary aerial reconnaissance and mapping is an established part of cultural resource management strategies. Adoption of ALS for research has generally followed once it is embedded in national archaeological management in some way, in part because the cost of purpose-collected survey generally will be beyond most researchers. However, until 2012 EUFAR (European Facility for Airborne Research) and the NERC

1  Interpreting archaeological topography: lasers, 3D data, observation, visualisation and applications ARSF (Natural Environment Research Council Airborne Research and Survey Facility) made it possible for a limited number of European and UK research projects with a proven commitment to ALS to acquire such directly-purposed data and these models may in time extend elsewhere. On the other hand, there are examples of archaeologically-commissioned projects which pioneer new uses or techniques for the data in regions without, or preceding, a broad CRM driven initiative for using ALS (e.g. Italy where research-driven applications have preceded widespread release and use of the Il Piano Straordinario di Telerilevamento Ambientale (PSTA), part of the national mapping). In either case the constant interchange between CRM and research communities as they work with ALS and other forms of 3D recording is essential to its continued development and increased use. This exchange of experience and knowledge can take place across disciplines on many scales. Early adopters of ALS working in Mediterranean Europe (e.g. Italy and Spain), where significant regional surveys are in their early stages and a number of research studies have been undertaken, are broadly part of the same research and CRM tradition as those working in Central and Northern Europe. Consequently there has been extensive communication and collaboration facilitated by organizations like AARG (Aerial Archaeology Research Group), EARSeL (European Association of Remote Sensing Laboratories) and Archaeolandscapes Europe and a level of standardization of data and metadata driven by the Inspire Directive. The rapid development and early work in these areas no doubt benefited from exchanges with colleagues in other areas of Europe, and is now making important contributions to the research community as researchers in these regions confront very different archaeological and land use conditions. The growing community of archaeologists using ALS in North and Central America, in contrast, has fewer opportunities (and perhaps fewer motivations given the historic separation of New and Old World archaeologies) to collaborate with European colleagues. Real, and in some cases perceived, differences between the state of preservation of surface and buried features, the types of sites dominating the survey record, the traditions of use of aerial imagery, and the dominant types of vegetation between New and Old World archaeologies

makes the transfer of relevant knowledge and experience more complicated (though see Burks 2010 for an American approach that will be familiar to many European workers). In spite of differences, it is hoped that exchanges between these communities will increase as ALS becomes a greater part of New World archaeological research and CRM practice, and that similar developments will occur in other regional archaeologies as appropriate.

Prospect Informal integration of ALS and other 3D digital data sources for archaeological recording has been taking place for ten years, and is spreading to new areas of archaeological practice. These processes can be challenging as they demand development of new approaches and work practices, and the ability to draw together complex data sources and interpretative frameworks. To maximise the added value of these processes for archaeology an open-minded approach is vital – one which is not constrained by demarcation of roles or expertise, but equally recognises specialist skills and archaeologically-based experience. In looking to the future, the most exciting developments will certainly be those that are not anticipated, but several areas of important further development can be identified. Developing and disseminating best practice is crucial and the ALS community already has an excellent track record in this area. Finding ways to include these usually very large and complex datasets in archives available to the wider research community is important to facilitate re-tasking of datasets, a topic discussed by Corns and Shaw and by Beck (this volume). The transfer of skills and experience can be a valuable short-cut to speed development, but should be progressed without assuming that problems will necessarily be uniform and best practices the same across regional disciplines. While lidar has been established as a basic part of the CRM process in regions where the data is widely available, and bespoke collections are justified in some management situations, now that the first set of technical challenges has been met its place in research projects is less clear and should certainly not be uncritically assumed (‘let’s do lidar’). Finding creative research-driven uses of 3D digital data will be a key part of the future of ALS and of similar data derived from other sources.



10

Rachel S. Opitz and David C. Cowley Technical developments will likely come from outside archaeology. The sheer size of a typical 3D dataset can put it beyond the resources of many, but the decreasing cost of digital storage and improved efficiency of processing software will make the technology more widely available. Archaeological applications will often demand classification and filtering processes beyond that which is delivered by standard software solutions (above; Opitz this volume), and improvements in these algorithms will also likely come from the wider ALS community. On the other hand, the archaeological community has taken the initiative in developing or adapting visualization techniques and 3D data management strategies and improvements in these areas will probably continue to come from within archaeology. The challenge of developing research agendas where 3D data will be useful and open new avenues will clearly have to come from within the archaeological and cultural heritage fields. A project underway in Cambodia as this volume is going to press encapsulates many of the issues discussed above and we are grateful to project director Dr Damian Evans (University of Sydney, Australia) for the following information, which is included here as a timely illustration of the many challenges of developing ALS use, the importance of research design and fitting to purpose. A consortium of eight teams has acquired ALS coverage over Angkor and two other areas, in an extension of work carried out over 20 years to map and analyse remains such as rice field walls, occupation mounds, temple sites, ponds, roadways and canals (Evans et al. 2007; Evans and Traviglia 2012). The project has addressed many issues, not least the bureaucracy of obtaining permissions from three different ministries in a process that took four months and went to Prime Ministerial level. Costs are significant and potential suppliers limited, while the constantly cloudy and/or hazy conditions and unpredictable weather are an added difficulty for planning. Fundamentally, however, the major obstacle to the project was the lack of precedent and the fear of gambling on an approach that has never been proven in the region. The project team hope that the Angkor mission will settle many of these issues and provide impetus for further missions in the region as there is enormous potential.

Origins and structure of this volume Origins This volume is rooted in discussions and presentations at the annual conference of the Aerial Archaeology Research Group (AARG) in 2010 and an international lidar workshop held at the Bibracte European Research Centre in 2011. The profile of lidar at AARG meetings has built steadily over the last few years, with a few papers in 2006 and 2007, followed by a full session at Ljubljana in 2008 in which the emphasis was very much on potential, and an extended session in Bucharest in 2010, with a strong emphasis on interpretation and a loose link to a session on more general issues of interpreting aerial remote sensed data. The work presented and issues raised at conferences very much set the tone for TRAIL 2011 (Training and Research in the Archaeological Interpretation of Lidar) held at Bibracte in March 2011. This drew together experts, practitioners and novices in a workshop framework, which ranged across hands-on training, workshops and general presentations. An extended lidar-themed session and keynote address at AARG 2011 in Poznań, Poland, in September 2011 continued this trajectory, with papers mainly concerned with ALS employed in an integrated landscape framework and the issues of interpretation. The change in emphasis from ALS as the new toy to serious applications and critical thinking which can be traced in this timeline represents a maturing understanding of the roles of ALS in archaeological prospection, interpretation and landscape archaeology. As the volume was developed in 2011, further papers were commissioned to expand the scope of the book to include closely related topics in 3D data and terrestrial-based survey. In general terms this timeline tracks an evolution of thinking from early presentations where the stress was partly on ‘look what you could do with this’ to a more recent emphasis on project results and reflections on practice and integration. It is this intellectual context along with the realisation of connections between various trends in 3D data that stimulated this volume. Structure The volume is divided into three main sections, following on from this introductory essay and a technical overview of airborne and terrestrial laser scanning in archaeology. Opitz’s overview

1  Interpreting archaeological topography: lasers, 3D data, observation, visualisation and applications provides background information for topics covered in the volume in recognition that some understanding of the processes of data collection and processing will help archaeologists to make more informed use of the data. The first group of papers (Towards understanding landscapes – lidar in context) provides a variety of perspectives on interpreting the landscape, from those rooted in ALS datasets (Doneus, Risbøl and Mlekuž) to discursive papers on field experience and observation and interpretation of aerial images (Halliday and Palmer respectively). In all cases there is an emphasis on understanding of, and critical reflection on, the source material and how to make best sense of it, drawing as appropriate on complementary information within integrated, holistic views of the landscape. In the following section (Working with lidar and 3D data) the emphasis shifts to an examination of processing, visualisation and manipulation of data. With lidar and other 3D datasets so dependent on processing and visualizations Kokalj et al. critically assess options and implications of different approaches for 3D datasets. Remondino’s paper reviews the state of the art of photogrammetry for archaeological survey, identifying advances in hardware and software that are giving this ‘old’ technology a new relevance as a source of 3D data. While multiple scales of analysis are implicit in many papers in this volume, Evans et al. explore microtopographies of objects in a study that highlights how Laser Scanning Confocal Microscopy of artefact surfaces is driving analyses of biography and identity. The theme of fitting technology and datasets to purpose is continued by Crutchley, who describes a variety of uses of lidar data by English Heritage. The majority of archaeological uses of ALS will, inevitably, be of ‘second hand’ datasets, collected for other purposes such as mapping and environmental modelling. This is likely to remain the case and the ‘formation’ of the rich lidar datasets for Ireland from a variety of sources is discussed by Corns and Shaw. This reflection and lessons learned from Ireland identifies issues such as managing expectations, how collaborations work and the potential for re-use of data. In the concluding paper Challis and Howard discuss the potential of intensity images (i.e. visualisations of the amplitude of returned laser pulses), a component of lidar data that is often ignored. The final section (Making meaning ful landscapes with lidar and being part of something

bigger) contains papers that explore the use of ALS in research projects and cultural heritage management. Collectively these highlight the diverse applications of ALS, and the need to consider survey design and intended purpose. Scale is a crucial consideration of Hesse’s paper on large-area prospection in Baden-Württemberg, which addresses head-on one of the major challenges of working efficiently with massive datasets, principally in the development of rapid ways of extracting information (algorithms) and the recognition that ‘certainty’ of interpretation will change with scale of analysis. The role of this data in cultural resource management strategies is discussed in an approach that should have wide-ranging implications. A completely different, heavily cultivated and open, landscape in the Mauguio of Southern France is discussed by Poirier et al. in a study that integrates historical maps, field walking survey and aerial photographs. Integration and identifying complementarities are central to Bennett et al., who discuss developments in airborne multisensor survey with the aim of promoting the full information content of topographic and spectral data to heritage professionals. Defining purpose and making appropriate, informed use of ALS are recurrent themes in papers throughout the volume, and are central to the contribution by Ainsworth et al. describing their experience using ‘ortholidar’ as a ground-based method for rapid survey of earthworks and upstanding remains. A different engagement with complex landscapes is presented by Davis et al. in a discussion of lidar survey in the Brú na Bóinne World Heritage Site, which highlights added value for landscape modelling and understanding. The final two papers look outwards from the ‘narrow’ concerns of archaeology to wider worlds of visualisation and research. Challis and Kincey discuss the potential of immersive visualisation using computer game engines for exploring sense of place, meaning and interpretation in landscape, highlighting the challenges about thinking laterally and imaginatively when trying the get the best from versatile datasets. This too, is a central theme in Beck’s discussion of collaboration and approaches to research, and its communication, explored with reference to heritage applications of ALS. Collaboration and communication are good themes to conclude on, and are particularly pertinent to ALS and other 3D data and their continuing development within archaeology. The

11

12

Rachel S. Opitz and David C. Cowley range of skill sets, perspectives and backgrounds represented in this volume highlight how important collaboration and effective communication are in this field. This volume highlights the relevance of multi-scaled topographic data to contemporary archaeological practice, and in a rapidly developing world of possibilities provides stimulating examples of thought and best practice.

Acknowledgments Our thanks to Kevin Barton, Ant Beck, Ralf Hesse, Billy MacRae and Ole Risbøl for comments on the text, to Rob Shaw for providing Figure 1.1, to Angela Gannon for providing Figure 1.2 and to Strat Halliday for providing Figure 1.3. Parts of the text have been presented at the First International Conference on Virtual Archaeology held at the State Hermitage Museum, St Petersburg, 4–6 June 2012. Attendance at this conference for D Cowley was funded by a grant from The Royal Society of Edinburgh, which is gratefully acknowledged.

References Bewley, R., Crutchley, S. and Shell, C., 2005. New light on an ancient landscape: lidar survey in the Stonehenge World Heritage Site. Antiquity 79, 636–47. Bowden M. 1999. Unravelling the Landscape: An Inquisitive Approach to Archaeology. Tempus: Stroud. Burks, J., 2010. Rediscovering prehistoric earthworks in Ohio, USA: it all starts in the Archives. In Cowley, D., Standring, R. and Abicht, M., (eds). Landscapes through the lens: Aerial photographs and historic environment. 77–87. Oxbow: Oxford. Cowley, D., 2012. In with the new, out with the old? Digital workflows and auto-extraction in remote

sensing archaeology. Paper presented at The First International Conference held at the State Hermitage Museum 4–6 June 2012. De Laet V., Mušič B., Paulissen E. and Waelkens M., 2008. Extracting archaeological features from very high resolution Quickbird-2 remote sensing imagery: A methodological approach based on the town of Sagalassos. In Degryse, P. and Waelkens, M., (eds). Sagalassos VI. Geo- and Bio-Archaeology at Sagalassos and in its Territory, 157–71. Leuven. De Laet, V., Paulissen, E., Meuleman, K. and Waelkens, M., 2009. Effects of image characteristics on the identification and extraction of archaeological features from Ikonos-2 and Quickbird-2 imagery: case study Sagalassos (southwest Turkey). International Journal of Remote Sensing 30(21), 5655–68. De Laet, V. Paulissen, E. and Waelkens, M., 2007. Methods for the extraction of archaeological features from very high-resolution Ikonos-2 remote sensing imagery, Hisar (southwest Turkey). Journal of Archaeological Science 34, 830–41. Evans, D., Pottier, C., Fletcher, R., Hensley, S., Tapley, I., Milne, A., and Barbetti, M., 2007. A comprehensive archaeological map of the world’s largest preindustrial settlement complex at Angkor, Cambodia. Proceedings of the National Academy of Sciences of the United States of America 104 (36),14277–82. Evans, D. and Traviglia, A., 2012. Uncovering Angkor: Integrated Remote Sensing Applications in the Archaeology of Early Cambodia. In Lasaponara, R. and Masini, N. (eds). Satellite Remote Sensing: A New Tool for Archaeology. 197–230. New York: Springer. Gannon, A., 1999. Challenging the past: the resurvey of Braidwood Hillfort. In Frodsham, P., Topping, P. and Cowley, D., (eds). ‘We were always chasing time.’ Papers presented to Keith Blood. Northern Archaeology 17/18, 105–11. Hesse, R., 2010. LiDAR-derived Local Relief Models – a new tool for archaeological prospection. Archaeological Prospection 17(2), 67–72. Trier Ø. and Pilø, L., 2012. Automatic detection of pit structures in airborne laser scanning data. Archaeological Prospection, 19(2), 103–21.

2 An overview of airborne and terrestrial laser scanning in archaeology Rachel S. Opitz The technologies of airborne and terrestrial laser are reviewed, providing brief histories and an introduction to some technical topics and comparisons between hardware, algorithms and processing techniques. Discussion is focussed on providing an accessible overview of how terrain models are influenced by factors including data-collection parameters, post-processing techniques, classification errors, and interpolation methods. These are presented to help develop confidence amongst researchers and cultural heritage professionals when interpreting terrain models and visualizations generated from lidar data. Keywords: lidar, ALS, TLS, History of Technology

Introduction This chapter provides a guide to the technologies of airborne and terrestrial laser scanning as they are applied in archaeology and background information for topics covered in contributions to this volume. Brief histories of each technology are given, along with an introduction to some technical topics and comparisons between hardware, algorithms and processing techniques. References to key technical texts and popular software are provided, with the recognition that these will date quickly in a rapidly moving field. The reader is also directed to the glossary of key technical terms on pp 266–268. Why provide a technical overview in a book intended primarily for a non-specialist archaeological audience? While it is certainly not necessary for an archaeologist using laser scanning data to have a detailed understanding of the technical aspects of data collection and processing, general knowledge of how the appearance of terrain models are influenced by choices made at the time of flight, postprocessing techniques, classification errors, and interpolation methods may prove useful for researchers and cultural heritage professionals,

and help to develop confidence when interpreting terrain models and visualizations generated from lidar data.

What is laser scanning? The term laser scanning describes any technology which accurately and repeatedly measures distance, based on a precise measurement of time, and aggregates these measurements into a collection of coordinates. These coordinates are normally stored as a point cloud, from which information on the morphology of the object being scanned may be derived. There are two methods commonly employed by laser scanners for measuring distance. The first measures time of flight: an individual short pulse of laser radiation is emitted from the scanner and the time it takes for the pulse to travel to the object being scanned and return to the instrument is measured. Scanners using this method are commonly referred to as Time of Flight (TOF) Scanners. In the second method a continuous beam of laser radiation is emitted from the scanner and the distance is calculated by measuring the phase shift between the emitted

14

Rachel S. Opitz

Figure 2.1: 3D digital documentation techniques by scale and object size (after Boehler 2001)

and received laser beams. Scanners using this method are called Phase Shift Scanners. Closely related to laser scanners are photogrammetric scanners, which project structured light over an object and record the position of the stripes or points of light from multiple camera positions, or use a single camera position but systematically move the object being recorded, usually on a turntable. Most airborne and mobile laser scanning systems use TOF technology. Terrestrial laser scanners designed for mid-range measurement projects typically use either TOF or Phase Shift based technology. Photogrammetric scanners are primarily used for close range scanning of small objects. The differences between these systems are reflected in significant variations in the complexity and density of data and the size/scale of the object being scanned (Figure 2.1) and therefore their primary applications (Table 2.1).

Crutchley, S. and Crow, P., 2009. The light fantastic: Using airborne laser scanning in archeological survey. Swindon: English Heritage. Glennie, C., 2007. Rigorous 3D error analysis of kinematic scanning LIDAR systems. Journal of Applied Geodesy 1, 147–57. Jones, D. (ed.). 2011. 3D Laser Scanning for Heritage (second edition) Advice and guidance to users on laser scanning in archaeology and architecture. Swindon: English Heritage. Payne, A., 2011. Laser Scanning for Archaeology. A Guide to Good Practice. http://guides. archaeologydataservice.ac.uk/g2gp/LaserScan_ Toc. Accessed 29 January 2012. Raber, B. and Cannistra, J., 2005.Lidar Guidebook: Concepts, Project Design, and Practical Applications. Park Ridge, IL: Urban and Regional Information Systems Association. Toth, C. and Shan, J., 2009. Topographic laser ranging and scanning, principles and processing. Boca Raton, FL: CRC Press, Taylor & Francis Group. Vosselman, V. and Maas, H-G., 2010. Airborne and Terrestrial Laser Scanning. Dunbeath: Whittles Publishing.

Essential reading and references

Barber, D.M., Dallas, R.W. and Mills, J.P., 2006. Laser scanning for architectural conservation, Journal of Architectural Conservation 12, 35–52. Boehler, W., Heinz, G., Marbs, A., 2001. The potential of noncontact close range laser scanners for cultural heritage recording. Proceedings XVIII CIPA Symposium, Potsdam, Germany. http://cipa. icomos.org/fileadmin/papers/potsdam/2001-11wb01.pdf.

Airborne lidar Airborne lidar is an active remote sensing technique, used to record the surface of the earth, documenting the topography of large areas of terrain and objects appearing on it. In cultural resource management (CRM) and archaeological research airborne lidar is used to

15

2  An overview of airborne and terrestrial laser scanning in archaeology Scanner Type Triangulation scanner Terrestrial TOF Scanner Terrestrial Phase based scanner Airborne Scanner (light aircraft)

Primary Applications Object scanning Architectural scanning Architectural scanning Landscape mapping

Airborne Scanner (helicopter)

Corridor mapping

Mobile Mapping

Urban modelling, coastal erosion monitoring

locate and map sites with preserved topography, particularly in wooded areas, and to provide detailed terrain models for landscape analyses and VR applications. Note that lidar and ALS are used more-or-less interchangeably in this volume, with LiDAR also used in other publications. Lidar derives from Light Detection And Ranging. History of the technology The earliest lidar remote sensing systems were used by NASA in the 1970s for mapping in the ice covered Arctic and Antarctic (Krabill et al. 1984), but despite a few research projects in remote sensing during the late 1970s and 1980s, uptake in disciplines outside geomatics did not happen until the 1990s. Indeed, the limitations in Inertial Measurement Unit (IMU) and Global Navigation Satellite System (GNSS) technologies, the high cost and bulkiness of data storage equipment needed on the aircraft, and the software and hardware necessary to manage large datasets in post-processing prevented the adoption of lidar for commercial mapping projects. However, in the mid-1990s georeferencing technologies improved significantly, and this, coupled with general improvements in software and data storage technologies made lidar practical for commercial and research applications. The popularization of airborne laser scanning, or lidar, followed in the late 1990s and early 2000s and is considered by many to represent the major development in airborne remote sensing and terrestrial survey in this period (Shan and Toth 2009). Airborne lidar systems may be divided into two broad types: discrete return and full waveform systems. Discrete return systems record the time (position) of a limited number of returned laser pulses whose amplitude (intensity) is greater than a set threshold. Full waveform systems effectively record the form of the entire returned waveform by regularly sampling the returned

Typical Accuracy Less than 1 mm 3–6 mm 5 mm 15 cm vertical 50 cm planimetric 8 cm vertical 20 cm planimetric 10–50 mm

pulse, and permit more detailed analyses. The earliest lidar systems in widespread use were based on discrete return recording, a technology which still dominates the airborne laser scanning industry at present (see below). In the mid2000s, following the increasingly widespread use of discrete return lidar systems, advances were made in the collection and exploitation of full waveform information. The acquisition of full waveform data can significantly improve the quality of the data’s classification, particularly in separating returns from the ground and low off-ground objects such as scrub vegetation, and provides additional information on the character of the targets. The first full waveform system was the Laser Vegetation Imaging Sensor (LVIS) produced by the NASA in 1999, which was followed in 2004 by the first commercial full-waveform system (Hug et al. 2004). While full waveform lidar is acknowledged to produce superior results for a number of applications, its commercial application remains limited at the time of writing. However, recognising this issue, many manufacturers produce a waveform digitizer to extend the capabilities of discrete return sensors, and many commercial data providers fit their aircraft with full-waveform capable scanners, although usually operating them in discrete return recording mode. Hardware improvements, including increases in pulse firing speed and waveform digitizers that increase the amount of data collected on each flight, have been fundamental to the increased uptake of lidar. Equally important have been advances in software for post-processing and visualization for commercial and research applications. The introduction of the LAS standard in 2003 was essential to the development of software for handling the large point clouds produced by laser scanning (Samberg 2007). The LAS standard is a binary file format developed and maintained by the American Society of Photogrammetry and Remote Sensing

Typical Range 0.1 to 1 m 0.5 to 100 m 0.5 to 100 m 1000 to 3500 m 50 to 250 m 100–200 m

Table 2.1: Comparison of scanner applications and accuracies. Adapted from Barber (2006)

16

Rachel S. Opitz

Figure 2.2: General schematic of a lidar system. With kind permission of Keith Challis

(ASPRS) to support the exchange of lidar data between users working with a variety of software programs. LAS standard files have open headers and include X,Y,Z position information and can contain information specific to lidar data such as intensity, scan angle, GPS time, return classification and point source following a proscribed format. Because the LAS standard is open, it can be used by both commercial and open source software and data is easily moved between them. A variety of software solutions are now widely available through both commercial and open source packages and most lidar projects will use multiple programs in their processing chain. Key open source packages specifically for airborne lidar include LAStools, developed by Martin Isenburg, and the libLAS library. Lidar handling and visualization tools are also now part of many GIS programs including GRASS, SAGA, ArcGIS and ENVI/IDL. Software

GRASS GIS http://grass.fbk.eu/ SAGA GIS http://www.saga-gis.org/ System for Automated Geoscientific Analysis BCAL Toolbar http://bcal.geology.isu.edu/tools-2/ envi-tools Boise Center AeroSpace Lab LibLAS http://liblas.org LAStools http://www.cs.unc.edu/~isenburg/lastools/

References

Hug, C., Ullrich, A. and Grimm, A., 2004. LITEMAPPER-5600 – a waveform digitising lidar terrain and vegetation mapping system. International Archives of Photogrammetry, Remote

Sensing and Spatial Information Sciences 36, 24–29 (Part 8/W2). Krabill, W.B.,Collins, J.G., Link, L.E., Swift, R.N. and Butler, M.L., 1984. Airborne laser topographic mapping results. Photogrammetric Engineering and Remote Sensing 50, 685–94. Samberg, A., 2007. An Implementation of the ASPRS LAS Standard. The Analyst XXXVI, 363–72. http://www.isprs.org/proceedings/XXXVI/3W52/ final_papers/Samberg_2007.pdf. Toth, C. and Shan, J., 2009.Topographic laser ranging and scanning, principles and processing.Boca Raton, FL: CRC Press, Taylor & Francis Group.

Fundamentals Airborne laser scanning systems are composed of a platform (i.e. aircraft), positioning and georeferencing equipment (i.e. GPS/IMU), the scanner itself and data recording systems (Figure 2.2). Airborne systems can employ a variety of platforms and types of sensors, dependant on the requirements of the project, including small fixed wing aircraft, helicopters and unmanned aerial vehicles (UAVs). Many of these platforms can also collect vertical photographs or multi- or hyper- spectral imagery to allow for the creation of fused or directly comparable data products. Solid-state pumped lasers are the most common type used in airborne systems because they emit very short, high powered pulses capable of achieving a high pulse repetition rate, which is important for rapid data capture and therefore a priority in commercial applications. Most airborne lidar systems work at wavelengths around 1.064nm, in the NIR range, although

2  An overview of airborne and terrestrial laser scanning in archaeology

Point Density Vegetation Penetration Area Covered

Fixed Wing Aircraft Circa 4–10 pts/m2 Limited, affected by flying height Large, e.g. 100 km2 per mission

 bathymetric lidar systems operate in the 532 nm range. Typical pulse durations (often referred to as widths) range from 4–10 ns and the scan rate is typically 100–150 kHz, or 100,000 to 150,000 pulses per second. Pulses are directed toward the target using a variety of systems, including oscillating mirrors, rotating mirrors, rotating prisms, and nutating (i.e. rocking or swaying) mirrors. Each system produces distinct patterns of returns from the ground, often referred to as the scan pattern.

Airborne Lidar Sensors A variety of lidar sensors are available for use in airborne mapping. Discrete return lidar systems are, at the time of writing, the most common type of sensor used to collect data over large landscape areas, and consequently the source of data for most archaeological projects. • Discrete Return lidar systems emit laser pulses and record the strength (intensity) and time of a limited number of returned pulses whose strength is over a certain threshold. Discrete return lidar usually has a small pulse footprint (c. 10 cm radius at nadir flying at 1000 m) and two to four returns per pulse are recorded. • Multiple Pulse in Air or MPiA lidar produces a result much like that of discrete return lidar, but is capable of collecting more points (nominally twice as many) in the same amount of flying time by firing laser pulses at a faster rate (i.e. a second pulse can be fired while the first is still in the air). MPiA systems are consequently popular for projects covering larger areas. In theory, the MPiA technology allows for a large number of pulses, but generating laser pulses at a high repetition rate is challenging in practice. However, one solution to overcome this limitation is to deploy multiple laser sensors, in a system frequently called multi-channel lidar. • Full wave form lidar systems represent the entire form of the returned waveform by sampling the intensity of the return signal at frequent and regular intervals, rather than just the strength and time of pulses over a certain threshold. Full waveform systems are useful for recording the

Helicopter Circa 40 pts/m2 Better than Fixed Wing due to lower flying height Medium, e.g. 35 km2 per mission

17 UAV ??? Circa 100 pts/m2 ??? Theoretically best Small, 2 km2 per mission

structure of the vegetation canopy, or in areas where differentiating between low vegetation and terrain is likely to be difficult. • Bathymetric lidar records water depth using two laser pulses emitted at different wavelengths, one in the infrared and one in the green spectrum. The infrared pulse represents the surface of the water and the green reflects from the underwater terrain. Bathymetric systems are used exclusively for off-shore or near-shore recording projects, and are sometimes flown in tandem with topographic lidar to provide higher point densities in near-shore areas.

Airborne platforms There are a variety of platforms such as fixed wing aircraft, helicopters, and UAVs, each of which has advantages and disadvantages that can have a major impact on the type of data collected. These factors include point density, survey time required, penetration rates through vegetation and accuracy. For example, while fixed wing aircraft are most appropriate for covering larger areas at lower point densities, helicopter mounted platforms can obtain a higher point density because the aircraft can fly more slowly and closer to the ground, but a smaller area is covered in each flight. Helicopter platforms are particularly popular for corridor mapping. The use of UAVs for airborne lidar collection is still experimental, as the weight of a lidar system and the sensitivity of the scanners to vibrations continue to prove problematic, but would be appropriate for recording very small areas at a high point density (Table 2.2).

Terrestrial lidar Terrestrial lidar is used here to encompass a wide variety of 3D recording technologies including time of flight laser scanners, fringe projection scanners, white light scanners, and triangulation scanners which usually operate from a tripod, arm-mounted or table-top platform. These technologies are variously used in archaeological

Table 2.2: Comparison of Fixed Wing, Helicopter and UAV platforms

18

Rachel S. Opitz research and CRM for documentation and modelling on scales from small objects to sites of several square kilometres. Terrestrial lidar has many applications in archaeology, and projects from the creation of virtual museums of scanned objects to monitoring the erosion of rock art are increasingly widespread.

History of the technology The first triangulation 3D laser scanner was developed in 1978, by the National Research Council of Canada. During the 1970s and 1980s scanner development occurred primarily within a research context (Mayer 1999). The miniaturization and integration of optical light sources and sensors developed in parallel, which was vital to making portable scanners practical. TOF terrestrial laser scanners appeared on the market in the late 1990s. Early scanners (c. 1997) featured external power and data storage modules and were difficult to take into the field. However, the integration of power supplies, data storage, cameras and GPS represent significant operational improvements, which took place beginning about 2002, and allow their deployment in a wide variety of contexts. The most recent (post2009) generation of scanners can collect full waveform data, using a waveform digitizer like that employed in aerial lidar systems. Like TOF scanners, photogrammetric scanners became commercially available in the mid-1990s and they have continued to develop increased spatial resolution and accuracy and integration with spectral sensors for colour data capture. Figure 2.3: (left) Dornier Do228-101 D-CALM (NERC ARSF), (centre) AS350BA helicopter (Aerotec LLC), (right) The Ammaia Project’s custom UAV. In all cases the laser scanner is mounted in the base of a customized aircraft

Sensors Many types of sensors are employed in terrestrial scanning in archaeology. Sensors may be selected for a project on the basis of portability or robustness for on-site scanning, resolution of the model produced, colour capture capabilities, data acquisition rates or budget (Figure 2.3). Some of the more common types of scanners are outlined below.

Triangulation scanners Triangulation scanners calculate the coordinates of points on the surface of an object by triangulating the position of a spot or stripe of laser light projected from the scanner onto the object, based on the known distance and angle between a camera in the scanner and the sensor emitting the spot or stripe of light. The object being scanned may be rotated on a turntable or the scanner may be mounted on a mechanical arm. Triangulation scanners typically perform badly in bright sunlight, so temporary shading is required for outdoor, daytime acquisitions. Triangulation scanners are therefore well suited to recording objects in the laboratory, projects where nighttime recording is practical, and recording of the interiors of structures. Triangulation scanners can produce data at sub-mm accuracy. Time of flight scanners Time of flight technology is used in mid-range scanners commonly used for architectural scanning and work on excavations. Like the airborne lidar system, this terrestrial scanning system is based on the measurement of the time of flight of a laser pulse, which is directed across the surface of an object using either an oscillating or rotating mirror. These scanners produce a point cloud with accuracy in the 3–6 mm range. TOF scanners can be used for recording at distances between a theoretical minimum of 0.5 m (in practice 1.5 m is more realistic as smaller distances to the scanning target introduce noise) to a maximum of about 300 m. Phase shift scanners Phase shift scanners emit laser pulses at alternating frequencies and measure the difference between peaks of the emitted and reflected signals at the two frequencies to calculate distance. Phase shift systems have a range of 0.5–80 m, although operation in the 1–50 m range is typical and avoids problems with noise. Phase shift systems typically operate at a faster rate than TOF scanners, i.e. TOF scanners operate at circa 50,000 points per second while phase scanners collect data at 500,000 points per second. Faster data acquisition rates make phase shift scanners appropriate for projects covering very large areas or working under time restrictions. Structured light scanners Structured light scanners project a series of regular patterns of light onto the object being scanned.

2  An overview of airborne and terrestrial laser scanning in archaeology One or two cameras are mounted at an angle slightly offset from the projector which records the pattern of light. A triangulation method is then used to calculate the position of each point in the structured light pattern, based on the known offset in distance and angle between the projector and the camera positions. Patterns may be projected as a single line of line of light, or as a series of stripes at various scales. Structured light scanners operate considerably faster than traditional triangulation scanners because they can record the entire area covered by the light pattern (the instantaneous field of view) simultaneously. Like traditional triangulation scanners, structured light scanners do not perform well in bright light and a shelter may be required for outdoor scanning during the day. Structured light scanners are appropriate for scanning relatively small areas and objects in the laboratory and for field recording of interiors and for work at night. Structured light scanners, like triangulation scanners, will theoretically produce data with sub-mm accuracy. Image Data Many terrestrial laser scanner systems can simultaneously collect high resolution digital photographs via an on-board or externallymounted camera. These images are assembled into a multi-image mosaic and the colour data extracted from the images is applied to the point cloud or stored as a texture. Colour data collected by terrestrial laser scanners can be used as an aid in classification and identification of objects within the scan, and facilitates the development of realistic visualizations. However, the use of colour data collected by terrestrial scanners can be challenging. Changes in lighting conditions as the scanner rotates or is moved to a new position, the presence of shadows and highlights, and differences in the colour range of each image can all lead to the need for colour balancing and other image processing before the colour data is useful for analysis or interpretation. This processing is usually done outside the scannerspecific software (e.g. Photoshop) after which the colour data can be reapplied. Furthermore, some classes of features are more recognizable when colour data has been removed and artificial lighting or shading schemes are applied to the polygonal mesh. In particular hillshade style directional lighting can be used to highlight small topographic changes in the modelled surface.

References

Mayer, R., 1999. Scientific Canadian: Invention and Innovation from Canada’s National Research Council. Vancouver: Raincoast Books. Toth, C. and Shan, J., 2009. Topographic laser ranging and scanning, principles and processing. CRC Press, Taylor & Francis Group, Boca Raton, FL. Vosselman, G. And Maas, H.G., 2010. Airborne and terrestrial laser scanning. Dunbeath, Scotland: Whittles Publishing.

Errors in terrestrial lidar surveys Sources of error in TLS surveys parallel some of those found in lidar surveys. Key sources of error are laser sensor (or camera) calibration errors (range measurement and scan angle), scanner (or camera) position errors, object surface characteristics, and scanning geometry characteristics. In common with lidar data, errors in the location of the scanner can cause misalignments between scans, equally, as with airborne lidar the divergence of the laser beam over longer ranges and in places with larger scan incidence angles can introduce vertical errors as a result of horizontal shifts. In TLS the problem of locating the scanner is simpler (no GPS/IMU error budget to account for), so scanning geometry and surface characteristics become a more important part of the total error budget. Some characteristics of the surface being scanned can be problematic and highly reflective, partially transparent and very rough surfaces can all introduce complex backscattering effects. References

Cuartero, A., Armesto, J., Rodríguez, P.J., and Arias, P., 2010. Error Analysis of Terrestrial Laser Scanning Data by Means of Spherical Statistics and 3D Graphs. Sensors 10, 10128–45. Lichti, D.D., 2010. Terrestrial laser scanner selfcalibration: Correlation sources and their Mitigation. Photogrammetry and Remote Sensing 65(1), 93–102. Lichti, D.D. and Licht, M.G.,2006. Experiences with Terrestrial Laser Scanner Modelling and Accuracy Assessment. IAPRS Volume XXXVI, Part 5, Dresden 25–27 September 2006. Reshetyuk, Y., 2010. A unified approach to selfcalibration of terrestrial laser scanners, Photo­ grammetry and Remote Sensing, 65(5), 445–56. Soudarissanane, S., Lindenbergh, R. and Gorte, B., 2008. Reducing the error in terrestrial laser scanning by optimizing the measurement set-up. IAPRS Volume XXXVII, Part B5, Beijing 2008.

19

20

Rachel S. Opitz

Post-processing for (Discrete Return) airborne lidar Lidar data in its raw form is a series of measurements of times and intensities of returned laser pulses. These time data are correlated with navigational information from the GNSS and IMU to calculate the geodetic position of each laser return. Adjacent strips of points are aligned to improve accuracy within the dataset and the final point cloud is adjusted to fit the coordinates of ground surveyed control points. These positions are typically stored as a point cloud where each point contains attribute information such as GPS time, intensity, scan angle and flightstrip number along with its X, Y and Z coordinates. Point clouds are usually manipulated and stored in the binary LAS or LAZ (compressed LAS) formats or, in the case of some legacy systems, in ASCII formats. Further processing is required to create products like DSMs, DTMs, and canopy models with which many users of lidar data work most frequently. Yet more processing is required to produce upstream products such as hillshaded terrain models, contours, intensity maps and other visualizations. At all stages assumptions are made, either explicitly by the individual processing the data, or within the software systems applied, which have a major impact on the derived data or visualisation. These are outlined below. Classification The first stage in the post processing chain is usually classification, an essential task when working with airborne lidar data. Lidar data is typically gathered across complete landscape blocks, which may include woodland, scrub and arable areas. One of the key advertised benefits of lidar is its ability to ‘see through’ tree cover as some returns will pass through gaps in the vegetation canopy, reaching the ground and allowing the creation of a bare earth DEM – making lidar an effective tool for survey in wooded and scrub areas. To accomplish this data must be classified (or filtered) to separate returns (e.g. from the ground and those from vegetation). Two types of classification errors are possible when identifying ground returns for the creation of a terrain model: the removal of points that should be retained (type 1) and the inclusion of points that should be removed (type 2). Aggressive settings for filtering parameters

(discussed below) will lead to the majority of errors being type 1 and less aggressive filtering will lead to type 2 errors. The application of an inappropriate classification algorithm or a poor choice of filtering parameters can have serious consequences for the appearance of archaeological features in the final terrain models. Overly aggressive algorithms have a tendency to remove small ‘peaks’ in the terrain and to smooth or flatten the terrain surface. Often these small peaks represent archaeological remains, but if they are classified as vegetation by the filtering algorithm they will not appear in the terrain model or in any subsequent upstream visualizations of the data and consequently the archaeological features will not be identified. Conversely, the inclusion of clumps of low vegetation returns in the ground class can result in false ‘features’ that look very much like archaeological remains (Doneus et al. 2008). For example, small bushes or piles of branches might be interpreted as low mounds and fallen trees might be mistaken for the bases of walls. Most classification algorithms were developed for applications and industries where a slightly smoothed ‘idealized’ terrain model is preferred over a noisy one and removing small localized terrain variations is not considered seriously problematic. Therefore archaeologists need to be careful in selecting a filtering algorithm and parameters, or at least be aware of the types of classification errors likely to be introduced by different filtering algorithms. This section reviews the major types of classification algorithms, noting their main assumptions and likely errors that would affect the appearance of archaeological features in the terrain models. To understand where errors occur in the terrain models and what the accuracy of a final product is likely to be under a given set of conditions, it is essential to grasp the way in which different classifiers and filters operate. Many archaeologists will be provided with lidar data as a bare earth DTM created using the points classified as ground, and will not have access to the original, raw or classified point clouds. In these cases, it remains important to understand how classification algorithms operate in order to recognize likely errors in the terrain models due to mis-classification. General assumptions The identification of ground points used to build a bare earth model begins with four assumptions.

2  An overview of airborne and terrestrial laser scanning in archaeology The first is that locally the ground will be represented by the lowest points when looking at a profile view of a lidar model. Secondly, it is assumed that the ground is a continuous surface and will be present, if not visible, everywhere in a rural area (classification of buildings in urban areas is a separate problem not discussed here). The third assumption is that returns from the ground are present at some minimal spacing in the lidar model. A final assumption is that returns from the ground are in some way different from other returns. While all of these assumptions introduce problems, the most significant is the assumption that returns from the ground are present at some minimum spacing in the model. Research (Rosso et al. 2003) indicates that in extremely dense vegetation 5–10% of the lidar returns, and in less dense vegetation 10–20% of the returns, should penetrate to ground. These statistics support the assertion that the ground should be definable, albeit at a reduced spatial resolution, for all areas of the lidar model. However, in practice it is possible that for some local areas there will be no ground returns and the ground will not be definable and the classification algorithm will mark the lowest vegetation points as ground. These areas of very dense vegetation cause the majority of the errors committed by classification algorithms. The density of the point cloud has an overall impact on the behaviour of all classification algorithms, as point spacing affects curvature and slope calculations on which many filters depend, and in particular it impacts on the success of filtering in areas of dense vegetation. Morphological filters Many early approaches to classification were based on terrain morphology, while more recent approaches apply various spatial statistics to separate terrain from vegetation. One of the earliest morphological filters (Kilian et al. 2004) applied a morphological opening operator (for details on how opening operators work see Dougherty 1992, 17–22) repeatedly over a raster terrain model, increasing the window size with each iteration. Opening removes all pixels in the regions too small to contain the structuring element. In other words, the opening operator removes small objects from the image. This process is repeated until all objects under a set size are removed. In this method, after the opening operation remaining points are weighted in a scheme where points closer to the lowest point

in the window are given higher weights and these points are classified as ground. The application of the opening operator is problematic for archaeologists because it removes details of the terrain model, possibly including archaeological features, in an attempt to create a smooth model of the ground. However, the second part of Kilian et al.’s strategy, sorting points into bands above a theoretical ground level, has been adopted by other methods. For example, the Progressive Morphological Filter (Zhang et al. 2003) incorporates elements from this earlier method. As with other approaches that depend on the opening operator, their algorithm results in a smoothed surface with small terrain features removed. To avoid the problem of smoothed terrain, Zhang et al. use this initial surface in the second step of their filtering process. Points are classified based on a maximum allowed elevation difference between the lidar points and the initial filtered surface. This is done iteratively and the allowed elevation difference and window size are varied. A popular morphological filter developed by Zaksek and Pfiefer (2006) incorporates first return data, using it to produce a trend surface based on the canopy surface. The kernel of their filter is a cone surface which can be generated in three possible ways: by synthetic function, by preserving important terrain features regarding the reference data or by minimizing classification errors. Every point is checked to determine if it lies above or beneath the cone and is classified as non-ground if any point in its neighbourhood is situated below the cone (Zaksek and Pfiefer 2006). Like most morphological filters these do not consistently remove low vegetation points and consequently are not ideal for archaeological projects. Slope and curvature filtering Sithole and Vosselman (Sithole 2001; Vosselman 2001) began a trend for the use of slope based filters. In their algorithm, if the slope exceeds a certain threshold then the highest point is assumed to be non-ground. Sithole (2001) modified the filter so the slope threshold varies with overall terrain conditions. Haugerud and Harding (2001) modified the slope based filters to create a curvature based filter, which takes into account that surfaces that include vegetation points tend to exhibit convexity everywhere while sloping terrain will exhibit convexity over the tops of hills and concavity at the bottoms of slopes. Curvature filters have been proved effective for

21

22

Figure 2.4: Ground points marked in orange, vegetation points marked in green. Even a small number of incorrectly classified vegetation points, as shown here, can cause the appearance of a slight mound in the interpolated data

Rachel S. Opitz heavily vegetated and sloping terrain and have been applied in archaeological projects.

and analogous approaches are common in some commercial software.

Adaptive surface filtering Kraus and Pfiefer (1998) developed a linear prediction method based on calculating the distance (residuals) from an average surface to the measured points. Each measurement is given a weight according to its distance from the smoothed surface. The real terrain surface is then computed based on the weights and orientation of the residuals. As with other techniques, this approach tends to leave low vegetation points but is an improvement on basic morphological filters.

Software The algorithms described above have been implemented in a number of software packages designed for working with lidar data. However, the specifics of implementation are often unpublished, a problem for scientific projects as noted by Beck (this volume).

Statistical index filtering Statistical index filters were developed for wooded areas containing steep slopes and dense complex under-storeys. Repeated surface estimation (REIN) is an example (Kobler et al. 2007) of this approach. This is a two stage method beginning by identifying negative outliers and most off-ground returns using morphological or curvature based filtering. In a second filtering and DTM generation stage, a DTM is produced from the first filtered point cloud using a algorithm which makes use of multiple ground elevation estimates at individual DTM points in a grid or a TIN (Triangulate Irregular Network) interpolated from surrounding ground returns. These elevation estimates are generated from multiple independent samples taken from the initially filtered point cloud. The input into the final filtering and DTM generation stage is a filtered point cloud containing mostly ground points scattered within the error band, some positive outliers and no negative outliers. Within the error band multiple estimates of the ground surface are made and the ground surface is determined. The REIN algorithm has been successfully employed in archaeological projects

Open source software packages incorporating classification tools include: • MCC-Lidar • Fusion • LASTools Commercial software packages commonly used for classification include: • TerraSolid • VR Mesh • MARS Software • LP360 Manual classification Because no classification or filtering process will be entirely accurate, manual editing of the point cloud classification is an essential step in the processing chain. Manual classification is usually carried out on a point cloud in which returns have been classified by an automated algorithm, viewing the points in profile or against background imagery and reassigning those incorrectly classified by the algorithm. Manual classification relies heavily on the experience of the operator. In areas with irregular terrain features and dense vegetation (a typical situation where archaeological remains are present!) it is not always clear whether an individual return belongs to the vegetation or the ground. An overall sense of the geomorphology and structure of the vegetation typical to an area will aid in making these judgement calls (Figure 2.4). In areas where there are few or no lidar returns from the ground but many returns from very low vegetation an operator may choose to include some vegetation returns in the ground class, assuming that the low vegetation closely follows the terrain, rather than leave large patches empty of returns resulting in an unrealistically flat surface interpolated across a gap in the point cloud. While carrying out manual classification on the entire lidar dataset may be beyond the resources (or interests) of many archaeological projects, careful checking and correction of the

2  An overview of airborne and terrestrial laser scanning in archaeology

23

classifications over areas of interest can help in the recognition of archaeological features and is fundamental to the creation of good surface models for visualization. References

Axelson, P., 1999. Processing of laser scanner data – algorithms and applications. ISPRS Journal of Photogrammetry and RemoteSensing, 54, 138–47. Doneus, M. Briese, C., Fera, M. and Janner, M., 2008. Archaeological prospection of forested areas using full-waveform airborne laser scanning. Journal of Archaeological science, 35, 882–93. Dougherty, E., 1992. An introduction to morphological image processing. SPIE Optical Engineering Press. Haugerud, R. and Harding, D.J., 2001. Some Algorithms for Virtual Deforestation (VDF) of Lidar TopographicSurvey Data. International Archives of Photogrammetry and Remote Sensing XXXIV(3), 211–7. Kilian, J., Haala, N., and Englich, M., 2004. Captureand evaluation of airborne laser scanner data. International Archives of Photogrammetry and Remote Sensing, XXXI, 383–8. Kobler, A., Pfiefer, N., Ogrinc, P., Todorovski, L., Ostir, K. and Dzeroski, S., 2007. Repetitive interpolation: A robust algorithm for DTM generation from aerial laser scanner data in forestedterrain. Remote Sensing of the Environment 108, 9–23. Kraus, K. and Pfiefer, N., 1998. Determination of terrain models in wooded areas with airborne laser scanner data. ISPRS Journal of Photogrammetry and Remote Sensing, 53(4), 193–203. Rosso, P., Ustin, S. and Hastings, A., 2003. Use of lidar to produce high resolution marsh vegetation and terrain maps. Three Dimensional Mapping from InSAR and LIDAR workshop. International Society for Photogrammetry and Remote Sensing. June 17–19. Sithole, G., 2001. Filtering of laser altimetry using a slope adaptive filter. International Archives of Photogrammetry and Remote Sensing 34, 203–10. Vosselman, G., 2001. Slope based filtering of laser altimetry data. International Archives of Photogrammetry and Remote Sensing, 33, 935– 42. Zaksek, K. and Pfiefer, N., 2006. An improved morphological filter for selecting relief points from a lidar point cloud insteep areas with dense vegetation. Technical report, Delft Institute of Earth Observation and Space systems, TU Delft, The Netherlands. Zhang, K., Cheng, S., Whitman, D., Shyu, M., Yan, J., and Zhang, C., 2003. A progressive morphological filter for removingnon-ground measurements from airborne lidar data. IEEE Transactionson Geoscience and Remote Sensing, 41(4), 872–82.

TLS post-processing The primary step in terrestrial laser scanning post-processing is the registration, or relative orientation, of point clouds derived from multiple scans. Registration is based either on survey targets included in adjacent scans or on features identified in post-processing as tie-features appearing in adjacent scans (Figure 2.5). Some TLS scanners incorporate dual axis tilt compensators which allow for resection and traverse calculations for the scanner position, and simplify the registration process. However, this is the exception rather than the rule, and registration during post-processing is the norm. Iterative closest point/patch (ICP) algorithms are used to match tie-features between scans (Besl and McKay 1992; Rusinkiewicz and Levoy 2001). With a sufficient number of tie-features a transformation for the point cloud from each scan can be calculated and scans can be aligned to a single coordinate system. ICP algorithms do not require exactly matched points or surfaces, and can work from approximately matched points and surfaces identified by the user to iteratively determine approximate matches between further points and small surface areas until a transformation is calculated which can be applied to the entire dataset. Once scans are registered together to create a coherent point cloud, classification, cleaning and feature extraction can occur. In airborne lidar the

Figure 2.5: Alignment between sections of TLS scan data can be accomplished by picking common points on natural features or targets. Here two matching points are marked with red and green push pins

24

Rachel S. Opitz automated processing of data from rural areas with ‘natural’ features with irregular geometries has been the focus of research for some time and automated algorithms exist to separate irregular terrain and vegetation and ‘natural’ features with irregular geometries do not pose significant problems. On the other hand, in TLS research has traditionally focused on the automated extraction of objects with well defined geometries (e.g. those common to urban environments such as bridges, road signs, and vertical walls) and while the automated classification and extraction of objects with regular geometries is increasingly practical, complex geometries typical of ‘natural’ objects remain beyond the capabilities of current algorithms. References

Besl, P.J. and McKay, N.D., 1992. A Method for Registration of 3-D Shapes. IEEE Transactions on Pattern Analysis and Machine Intelligence 14(2), 239–55. Remondino, F., 2011, Heritage Recording and 3D Modeling with Photogrammetry and 3D Scanning. Remote Sensing 3, 1104–38. Rusinkiewicz, S. and Levoy, M., 2001. Efficient Variants of the ICP Algorithm. Third International Conference on 3D Digital Imaging and Modeling (3DIM 2001). http://www.cs.princeton.edu/~smr/papers/fasticp/ fasticp_paper.pdf. Accessed 29 January 2012.

Post-processing for terrestrial and airborne lidar: interpolation and visualization Interpolation Lidar data may be viewed as a 3D point cloud coloured by schemes based on elevation, intensity, etc.; the point cloud may be filtered by class so that only the ground points are displayed (or only the ground points and low vegetation, or ground points and buildings, etc.) and archaeological interpretations can be based on a reading of the point cloud. While it is possible to identify some archaeological features in the point cloud, the creation of shaded surface models is an essential part of many archaeological projects. Using such surfaces for archaeological interpretation in conjunction with the point cloud is favoured because the surface format greatly increases the ways in which the data may be visualized and provides a familiar look for interpreters accustomed to working with aerial photographs. Both aerial and terrestrial laser scan

datasets may be interpolated to create surfaces. Aerial lidar is usually interpolated to form TINs or DEMs and DSMs stored in a raster format (e.g. GRID or geoTIFF), while terrestrial lidar is usually interpolated into a polygonal mesh and stored in a 3D geometry format such as obj, 3ds, or x3d which support more complex 3D geometries. A TIN (Triangulated Irregular Network) consists of nodes connected by edges to form continuous, non-overlapping triangular facets. The nodes contain the X, Y and Z values derived directly from the point cloud and facets and edges contain interpolated values. A raster DSM or DEM surface is a grid of pixels of a defined size in which each pixel is assigned the Z value (maximum, minimum or mean) of the point(s) which fall within it, or these values may be interpolated based on all points within a locally defined window. Any empty pixels where no points are located have values interpolated based on neighbouring pixels. Interpolation methods used to generate raster surfaces include IDW, spline, kriging and natural neighbour. At the point spacing typical for airborne lidar the choice of interpolation method is not particularly important. TINs are the most commonly used interpolation method in software performing on-the-fly rendering of terrain surfaces. Rasters are typically the basis for upstream visualizations such as hillshades or elevation colour-coded surface models. The process of surface or mesh creation from a point cloud is complex. In addition to choices of interpolation methods, smoothing, hole filling, fairing and deformation, de-striping, and any number of other mesh/surface processing steps will likely need to be applied to achieve a continuous surface. The application of these mesh or surface processing steps will depend on the end goals of a project (metric accuracy vs. visual appeal) and the time available for post-processing as this stage usually involves substantial manual intervention. Visualization The visualization of lidar data is discussed extensively in contributions to this volume and will be dealt with only briefly here. Visualizations of lidar data are normally carried out in a geographic information system (GIS) or computer aided design (CAD) program, a modelling program such as MeshLab or Blender, or in a virtual reality environment like

2  An overview of airborne and terrestrial laser scanning in archaeology Unity. The choice of visualization environment will depend very much on the task at hand. Mapping projects based on lidar data will likely work and create their visualizations in a GIS or CAD environment, while terrestrial scanning data is more often manipulated and visualized in a modelling program and VR environments are currently primarily used to view ‘finished’ visualizations, rather than for their creation. While advances in lidar processing, meshing and classification algorithms are primarily taking place within the lidar industry and improving the core products available to archaeologists, new visualization techniques are rapidly being developed (or re-purposed from diverse fields) by the archaeological lidar community. This focus on visualization techniques is linked to a growing emphasis on model (image) interpretation, rather than technical problems or potentials of lidar data collection (see Doneus and Kühteiber this volume). This field, perhaps above others, is rapidly evolving because it is the focus of methodological development within the archaeological lidar community. The rendering of large, dense datasets usually requires large amounts of RAM and a high performance graphics card, and these technical requirements compound the methodological challenges of developing appropriate visualizations (Figures 2.6, 2.7 and 2.8).

Figure 2.6: Machu Picchu Scanning Project: example of a source photograph (top left); Meshed Scan Data (top right); Sketchup Rendering (bottom left); Vue Rendering (bottom right). Images reproduced with kind permission of CAST, U. Arkansas

Software Unity unity3d.com/ Meshlab meshlab.sourceforge.net/ Blender www.blender.org/ Rapidform www.rapidform.com/ References

Bater, C.,Nicholas, W., and Coops, C., 2009. Evaluating error associated with lidar-derived DEM interpolation, Computers and Geosciences 35(2), 289–300. Guo, Q.,Li, W., Yu, H. and Alvarez, O., 2010. Effects of topographic variability and lidar sampling density on several DEM interpolation methods. Photogrammetric Engineering & Remote Sensing 76(6), 701–12.

Errors in airborne lidar surveys A general understanding of sources of error in a lidar survey is important for archaeologists, since relatively small changes in elevation may be of interest and these changes may be near or within the error budget, or latitude in accuracy, of a lidar survey. Typical surveys will have a vertical

Figure 2.7: Breuckmann HE scan of a an unpublished Greek inscription (University of Mississippi Museum – University Museum Inventory No. 77.3.681) coloured by curvature. (Image reproduced with kind permission of Katie Simon and Stephanie Sullivan, CAST, U. Arkansas)

25

26

Rachel S. Opitz

Figure 2.8: Archaeological features identified in airborne lidar data. (MSHE Ledoux/ ModeLTER)

earthworks and subtle terrain variations caused by archaeological features are concerned. The total error for a lidar system is the sum of the errors from each subsystem, namely the laser rangefinder, the GPS and the IMU orientation errors. These sources of error and the calculation of error budgets are discussed by Baltsavias (1999) and Habib et al. (2008), but the main error sources can be summarized as:

Figure 2.9: Vertical errors as the result of horizontal shift. After Hodgson and Bresnahan (2004)

• Platform navigational errors • GPS/IMU navigational errors • Laser sensor calibration errors (range measure­ ment and scan angle) • Timing resolution • Boresight misalignment • Terrain and near-terrain object characteristics

error budget of 20–30 cm. The largest errors are usually in areas of strip overlap, where alignment problems are most apparent, and in these areas the errors can be of direct relevance where low

Errors may be either vertical (along the Z axis) or planimetric (shifts in the XY plane) and while the two are obviously related they are usually quantified separately in accuracy

2  An overview of airborne and terrestrial laser scanning in archaeology

27 Figure 2.10: Boresight misalignment or GPS/IMU navigational errors can lead to strip alignment problems. Here pitched roofs from adjacent strips (points coloured by strip) do not quite match up

reports. In commercial applications accuracy analyses usually report vertical accuracy, while planimetric accuracy (XY) is treated as secondary (Figure 2.9). Errors may be described as random, systematic or terrain dependent. The dominant source of random error is position noise from within the GPS/IMU navigational system, which will lead to related noise in the derived point cloud (Figure 2.10). Such point cloud coordinate errors are independent of the flying height, scan angle and the terrain being scanned. Systematic errors include those internal to the laser scanner (errors in range measurement, boresight misalignment, lever arm offset and mirror angle are common) and some errors from the GPS/IMU system (INS initialization and misalignment errors and multi-path returns). These errors will appear throughout the dataset. Terrain dependent errors are introduced through the interaction of the laser beam with the terrain and off-terrain objects. In steeply sloping terrain or areas with off-terrain objects, and at higher scan angles, beam divergence may be increased and result in vertical errors due to horizontal positional shift (as shown in Figure 2.11). References

Baltsavias, E., 1999. A comparison between photogrammetry and laser scanning. ISPRS Journal of Photogrammetry and Remote Sensing, 54(1), 83–94. Habib, A.F., Al-Durgham, M., Kersting, A.P. and Quackenbush, P., 2008. Error Budget of Lidar Systems and Quality Control of the Derived Point Cloud. In ISPRS Congress, Beijing, China, 3–11 July 2008, 37Part B1, 203–9. Hodgson, M.E. and Bresnahan, P., 2004. Accuracy of airborne Lidar-Derived Elevation: Empirical Assessment and Error Budget. Photogrammetric Engineering and Remote Sensing, 70(3), 331–39. ASPRS Horizontal Accuracy Reporting Guidelines. http://www.asprs.org/a/society/committees/ standards/Horizontal_Accuracy_Reporting_for_

Figure 2.11: The scan parameters of swath width, laser footprint and scan angle all affect the quality of the collected data

Lidar_Data.pdf. Accessed 20 January 2012. ASPRS Vertical Accuracy Reporting Guidelines. http://www.asprs.org/a/society/committees/lidar/ Downloads/Vertical_Accuracy_Reporting_for_ Lidar_Data.pdf. Accessed 20 January 2012.

Sources of error: GPS/IMU Accurate systems to record the precise location and attitude of the scanner as it is transported across the landscape are vital. The GPS (Global positioning system) in the aircraft provides data on its location while the IMU records its pitch and yaw. For all airborne platforms GPS/IMU system errors can account for a significant portion of the total error in the final calculated coordinates of the lidar returns. To accurately measure the location of the GPS sensor in

28

Rachel S. Opitz

Table 2.3: GPS system accuracy levels (After Toth 2010, Table 1)

GPS system type Differential with base stations Differential with network Satellite based differential Real time kinematic (RTK) Virtual reference systems

Accuracy cm cm sub-m cm cm



the aircraft reference base stations are needed. The aircraft GPS should not be too far from a base stations at any point in the survey (i.e. short baseline lengths are best) so a network of GPS base stations should be distributed throughout the survey area. Further, it is good practice to carry out a quality control survey during or after the lidar survey to ensure that the coordinates generated by the lidar survey match the coordinates recorded in an onthe-ground survey. Therefore, quality control surveys are conducted in which differential GPS surveyed points or RTK (real time kinetic GPS) profiles (Habib and Van Rens 2009) are collected throughout the study area in a variety of types of terrain (e.g. flat, vegetated, sloping, hard surfaces) and are compared with the lidar generated points or profiles of the same locations. While it is unusual in practice, corrections to the lidar data may be made on the basis of the quality control survey data. Since the error budget of any integrated GPS/IMU-based system is mainly defined by the GPS performance, the GPS positioning method applied in the georeferencing system has a direct bearing on the accuracy of the lidar point cloud. While, there are a variety of possible GPS configurations available (Table 2.3), in practice most data collectors employ the differential GPS with base station method.This method provides good accuracy near base stations, but over larger distances (> 50 km) accuracy decreases. Some data collectors employ the network-based differential system because the accuracy of these systems is not dependent on distance and, in areas with good networks of Continuously Operating Reference System (CORS) instruments accuracy is good. However, good CORS networks are not available everywhere, and at present their use is largely limited to parts of North America and Europe. GPS and base station systems normally provide data with accuracy between 4 and 7 cm range for shorter baselines. Within a good CORS, a network-based differential GPS system can supply data at accuracy levels of 2–3 cm.

Sources of error: airborne scanner parameters The basic parameters of a lidar system also have a direct impact on ground point density and the accuracy of measurements. • The scanning frequency, or pulse repetition rate, is the number of pulses emitted by the laser scanner per second. Present-day systems support frequencies of up to 500 kHz (500,000 pulses per second) for discrete return systems and 120 kHz for full waveform systems, and can be operated at lower frequencies. The density of the final point cloud is correlated to the scanning frequency. Lower point densities equate to a lower resolution DTM. In vegetated areas, lower resolution datasets may be more prone to classification errors. • The pulse footprint width can affect the horizontal and vertical accuracy of the return value and the real spatial resolution of the data. The pulse emitted from a laser scanner will diverge slightly (as opposed to a true laser system where the beam remains coherent) resulting in a cone-shaped pulse, with the wider base of the cone at the target (the ground or near ground objects). The pulse diameter increases as the pulse travels away from the sensor. At a typical flying height of 1000 m the pulse diameter, or footprint, will be around 30 cm. Larger footprints (the result of greater pulse divergence) correspond to lower accuracy because the signal to noise ratio is decreased and it becomes more difficult to identify peaks in the waveform, and because more objects and/or terrain variations are encountered by a single pulse. • Scan angle, or half-angle, is the angle between the zenith, directly under the scanner and the outermost position to which a pulse is directed. Most lidar systems will allow for a scan angle of 60 degrees, or a half angle of 30 degrees. For applications where penetration through vegetation is important a narrower scan angle is recommended, typically less than 40 degrees, or a half angle of 20 degrees. Larger scan angles mean that pulses on the outside of the swath will encounter flat terrain at an angle, resulting in a larger footprint and decreased accuracy, and in vegetated areas are more likely to encounter trees and other impenetrable objects resulting in fewer ground returns. • The swath width is defined by the scan angle and flying height. Flying at greater heights and with a larger scan angle with result in a wider swath, and a larger area may be covered in less time. However, an increase in flying height will decrease the number of returns from the

2  An overview of airborne and terrestrial laser scanning in archaeology

29

ground under vegetation canopy, resulting in a decrease in accuracy in the terrain model in these areas.

References

Habib, A., 2010. Airborne Lidar Mapping. In Bossler, J., (ed.). Manual of Geospatial Science and Technology. 2nd edition. Boca Raton, FL: CRC Press, Taylor and Francis. Habib, A. and Van Rens, J., 2009. Quality Assurance and Quality Control of LiDAR Systems and Derived Data. http://www.asprs.org/a/society/committees/ lidar/AKAM_LiDAR_Calibration.pdf NERC ARSF. Mission planning. http://arsf.nerc. ac.uk/howtoapply/planning.asp. Accessed 14 December 2011. Toth, C., 2010. Airborne lidar technology: the state-of-the-art and future trends. Latin American Remote Sensing Week Regional ISPRS Conference. 4–8 October 2010, Santiago, Chile. http://www. lars.cl/biblioteca/LARS_Thot.pdf. Accessed 28 January 2012. Wehr, A., and U. Lohr, 1999. Airborne laser scanningan introduction and overview. ISPRS Journal of Photogrammetry and Remote Sensing, 54, 68–82.

Data management Because lidar datasets are typically very large, strategies for data management are important. Lidar data is collected in long strips or swaths (Figure 2.12) which overlap by approximately 20%. Many lidar surveys used by archaeologists will cover an area on the scale of 50–200 km2. A lidar dataset collected at a nominal resolution of 8 pts/m2 will contain approximately 5,000,000 points in a 0.5 km2 area and result in a 130MB LAS file. Datasets covering hundreds of square kilometres, resulting in many gigabytes of data, will be unwieldy if read into memory on an average computer in their entirety and older graphics cards may struggle to render large point clouds fluidly. Tiling a lidar dataset is an effective way of organizing the data for visualization, processing and analysis and is standard practice in the lidar industry. Dividing the dataset into regularly sized squares (e.g. 1km2 or 0.5km2) will allow the processing of individual tiles on a desktop or laptop computer. Lidar data can be stored as a point cloud in ASCII format as a human-readable list of coordinates and attributes. This format maximizes compatibility with a variety of GIS, modelling and VR software but file sizes are very large. The binary format LAS 1.0 standard was introduced in 2003 and its file sizes are typically

35–80% smaller than ASCII point clouds. The LAZ format is a compressed LAS format and further reduces the file size to 7–20% of the LAS size. However, most software does not currently read the LAZ format and therefore it is, for the moment, primarily an archival format. Lidar data is often transformed into rasters or meshes to create continuous surfaces (see interpolation above). It should be noted that the transfer from point cloud to surface usually involves some minor loss of spatial resolution and geometric fidelity. Mirroring the tiling schemes for point clouds, rasters are often tiled for efficient loading and processing. Raster catalogues are therefore a common means of organizing and managing tiled raster collections. In parallel, meshes produced in terrestrial scanning projects are often divided into regions of a regular size or based on natural breaklines in the object being scanned in order to facilitate data processing and visualization.

Dissemination and Metadata WebGIS dissemination Sharing lidar data and making it publicly available presents significant challenges because of the volumes of data involved. Dissemination to the public and the research community at large is ideally carried out via a webmapping

Figure 2.12: Adjacent flightstrips 181 and 182, coloured in green and red respectively, overlap by about 20%

30

Rachel S. Opitz service. Two regional lidar surveys illustrate possibilities for this kind of dissemination. The Murcia lidar survey in Spain is hosted on the web using a platform developed by Dielmo3D, which allows users access to the point cloud data, areas of which may be interactively selected and downloaded. This system does not provide ready-made raster visualizations of the data, but rather leaves it to the user to perform upstream tasks. The Jura Canton in Switzerland takes a different approach. Their system has integrated their lidar data into a webGIS along with aerial photographs, topographic maps, cadastres and other datasets. They provide access to basic elevation and hillshaded rasters and a canopy height model. While this approach does not allow users to access the raw data, it is in many ways more intuitive and the data is easily interpreted through the webGIS. Several services now exist to provide information on the locations of existing lidar surveys and encourage sharing and redistribution of lidar datasets. Lidar-Online by Dielmo3D is based on a social network model, while the OpenTopography network (www. opentopography.org/) focuses on US based datasets and cooperates closes with the USGS (United States Geological Service). The NERC ARSF (arsf.nerc.ac.uk/) also provides a service to locate and request access to their existing datasets. These data and metadata redistribution schemes will play an important role in the use of lidar in the archaeological community by providing access to relatively low cost existing datasets for a growing number of regions around the world. Mesh and 3D object dissemination While aerial lidar is primarily shared via webGIS platforms, the results of terrestrial or object scanning are usually disseminated as meshes or 3D polygonal objects. These files may simply be downloaded from a server, may exist within a virtual museum or collection environment (e.g. The Virtual Hampson Museum), or may be hosted and interacted with via an online game engine (e.g. Unity web player). Initiatives to support the dissemination of 3D cultural heritage objects derived from terrestrial or object scanning include the 3D COFORM Consortium, whose mission is “to establish 3D documentation as an affordable, practical and effective mechanism for long term documentation of tangible cultural heritage” and the E-Curator Project, a group developing post-processing workflows to share

the results of the scanning of museum objects over the web. Challenges in sharing of terrestrial and object scanning results over the web include compatibility between a variety of software and file formats used by producers and consumers of the models, the size of the datasets and associated issues of compression and subsetting of models, and the inclusion of consistent semantic information with each model. Archival and delivery formats Preparing lidar datasets for archiving or redistribution to other researchers or users requires a choice among file formats. A simple ASCII text file containing fields for XYZ coordinates, intensity information, colour data, and associated information such as scan angle, GPS time and return number is often used for archival purposes because of the widespread ability of software to read ASCII files and the human-readable nature of this data format. However, ASCII files are significantly larger than other available formats, which might be used for transfer and are read by an increasing number of software packages. Aerial lidar data can currently be stored as a point cloud in the LAS, LAZ, e57 or MrSID formats, and interpolated surfaces and their derivatives are stored as rasters in ASCII, GRID, geoTIFF or other similar formats. Terrestrial scan data is typically stored and exchanged as points in the Leica PTS or PTX formats, and as a meshed surface in OBJ, VRML, X3D and PLY formats, among others. Metadata standards for lidar data are rapidly evolving. Because lidar data has applications in many communities, when preparing metadata it is important to consider the needs of a variety of users. At a minimum, metadata for lidar datasets should include: • • • • • • • • • • •

The file name The date or dates of survey The specifics of the scanning system A description of the survey area A description of the project A scan number (out of xx scans) The number of points in the file The nominal point density The nominal scanning distance to the target The data collection conditions The flight plan (for aerial surveys)

Such standards are suggested by the ADS Guides to Good Practice for Laser Scanning (Payne 2009). The core archaeological digital

2  An overview of airborne and terrestrial laser scanning in archaeology archives, The Digital Archaeological record (www.tdar.org) and the Archaeology Data Service (http://archaeologydataservice.ac.uk/), are not specifically geared toward the hosting of massive datasets like those created through laser scanning, although both organizations are actively working to incorporate laser scanning and other 3D data into their systems. Further, the interdisciplinary potential of lidar datasets mean that an archaeological archive may not be the most appropriate venue for deposit. Rather, it may be preferable to archive the laser scanning data with a lidar-oriented service, with links to the project’s archaeological archive. References

Austin, T. and Mitcham, J., 2007. Preservation and Management Strategies for Exceptionally Large Data Formats: ‘Big Data’. Archaeological Data Service and English Heritage. http://ads.ahds. ac.uk/project/bigdata/index.html. Accessed 20 January 2012. Brown, D., 2007. Archaeological Archives. A guide to best practice in creation, compilation, transfer and curation. AAF. http://www.britarch.ac.uk/ archives/Archives_Best_Practice.pdf. Accessed 20 January 2012. Payne, A., 2009. ADS Guides to Good Practice. Laser

Scanning. http://guides.archaeologydataservice. ac.uk/g2gp/LaserScan_Toc. Accessed 20 January 2012.

E-Curator http://www.ahessc.ac.uk/e-curator# technologies Inspire Directive http://inspire.jrc.ec.europa.eu/ Jura Regional Lidar Survey http://geoportail. jura.ch/ Lidar Online https://www.lidar-online.com/ Murcia Regional Lidar Survey http://www. cartomur.com/ Open Topography www.opentopography.org/ The Archaeological DataService http://archaeology dataservice.ac.uk/ The Digital Archaeological Record http://www. tdar.org The Virtual Hampson Museum http://hampson. cast.uark.edu/ Unity http://unity3d.com 3D COFORM www.3d-coform.eu

Acknowledgements My thanks to Fred Limp and Žiga Kokalj for their assistance and comments – all errors are my own.

31

3 Airborne laser scanning and archaeological interpretation – bringing back the people Michael Doneus and Thomas Kühteiber As with every other prospection method, data provided by Airborne Laser Scanning (ALS) have to be interpreted to add archaeological value. Interpretation is considered in a holistic way, where the whole chaîne opératoire from project planning to the final filtered ALS dataset has to be evaluated to estimate the archaeological potential. Decisions upon the most useful combination of visualisation techniques, interpretive mapping (usually done in a GIS-based environment), and ground inspection are strongly interrelated and should not be done in isolation. Additionally, integrated with other sources of information (e.g. other prospection methods, historical documentary research) the detailed topographical information from ALS can help to deepen understanding of the archaeology in its landscape setting on multiple scales.   This is demonstrated by studying the religiously-motivated choice of location for a Carmelite friary. Going beyond standard descriptive mapping, the choice of location is interpretively explained combining etic and emic observations based on multi-scale ALS information and the inclusion of historical context and perception variables. The case study demonstrates that we should not attempt to distinguish between true and false interpretations, but between good and bad ones. This distinction has to be based on the arguments used and whether the interpreter was following scientific rules. A good interpretation should integrate information from multiple sources generating archaeological information which extends beyond the obvious. Keywords: interpretation, visualisation, ALS, archaeological prospection, lidar, post-medieval, friary, landscape archaeology

Introduction During the last decade archaeological applications of airborne laser scanning (ALS) using the technology of light detection and ranging (lidar/ LiDAR: Hug et al. 2004; Wagner et al. 2006) have been increasing, and there are now many archaeological publications reporting on use of ALS-derived data. As with most methods that are introduced to archaeology, the main topics of the first wave of publications are methodological and focus on types of applications. This is especially true for ALS because, at least in Europe, many countries have policies for total scanning of their territory, which gives archaeologists easy, and

often free, access to ALS-derived digital terrain models (DTM), collected country-wide by order of governmental organisations. As country-wide lidar derived DTMs become available, systematic archaeological interpretation of large datasets has started in various institutions (e.g. Schmidt et al. 2005; Doneus et al. 2007; Boos et al. 2008; Challis et al. 2008; Opitz 2009; Bofinger and Hesse 2010; Ainsworth et al. this volume; Crutchley this volume). At the same time, the problems of interpretation have only recently become more of a focus for discussion in recent conference-sessions and publications (e.g. Doneus et al. 2008b; Campana et al. 2010;

3  Airborne laser scanning and archaeological interpretation – bringing back the people AARG 2010 and 2011; papers in this volume). However, the number of publications dedicated to the topic of interpretation is still comparatively small. This article explores the topic of ALS-based archaeological interpretation and in short presents the various stages/levels of interpretation that go beyond the mere description of geometric shapes and their geographic location. The potential of ALS-derived DTMs for a ‘deep’ interpretation will be illustrated using a case study of a friary in the eastern part of Austria (parts of this paper are taken from Doneus and Kühtreiber 2012).

ALS and interpretation As with all other prospection methods, archaeological interpretation of ALS data has to go beyond mere geometrical description of what can be discerned in the visualisations of the data. Interpretation has to add to our archaeological knowledge by helping the archaeologist to explain and understand past societies on the basis of their material remains. Here, one is immediately confronted by a problem of scale. Macro-structures (societies, cultures, settlement patterns) are recognized and reconstructed with the help of micro-elements (artefacts, ecofacts e.g. Hinz 2008, 33), but often explained using factors from the macro-level (mainly environmental data like climate, hydrology, relief and pedology). However, different sets of factors would appear to be relevant depending on the scale of the investigated area. The scale adopted for research also marks a significant difference between functional-processual landscape archaeology and postprocessual approaches. While processualists are chiefly interested in an etic perspective of society, of general systemic relations and of macro-structures, and attempt to filter individual phenomena out of their statistical analyses as ‘deviations’, ‘blips’ or ‘background noise’ (Hodder 2000, 26), post-modern approaches criticize the neglect of the individual (Thomas 1993, 26; Chadwick 2004) and emphasise an emic approach. Despite their one-sided narrow perspectives, both approaches seem to be relevant and should be integrated. In doing so, archaeologists have to introduce both individual agents and their social and environmental context into their interpretations. Depending on the specific purpose at hand,

ALS data can be used to derive precise models of the terrain (bare ground) as well as of the surface (i.e. the terrain plus all objects) or a mixture of both (as often used in archaeology, where remains of walls and buildings are kept in the terrain dataset). Thus ALS-derived data provides archaeologists with detailed and multiscaled topography which is a powerful resource for interpretation. Similarly to working with aerial photographs, interpretation of ALS data can therefore be done on different scales: both the individual feature and the topographical landscape setting. In that way, both micro- and macro-levels are integrated into an explanatory model, which in sociology is called ‘deep explanation’ (Opp 2004, 64). However, to understand and evaluate the results of an archaeological interpretation of ALSderived data, it is necessary to consider the whole workflow from the design of a scanning project to its final results. The whole process of ALS can be visualised as a long chain of individual workflow steps. Each link of the chain represents decisions or processes that influence the final result and therefore must be documented in its metadata. Also, the level of control the archaeologist can exercise on each link in the process chain will vary greatly. The process chain can be subdivided into various groups, here divided into three: 1. Evaluation of metadata a.  Data acquisition b.  Data processing 2. Detailed interpretation a.  Visualisation b.  Interpretative mapping c.  Ground inspection 3. ‘Deep’, integrated multi-scale interpretation These groups are not sharply bounded, but have large overlaps and interrelate strongly with each other. The process of detailed interpretation should not be done in isolation, but rather integrate information from different prospection methods, excavation, environmental studies and the like. However, experience shows that this is often not the case. Therefore, ‘integrated, multiscale interpretation’ is included here to stress its importance. Evaluation of metadata It may seem unnecessary to include the data acquisition phase in the interpretation process, but as is the case for all other prospection methods, the parameters chosen have an important impact on the interpretability of

33

34

Michael Doneus and Thomas Kühteiber the resulting data (Opitz this volume). The key factors for ALS are choice of sensor, time of flight, flying height above ground, scanner field of view, and point density (for detailed assessment see Doneus and Briese 2011). Only in the rare case of an archaeologically-commissioned scanning project, will the archaeologist be able to exert full control on the data acquisition (although even in such a case the date of flight is usually a compromise allowing for local flying regulations, aircraft availability, climatic conditions and local weather). The inevitability of compromise is also true for data processing steps, which involve echo detection (in case of full waveform scanners), georeferencing, strip adjustment, filtering (choice of software and filtering strategy), and data export and terrain modelling (Doneus and Briese 2011). In forested areas, the filtering process is of crucial importance for the archaeological interpretability of a DTM and should therefore be explicitly documented (i.e. software used and its version, as well as parameter settings). When working with general purpose data, these factors are of course given; nevertheless, it is possible to evaluate the archaeological potential of the data and aid the interpretation by inspecting the metadata. Data are usually exported as .las-files, which has become a standard. Depending on the primary filtering undertaken by the data supplier, the file specification will allow a user to chose which points to use for terrain modelling (e.g. terrain, low vegetation, high vegetation, buildings). Again, the process of terrain modelling is not straightforward and depending on the software/algorithm, results can differ to a certain degree (for an overview see Conolly and Lake 2006, 90ff.). Unfortunately, information on the software, algorithms and parameters used are usually not part of metadata when general purpose data are provided as premodelled ASCII-grids with regular grid-spacing (often 1 m). Detailed Interpretation Up to this point, when using general purpose data archaeologists have had more or less no influence on the resulting point cloud, which is one of many possible representations of the terrain under investigation. Whether it is suitable for a specific archaeological purpose can only be judged by evaluating the metadata. Detailed interpretation as the next step is usually used synonymous with ‘interpretation’ and comprises

visualisation, interpretative mapping and ground inspection, which should not necessarily be seen as having a linear relationship; they are rather performed in an integrated, iterative way (e.g. Ainsworth et al. this volume). To access the archaeological information content of the point cloud, several techniques can be used to derive geomorphometric models and visualise them (for an overview, see Hengl and Reuter 2009). This topic is still young and visualization of ALS-derived terrain models (see Challis and Kincey this volume) has been identified as a future challenge for archaeological interpretation (Millard et al. 2009), though literature on the topic is growing (Bewley et al. 2005; Devereux et al. 2008; Doneus and Briese 2006; Humme et al. 2006; Hesse 2010; Kokalj et al. 2010; Kokalj 2011). A key point is that if the archaeologist can get hold of the filtered point clouds in the form of .las-files or ASCII grids, they can fully control the visualisation. In this case, representations of the data as contour lines, slope image, various shadings, PCA, skyview factor, local relief model and the like can be calculated, adding to the interpretation process and should be applied in combination whenever possible. Archaeological interpretative mapping is certainly the main step of the interpretation workflow and typically takes place in a GISbased environment (though see Ainsworth et al. for discussion of using ‘ortholidar’ as a field tool). ‘Potential’ archaeological sites and features will be identified and drawn as lines or polygons. Interpretation is a subjective process and confidence in the product will vary from site to site (and between individuals). There are many natural and recent features that can affect the interpretation of ALS data, and to distinguish a pile of wood from a barrow can be a very difficult task when working only with ALS data. Hence, ‘non-sites’ may be identified as potential archaeological sites or, on the other hand, natural meaning can be given to features which are in fact archaeologically relevant. Therefore, to evaluate the accuracy of this identification process, these ‘potential’ sites will need visiting to ascertain their character. These visits are essential to the production of an image interpretation key, which should help to improve future interpretation. Integrated, multi-scale interpretation Although an interpretative map may mark the

3  Airborne laser scanning and archaeological interpretation – bringing back the people end of an ALS-based interpretation, there is additional benefit in the data, namely the multiscale detailed topographical information, which can help to deepen our understanding of the archaeology in its landscape setting. Introducing the terms ‘landscape’ and ‘landscape archaeology’, it should be made clear that these terms are defined here on the basis of concepts of space taken from the philosophy of nature: ‘physical space’ as physical entity, and ‘cognitive space’, which is filtered through human perception or imagination (Hartmann 1980). Thus landscape has a meaning and exists both physically (set by nature and, in a historical process and in complex interaction, knowingly and unknowingly shaped by society) and as a concept of its residents (or dwellers). As discussed above, interpretation should include both micro- and macro-levels. Therefore, a locational analysis (to mention a typical application) should not be based solely on environmental factors (slope, distance to water, soil quality and the like) from the macrolevel, but has to include human agency and individuals at the micro-level. In order to extend interpretation beyond interpretative mapping and pure description, it is therefore important to address the notion of cognitive space. Hodder (2000) suggests a narrative approach, in which, while beginning with an individual, a link with the macro-level is established. In our opinion such narratives should include both etic and emic aspects, to explain and understand formation and change of social structures in an inter-subjective and comprehensible way. That means that an approach must be sought which can help to overcome the dualism between micro- and macro-structures and which will be true to scientific principles. For decades now there have been approaches in sociology that root social phenomena in individual actions, and which are grouped together as ‘methodological individualism’ (or the ‘structural-individualistic approach’) – a term coined by Karl Popper (2003) to mean that explanations cannot be exclusively based on macro-structures (Esser 2002, 27). Knut Petzold (2007) has recently introduced this approach to archaeology, whereby explanations of social phenomena can and must be drawn from a multitude of individual actions (Opp 2004, 43). This results in ‘deep’ explanations with a high informational content, which can be described as ‘interpretive explanation’. In a recent publication (Doneus and Kühtreiber 2012) the

idea of applying the structural-individualistic approach to landscape archaeology is set out in detail. The following case-study is taken from this publication and illustrates the chain of interpretation presented so far.

Case Study: the siting of the friary St Anna in der Wüste The remains of the friary complex of St Anna in der Wüste (St Anna in the Wilderness) lie in the Leithagebirge (Leitha Hills) on the border between the Austrian federal regions of Lower Austria and Burgenland (Figure 3.1), which was the earlier border between the Holy Roman Empire and the Kingdom of Hungary. The area of this former friary of the Discalced Carmelites is wooded, mainly with deciduous trees (oak and beech – Figure 3.2), and was part of an area scanned in 2006 and 2007 at a high resolution and according to archaeological principles as part of the project LiDAR-supported archaeological prospection in forest areas (Doneus et al. 2008a). The ruins of the friary are prominent above-ground traces and survive in the contemporary landscape alongside old trackways, former quarries and boundary markers. Extensive surviving written and pictorial source materials make the friary and its surroundings an ideal subject for a landscape archaeological interpretation that integrates the emic and etic aspects mentioned above. An interpretation will be offered of the position and above all the boundaries of the friary. Airborne Laser Scanning at St Anna in der Wüste The airborne laser scan was acquired for archaeological purposes to defined parameters (Table 3.1). Archaeologically speaking, using a full waveform scanner, the quality of the resulting DTM could be enhanced, especially in areas with low and dense vegetation (Doneus and Briese 2011, Figure 3.3). The scan date in late March/early April ensured that the deciduous trees and most of the understory were still without leaves. The point density is a function that mainly relies on the pulse rate, field of view, flying height above the terrain, overlap between two neighbouring scan stripes and speed of the aircraft. The parameters were chosen to allow a point density of roughly 7 points per square metre. This would – after filtering – result in a

35

36

Figure 3.1: Location of St Anna in der Wüste. © Martin Fera

Michael Doneus and Thomas Kühteiber

final DTM with a grid width of 0.5 m (Figure 3.3), which seemed to be a reasonable costbenefit for a footprint of roughly 30 cm. The large overlapping area (50%) of two neighbouring strips was an advantage during the process of strip-adjustment for advanced georeferencing (see also Doneus and Briese 2011). A large overlap ensures that every object on the ground will be hit at least twice from two view points, and that there is a good chance that some of the oblique laser pulses will hit the ground below conifers where the almost vertical laser pulses of a system operating with a narrow scan angle might not get through. For classification of the ALS points into terrain and off-terrain points (so called filtering), the software SCOP++ (see Kraus and Pfeifer 1998; Kraus and Otepka 2005) was used (for a description see Doneus et al. 2008a), where parameters were set to fit our archaeological purpose (i.e. remains of walls and buildings were more or less kept in the terrain dataset). Simple shaded relief has become a standard visualisation

of ALS derived DTMs (Figure 3.4), and while it is easily read, its reduced information content is a major drawback (Devereux et al. 2008; Doneus and Briese 2006, Fig. 2). Recent publications have suggested other more or less elaborate techniques to display relief information for archaeological interpretation (see above; Opitz this volume; Kokalj et al. this volume). From these, Local Relief Model (LRM, Hesse 2010), a combination of slope and hillshade (Doneus and Briese 2011, 66ff), and Openness (Yokoyama et al. 2002) which is similar to Sky View Factor (Kokalj 2011), were used. The computation of LRM and Openness is not straightforward. Both are calculated applying various processing steps, where a kernel is used to derive statistical parameters from the original DTM. The resulting visualisations will differ depending on the kernel size,. Therefore, different topographic settings (relief, size and structure of objects) will need different parameters to compute LRM and Openness. In addition, the visualisations are not immediately

3  Airborne laser scanning and archaeological interpretation – bringing back the people

37 Figure 3.2: Vertical photograph of the study area, July 2000. ©bmlvs/ luaufklsta

 Data Acquisition Purpose

Archaeology

Time of Data Acquisition

26 March – 12 April 2007

Equipment

RIEGL Airborne Laser Scanner LMS-Q560

Scanning System

Full Waveform Scanner

Point-Density (pt per sq m)

ca. 7

Scanner Type

Full-Waveform

Scan angle (whole FOV)

45°

Flying height above ground

600m

Footprint

ca. 0.3m

Speed of aircraft (TAS)

36 m/s

Laser Pulse Rate

100,000 Hz

Scan Rate

66 Hz

Strip Adjustment

Yes

Filtering

Robust interpolation (SCOP ++)

DTM Modelling

Advanced Kriging



comprehensible, and can only be interpreted correctly when the computation process is understood and taken into consideration. This is evident, for example, in the case of a terrace

or road which is cut into a slope (Figure 3.5). The LRM will display a sequence of positive and negative relief, which can be also read as an earthwork bank accompanied by a ditch. In

Table 3.1: Parameters of the ALS flights used in this paper

38

Figure 3.3: St Anna in der Wüste – perspective view of filtered DTM with surrounding wall (white)

Michael Doneus and Thomas Kühteiber

this situation, a combination of positive and negative openness gives a more immediately understandable image, where ridges are displayed as dark and crests as white lines (Figure 3.6). In areas of shallow relief, application of LRM produces clear and distinct visualisations, as seen in the case of highly eroded round barrows (Figure 3.7). While in the slope-hillshadecombination (Figure 3.7a) only nine barrows are visible, the interpretation of the LRM suggests at least 15 (Figure 3.7b). This demonstrates the importance of combining various visualisation techniques, since all of them filter and display information content in a different way. Mapping and interpretation The visualisations (a combination of hillshade and slope, LRM, Openness) were loaded into a GIS system and used as a basis for mapping. Archaeological features were drawn as polygons and a table attached, storing information on feature-interpretation, relief of feature, the name of the interpreter, and date of interpretation. While earthworks, buildings, walls, field-systems, terraces and pathways are often obvious, smaller structures are less so. On basis of the DTM it is often impossible to distinguish natural and recent features from archaeologically relevant

structures. Therefore, field visits are a necessary component of any detailed interpretation. Using a GPS, mapped features can be visited and identified. During site visits to St Anna, additional information was gathered which otherwise could not be derived from the ALS data, including construction materials (stone type and mortar), mason’s marks and artefacts which helped advance interpretation and provided dating evidence. In the field, small DTMfeatures could be identified as dense bushes, fallen trees, stones, tree stumps and the like. This information, as well as the general appearance of earthwork structures was documented using photographs. The georeferenced photographs (the camera was connected to a GPS) were imported to ArcGIS (Figure 3.8), supporting the production of an interpretation key, which was used to refine interpretation and which will be a valuable basis for future projects. However, field visits were often inconclusive. Thus, while it may be easy to distinguish a barrow from a large tree stump, and in that way falsify an assumed archaeological feature, in most cases, the situation is less obvious. It may be impossible to tell a barrow from a pile of stones or a geological outcrop without application of further prospection techniques

3  Airborne laser scanning and archaeological interpretation – bringing back the people A

B

C

D

(e.g. GPR) or invasive methods. The same is true for features in extremely shallow relief, which are not recognizable in the field although they have been visualised in the DTM. A field inspection therefore should not be regarded as ‘ground truthing’, as it will never be able to inform us about the ‘truth’. Field observation is an additional prospection technique, which can add to our contextual knowledge and support further interpretation. Interpretation is context based. That is to say that interpretations are provisional, and depending on our understanding of the historical, topographical, environmental and research- context, may change. As a consequence, an interpretation of ALS data is never completed. It is an experience- and context-driven estimation on the archaeological content of a scanned area

that survives in relief. It is therefore not useful to distinguish between true and false interpretations, but between good and bad ones (see Palmer this volume). This distinction has to be based on the arguments used and whether the interpreter was following scientific rules. A good interpretation should generate archaeological information, which extends beyond the obvious. Visualisation, interpretative mapping, topographic analysis, field visits and historical documentary research for the project were done in an interactive and iterative way. A description of the physical space in its present condition, which is outlined in the following, served as starting point for the ‘deep’ interpretation. The physical remains Most side valleys in the Leithagebirge are gorges

39 Figure 3.4: Visualisation of the filtered DTM. A: hillshade; B: combination of hillshade and slope; C: positive and negative openness; D: local relief model and hillshade

40

a

Michael Doneus and Thomas Kühteiber

b

c

Figure 3.5: Road cut into a slope. While the situation is evident in the (a) hillshade, the (b) LRM – as a result of the computation process – displays an ambiguous sequence of positive and negative relief.This is the result of the computation process, where the large-scale relief (a low-pass filtered and purged DTM – see Hesse 2010, Fig. 1) is subtracted from the original DTM, as demonstrated in the cross-section (c)

3  Airborne laser scanning and archaeological interpretation – bringing back the people

41

with a v-shaped cross-section, but around the friary complex the valley widens from east to west along the Arbach stream. This little valley is flanked by two ridges, which rise about 50 m above the Arbach (Figure 3.3). The little valley basin today is occupied by the remains of the St. Anna in der Wüste friary complex, abandoned in 1783, which lies at the foot of the so-called Schlossberg (Castle Hill). Schlossberg is the site of a prehistoric hilltop settlement, repeatedly occupied between the Later Neolithic and the Hallstatt periods (Melzer 1980, 77). A group of low-relief, presumably Early Iron Age, round barrows is spread across the north-eastern side of a small hill some 400 m south of the earthwork (Figure 3.7). Scharfeneck Castle was erected within the hillfort in the late 14th century. Since the A

B

Figure 3.6: Situation of Figure 3.5 as seen in a combined visualisation of positive and negative openness. Ridges are displayed as dark and crests as white lines

C

Figure 3.7 (above): Highly eroded round barrows. A: Combination of slope and hillshade; B: combination of LRM and slope; C: interpretation

Figure 3.8 (left): Detailed interpretation involves mapping and site visits. Photographs taken on site are re-imported into ArcGIS to support further interpretative mapping

42

Michael Doneus and Thomas Kühteiber

Figure 3.9: DTM. Colour-coded combination of hillshade and slope of the area around the friary complex: a friary. – b hermitages. – c ponds. – d ruined farm. Light source to the northeast

Middle Ages the locality has been further shaped by the quarrying of limestone for building stone and the manufacture of mortar. An aerial photograph from June 2003 shows the thickly-wooded character of the area (Figure 3.2), comprising mostly mixed oak and beech, with thick undergrowth in some areas. The immediate vicinity of the friary is largely free of tree cover. The relatively late foundation of a noble residence – the first reference to a noble family there is 1383, while the castle is first mentioned in 1417 – can probably be explained by its location on the border of the Duchy of Austria and the Kingdom of Hungary: The political affiliation of the place changed several times over the course of the centuries (Lampel 1899, 49, 51; Mochty 1998). The castle was abandoned in the

second half of the 16th century, after lightning had partially destroyed the keep in 1555. In 1683 the ruin served the local population as a place of refuge from Turkish troops, who laid siege to Vienna and laid waste large parts of eastern Austria as the Ottoman Empire expanded eastwards. Following the desertion of the castle in the 16th century, a hermitage of the Discalced Carmelites named St Anna in der Wüste was founded in 1644 on the initiative of the Imperial widow Eleonora (Schatek 1938, 1–38; Aguinaga 1993, 6–52; Mayer 1900, 85–91). The remains of the friary, arranged like a Carthusian complex around a church are clearly visible in the central part of the valley floor (Figure 3.2). The shaded plan of the terrain model shows further details (Figure 3.9). Individual hermitage buildings,

3  Airborne laser scanning and archaeological interpretation – bringing back the people most of which survive only as foundations (Figure 3.10), can be clearly seen in the plan (Figure 3.9). Some of the surrounding linear features are former tracks, others are probably field boundaries. Ponds beside the Arbach (Figure 3.9c) and remnants of farm buildings on the valley slope north of the river (Figure 3.9d) are clearly recognizable remains of the hermitage’s economic base, which may have utilized pre-existing structures from the noble residence. Among the known structures the remains of limestone quarrying and processing are most common, unsurprising since the easily workable Leitha limestone was used for dressed stone and mouldings over a wide region from at least the later Middle Ages onwards (Starzer 1900, 7.17). Numerous quarries of varying size are scattered over the entire area, including a large abandoned quarry north of the ruined friary farm mentioned above. Its structure can be seen more or less clearly in the terrain model and several later tracks cut across it. A lime kiln situated a little north of the farmyard has been partially excavated for presentation to the public and is very clear in the terrain model for this reason. The remains of another kiln can be found not far away, but this has not been prepared for display. Important insights about the spatial concept

43

behind the friary complex can be gained from a bird’s eye view engraving of the friary by J. Martin Lerch (Schatek 1938, 2nd Section, 32–38). It is clearly a view of the physical space, but on closer inspection there are distinct differences in the form and position of individual structures compared to the ALS-based digital terrain model (DTM). In contrast to modern mapping, the engraving cannot be seen as an exact geometric representation of the physical space, but rather reflects the cognitive space of the artist or his

Figure 3.10 (above): The foundations of one of the hermitages

Figure 3.11 (left): Engraving of the friary St Anna in the Wilderness by J. Martin Lerch, 1689

44

Michael Doneus and Thomas Kühteiber

Figure 3.12: View of the entire friary complex with the course of the wall added

patron. The central picture of the engraving shows the friary, as restored after the Turkish war of 1683, as part of an idealized landscape encircled by orchards and surrounded by woods, thus creating a sense of ‘wilderness’ as a place of retreat (Figure 3.11). The idealized course of the enclosing wall is particularly apparent when compared to the terrain model. The actual course of the wall only bulges out significantly around the burial mounds south of the castle, but the engraving shows two symmetric arcs, which together with the hollow of the Arbach valley form the shape of a heart. The heart image is repeated in the middle of the picture in the shape the meadows surrounding the friary (Schatek 1938, 2nd Section, 37; Aguinaga 1993, 15). It is highly likely that the Sacred Heart motif is in this way being written into an imagined monastic

landscape of ‘wilderness’. This motif drew on the heart metaphor employed in the New Testament and by Augustine, and was particularly popular during the Counter-reformation, especially after the introduction of the Sacred Heart holiday in 1672 (Fassbinder 2003, 277). Interpretation of the location and of the course of the wall By now it is clear that we are dealing with a landscape used and interpreted in an explicitly religious framework. However, the monks were self-sufficient, reliant on their agricultural products, and so environmental factors seem to be important to explaining the location of the monastery close to a small stream on flat terrain sheltered from the wind. This explanation is obvious, but does not go very deep. In landscape

3  Airborne laser scanning and archaeological interpretation – bringing back the people

45 Figure 3.13: St. Anna in the Wilderness: Total viewshed map (50 m raster width) of the area based on a wide-ranging DTM with raster width of 10 m

archaeological terms, it does not consider cognitive space and individual decisions. In other words, explaining the location of the monastery only through a standard repertoire of environmental factors will generate standard explanations. In the area under research, there would be many potential locations with the same environmental characteristics. The question one should therefore ask is that given self-sufficiency as a decisive factor for the location of the monastery, what made the monks choose the actual location? To answer these questions we have to start with the obvious discrepancy between the course of the wall in reality and its portrayal in the engraving by Martin Lerch. Albert Schatek (1938, 2. 37) pointed out many years ago that the complex’s boundary wall was “too much of a heart shape”, seeing this as a mistake in

the illustration. In fact – as pointed out above – it is precisely this deviation which reveals the ideological meaning of the course of the wall (the Sacred Heart motif ). If this concept really was central however, then why was the friary complex not actually enclosed by a heart-shaped wall? Figure 3.12 shows that the wall is only vaguely heart-shaped. The wall does form a point in the west and there is a small indentation on the east side (beside the ruin of Scharfeneck castle), which divides the two ‘heart chambers’, but the indentation does not lie in the middle, but somewhat to the south. The protuberance enclosing the probable Iron Age burial mounds in the southwest corner of the territory is on the other hand highly visible. This is a striking deviation from the heart shape, although it would definitely have been avoidable without

46 A

Figure 3.14: A: Cumulative viewshed map, calculated from the seven hermitages with a position height of 1.5 m. Based on an ALS datasourced DTM (1 m raster width), target height 1.5 m. B: Viewshed map calculated from the top of the church tower on the basis of an ALSdata-sourced DTM (2 m raster breadth) and a target height of 1.5 m. The enclosing wall of the friary complex is shown by a red line

Michael Doneus and Thomas Kühteiber B

great difficulty. The presence of this protuberance proves, in fact, that this deviation from a heart shape was deliberate, and that there must instead be another underlying principle behind the course of the wall. The documentary sources tell us that the course of the wall was fixed during an inspection of the wooded area at a time of year when the vegetation was at a minimum and that that course was deliberately chosen. The sources do not reveal which motives were most influential, but one can refer to the rules of the Carmelite order, which insist on seclusion from worldly things – ideologically and in spatial fact. Based on this assumption, we can therefore attempt to explain the choice of location from the data available to us. Seclusion can be easily modelled from spatial distance and from the views available. The meaningfulness of visibility to the Carmelites can be seen in the idea that the seven hermitage units distributed around the friary should be visible from the central complex. The visibility properties of a terrain can be measured through a total viewshed map (Llobera 2003, 33–6). The total viewshed map (Figure 3.13) was calculated on the basis of a detailed dataset (10 m raster width) from the Federal Department of Calibration and Surveying (Bundesamt für Eichund Vermessungswesen – BEV), which covered the entire Leitha Hills and included a 10 km

wide buffer zone. The buffer zone was necessary in order to generate viewshed maps with a maximum radius of 10 km without statistical anomalies at the edges (Wheatley and Gillings 2000, 11). A total of about 10,000 individual viewshed maps were generated in a raster with 50 m intervals, and are combined in a total viewshed map. This map shows how many other points of the 50 m raster are visible from every single point of the raster. When a total viewshed map is superimposed on the landscape (Figure 3.13), it becomes very clear that the central part of the friary complex is secluded. The total viewshed of raster elements in the friary area is generally around 100 (that means the friary is visible from around 100 points in the total viewshed map raster). In comparison, the ruined castle is highly visible with values of around 1,600. This shows that friary complex does indeed lie in a low visibility area. This is not obvious on site, because of the heavy forestation. Nevertheless, the idea of seclusion can be modelled and demonstrated in a GIS, thus confirming the historical information. The course of the wall and in particular of the protuberance around the burial mounds to the southwest have not yet been explained however. An explanation only emerges when the visibility factor is subjected to further calculations as part of an explorative data analysis. The first assertion to be analysed is that the seven

3  Airborne laser scanning and archaeological interpretation – bringing back the people hermitages were located in sight of the friary And this was checked with a cumulative visibility map (Wheatley 1995). The calculations were carried out on the basis of a terrain model derived from ALS-data with a resolution of 2 m. The height of observation and the target height were both fixed at 1.5 m (eye level). The visibility map shows which areas can be seen from all hermitages, and logically, all hermitages are in turn visible from those areas. This result would not be applicable with a target height set at ground level (0 m) (Gillings and Wheatley 2001, 32). The cumulative visibility map can be seen in Figure 3.14 (left). For reasons of simplicity those areas visible from all seven hermitages are coloured green, while invisible areas are red. The map shows that, alongside a large part of the meadows and of the northern woods, parts of the friary and of the church tower really were visible from all hermitages. Another phenomenon is also apparent however, that the invisible area matches the course of the wall to a certain extent. This finding led to further visibility tests. The height of the wall was measured in sections on site and fed into the ALS-data derived terrain model with a 2 m raster width. Ultimately, it was shown that a visibility map with a target height of 1.5 m measured from the top of the friary’s church tower reproduced the course of the wall almost exactly (Figure 3.14 right). Even the protuberance deviating from the potential heart shape seems explainable, as this section is also part of the area from which the top of the friary’s church tower can be seen. The course of the wall can thus be explained by the desire to be fully isolated from the surrounding world, a result which fits the written information about the preferences of the Carmelites. How it was possible to demarcate the area visible from the friary so exactly is not entirely understandable from a present-day vantage point. The written sources tell us nothing in this case. The cutting of visibility lanes through the woods around the friary and the staking out of points by which the course of the wall could later be followed is a possibility. Equally possible is a simple rule, which coincides with our observations on site, that the wall would as far as possible not have been built on the hilltops, but instead, as seen from the friary, a certain distance behind them. In reaching a conclusion a question remains, which is particularly relevant to landscape archaeology. In this particular case, would

one come to a similar interpretive explanation without written sources? The answer is a qualified ‘yes’, for even without knowledge of the social context, total viewsheds would imply that deliberate isolation was involved. The reason for that isolation would however be much more difficult to establish without the information available to us from the written records. Seclusion and not being seen can, to name a few examples, be religiously motivated, result from a need for protection, express a deliberate social distance to political elites or to subordinates, or be enforced by those in power. The location factor of ‘remoteness’ would be recognizable, but the choice of one of the possible explanations cannot be made on the basis of a topographic GIS analysis itself, but must instead take account of the available information about the social context.

Conclusion Interpretation is a central element of archaeological prospection. In most cases it is equated with the procedure of mapping (or photo-reading in the case of aerial photographs). However, one must be aware that all links of the workflow chain have an impact on the final interpretation. This starts with data acquisition (decisions on date, instrumentation etc.), and continues with processing and visualisation, where all decisions taken will already have an effect on the final interpretation. Thus, to be able to understand and evaluate the result of an archaeological interpretation of ALS-derived data, it is necessary to consider the whole workflow from the design of a scanning project to its final interpretative results. Interpretative mapping, which usually stands for interpretation, is therefore neither the only part of the total interpretation workflow, nor is it the final one. Despite the fact that ALS can document only a limited range of archaeological remains (Doneus et al. 2008a, 891), ALS has an additional benefit. It offers, as is true for aerial archaeology, the possibility to interpret a scene at various scales – from the individual object to the larger topographic setting – and in that way can be used to combine the micro- and the macro-level in an explanatory model. Finally, ��������������� it is important to stress that an����������������������� interpretation of ALS data is never completed. Being experience- and context-driven, it is only useful to distinguish between good and bad interpretations. This

47

48

Michael Doneus and Thomas Kühteiber distinction has to be based on the arguments used and whether the interpreter was following scientific rules. A good interpretation should generate archaeological information, which extends beyond the obvious.

Acknowledgements This research has been supported by the Austrian Science Fund (FWF) under project number P18674-G02. The authors thank the Hydrologie und Geoinformation Department of the County of Lower Austria for their support. Parts of this paper were taken from Doneus and Kühtreiber 2012, which had been translated from German by Paul Mitchell. The Ludwig Boltzmann Institute for Archaeo­ logical Prospection and Virtual Archaeology (archpro.lbg.ac.at) is based on an international cooperation of the Ludwig Boltzmann Gesell­ schaft (A), the University of Vienna (A), the Vienna University of Technology (A), the Austrian Central Institute for Meteorology and Geodynamic (A), the office of the provincial government of Lower Austria (A), Airborne Technologies (A), RGZM-Roman- Germanic Central Museum Mainz (D), RAÄ-Swedish National Heritage Board (S), IBM VISTAUniversity of Birmingham (GB) and NIKUNorwegian Institute for Cultural Heritage Research (N).

References Aguinaga, J., 1993. Die Karmeliteneinsiedelei ‘St. Anna in der Wüste’ zu Mannersdorf am Leithagebirge (1644–1783) = The Carmelite hermitage ‘St. Anna in theWilderness’ at Mannersdorf am Leithagebirge (1644–1783). Master’s thesis, Wien. Bewley, R.H., Crutchley, S. and Shell, C., 2005. New light on an ancient landscape: LiDAR survey in the Stonehenge World Heritage Site. Antiquity 79, 636–47. Bofinger, J. and Hesse, R., 2010. Mit dem Laser durch das Blätterdach – Der Einsatz von Airborne-Laserscanning zur Entdeckung von archäologischen Geländedenkmalen. In ������������� Bofinger, J. and Merkl, M., (eds). Mit Hightech auf den Spuren der Kelten. Landesamt für Denkmalpflege: Esslingen. 70–89. Boos, S., Hornung S., Jung P. and Müller H., 2008. Erfassung und Analyse von Lasercandaten zur Unterstützung der Landschaftsarchäologie. Publikationen der DGFP 17, 219–27. Campana, S, Forte, M. And Liuzza, C., (eds). Space, Time, Place – Third International Conference

on Remote Sensing in Archaeology. BAR Int Ser. 2118. Chadwick, A.M., 2004. ‘Geographies ��������������������������� of sentience’ – an introduction to space, place and time. In Chadwick, A.M. (ed.). Stories from the Landscape: Archaeologies of Inhabitation. ���������������������� Archaeopress, Oxford.� 1–31. Challis, K., Kokalj, Z., Kincey, M., Moscrop, D. and Howard, A.J., 2008.Airborne Lidar and Historic Environment Records. Antiquity 82, 1055–64. Conolly, J. and Lake, M., 2006. Geographical information systems in archaeology. Cambridge Manuals in Archaeology: Cambridge. Devereux, B.J., Amable, G.S. and Crow, P., 2008. Visualisation of LiDAR terrain models for archaeological feature detection. Antiquity 82, 470–9. Doneus, M. and Briese, C., 2006. Full-waveform airborne laser scanning as a tool for archaeological reconnaissance. In Campana, S. and Forte, M., (eds). From Space to Place. Second International Conference on Remote Sensing in Archaeology. Proceedings of the 2nd international workshop, CNR, Rome, Italy, December 2 – 4, 2006. Archaeopress: Oxford. 99–106. Doneus, M. and Briese, C., 2011. Airborne Laser Scanning in Forested Areas – Potential and Limitations of an Archaeological Prospection Technique. In Cowley D.C., (ed.). Remote Sensing for Archaeological Heritage Management, proceedings of an EAC Symposium, Reykjavik, Iceland, 25–27 March 2010, EAC Occasional Paper No. 5, Archaeolingua. Doneus, M., Briese, C., Fera, M., Fornwagner, U., Griebl, M., Janner, M. and Zingerle, M.-C., 2007. Documentation and analysis of archaeological sites using aerial reconnaissance and airborne laser scanning. In Georgopoulos, A., (ed.). AntiCIPAting the Future of the Cultural Past. Proceedings of the XXIst International Symposium CIPA, Athens 2007. 275–80. Doneus, M., Briese, C., Fera, M. and Janner, M., 2008a. Archaeological prospection of forested areas using full-waveform airborne laser scanning. Journal of Archaeological Science 35, 882–93. Doneus, M., Briese, C. and Kühtreiber, T., 2008b. Flugzeuggetragenes Laserscanning als Werkzeug der archäologischen Kulturlandschaftsforschung. Das Fallbeispiel „Wüste“ bei Mannersdorf am Leithagebirge, Niederösterreich. Archäologisches Korrespondenzblatt 38 (1), 137–56. Doneus, M. and Kühtreiber, T., 2012. Landscape, ��������������� the Individual and Society: Subjective Expected Utilities in a Monastic Landscape near Mannersdorf am Leithagebirge, Lower Austria. In Mehler, N., (ed.). Historical Archaeology in Central Europe. Society of Historical Archaeology Special Publications (in press). Esser, H., 2002. ���������������������������� Situationslogik und Handeln (Situationallogicandactivity). Soziologie: spezielle Grundlagen Vol. 1. Campus-Verlag: Frankfurt/ Main.

3  Airborne laser scanning and archaeological interpretation – bringing back the people Fassbinder, St., 2003. Wallfahrt, Andacht und Magie. Religiöse Anhänger und Medaillen; Beiträge zur neuzeitlichen Frömmigkeitsgeschichte Südwestdeutschlands aus archäologischer Sicht (Pilgrimage, Devotion andmagic. ���������� Religious pendants and medallions; Archaeological contrib­ utions to the modern-period history of piety in southwestern Germany), ZeitschriftfürArchäologie des Mittelalters, Beiheft 18, Habelt: Bonn. Gillings, M. and Wheatley, D., 2001. Seeing is not Believing. In Slapsak, B., (ed.). On the good use of geographic information systems in archaeological landscape studies. 25–36. Office for Official Publications of the European Communities: Luxembourg. Hartmann, N., 1980. Philosophie der Natur. Abriß der speziellen Kategorienlehre. ������������������� de Gruyter: Berlin. Hengl, T.and Reuter, H., 2009. Geomorphometry. Concepts, software, applications. Amsterdam: Elsevier (Developments in soil science, 33). Hesse, R., 2010. LiDAR-derived Local Relief Models – a new tool for archaeological prospection. Archaeological Prospection 17(2), 67–72. Hinz, M., 2008. Rezension. Knut Petzold, Soziologische Theorien in der Archäologie. Konzepte, Probleme, Möglichkeiten (Saarbrücken 2007). Rundbrief Arbeitsgemeinschaft TheorieArch 7 (2), 33–39. Hodder, I., 2000. Agency and individuals in longterm processes. In Dobres, M-A. and Robb, J.E., (eds). Agency in archaeology. 21–33. Routledge: London/New York. Hug, C., Ullrich, A. and Grimm, A., 2004. Litemapper5600 – A Waveform-Digitizing LIDAR Terrain and Vegetation Mapping System. In Thies, M., Koch, B., Spiecker, H. and Weinacker, H., (eds). Laser-Scanners for Forest and Landscape Assessment. Proceedings of Natscan, Laser-Scanners for Forest and Landscape Assessment – Instruments, Processing Methods and Applications. 24–9. Humme, A., Lindenbergh, R. and Sueur, C., 2006. Revealing Celtic fields from lidar data using Kriging based filtering. In Maas, H.G. and Schneider, D. (eds). Proceedings of the ISPRS Commission V Symposium ‘Image Engineering and Vision Metrology’. Dresden. Kokalj, Ž., Zakšek, K. and Oštir, K., 2011. Application of sky-view factor for the visualisation of historic landscape features in lidar-derived relief models. Antiquity 85, 263–73. Kraus, K. and Otepka, J., 2005. DTM Modelling and Visualization – The SCOP Approach. In Fritsch, D., (ed.). PhotogrammetricWeek. Heidelberg. 241–52. Kraus, K. and Pfeifer, N., 1998. Determination ����������������� of terrain models in wooded areas with airborne laser scanner data. ISPRS Journal ofPhotogrammetryand Remote Sensing 53(4), 193–203. Lampel, J., 1899. Erörterungen und Materialien zur Geschichte der Leithagrenze (Questions and materials in thehistoryofthe Leitha border). Blätter des Vereins für Landeskunde von Niederösterreich 33.

Llobera, M., 2001. Building Past Landscape Perception With GIS: Understanding Topographic Prominence. Journal of Archaeological Science 28, 1005–14. Llobera, M., 2003. Extending GIS based analysis: the concept of visualscape, International Journal of Geographic Information Science 17, 25–49. Mayer, A., 1900. Die Karmeliter-Eremie St. Anna in der Wüste (The Carmelitehermitage “St. Anna in theWilderness). Blätter des Vereines für Landeskunde von Niederösterreich XXXIV, 120–37. Melzer, G., 1980. Verzeichnis der archäologischen Fundstellen in Au am Leithaberge, Hof am Leithaberge, Mannersdorf am Leithagebirge und Sommerein (Directory ofarcheologicalsites in Au am Leithaberge, Hof am Leithaberge, Mannersdorf am LeithagebirgeandSommerein). In Katalog des Museums Mannersdorf, Kultur- und Museumsverein Mannersdorf am Leithagebirge. Museum Mannersdorf am Leithagebirge, 55–99. Millard, K., Burke, C., Stiff, D. and Redden, A., 2009. Detection of a low-relief 18th-century British siege trench using LiDAR vegetation penetration capabilities at Fort Beauséjour–Fort Cumberland National Historic Site, Canada. Geoarchaeology 24(5), 576–88. Mochty, C., 1998. Die Marktgemeinde Hof am Leithagebirge von ihren Anfängen bis 1600 (The township Hof am Leithagebirge from its beginnings until 1600). In Mochtyand, C. and Bezemek, E., (eds). Die Marktgemeinde Hof am Leithagebirge im Wandel der Zeit (The township Hof am Leithagebirge through the years). 39–57. Marktgemeinde Hof am Leithagebirge. Opitz, R., 2009. Integrating lidar and geophysical surveys at Falerii Novi and FaleriiVeteres (Viterbo). Papers of the British School at Rome 77, 1–27, 335–43. Opp, K.-D., 2004. Die Theorierationalen HandelnsimVergleich mit alternativen Theorien Soziologie (The theory of rational action in comparison to other theories). In Gabriel, M., (ed.). Paradigmen der akteurszentrierten Soziologie (Paradigms of protagonist-orientated sociology). 43–68. VS Verlag für Sozialwissenschaft: Wiesbaden. Popper, K.R., 2003. Die offene Gesellschaft und ihre Feinde. Bd. 2: Falsche Propheten. Hegel, Marx und die Folgen (The open society and its enemies. 2: Hegel and Marx). Gesammelte Werke, Mohr Siebeck: Tübingen. Petzold, K., 2007. Soziologische Theorien in der Archäologie: Konzepte, Probleme, Möglichkeiten. Univ., Diplomarbeit Leipzig, 2005. VDM-Verl. Müller: Saarbrücken. Schatek, A., 1938. Führer durch die Wüste der Karmeliten bei Mannersdorf am Leithagebirge (Guide to the wilderness of the Carmelites near Mannersdorf am Leithagebirge). Wien. Schmidt, S., Bofinger, J., Keller, R. and Kurz, S., 2005. LIDAR – High resolution raster data as a survey tool. In Figueiredo A. and Leite Velho, G., (eds).

49

50

Michael Doneus and Thomas Kühteiber The world is in your eyes. Proceedings of the XXXIII CAA2005 Conference. Tomar. 255–60. Starzer, A., 1900. Mannersdorf am Leithagebirge und Umgebung (Mannersdorf am ���������������������� Leithagebirge and its surroundings). Wien. Thomas, J., 1993. The Politics of Vision and the Archaeologies of Landscape. ���������������������� In Bender, B., (ed.). Landscape. 19–48. Berg: Oxford. Wagner, W., Ullrich, A., Ducic, V., Melzer, T. and Studnicka, N., 2006. Gaussian decomposition and calibration of a novel small-footprint fullwaveform digitising airborne laser scanner. ISPRS Journal of Photogrammetry and Remote Sensing 60 (2), 100–12. Wheatley, D., 1995. Cumulative viewshed analysis: a

GIS-based method for investigating intervisibility, and its archaeological application. In Lock, G. and Stancic, Z., (eds). Archaeology and Geographical Information Systems: A European Perspective. 171–85. Taylor and Francis: London. Wheatley, D. and Gillings, M., 2000. Vision, perception and GIS: developing enriched approaches to the study of archaeological visibility. In Lock, G., (ed.). Beyond the map. 1–27. IOS Press [u.a.]: Amsterdam. Yokoyama, R�������������������������������������� .,������������������������������������ Sirasawa, M������������������������ . and������������������� Pike, R����������� .���������� J.�������� , ������ 2002��. Visualizing topography by openness: A new application of image processing to digital elevation models. �� Photogrammetric Engineering & Remote Sensing 68 (3), S. 257���� –��� 65.

4 Cultivating the ‘wilderness’ – how lidar can improve archaeological landscape understanding Ole Risbøl Lidar has been well received by archaeologists since it emerged as a useful remote sensing method for archaeology a decade ago and the first part of this paper describes some aspects of the introductory phase. So far, the emphasis has been on improving technical parameters and processing for archaeological purposes, but a recent trend towards using lidar to add to cultural historic understanding of larger areas is observed. Thus the potential benefit from using lidar in landscape archaeology is discussed in the paper with a certain focus on archaeology as a key agent in landscape understanding, stressing the importance of knowledge about all zones of the landscape including those usually perceived as natural. The potential of lidar to demonstrate the often comprehensive human impact on wooded areas is discussed, i.e. ground that traditionally has received less attention during survey projects. The resulting biased survey coverage might have severe consequences for the understanding of how humans have used all parts of the landscape. This is illustrated by a case study of a forested valley in a Norwegian municipality. The cultural historic understanding of this municipality, and the region where it is located, has changed significantly since some major archaeological surveys also covered large wooded areas which were ignored by previous surveys. Keywords: lidar, landscape archaeology, woodland, outfield, survey bias, cultural historic understanding

Introduction Lidar has been available to archaeologists for approximately a decade now, and it should no longer be regarded as ‘new technology’. Ten years of application is long enough to provide the means to evaluate how lidar has been implemented, describe some major trends in the application of this method and point out some unexploited potential the use of lidar offers the archaeological community. This paper takes a Norwegian perspective and is based on our own experiences and the challenges presented by practices and results from international lidar research and development. The main focus will be on the use of lidar in wooded areas. The first part of this paper is devoted to a review of some aspects of the introduction of lidar to

archaeology and discusses why it has been so well received by the archaeological community. This is followed by a discussion about the (potential) contribution of lidar to landscape archaeology with an emphasis on the prevailing biased knowledge about human impact on landscape – especially lack of knowledge about outfield areas (i.e. areas beyond the core of prehistoric and later agrarian settlement) – and how this can be related to the survey method used when creating the Norwegian national cultural heritage database.

A fascinating method The potential of airborne laser scanning for archaeological purposes was first recognised at the

52

Ole Risbøl

Figure 4.1: A section of a long row of pitfalls for elk situated in Grimsdalen, Dovre municipality, southeast Norway. The left hand image is a DTM where the pit-falls appear clearly in contrast to the same area on the right hand orthophotograph where no cultural remains are visible

beginning of the new millennium (see Doneus and Briese 2011) and since then growth in use has been exponential as ever more archaeologists in an ever wider number of countries have taken lidar into use in their work. One of the reasons for this is that lidar is an extremely powerful method which is readily understandable (i.e. due to its visual power). Not all, but many and probably the majority of archaeologists who have started using lidar, are from what we could classify as the archaeological remote sensing community and are accustomed to handling aerial photographs and satellite images and used to dealing with archaeology from a high distance above ground. For many archaeologists it has probably been a natural step to use lidar in addition to aerial photographs in their professional work. This is underlined by studies comparing the pros and cons of aerial photographs and lidar (Challis 2008; Crutchley 2009). Undoubtedly lidar also has attracted archaeologists with no or only limited experience of remote sensing who have found the opportunity to work with detailed 3-D elevation models of landscapes fascinating and helpful. One major fascination of lidar is the use of filtering techniques that remove vegetation from the digital surface model (DSM) in order to create a digital terrain model (DTM) – an operation that has a sense of a ‘conjuring trick’ with wide appeal to both professionals and the broader public (Figure 4.1). The filtering options allowing the elimination of trees and thereby exposing the bare earth for visual inspection is probably the greatest advantage of lidar compared to other remote sensing methods. My first acquaintance with lidar was when by chance I came across a paper by the geographer Benoît

Sittler (2004) who used lidar to detect ridge and furrow cultivation remains in a wooded landscape in Central Baden in southwest Germany. I was immediately fascinated by the technique and the possibilities this could offer for improving the archaeological inventory in Norway where almost 40% of the land is covered by forest. Thus the first archaeological lidar project in Norway was carried out in 2005 (Risbøl et al. 2006). This description is probably recognisable by other archaeologists who have been introduced to and started using lidar at some stage. My point is not to uncritically exalt lidar as an infallible method but to describe how the introduction of new techniques and methods might be received by (at least parts of ) the archaeological community. When new technologies are introduced to archaeologists they might be met with passivity, or sometimes even rejection, but archaeologists are generally characterized by openness towards new developments and a pro-active approach to new technologies. The introduction of lidar is similar to the introduction of GIS for instance, where the initial stage also was characterized by curiosity and fascination by those archaeologists who were genuinely interested in the possibilities new technologies represented and actually started using them. This can be exemplified with some of the very first published examples of the use of lidar in archaeology that basically show some nice images to illustrate the potential but also point out future possibilities (Holden 2001; Motkin 2001). This stage was soon replaced by a second stage where the limitations and challenges of the new techniques or methods emerged and triggered attempts to improve these. Within lidar research this has mainly concentrated on

4  Cultivating the ‘wilderness’ – how lidar can improve archaeological landscape understanding introduction and development of improved filtering techniques, such as those based on new sensors providing full-wave-form data (Doneus et al. 2008a) or by the use of kriging based filtering (de Boer et al. 2008). Others have tested or developed different ways of visualising by creating hillshaded representations of DTMs from multiple view directions (Devereux et al. 2008), or by the use of approaches like Local Relief Models (Hesse 2010) or sky-view factor (Kokalj et al. 2011). Thus the main goal has largely been to improve the basis for interpreting DTMs in archaeology. A recent example (Challis et al. 2011) presents a comparative study of a range of different techniques for visualisation and analysis of lidar data used for archaeological purposes. A second example (Lasaponara et al. 2011) has studied the processing of full-waveform lidar data using threshold-based algorithms (see Beck this volume). To improve the outcome of lidar data by addressing its limitations and challenges has dominated lidar research and development so far and has often been carried out by archaeologists and technicians in cooperation. The efforts to continuously improve the archaeological outcomes of lidar will, as with other techniques and methods used in archaeology, carry on as an important part of the research agenda. More recently a third phase has started to emerge amongst archaeologists using lidar to explore the contribution of this new technology to archaeological understanding mainly at a landscape level. This also comprises considerations about the theoretical implications of the use of lidar. This third phase will not replace the emphasis on the technical improvement of lidar for archaeology but should emerge as an essential research strand. So far only a few examples are published where the direct impact of the use of lidar has been demonstrated to dramatically change the archaeological understanding of cultural environments or landscapes. A paper by Doneus et al. (2008b) is one such example where lidar data was used to improve the understanding of the whole design of an Austrian monastery complex and to analyse it in a landscape context, providing a new cultural historic understanding of the area. While such examples are rare, the presentation of lidar at recent conferences and seminars and papers in this volume illustrate a move towards an increased use of lidar to contextualise archaeology and thereby improve archaeological landscape understanding. Lidar is often not the only technique used to improve

archaeological landscape understanding, but is one of a range of methods used in the emerging trend of multi-method approaches to landscape archaeology, perhaps best exemplified by the on-going work in the project ‘Archaeological Prospection and Virtual Archaeology’ by the Ludwig Boltzmann Institute (http://archpro.lbg. ac.at). Such approaches can include other remote sensing techniques like aerial photographs, satellite and hyper-spectral imaging in addition to geophysics like ground penetrating radar and magnetometer (see e.g. Campana 2011). The advantage of such highly technological archaeological projects is the possibility to work with extensive areas and with the different environmental zones of a landscape, whether they are arable land, pasture-land or forested areas. Lidar can be used in all these landscape categories, but its most profound contribution to remote sensing based archaeology is probably the opportunity it gives to examine those parts of the landscape that are covered with trees which prevent visual accessibility to the ground (see Ainsworth et al., Bennett et al., Challis and Howard, and Poirier et al. this volume for nonwoodland examples of lidar application). Passive remote sensing methods like aerial photographs and satellite images do not allow for this, and quite a large proportion of lidar projects carried out so far have taken this advantage and explored the archaeology of forested areas from the air (e.g. Sittler and Schellberg 2006; Devereux et al. 2005; Risbøl et al. 2006; Doneus and Briese 2006; Bofinger et al. 2006).

Landscape archaeology Lidar scanning projects usually cover large geographical areas and therefore provide a good basis for working with archaeology at a landscape level. While landscape is used in many ways (see Mlekuž this volume) the two basic meanings ascribed to the landscape concept in the Western world are as a specific geographical area on the one hand and as depiction on the other (Olwig 1993, 1996). The first understanding is rooted in early Anglo-Saxon history, while the other is attributable to the emergence of landscape painting in seventeenth-century Dutch pictorial art. Scholars have pointed out that the latter is the dominant understanding in much contemporary landscape research and planning (Howard 2004). Thus, an outside view of landscape giving priority

53

54

Figure 4.2: A landscape like this from Åmot municipality in southeast Norway might at a first glance appear as unmarked by human activity and experienced as natural, but in reality there are considerable cultural remains present, which tells us a different story (Photo: A. Kjærsheim)

Ole Risbøl

to aesthetic appearance and neatness will often be predominant when landscapes are evaluated. The emphasis on amenity and visual qualities of landscapes has been criticized for not being able to cope with the historic time depth of the landscape and for being too partial, looking only at the exterior landscape and discriminating against its inherent, less visible qualities (Barrett 1999; Layton and Ucko 1999). Precedence is often given to form while long-term processes and changes caused by anthropogenic landscape use are disregarded. This is met with criticism from those who plead for an understanding of landscape values based on depth of vision, arguing that landscape is too important and interesting to be represented as scenery confined to a particular time and place (Bender 1993; Ermischer 2004). Awareness of the complexity of landscape understanding is realised to a greater extent today and calls for integrated approaches. Indeed, a lack of inter-disciplinary studies and the absence of a holistic approach to landscape research and management could for many years explain the imbalance in processes leading to landscape assessments (Sarlöv-Herlin 2004). An awareness that the components of a landscape are more than the sum of its parts has prevailed (Dormor 1999) and is in line with international trends in recent years, where an integrative understanding including environment, nature and culture is now more predominant in landscape approaches (Thomas and Wells 1999; Scazzosi 2004). Still,

many challenges connected to landscape research and management remain. From an archaeological perspective one of the major challenges is lack of knowledge about long-term human impact, as time-depth is often left out. In addition, the spatial dimension of human impact that can be comprehensive, but more or less visible, is another contribution that archaeology can bring to landscape studies, along with knowledge about long-term change and continuity in landscape use (Fairclough 2003). Despite the prevailing emphasis on interdisciplinarity these perspectives are often ignored by different traditions in landscape study which often neglect evidence for human presence. Traditionally this applies mainly to forest and out-field areas which may be categorized as ‘untouched nature’ (Figure 4.2). In many countries archaeological records are often heavily biased resulting in an uneven knowledge about the different parts of a landscape. In rural Norway this bias is seen between agricultural lands where archaeological inventories of visible remains and monuments are more comprehensive than in forested and other outfield areas. This presents a basic paradox, since the better preservation of monuments is in non-agricultural areas. This type of fundamental bias has a major impact on the way in which landscapes are understood and managed (see also Cowley 2011). In this respect lidar has an important role and the limitations of inventories in wooded areas is routinely identified as a major reason for implementing lidar (Devereux et al. 2005; Doneus and Briese 2006; Risbøl et al. 2006; Crow et al. 2007; Gallagher and Joseph 2008; Georges-Leroy 2011; Rutar and Črešnar 2011). Knowledge about human impact on landscape is important and archaeology is a key agent in bringing this aspect into landscape understanding. The long-term use of the landscape is often masked and is seldom taken into account in visually based landscape designations (Lambrick 1992; Lapka and Cudlinova 2003), a factor that prevents the articulation of complex landscape development and carries the danger of creating only one picture of a particular period (Fairclough 1999). The practice of continually emphasising visual aesthetic qualities of the landscapes which in reality consist of a long history of complex and ever-changing land-use modes can be problematic. The result is that one layer of landscape history may be given priority at the expense of others (Howard 2004, 426).

4  Cultivating the ‘wilderness’ – how lidar can improve archaeological landscape understanding Treating landscape merely as scenic, emphasizing the visual qualities, is quite common in landscape assessment efforts and involves the danger of excluding time-depth and change (Fairclough 1999; Howard 2004). To freeze a certain landscape picture is problematic when judged from a landscape history perspective (Sarlöv-Herlin 2004). Landscape is by definition not static but an ever-changing organism, a fact that calls for a holistic approach embracing natural, cultural and cognitive values (Darvill 1999; Knapp and Ashmore 1999), in addition to time depth values. Landscape changes are caused by environmental influence as well as human impact and changes in thinking, all in the long passage of time (Coones 1992; Widgren, 1999). To take into account time depth in the landscape is important because it can throw light on extensive land history, how people related to the land in the past and how it was altered by human activity (Macinnes and Wickham-Jones 1992). This is also important knowledge when one wants to understand how landscapes appear today in terms of long-term change and continuity. To be able to do so, better knowledge about all parts of the landscape is required. Another concern of treating landscape merely as scenic is the risk that a staged landscape will offer an idealized picture that communicates (consciously or otherwise) a certain ideology and subjective comprehension of how history is to be presented (Cooney 1999; Knapp and Ashmore 1999). It is often the ideal picture with its symbolic meaning that is appreciated rather than the holistic landscape history, a conspicuous tendency that coincides with the way rural landscapes often are managed today – as an image gazed at from the outside (Widgren 1999; Lapka and Cudlinova 2003). However, usually landscapes contain more aspects than are immediately visually comprehended. Thus, rural landscapes are usually appreciated for their immediate visual amenity at the same time as they are embedded with more or less visible traces of chronologically diverse and varied human exploitation (Ermischer 2004), giving these landscapes a certain importance as holders of cultural history values. But in many cases these values are not recorded – a situation that prevents an inclusion of long-term human impact in landscape assessments and understanding. From a cultural historian’s point of view, the way in which the long-term use of landscapes

is expressed in the landscape by its remnant traces is of utmost importance because these are the physical evidence of a landscape under continuous change throughout time. The everchanging use of the landscape creates layers of evidence with mutual relationships. These layers and their relationships give the opportunity to understand complex social and economic development from a long-term perspective (Muir 2003). Implementing this view in landscape management efforts and understanding requires knowledge about remains in the landscape and an approach to landscape that takes into account both space and time. Landscapes consist of many layers of history and basically landscapes have been created by different strategies throughout history, leaving traces now somehow embedded in the landscape (Allen 1999; Scazzosi 2004). As Fairclough (2003, 296) puts it, landscape is an archive of information. Dealing with landscape presents a demanding challenge to comprehend all the facets of a complex history that has horizontal and vertical dimensions – both time and space – a statement that can be further elaborated by a third dimension: the process of landscape change (Ermischer 2004). How these values are understood and emphasized is of vital importance to how landscapes are eventually percieved and managed (Sarlöv-Herlin 2004). Biased survey coverage leads to challenges when single monuments or remains are to be understood within a wider context. The notion of context became increasingly important with the emergence of post-processual archaeology in the later part of the 1980s and emerged as a response to positivism as the dominating idea in processual archaeology. According to Ian Hodder (1986, 1987) the concept of ‘context’ can be used to describe both the theory about archaeological interpretation as well as the interpretive practice. The meaning of archaeological data can only be understood through its context. Context can be studied on a micro or a macro level where the latter is about identifying relationships on a broader landscape level in a way that will inform us about social and economic conditions in a prehistoric society. In practice this means to study the spatial and chronological distribution of cultural monuments and remains as well as their mutual relationships. In order to be able to do so we need data – on a micro level from archaeological excavations and on a macro level from archaeological surveys. As Gary Lock (2003, 12) has stated: “It is only through

55

56

Ole Risbøl contextual relationships that meaning can be constructed and this is likely to be more successful where data are more richly networked”. Of course data used for such networking on a landscape level comprise more than can be provided by the use of lidar. This is an obvious conclusion since lidar has important limitations, such as in woodland areas with dense under-storey vegetation that may be impossible to filter, while its potential for revealing sub-surface archaeology through intensity values (Challis and Howard this volume) is clearly challenging. The notion of context is also important within cultural resource management. In accordance with land-use planning the archaeologists’ task is often to evaluate larger entities like cultural environments or single cultural monuments or remains and attach priorities to them. Such eval­ uations require adequate information about the presence of archaeological remains, monuments and sites in the landscape and the environmental context in which they are situated. The heavily biased archaeological inventories which are the reality in many countries and regions will often be an obstacle in terms of carrying out good knowledge based evaluations that consider the spatial and chronological distribution of monuments, remains and sites and the relation­ ship between these on a satisfactory level. The argument for improved surveys and inventories to support solid landscape under­ standing is more in line with a positivist framework within processual archaeology than with the succeeding post-processual understanding that focusses on cognitive aspects of humans’ relations to landscape. Post-processual landscape understanding emphasizes the intangible parts of this relation within a phenomenological approach to landscape studies. The postprocessual approach to landscapes is challenging for archaeology which is a discipline firmly rooted in the tangible ���������������������������������� – artefacts, �������������������������������� sites, monuments and other remains. To include cognitive ��������������������� aspects of human relations to landscape is perhaps even more difficult in technological approaches to landscape studies based on GIS analysis, for instance (Lock 2001) or the use of lidar. This problem was one of the first critiques of the use of GIS in archaeology from a post-processualistic point of view – a criticism that was met by the implementation of new approaches to GIS-based landscape studies such as view-shed analysis (Wheatley and Gillings 2002) and the inclusion of the concept movement in visual analysis

(Llobera 2000). Lidar offers an interesting, but so far almost unexplored, potential to develop the field of visual archaeological landscape analysis based on lidar-generated 3D-models (see Challis and Kincey this volume). However, there are still some great challenges in linking new technology and the post-processual approach to landscapes with its emphasis on intangible dimensions. But, irrespective of theoretical standpoint, whether processual or post-processual, knowledge about human activity in the landscape is required.

Cultivating the ‘wilderness’ L����������������������������������������������� idar is one of more important tools to provide more cost-effective information about longterm human impact on landscapes. For outfield areas, especially those that are wooded, the ability of lidar to eliminate vegetation makes it an immensely powerful tool to redress major biases in archaeological inventories of such areas, and absolutely fundamental for landscape understanding and management. The relationship between the state of inventories and landscape archaeology is explored through a case study to show the significance of improved archaeological knowledge in all parts of the landscape based on thorough archaeological surveys. The backbone of Askeladden, the Norwegian national cultural heritage database, is data collected during a national survey campaign conducted from 1963 to 1994 (Skjelsvik 1998). Prior to that, attempts had been made to carry out national archaeological surveys since the early 20th century, but it was not until such surveys were linked to the general national mapping campaign (launched in the early 1960s to map large parts of the country (40–50%) at 1:5000 and 1:10,000) that archaeological surveys were carried out in a more systematic way (Skjelsvik 1978). The use of airborne photogrammetry (which began in 1936 in Norway) made it possible to map large parts of the country. The incentive to carry out the general national mapping campaign was rooted in a need to gain better control of the economic resources of Norway and is reflected in its name – the Economic Map Services (EMS). In archaeology, stereoscopic pairs of photos were used in order to map single monuments or remains, providing accurate locations in contrast to previous surveys. Even though the national survey campaign was a quantum leap for Norwegian archaeology

4  Cultivating the ‘wilderness’ – how lidar can improve archaeological landscape understanding and cultural heritage management two major flaws have to be considered if the potential use of the database is to be critically evaluated. The first is that coverage is not comprehensive, and of 435 municipalities, 72 were not surveyed at all. Since the national campaign stopped in 1994 archaeological surveys are no longer conducted in a systematic manner, but are now undertaken for specific reasons arising from development and carried out by archaeologists at a regional level (Holme 2001, 58). This means that areas where there is no development threat usually not are surveyed. The second flaw is more important in this context and connected to the methodology by which the EMS surveys were carried out. They were mainly based on information from written sources and landowners and did not undertake systematic terrain walking (Larsen 1990, 49; Larsen and Sollund 1993). As a consequence of this, surveys were mainly carried out in settled areas while outfield areas were given a lower priority. Furthermore this priority was due to both the economically confined scope and limited knowledge about the magnitude of human impact on out-lying land. In addition, the legislative framework had an impact, specifically on the perceived antiquity of remains worthy of protection. According to the Norwegian Cultural Heritage Act the law automatically protects all monuments, remains and sites pre-dating the Reformation in 1537. So, for decades the lack of dating for many cultural remains in outfield areas (i.e. iron-production sites, charcoal pits etc.) resulted in a downgrading of these and a focus on well-dated monuments (i.e. grave mounds, grave cairns, menhirs etc.) found mainly in settled areas (Larsen 1990, 57). Thus, the most common archaeological remains in outfield areas (Figure 4.3: iron production sites, charcoal pits and pits for catching elk and reindeer) were not mentioned in the first editions of the Cultural Heritage Act and were not accounted for until a revision of the law in 1978 (Larsen and Sollund 1993; Trøim 1999, 140). Archaeological projects conducted in outfield areas from the 1980s resulted in more knowledge about a range of cultural remains from outlying land. A major boost in outfield archaeology came with some large projects concerned with the development of waterpower schemes and the establishment of military exercise and shooting ranges from the late 1980s (Jacobsen and Larsen 1992; Narmo 2000; Risbøl 2005). Since, a

57 Figure 4.3: An example of a slap heap indicating pre-industrial iron production as a representative for a typical out-field cultural remains in Åmot municipality, southeast Norway (Photo: A. Kjærsheim)

representative selection of outfield remains has been entered to the national database giving us a chance to dimly perceive the distribution and dimension of human impact on landscape. But still the overall picture is a lack of surveys from the greater part of outfield areas. The limited archaeological work in large parts of the landscape has consequences for both management and our knowledge of landscape history in certain areas. This can be illustrated by a case study from Åmot municipality in the Østerdalen valley in southeast Norway, concentrating on monuments and remains from the Iron Age and early medieval period (c. 2500–500 BP). Åmot is a rural, forest dominated community covering an area of 1340 square kilometres. As much as 75% is forest, interspersed with mires (12%) and water bodies (3.5%), built-up areas (0.7% including infrastructure) and other areas (7.1%), with only 1.7% as arable land. In Åmot surveys for Economic Map Services were carried out in 1977–78 and 1986, resulting in the mapping of 200 sites with a total of 680 single remains or monuments (Figure 4.4). Of these 509 are either the remains of rural agricultural settlement (278 grave mounds, grave and clearance cairns) or non-agrarian activities carried out in outfield areas (231 pit-falls, charcoal pits, iron production and tar production sites). The two groups are almost equal in number but do not by any means reflect the proportional relation between agrarian and outfield land. This was more or less the situation prior to two large archaeological projects prompted by the establishment of military ranges covering almost 300 square kilometres of outfield land in Åmot in 1994–7 and 1999–2004 respectively. These projects gave archaeologists the opportunity to thoroughly survey large, connected areas and increase knowledge about prehistoric land-use on a landscape level in a

58

Ole Risbøl

Figure 4.4 (right): The 200 sites containing a total number of 609 cultural remains or monuments surveyed prior to the large outfield projects. © Norwegian mapping authority

Figure 4.5 (far right): The 3799 sites containing a total number of 3955 cultural remains or monuments after the large survey projects. © Norwegian mapping authority

forest dominated region like Østerdalen and contributed a large number of new registrations to the national cultural heritage database. If we compare the figures from the EMS survey with the data in the present database the situation is as follows. Quantitatively the number of records has increased to a total of 4624 sites composed of 5387 individual monuments and remains. Of these a total of 377 can be related to Iron Age and early medieval settlement, while 3578 are remains from outfield production belonging to the same period of time. These numbers better reflect contemporary percentage distributions of arable and outfield land-use cover in Åmot. It also proves another prehistoric activity pattern than the one that could be established prior to the large total-coverage survey projects (Figure 4.5). The improved survey situation has resulted in increased knowledge about how this area was used during prehistoric and medieval times in a way that puts it in a completely new light. Two brief but important examples are given here. Prior to the surveys (and succeeding excavation projects) Åmot and the whole region was perceived as a society based primarily on ordinary agrarian activities without giving the exploitation of outfield resources the attention it requires to comprehend the specific character of this region. Late Viking age and early medieval iron production has been proven to be protoindustrial in character and of a scale that exceeded

local needs at many times (Rundberget 2009). Furthermore the improved knowledge has shed light on questions concerning the presence of two ethnic groups in this region – the Norse and Sámi people, who had different subsistence strategies and thus made different impacts on landscape (Bergstøl 2004). This study demonstrates the biases of the national cultural heritage record and provides a powerful argument for lidar as a potential method for improving this situation. Lidar is a cost-effective way to acquire more and better information about human land-use in prehistoric and historic times and thereby to support landscape archaeology. In 2005 the first Norwegian lidar project was carried out in an area near Åmot with similar landscape and archaeological remains (Figure 4.6). A subsequent study proved that 82% of the total amount of visible remains in a section of the scanned area could be detected with lidar (Risbøl 2010). The large archaeological projects carried out in Åmot were conducted without lidar and similar percentage returns might have been expected from a more cost-effective lidar-based survey. However, it is important to stress that the goal of implementing lidar, or other new technologies, is not just a matter of putting more dots on the map or improving the quantitative content in the national database. It is even more important to emphasise that the opportunity to work on a landscape scale can generate different stories

4  Cultivating the ‘wilderness’ – how lidar can improve archaeological landscape understanding

59 Figure 4.6: A DTM showing two parallel slag heaps from late Viking Age/ early Medieval iron production. In the surrounding area more charcoal pits can be identified. These remains are situated in Elverum municipality, southeast Norway

based on more knowledge about human impact on the entire landscape in a region instead of giving priority to only some areas in that region. Archaeological landscape studies show that the Iron Age and early medieval society in Åmot was not a peripheral agrarian community living in isolation but was integrated in complex national, and perhaps even international, networks of trade and transport, also raising questions about control and power. The environment of the Åmot Iron Age and early medieval population was not untouched wilderness but rather a protoindustrial landscape marked by major resource exploitation very different from preceding and later periods.

Conclusion Lidar emerged as a new remote sensing method for the benefit of the archaeological community a little more than a decade ago. Archaeologists with an interest in new technologies and remote sensing have eagerly adopted lidar and added it to their archaeological toolbox. In the first decade of archaeologically led research and development of lidar there has been a focus on improving the technique in order to optimize its use for archaeological purposes. More recently lidar generated data has increasingly been applied to improve the cultural historic understanding of larger areas or landscapes. In this paper some

challenges concerning archaeology’s role in landscape understanding have been pointed to, emphasizing long-term human impact and the demonstration of continuous landscape change and continuity as the most important contribution archaeology can offer to landscape assessment and understanding. Archaeologists’ involvement in landscape discourses is dependent on knowledge about landscapes as scenes for human action as well as the theoretical approaches constituting our way to understand these landscapes. Empiricism is needed whether our approach to landscape understanding is rooted in the ‘objectivity’ of a processual tradition or in a post-processual framework giving the cognitive aspects of landscape understanding more attention. As the Norwegian case presented here shows the foundation for working with landscape history can be heavily biased due to priorities and professional traditions; these have major consequences for the ability to contextualize archaeology on a landscape level and thereby improve our cultural historic understanding. In a nature versus culture dichotomy understanding of landscape, the forest and outfield areas are perceived primarily as nature while they in reality often are heavily marked by human impact. Lidar offers an efficient method for improving our knowledge on the character and extent of human interaction with all kinds of landscapes – forested or otherwise.

60

Ole Risbøl

References Allen, T., 1999. The management of the rural landscape: a sense of place. In Grenville, J. (ed.). Managing the Historic Rural Landscape, 163–72. London/New York: Routledge. Barrett, J.C., 1999. Chronologies of landscape. In Ucko, P.J. and Layton, R. (eds). The Archaeology and Anthropology of Landscape, 21–30. London/ New York: Routledge. Bender, B., 1993. Landscape: Politics and Perspectives. Berg: Providence, R.I. Bergstøl, J., 2004. Creols in Iron Age Norway? In Hakenbeck, S.E. and Matthews, S.G. (eds).Reconsidering Ethnicity. Material Culture and Identity in the Past. Archaeological review from Cambridge, 19.2, 7–24. Department of Archaeology: Cambridge. Bofinger, J., Kurz, S. and Schmidt, S., 2006. Ancient maps – modern data sets: different investigative techniques in the landscape of the Early Iron Age princely hill fort Heuneburg, Baden-Württemberg. In Campana, S. and Forte, M. (eds). From Space to Place: 2nd international conference on remote sensing in archaeology (Proceedings of the 2nd international workshop, CNR, Rome, Italy, December 2–4, 2006), 87–92. Oxford: Archaeopress. Campana, S., 2011. ������������������ ‘Total Archaeology’ to reduce the need for Rescue Archaeology: The BREBEMI project (Italy). In Cowley, D.C. (ed.). Remote sensing for archaeological heritage management. Europae Archaeologia Consilium (EAC). EAC Occasional Paper No. 5, 33–42. Brussel. Challis, K., Kokalj, Ž., Kincey, M., Moscrop, D. and Howard, A. J., 2008. Airborne lidar and historic environment records. Antiquity 82, 1055–64. Challis, K., Forlin, P. and Kincey, M., 2011. A Generic Toolkit for the Visualization of Archaeological Features on Airborne LiDAR Elevation Data. Archaeological Prospection 18(4), 279–89. Coones, P., 1992. The unity of landscape. In Macinnes, L. and Wickham-Jones. C.R., (eds). All Natural Things. Archaeology and the Green Debate, 22–40. Oxford: Oxbow Books. Cooney, G., 1999. Social landscapes in Irish prehistory. In Ucko, P.J. and Layton, R., (eds). The archaeology and anthropology of landscape, 46–64. London/ New York: Routledge. Cowley D.C., 2011. Remote sensing for archaeology and heritage management – site discovery, interpretation and registration. In Cowley, D.C. (ed.). Remote sensing for archaeological heritage management. Europae Archaeologia Consilium (EAC). EAC Occasional Paper No. 5, 43–55. Brussel. Crow, P., Benham, S., Devereux, B.J. and Amable, G.S., 2007. Woodland vegetation and its implications for archaeological survey using LiDAR. Forestry 80(3), 241–52. Crutchley, S., 2009. Shedding light on the past: Using airborne LIDAR to understand ancient landscapes. In Bender, O., Evelpidou, N., Krek,

A. and Vassilopoulos, A., (eds). Geoinformation Technologies for Geocultural Landscapes: European Perspectives, 87–103. London, New York, Leiden: Taylor & Francis. Darvill, T., 1999. The historic environment, historic landscapes, and space-time-action models in landscape archaeology. In Ucko, P.J. and Layton, R., (eds). The archaeology and anthropology of landscape, 104–18. London/New York: Routledge. de Boer, A.G., Laan, W.N.H., Waldus, W. and van Zijverden, W.K., 2008. LIDAR-based surface height measurements: Applications in archaeology. In Frischer, B. and Dakouri-Hild, A. (eds). Beyond Illustration: 2D and 3D Digital Technologies as Tools for Discovery in Archaeology, 76–84. Oxford: Archaeopress. Devereux, B.J., Amable, G.S., Crow, P. and Cliff, A. D., 2005. The potential of airborne lidar for detection of archaeological features under woodland canopies. Antiquity 79, 648–60. Devereux, B.J., Amable, G.S. and Crow, P., 2008. Visualisation of LiDAR terrain models for archaeological feature detection. Antiquity 82, 470–79. Doneus, M. and Briese, C., 2006. Digital terrain modelling for archaeological interpretation within forested areas using full-waveform laserscanning. In Ioannides, M., Arnold, D., Niccolucci, F. and Mania, K. (eds). The 7th International Symposium on Virtual Reality, Archaeology and Cultural Heritage VAST, 155–62. Aire-La-Ville: The Eurographics Association. Doneus, M., Briese, C., Fera, M. and Janner, M., 2008a. Archaeological prospection of forested areas using full-waveform airborne laser scanning. Journal of Archaeological Science 35/4, 882–93. Doneus, M., Briese, C. and Kühtreiber, T., 2008b. Flugzeuggetragenes Laserscanning als Werkzeug der archäologischen kulturlandschaftsforschung. Archäologisches Korrespondenzblatt 38, 137–56. Doneus, M. and Briese, C., 2011. Airborne laser scanning in forested areas – potential and limitations of an archaeological prospection technique. In Cowley, D.C. (ed.). Remote sensing for archaeological heritage management. Europae Archaeologia Consilium (EAC). EAC Occasional Paper No. 5, 59–76. Brussel. Dormor, I., 1999. Current planning policies and legislation for historic rural landscapes. In Grenville, J. (ed.). Managing the Historic Rural Landscape, 44–56. London/New York: Routledge. Ermischer, G., 2004. Mental landscape: landscape as idea and concept. Landscape Research 29(4), 371–83. Fairclough, G., 1999. Protecting the cultural landscape: national designation and local character. In Grenville, J. (ed.). Managing the Historic Rural Landscape, 27–39. London/New York: Routledge. Fairclough, G., 2003. ‘The long chain’: archaeology, historical landscape characterization and time

4  Cultivating the ‘wilderness’ – how lidar can improve archaeological landscape understanding depth in the landscape. In Palang, H.G.F. (ed.). Landscape Interfaces, 295–318. Dordrecht: Kluwer Academic Publishers. Gallagher, J.M. and Josephs, R.L., 2008. Using LiDAR to detect cultural resources in a forested environment: an example from Isle Royale National Park, Michigan, USA. Archaeological Prospection 15, 187–206. Georges-Leroy, M. 2011. Airborne Laser Scanning for the management of archaeological sites in Lorraine (France). In Cowley D.C. (ed.). Remote sensing for archaeological heritage management, Europae Archaeologia Consilium (EAC). EAC Occasional Paper No. 5, 229–34. Brussel. Hesse, R., 2010. LiDAR-derived Local Relief Models – a new tool for archaeological prospection. Archaeological Prospection 17(2), 67–72. Hodder, I., 1986. Reading the past. Cambridge: Cambridge University Press. Hodder, I., 1987. The archaeology of contextual meanings. New directions in archaeology. Cambridge: Cambridge University Press. Holden, N., 2001. Digital airborne remote sensing: the principles of LiDAR and CASI. AARGNews 22, 23. Holme, J., (ed.). 2001. Kulturminnevern. Lov, forvaltning, håndhevelse. Oslo: Økokrim. Howard, P., 2004. Spatial planning for landscape: mapping the pitfalls. Landscape Research 29(4), 423–34. Jacobsen, H. and Larsen, J.H., 1992. Dokkfløys historie. Dokkfløy fra istid til kraftmagasin. Lillehammer: Gausdal kommune og Oppland energiverk. Knapp, A.B. and Ashmore, W., 1999. Archaeological landscapes: Constructed, Conceptualized, Ideational. In Ashmore, W. and Knapp, A. B., (eds). Archaeologies of Landscape: Contemporary Perspectives, 1–30. Massachusetts & Oxford: Blackwell Publisher. Kokalj, Ž, Zakšek, K. and Oštir, K., 2011. Application of the sky-view factor for the visualization of historic landscape features in lidar derived relief models. Antiquity 85(327), 263–73. Lambrick, G., 1992. The importance of the cultural heritage in a green world: towards the developement of landscape integrity assessment. In Macinnes, L. and Wickham-Jones C.R., (eds). All Natural Things. Archaeology and the Green Debate, 105–26. Oxford: Oxbow. Lapka, M. and Cudlinova, E., 2003. Changing landscapes, changing landscape’s story. Landscape Research 28(3), 323–28. Larsen, J.H., 1990. Om desimeringen av våre fornminner. Noen resultater fra arbeidet med registrering av fornminner for Det økonomiske kar tverket i 1980-årene. Universitetets Oldsaksamlings Årbok 1989/1990, 47–60. Larsen, J.H. and Bøe Sollund, M-L., 1993. Fornminneregisteret ved Oldsaksamlingen. Universitetets Oldsaksamlings Årbok 1991/1992, 37–50. Lasaponara, R., Coluzzi, R. and Masini, N., 2011.

Flights into the past: full-waveform airborne laser scanning data for archaeological investigation. Journal of Archaeological Science 38(9), 2061–70. Layton, R. and Ucko, P.J., 1999. Introduction: gazing on the landscape and encountering the environment. In Ucko, P.J. and Layton, R., (eds). The archaeology and anthropology of landscape, 1–20. London/New York: Routledge. Llobera, M., 2000. Understanding movement: a pilot model towards the sociology of movement. In Lock, G. (ed.). Beyond the Map. Archaeology and Spatial Technologies, 1–27. NATO Science Series A: Life Sciences. Amsterdam: IOS Press. Lock, G., 2001. Theorising the practice or practicing the theory: archaeology and GIS. Archaeologia Polona 39, 153–64. Lock, G., 2003. Using Computers in Archaeology: Towards Virtual Pasts. London: Routledge. Macinnes, L. and Wickham-Jones, C.R., 1992. Time-depth in the countryside: archaeology and the environment. In Macinnes, L. and WickhamJones, C.R., (eds). All Natural Things. Archaeology and the Green Debate, 1–13. Oxford: Oxbow Books. Motkin, D., 2001. An assessment of LIDAR for archaeological use. AARGNews 22, 24–5. Muir, R., 2003. On change in the landscape. Landscape Research 28(4), 383–403. Narmo, L.E., 2000. Oldtid ved Åmøtet. Østerdalens tidlige historie belyst av arkeologiske utgravinger på Rødsoen i Åmot. Åmot: Åmot Historielag. Olwig, K.R., 1993. Sexual Cosmology: Nation and Landscape at the Conceptual Interstices of Nature and Culture; or, What does Landscape Really Mean? In Bender, B., (ed.). Landscape: Politics and Perspectives, 307–43. Providence/Oxford: Berg. Olwig, K.R., 1996. Recovering the Substantive Nature of Landscape. Annals of the Association of American Geographers 86(4), 630–53. Risbøl, O., 2005. Kulturminner i Gråfjell – kultur­ historiske resultater fra registreringsprosjektet. In Stene, K., Amundsen, T., Risbøl, O. and Skare, K., (eds). ‘Utmarkens ���������������� grøde’. Mellom registrering og utgravning i Gråfjellområdet, Østerdalen. Varia 59, 5–26. Oslo: KHM – Kulturhistoriske Museum, Universitetet i Oslo. Risbøl, O., 2010. Towards an improved archaeological record through the use of airborne laser scanning. In Forte, M., Campana, S. and Liuzza, C., (eds). Time and place: 3rd International Conference on Remote Sensing in Archaeology, 105–12. Oxford: Archaeopress. Risbøl, O., Gjertsen A.K. and Skare, K., 2006. Airborne laser scanning of cultural remains in forests: some preliminary results from a Norwegian project. In Campana, S. and Forte, M., (eds). From Space to Place: 2nd International Conference on Remote Sensing in Archaeology, 107–12. Oxford: Archaeopress. Rundberget, B., 2009. Reflections of the ������� ‘Utmark’, today and yesterday, pointed up through iron production in Gråfjell. In Holm, I., Stene, K. and

61

62

Ole Risbøl Svensson, E., (eds). Liminal landscapes. Beyond the concepts of “marginality” and “periphery”. Oslo Archaeological Series 11, 67–85. University of Oslo. Rutar, G. and Črešnar, M., 2011. Reserved optimism: preventive archaeology and management of cultural heritage in Slovenia. In Cowley, D.C., (ed.). Remote sensing for archaeological heritage management. Europae Archaeologia Consilium (EAC). EAC Occasional Paper No. 5, 259–64. Brussel. Sarlöw-Herlin, I., 2004. New challenges in the field of spatial planning: landscapes. Landscape Research 29(4), 399–411. Scazzosi, L. 2004. Reading and assessing the landscape as cultural and historical heritage. Landscape Research 29(4), 335–55. Sittler, B., 2004. Revealing historical landscapes by using airborne laser scanning. A 3-D model of ridge and furrow in forests near Rastatt (Germany). In Thies, M., Koch, B., Spiecker, H. and Weinacker, H., (eds). Laser-Scanners for Forest and Landscape Assessment (Proceedings of the ISPRS Working Group 8/2, Freiburg, Germany, October 3–6, 2004; Internatinal Archives of Photogrammetry, Remote Sensing and Spatial Information Sciences Volume 36, Part 8/W2), 258–61. ISPRS – International society of photogrammetry and remote sensing, XXXVI, part 8/W2: Freiburg, Germany. Sittler, B. and Schellberg, S., 2006. The potential of LIDAR in assessing elements of cultural heritage

hidden under forest canopies or overgrown by vegetation: Possibilities and limits in detecting microrelief structures for archaeological surveys. In Campana, S. and Forte, M., (ed.). From space to place: 2nd international conference on remote sensing in archaeology, 117–22. Oxford: Archaeopress. Skjelsvik, E., 1978. Registrering av fornminner for det økonomiske kartverket i Norge 1963–1977. Fornvännen – Tidsskrift för svensk antikvarisk forskning 73(1978/2), 134–41. Skjelsvik, E., 1998. Registrering av fornminner for Det økonomiske kartverket 1963–1994. Oslo: NIKU – Norsk institutt for kulturminneforskning. Thomas, R.C. and Wells, D., 1999. Nature conservation and historic properties: an integrated approach. In Grenville, J., (ed.) Managing the Historic Rural Landscape, 149–62. London, New York: Routledge. Trøim, I., 1999. Arkeologisk forskning og det lovregulerte fornminnevernet. En studie av fornminnevernets utvikling i perioden 1905–1978. Varia 49. Oslo: Universitetets Oldsaksamling, Universitetet i Oslo. Wheatley, D. and Gillings, M., 2002. Spatial technology and archaeology. The archaeological application of GIS. London and New York: Taylor & Francis. Widgren, M., 1999. Is landscape history possible? Or, how can we study the desertion of farms? In Ucko, P.J. and Layton, R., (eds). The archaeology and anthropology of landscape, 94–103. London/New York: Routledge.

5 I Walked, I Saw, I Surveyed, but what did I see?... and what did I survey? Stratford Halliday Everyone sees the landscape slightly differently and the capacity of the individual to find, observe and record archaeological features is tempered by the evolution of knowledge in general, and their knowledge and experience in particular. Two case-studies are presented to show how these factors operate, one dealing with an individual monument, the other an area of landscape. It is suggested that these factors are not restricted to survey in the field, but equally affect the interpretation of lidar data. Both depend on an engagement with the macro and micro topography of the landscape, irrespective of whether this takes place in the field or on a computer screen. Keywords: field survey, observation, landscape recording, experience, interpretation, archaeological knowledge

Introduction Survey is easy. Anyone can do it. You just go out into the countryside and record what you see. Well yes…and no. There is little doubt about the basic premise, whether you are walking intuitively looking for something in particular or simply sticking to fixed transects, but after thirty plus years in the field for the Royal Commission on Ancient and Historical Monuments of Scotland (RCAHMS) I am equally certain that in either case different people often see different things, or at the very least see the same things differently. Indeed, over the years, my own repeated visits to favourite sites tell me that I am no different, my own view and understanding often shifting from visit to visit, sometimes in only subtle and minor ways, but on others in gross measure. And then there is the ‘record’ itself, covering a multitude of sins from a spot on the map to a topographical description or a drawn plan. I cannot be alone in discovering, survey staff in hand, that the site I wrote a description for on one occasion has turned into a rather different beast on another. Traditional survey is not a technical exercise in right and wrong, it is a subjective engagement with the

topography in the formation of a spectrum of individual and consensual opinions, in which the parameters of good and bad observation are simply shades of grey. To take as an example Raeburnfoot, a little north of Eskdalemuir, Dumfriesshire, in southern Scotland. How was it that in 1979 two of us failed to find the Neolithic bank barrow that crosses the valley obliquely from northeast to southwest just above the Roman fort? We examined one of the terminals on Tom’s Knowe for goodness sake, and like the Ordnance Survey (OS) surveyor who had found it before us concluded that perhaps it was an unusual ditched barrow. Some ten years later, two of our colleagues found the other terminal on the ridge above the fort, and tracked the body of the mound with its shallow flanking ditches southwards in the belief that it was probably a roadway leading up to some other Roman installation. It was left to me to make the connection with Tom’s Knowe, on the other side of the valley, but it took a visit by another colleague, Gordon Maxwell, to open our eyes to what we had really found. On the one hand schooled in a lifetime of Roman earthworks in Scotland, on the other the guiding hand behind

64

Stratford Halliday the first systematic aerial survey of cropmarks in Scotland, Gordon brought a completely different perspective. We duly drew and mapped the bank barrow in its landscape (RCAHMS 1997, 92, fig. 83; 108–9), and yet ten years later I was back again, this time to look at the earthworks of a Roman temporary camp that another of our colleagues had happened across on aerial photographs of the same area (Jones and McKeague 2009, 125). Its perimeter cuts across the bank barrow, we had even mapped part of one of its sides, but we had failed to recognise its significance. Survey is easy…isn’t it? So how could I have made such a meal of this tiny fragment of the Scottish landscape? I am not going to supply an answer to that question just now, but these are the issues I want to explore in this essay, drawing on two case-studies, one representing the particular – a single fort known as Corsehope Rings – the other the general – the landscape of Village Bay on Hirta, St Kilda. Both case-studies are constructed against a historical framework in which record and interpretation have changed over time, though the visible archaeology has remained essentially the same. Both are asking the same question. Why is it that separate people pausing in the same vista look at different things? They see the same texture of stalks, stems, trees, stones and living rock as far as the eye can see, but what do they actually look at? And of course surveyors are constantly on the move, their eyes assessing and reassessing a continually shifting tableau as the light and perspective alters with every step. In effect they are selectively screening the landscape, filtering out and discarding the majority of what they see to reduce it to a relatively small number of triggers that involve a second glance. These are what they actually look at, and most are probably immediately dismissed as still more is filtered out to leave a tiny rump of archaeological features. These, then, become the focus of their recording, and if they are really honest with themselves, these, this tiny fraction, are the only things they have really looked at – and it all happened in a twinkling of an eye. But the question remains, why this fraction and not some other?

Corsehope Rings, Midlothian This fort is one of a small number across southeast Scotland defended by multiple lines

of relatively low and inconspicuous ramparts. It is probably for this reason that it does not appear on Roy’s Military Survey drawn up in the mid-18th century, which instead shows a now severely reduced univallate fort on the opposite side of the valley to the east (Roy 1747–55). On the west Roy has nothing, and though this cannot be taken too literally, it is a reasonable conclusion that at that time the spur on which the fort stands was mainly in rough pasture. By 1853, however, when the first edition of the OS 6-inch map was surveyed (Edinburghshire 1854, xxiii), the eastern end of the spur above the main valley had been enclosed by field boundaries, and the interior of the fort and large patches on the flanks of the hill to the southeast and south also seem to have been improved, albeit that this is only in the sense that they are not shown as rough pasture. The depiction of the fort itself shows four ramparts, broken by two narrow gaps on the southwest and west-southwest respectively. A strip of trees dropping down from immediately outside the outer rampart on the southeast had yet to be planted, and an enclosure carried concentrically round the whole fort is not shown, though this is perhaps no more than a reflection of the small scale of the depiction. The first edition of the 25-inch map drawn up in 1892 represents this enclosure with a continuous line (Edinburghshire 1894, sheets xxi.9 & 13), which on the south flank of the fort is taken by a ruined wall that was relatively newly built when William Galloway surveyed the fort in 1879 (Galloway 1880, 259–60). Elsewhere, however, its perimeter can be traced only intermittently as a low turf bank. In 1879 this was sufficiently decayed that Galloway mistook it for a component of the defences at the northeast entrance (1880, 257–8). This minor detail aside, Galloway’s description and his measured plan are a model of good observation and recording (Figure 5.1). Trained as an architect, he was familiar with the disciplines of measured survey and triangulated the line of the inner rampart with a chain from a central baseline to plot the shape of the fort. He also managed to render correctly the positions of the rampart terminals at the entrances on the northeast, southeast and southwest respectively, and seems to have measured a series of transects across the rest of the defences. He was certainly sufficiently familiar with the defences that he recorded the variations in their overall depth, noted their topographical course, and to the south

5  I Walked, I Saw, I Surveyed, but what did I see?...and what did I survey?

65 Figure 5.1: Corsehope Rings, Midlothian. In 1879 William Galloway triangulated the shape of the interior from points set out at intervals of 50 feet along a baseline through the centre of the interior, but he also incorporated two larger triangles to check the overall result. DP110698 © Crown Copyright: RCAHMS

of the northeast entrance identified traces of an intermediate line between the second and third ramparts. At the southwest end of the interior he also depicted two large round-houses, though he had little understanding of quite what he was recording (1880, 260). A little over ten years later, in 1892, the same year that OS surveyors produced their new depiction for the 25-inch map, the fort was visited by David Christison and Frederick Coles, who supplemented Galloway’s record with two measured profiles across the defences (Christison 1895, 117–18). On 15th July 1913 Graham Callander, recently appointed as an RCAHMS investigator, visited the fort in the preparation of the County Inventory for Midlothian, though the published plan was not surveyed until 26th May 1915 (RCAHMS 1929, 76–8, no. 108; RCAHMS MLD17/1). This new plan, probably prepared with another new appointment to the RCAHMS staff, Charles Calder, was drawn with a plane table from a single station at the northeast end of the interior. Most of the measured points lie along the innermost rampart, but apart from a rather angular depiction of the western

quarter they captured many irregularities in the shape of the fort that had escaped Galloway. Callander’s description written in 1913 draws attention to ‘the faint traces of dug-out hollows of indeterminate character’ within the interior (RCAHMS 1929, 78), but no attempt was made to resolve these on the plan. Indeed, had Galloway not depicted two of them, as did the OS surveyors on the OS 25-inch map, it is reasonable to wonder whether they would have been mentioned. The measured points on the depiction of the defences prepared two years later show that on that occasion they examined the northeast end quite closely, but recorded progressively fewer measurements onto the outer defences as they got further away from the survey station, and probably, therefore, examined the southwest end in correspondingly less detail. The next recorded survey was carried out by Robert Stevenson, Keeper of the National Museum of Antiquities of Scotland (Stevenson 1949). In 1940 Stevenson had embarked with Kenneth Steer, then a young investigator of the RCAHMS and later to become its Secretary (Chief Executive), on an excavation of a round-

66

Stratford Halliday

Figure 5.2: Corsehope Rings, Midlothian, looking eastwards. The palisade trenches cut into the ramparts are most clearly visible on the right of the tracks extending through the entrance in the foreground. Traces of numerous timber roundhouses can be seen within the interior SC1004936 © Crown Copyright: RCAHMS/aerial.rcahms. gov.uk

house in an earthwork enclosure at Braidwood, also in Midlothian. Unlike the stony banks that defined most round-houses then recognised, this comprised a shallow ring-ditch and was one of four shown in the interior on the plan in the County Inventory (RCAHMS 1929, 156–7, No. 206). Stevenson returned to Braidwood in the summer of 1947 to strip the rest of the round-house and to extend the excavation over an adjacent palisade trench, part of which is visible forming a groove in the turf (see Gannon 1999, 109, fig. 2). While Steer is usually credited with the first recognition of palisaded enclosures as minor earthworks on the Border hills (Steer 1949), on one occasion Stevenson hinted to me in conversation that the connection was first drawn at Braidwood. Be that as it may, there was a growing awareness through survey and excavation in Roxburghshire (RCAHMS 1956;

Piggott 1949) that many of the settlements contained earthwork remains of timber roundhouses. Mrs C.M. Piggott (later remarried to become Margaret Guido) drew Stevenson’s attention to the round-houses in Corsehope Rings, and he plotted no less than seventeen onto the RCAHMS plan (Stevenson 1949, 2, fig. 2). Richard Feachem, another RCAHMS investigator, visited the fort (Feachem 1963, 138; 1965, 196), but the traces of palisade trenches that are also visible eluded them all and were to pass unnoticed until the summer of 1979 when the fort was photographed from the air one summer evening by Mike Brookes of Historic Scotland and Peter Hill (see Figure 5.2). The easiest line of approach to the fort mounts the spur to the east and arrives at the northeast entrance. This, however, is the least informative way to observe the earthworks, and a better route

5  I Walked, I Saw, I Surveyed, but what did I see?...and what did I survey? skirts the southeast flank of the fort between the outermost rampart and the later drystone enclosure wall. This comes past a quarry that had been driven across the defences on the south before Galloway’s day, and round into the southwest entrance on the same axis as the aerial photograph in Figure 5.2. For a first-time visitor this approach conveys something of the character of the belt of defences, stepping back up the slope in a series of low tiers over a distance of some 30 m. The presence of shallow external ditches becomes more obviously apparent between the quarry and the entrance, and the keen eyed will start to wrestle with the minor scarps forming the outer lip of the ditch, a low counterscarp bank and a palisade trench, the latter, as can be seen on the aerial photograph on the right of the entrance, cutting obliquely across the counterscarp bank to return around the head of the ditch terminal and ride up onto the crest of the rampart. As an exercise in field observation, the battle is to get them to engage with what they can see and what they are actually looking at, simply to distinguish between the minor scarps making up the different features – as often as not they will try to leapfrog the components of observation to go directly to interpretation. Walking up between the terminals, deep tractor ruts distract the eye, obscuring most of the detail on the left-hand side, but on the right the groove of a second trench rides up onto the next (third) rampart. This is set towards its rear and for much of its length is identifiable only from the acute sharpening of the scarp forming the back of the rampart. Christison portrays this feature in his section drawn across the defences on the northwest (1895, 18, fig. 5), though he did not understand its significance. The next (second) rampart displays this same profile, but a much clearer groove can also be seen extending parallel to the ruts of the track to link its terminal to the innermost rampart, apparently forming one side of an entrance passage. The innermost rampart terminal extends a little beyond the line of this trench, but the body of its mound also displays yet another groove running back along its leading edge, while the inner scarp has the same sharpened profile encountered in the outer ramparts. Almost certainly there is yet another palisade trench here. To the southeast (right on Figure 5.2) this latter eventually detaches itself from the inner scarp of the rampart to swing in a gentle arc round the rear of one of the large ring-ditch houses noted by Galloway, before

doglegging round the next round-house and returning in a gentle arc to the rampart on the northwest side of the interior. The easternmost transit of this groove appears to pick up the same line as another groove, in this case with a low bank on its northeast lip. This cuts in a shallow arc across the southwest end of the interior and seems to represent yet another enclosure, though whether the earthworks of a palisade or simply a post-medieval fence is not certain. Accompanying banks of upcast are a fairly common feature of palisade earthworks, as here usually occurring on the outer lip of the trench, but within Corsehope Rings there is also another broader groove with a much bolder bank swinging obliquely across the northeast part of the interior. This apparently fades away at its east end just short of the inner rampart, but at the point where it disappears it is aligned on the corner of the plantation strip, hinting that it is a much later boundary and thereby raising a question about the enclosure cutting off the southwest end of the interior. Despite the palimpsest of round-houses visible within the interior, it is also clear that the interior has not entirely escaped cultivation. This is given away by three wide-spaced grooves cutting across it from northwest to southeast, and traces of a headland about 3 m wide running concentrically within the innermost rampart along the length of the northwest side. This ploughing, which is stratigraphically the last episode of activity recorded in the field archaeology, is presumably the improvement to the pasture identified on the first edition OS 6-inch map, but these two sub-divisions of the interior are possibly evidence of other phases of post-medieval land-use. To all intents and purposes the only changes that have occurred since the interior was cultivated are the ruts from tractors driving through the entrances and various bits of erosion caused by grazing beasts and burrowing animals.

Village Bay, Hirta, Western Isles The drystone storehouses known as cleits (Figure 5.3) have been a part of the St Kildan landscape since before the late 17th century (Harman 1997, 161), though many of those standing in the fan of crofts above the shore in Village Bay (Figure 5.4) are much more recent (see Stell and Harman 1988, 7). The crofts themselves were only brought into being under the influence of

67

68

Stratford Halliday

Figure 5.3: A typical stone cleit standing on a pedestal of uncut turf above Village Bay, St Kilda. © Stratford Halliday

Figure 5.4: A general view of Village Bay, St Kilda. © Jill Harden

Neil MacKenzie, the resident minister 1830–44, who was also responsible for reorganising the old township along what is known as The Street. The first record of the new arrangement is provided by a plan drawn up by H. Sharbau, initially in 1858, but with additions in 1860–1, including the first of the sixteen gabled cottages that became the principal dwellings on the crofts

until the islanders were finally evacuated in 1930. With all the caveats that the use of such plans entail, this is an invaluable tool in understanding the history of the existing buildings, yards and gardens along The Street and underpins the analysis of a survey carried out in 1983–6 by the RCAHMS (Stell and Harman 1988). An example of fine penmanship, this plan shows

5  I Walked, I Saw, I Surveyed, but what did I see?...and what did I survey?

69 Figure 5.5: An extract from one of the RCAHMS pencil survey drawings of Village Bay, St Kilda, prepared in 1983–6 at 1:500 © Crown Copyright: RCAHMS

all sorts of minor scarps and boulders to create a variable texture that reasonably fairly represents the different crofts as they are today. This is amongst the earlier surveys of landscape carried out by the RCAHMS and was drawn on plane tables at a scale of 1:500 (Figure 5.5). This scale was more usually used for site plans of small groups of buildings rather than large townships, and it is a sign that this particular landscape plan essentially evolved out of a smaller site plan. This is a history we should be grateful for, showing far

more detail than if it had been drawn at 1:1000, then the typical scale for a large township plan. Had it been initiated few years later the outcome would probably have been no more than a line drawing for depiction at 1:2500. Nevertheless, this survey has provided the starting point for most archaeologists working in Village Bay since, amongst them Andrew Fleming, whose book St Kilda and the Wider World: Tales of an Iconic Island (2005) has lain to rest many of the myths that have grown

70

Stratford Halliday up around the islands. He emphasises from the outset that the sea is as much a highway as a barrier, a cultural connector rather than a separator, and thus a linkage between the islanders and their cultural cousins elsewhere in the Western Isles. In terms of survey, his most important contribution to the archaeology of St Kilda is probably his recognition of the stone tools – hoe blades, Skaill knives (simple flake stone tools) and hammer stones – reused as pinnings in drystone walls and lying discarded on clearance heaps (Fleming 2005 39–51). They caught his eye. Using the RCAHMS plan he went on to systematically plot their distribution across Village Bay and to identify the quarries from which the raw materials had been won. Would I have noticed them if I was mapping the bay? One or two perhaps, but probably not in any numbers, and without the tools I doubt I would have gone on to register the significance of the line of quarry scars visible across the face of Mullach Sgar overlooking the bay on the west. Fleming brought a new eye and a different working experience to the island, and uncovered an entirely new dimension to its landscape, which he attributes to a Neolithic occupation. Other commentaries that Fleming offers relate more directly to the interpretation of features depicted on the RCAHMS plan. The remains of the old boundaries immediately above the head-dyke – what Fleming calls the Tobar Childa field-system – as the remains of an earlier pattern of enclosure is a classic bit of archaeological survival, though their attribution to the Bronze Age perhaps owes something to his experience elsewhere (2005, 51–4). The old tracks, so labelled on account of an earlier interpretation of the archaeology, are patently the remains of old boundaries, as Mary Harman also realised (1997, 79), though his tentative dating to the Norse Period is not particularly convincing (2005, 65–6). The position of the old township, as he observes, is clearly betrayed by the traces of the robbed footings of buildings lying to either side of The Street to the east of the burial-ground (2005, 132–8). In contrast, the impression of a cluster of ruins on the RCAHMS survey immediately below the head-dyke to the northwest is probably a product of the survey scale and a consequence of the detail that scale demanded; confronted with a sea of stones and boulders, some of which appear to be ordered into lines, there are as many solutions here as surveyors. For what it is worth, apart from a few relatively clear robbed structures,

I would have been more ruthlessly selective. My own fieldwork on St Kilda came about through a new RCAHMS project on the archipelago to map all the cleits. I was given the task of checking over the earlier plan of Village Bay. For three days I worked my way across this landscape, led by the earlier plan and thus within the framework of my predecessor’s observations. Finally, however, the penny dropped. The Tobar Childa field-system extends throughout the area that was laid out in crofts below the head-dyke, right down to the sea. You can see it if you look hard enough. In addition to several lengths of old dykes already recorded on the RCAHMS plan, the clues lie in the way so many of the cleits within the crofts appear to perch on grassy pedestals, to use the same description that Fleming applied to some of the cleits standing in areas of turf-stripping outside the head-dyke (Figure 5.3; Fleming 2005, 95). Set across the leading edges of earlier fields, these cleits preserve the original profiles of several underlying lynchets (Figures 5.6 and 5.7). If nothing else, they bring into sharp focus the scale of earth moving that occurred in the course of MacKenzie’s improvements, in places digging away up to 1 m of soil. Where the original profiles of some of these lynchets are fossilised beneath cleits, their crests have migrated anywhere between 5 m and 10 m back uphill to become barely perceptible breaks of slope. Other fossils are preserved on the slope below The Street and are revealed in differences in height along the lines of the croft boundaries, while the topography of The Street itself, where the ground is consistently at a higher level to the rear of the cottages, almost certainly preserves a series of earlier field boundaries. In other words, the position of The Street was partly determined by the layout of the earlier field-system. It is more than likely that a trackway already existed on this line. Some of the features containing burnt stones, which Fleming speculated might be burnt mounds (2005, 54–6), are further fragments of these fields, though the incorporation of so much burnt stone in the field soils suggests that they may have modified the topography of an earlier settlement mound. This is not the place to discuss the wider implications of these new observations. Suffice it to say that it seems inherently likely that Fleming’s Tobar Childa field-system, his supposedly Norse head-dykes and the fossils within the compass of the head-dyke are all components of the

5  I Walked, I Saw, I Surveyed, but what did I see?...and what did I survey?

71 Figure 5.6: Village Bay, St Kilda. The large stones incorporated into the foot of this cleit formed part of an earlier boundary dyke that has otherwise been robbed away. © Stratford Halliday

Figure 5.7: This cleit in the croftlands of Village Bay, St Kilda, preserves the original profile of an earlier lynchet, while the boundary of a later garden plot can also be seen extending off-picture to the left. © Stratford Halliday

field-system that was replaced by the crofts during MacKenzie’s improvements. It may have accreted over several centuries, indeed the lines of the dykes imply an expansion up onto the slope above the head-dyke below Conachair, and round eastwards onto the foot of Oiseval.

Despite the air of deep antiquity presented by the shattered walls above the head-dyke, there is no compelling evidence that any of them should be Bronze Age. The important relationships are within the improved ground, namely the two dykes that ride up onto the corbelled building

72

Stratford Halliday

Figure 5.8: At Greenborough, Scottish Borders, the rectangular outline of the timber stockade enclosing this small settlement, which probably dates from the mid 1st millennium BC, is traced by a shallow groove in the turf. Traces of what may be contemporary cultivation rigs (cord rig) lie immediately adjacent on the left. SC 1271344 © Crown Copyright: RCAHMS/aerial.rcahms. gov.uk

known as Calum Mor’s House (Stell and Harman 1988, 44, no. 13) and the way the burial-ground appears to be set into one of the earlier fields. An educated guess would suggest these relate to a medieval or later history. The important point, however, is that every surveyor that has worked here has seen the same things, but as a result of Fleming’s work and the new RCAHMS survey we are looking at them in different ways and are still recognising new subtleties in this remarkable landscape.

Discussion In both these case-studies the archaeological evidence has remained essentially unchanged throughout the span of archaeological investigation, yet the observation of the detailed surface topography has changed dramatically, with far-reaching implications for the way we interpret on the one hand Border forts, and on the other a Western Isles township. At the root of this evolution must lie the way knowledge has been exercised by our cast of leading characters. In the case of Corsehope Rings, it would be totally unreasonable to criticise Galloway, Christison and Callander for their

failure to observe the palisade trenches, though it was certainly Galloway’s assumption that the ramparts had been crowned with timber stockades (1880, 259); they had no knowledge of such minor earthwork remains, which would not become part of the archaeological lexicon until the 1950s, so they could not have had any expectation of finding traces of them in the turf. Even when Steer first published his discoveries in Roxburghshire, the general belief was that palisade trenches would be found from the air rather than on foot, the notable exception being Greenbrough Hill (Figure 5.8), which he considered so small that it was beyond the resolution of the available aerial photographs (Steer 1949, 66). Callander might have done better with the ring-ditch houses in Corsehope Rings, but the vegetation would have been at its worst when he visited in July. Stevenson found only the round-houses, but then that is what he had gone to find. At a guess he would have approached through the northeast entrance where the evidence of the palisading is not particularly clear on the ramparts and, with the RCAHMS plan of the earthworks in his hand, there was no reason for him to walk the circuit of any of the ramparts. Peter Hill of course was using oblique evening lighting to look for such things from the

5  I Walked, I Saw, I Surveyed, but what did I see?...and what did I survey? air. Would we have found them on foot? I would like to think so, but then courtesy of the work of Steer, Feachem, the Piggotts, Stevenson and, in Northumberland, George Jobey, our incremental knowledge is so much greater than theirs and we would have been specifically looking for traces of timber enclosures and round-houses. This sort of overall knowledge supports an evolving range of recognisable archaeological features, though this is not necessarily the subset of knowledge carried by every individual who sets foot in the field. Nevertheless, the latter brings together his or her own subset of this archaeological knowledge and a cumulative practical experience of their own. This latter is a dangerous arena, in which the acquisition of untutored experience can lead to poor observation and misleading interpretations, yet tutored can constrain and channel observation and understanding along what are essentially predetermined and conventional paths. Yet in the first instance we all have to learn how to distinguish between what is natural and what may be artificial, though again thirty years has taught me that there is no dogmatic division here, if only because the earthworks we seek to record have been subjected to a wide range of natural processes which may have dramatically modified their forms. The accumulation of this sort of experience, however, is fundamental to the visual screening process I elaborated in the introduction. What first catches your eye is something that seems to break your perception of the natural line of the countryside, which of course may have been heavily modified by centuries of farming. It stands to reason that the less you understand of the natural processes that have formed different landscapes, and the processes that continue to modify them, the less you will select out for a second glance. It is the archaeological experience that kicks in as we go into that next stage of screening. This is where we resolve whatever has caught our eye into a three dimensional shape, which we then reference back into our personal experiences of visiting different types of monuments, with their links back into the accumulated knowledge of the discipline. If you have visited several palisaded settlements, better still if you have been on the staff to survey one, you are far more likely to recognise one in the field. You will have engaged with the presence of a bank of upcast and how under some circumstances this may have obscured all traces of the trench itself, indeed

how all trace of either may disappear in some parts of the circuit. But that experience will have told you to walk along the feature, to examine it from different angles. In the case of a fort like Corsehope Rings, it will have told you that you do not walk straight into the interior. First you walk the inner rampart of the defences looking both in and out; before you are done, you will have not only walked the circuit of every rampart, but systematically quartered the interior, making sure you see all of it and preferably from a range of different angles. But what if you don’t have the practical experience of a particular type of monument? Chances are that unless it is blatantly artificial, it will become a casualty of the second stage of the screening process. Again drawn from my own experience, take the burnt mounds of Wigtownshire. Keith Blood of the OS had first shown us examples of these monuments in Sutherland in 1977, but we were singularly unsuccessful in discovering any for ourselves until 1986, despite having worked in parts of Dumfriesshire where they probably exist in some numbers (RCAHMS 1997, 100–102). In 1985–6, however, in the course of what would now be termed a walkover survey of a block of 150 square kilometres around Glenluce we identified a total of 75, but not until after we had consulted Northern Irish colleagues standing on one evidently artificial and classically shaped example. Most were identified in the last 50 square kilometres we surveyed, coming in a wide range of shapes and sizes, and some entirely buried beneath peat (Halliday 1990, 61). There can be little doubt that we missed many more across the rest of the area, and indeed in the neighbouring parts of southern Ayrshire. I have made a point of passing this experience on to new colleagues, because to some extent surveyors find what they look for. This, I am afraid, is an unpalatable truth, but it is very difficult to go into any landscape and look for the entire range of known types of archaeological feature at the same time, including the types you have never seen before. Different features may occupy particular topographical locations and these will require a deliberate process of assessment in the way you see your landscape to make sure you go and stand on them. Palisades and timber round-houses are cases in point; unless you put yourself within 10 m you may not see them. Things like burnt mounds are often tucked away in swamps you would not necessarily choose to visit, so if you do not look

73

74

Stratford Halliday for them you probably will not find them. Being able to draw on someone else’s working experience is clearly an important aspect of learning the surveyor’s craft, but not in such a way that this form of knowledge is simply being cloned from one to another. This is the point about the Village Bay landscape. Within the cocktail of skills that make up the best surveyors is the inclination to challenge everything that they both see and look at, testing received wisdoms and challenging themselves to look at the field evidence of the past in different ways. Fleming’s observations of the stone tools, for example, gave the landscape of Village Bay a completely new dimension. In my own case I narrowly escaped simply adding and subtracting a few scarps on the existing RCAHMS plan, essentially titivating the veneer of little garden plots that are visible on the majority of the crofts, but without arriving at any understanding of the real processes that had been operating in that landscape and the way different features from the earlier organisation were manifesting themselves. Village Bay is an UNESCO World Heritage Site and is due for a lidar survey as part of the Scottish Ten Project (www.scottishten.org), but at the time of writing this has not been completed. Had this data been available it is perhaps worth asking briefly what role it might have played in elucidating the observations and interpretation set out above. I fully recognise that until the data is made available any case I might make here is untested, but instinctively I do not believe that we should expect that the creation of a digital skin for the landscape will of itself provide any new insight. It is simply another tool in the surveyor’s armoury, albeit a tool that is immensely powerful and flexible, and capable of interaction with other digital datasets. The parallel is surely with the development of stereoscopic photogrammetry, which in its day was said to spell the end for the OS field surveyor. In the event, the surveyor on the ground proved as essential as ever, on the one hand checking aspects of the photo interpretation, on the other filling-in gaps where the surface of the ground had been masked from the aerial camera. Lidar, it seems, is not altogether immune to some of these same problems (see Ainsworth et al. this volume), while the resolution of the data commissioned and the way in which it is subsequently processed and visualised (see Kokalj et al. this volume) will determine whether or not the fine detail of a site such as Corsehope

Rings is visible. One might anticipate that combinations of unenclosed timber roundhouses and certain types of surface vegetation will prove as challenging on the computer screen as they are in the field. The creation of a digital model of the landscape, however, creates the opportunity to shift some measure of the interpretation that might normally take place in the field into the office. But if that is the case, the process of what you see and what you look at will still be driven by similar considerations to the ones I have already outlined, with the added twist of a new set of scales and perspectives to the viewpoints. Furthermore, any automated feature-recognition programmes will have to be reviewed by the operator and will be prey to all the same constraints of knowledge and experience. There are specialist skills to be learned, but in the final analysis, the solutions will always lead back into the field; can you see it, what does it look like, how does it change our perception of the features round about? While all the features that have been observed in Village Bay will probably appear in the lidar data, and may be flagged up by visualisations, this is some way short of recognising them as features that inform the understanding of this landscape and its evolution.

Conclusion Nobody will be surprised that knowledge and experience are such vital ingredients in determining what we see and look at in the landscape, but experience cannot be acquired in isolation. There really is no substitute for having to explain to a sceptical colleague what you think you are looking at, to break it down into its components and argue the validity of each lump and bump en route to some consensus. But to get the best from survey as a research tool, to use it in all its forms as the first line of attack to find out about the past, this has to come with imagination and breadth of view: imagination to try new approaches and to think in different ways, such as the applications of lidar; breadth of view to provide context and relevance to the results. But what sets the best surveyors a little apart is that they examine what they do introspectively and with some humility, continually learning from the experience. They also have the discipline to push themselves into odd and unlikely corners of the landscape at the end of a long and tiring day.

5  I Walked, I Saw, I Surveyed, but what did I see?...and what did I survey? To them every day in the field is an opportunity to learn something new, and once they stop learning they will probably give up. So how did I do in Eskdalemuir? At face value, not very well, but I understand how the style of survey that we initially conducted thirty years ago failed to find the bank barrow above the Roman fort. A rapid walkover, we were essentially contouring along large sections of valleys to achieve maximum command of the ground for the greatest economy of effort in the limited time available. We must have missed the terminal by a whisker, lost it on a zig or a zag, and were just unlucky to cross its line where there is barely any relief to either the bank or the flanking ditches. In my defence, at least I remembered the ‘odd barrow’ to find the southwest terminal. And the Roman temporary camp? Well that remains one of the most challenging earthworks I have ever seen, the more so simply for the sheer size of the enclosure and the variations in the character of its perimeter (see Jones and McKeague 2009, 128–31). If the mapping scale survey of the post-medieval field-system we were carrying out twenty years ago had taken in the whole ridge we might at least have been in with a shout, but it did not. Somewhere serendipity and luck have to play their part in survey, right place, right time, shaft of light and a revelation, but not on that occasion. But dear reader, the National Grid Reference is NY 2500 9959, why not have go yourself. Put the plan in your back pocket and see how you get on without it. Alternatively commission some high resolution lidar before you go and who knows what else might turn up. Its ten years on now and I am probably due another visit.

References Christison, D., 1895. The forts of Selkirk, the Gala Water, the southern slopes of the Lammermoors, and the north of Roxburgh. Proceedings of the Society of Antiquaries of Scotland 29 (1894–5), 108–79. Feachem, F.W., 1963. A Guide to Prehistoric Scotland. Batsford: London.

Feachem, F.W., 1965. The North Britons; The Prehistory of a Border People. Hutchinson: London. Fleming, A., 2005. St Kilda and the Wider World: Tales of an Iconic Island. Windgather Press Ltd: Macclesfield. Galloway, W., 1880. Notice of a camp on the MidhillHead, on the estate of Borthwick Hall, in the parish of Heriot, Midlothian, the property of D. J. Macfie, Esq. Proceedings of the Society of Antiquaries of Scotland 14 (1879–80), 254–60. Gannon, A.R., 1999. Challenging the past: the resurvey of Braidwood Hillfort. In Frodsham, P., Topping. P. and Cowley, D., (eds). ‘We were always chasing time.’ Papers presented to Keith Blood. Northern Archeology 17/18 (special edition), 105–11. Halliday, S.P., 1990. Patterns of fieldwork and the distribution of burnt mounds in Scotland. In Buckley, V., (compiler). Burnt Offerings: International Contributions to Burnt Mound Archaeology, 60–1. Wordwell Ltd: Dublin. Harman, M., 1997. An Isle called Hirte: History and Culture of the St Kildans to 1930. Maclean Press: Isle of Skye. Jones, R.H. and MacKeague, P., 2009. A ‘Stracathro’gated temporary camp at Raeburnfoot, Dumfries­ shire, Scotland. Britannia 40 (2009), 123–36. Piggott, C.M., 1949. The Iron Age settlement at Hayhope Knowe, Roxburghshire, excavations 1949. Proceedings of the Society of Antiquaries of Scotland 83 (1948–9), 45–67. RCAHMS, 1929. Tenth Report, with Inventory of Monuments and Constructions in the Counties of Midlothian and West Lothian. HMSO: Edinburgh. RCAHMS, 1956. An Inventory of the Ancient and Historical Monuments of Roxburghshire, with Fourteenth Report. HMSO: Edinburgh. RCAHMS, 1997. Eastern Dumfriesshire: an archaeological landscape. HMSO: Edinburgh. RCAHMS MLD17/1. Pencil plane table survey. Roy, W., 1747–55. Military Survey of Scotland. Available online at National Library of Scotland (htttp://maps.nls.uk/roy/index.html). Steer, K., 1949. The Identification of Palisaded Enclosures from Surface Indications. Proceedings of the Society of Antiquaries of Scotland 83 (1948–9), 64–7. Stell, G. P. and Harman, M., 1988. Buildings of St Kilda. HMSO: Edinburgh. Stevenson, R. B. K., 1949. Braidwood fort, Midlothian: the exploration of two huts. Proceedings of the Society of Antiquaries of Scotland 83 (1948–9), 1–11.

75

6 Reading aerial images Rog Palmer

Our eyes, brain, accumulated knowledge and experience guide us through life and also through the processes of interpretation of aerial images from any source. Awareness of these elements and their effects on interpretation are discussed through the ways in which we see things. A case study at Hambledon Hill, Dorset, explains some of the relationships between observation and interpretation and returns to the theme of knowledge and experience, noting how these are confined by contemporary theory and practice. Non-archaeological features are also visible in aerial images and lidar and the relevance of these to interpretation, management and conservation is noted. Keywords: knowledge, experience, vision, perception, subjectivity, observation, interpretation, Hambledon Hill, natural features

“Whatever the form of photographic or visual record from the air, the end is to make it an instrument of purposeful archaeological policies; the means lies in a clear understanding of what evidence to look for and how to find it” (Bradford 1957, 11).

Introduction and definitions The eye-brain process in humans is a fantastic combination that enables most of us to make our way through the world in a relatively trouble-free manner. We see, we comprehend and we make decisions based on a combination of events and how they relate to our experience of the world. Take the simple act of crossing a road, where children in Britain were taught to use a mantra of ‘look left, look right then look left again’. If our eyes perceived no traffic then our brain considered it was safe to cross the road. Accident statistics show that this simple process works pretty well, until we go to a country where the traffic drives on the ‘wrong side’ of the road. Many of us have encountered this and have stood at the side of the road wondering which way to look. Our experience has let us down and, for our own safety, we need to learn

a new set of rules. This example illustrates that each of us has an individual view of the world derived from our past and our experiences. This accumulation of knowledge and experiences also has direct relevance to the way we interpret aerial images, or field remains (Halliday this volume), and for that matter any ‘visual’ information for archaeological purposes, including lidar. However, the central role of knowledge and experience is not routinely recognised, especially in some research areas. For example, the application of satellite images for archaeology (e.g. Parcak 2009) is often focused on technical image manipulation and not on archaeological interpretation, an emphasis that has played a large part in setting satellite images apart from lower-altitude photographs. To some extent the use of airborne laser scanning, or lidar, has followed a similar route into archaeology and, with the exceptions of people who worked with aerial images before embracing lidar, it is often the manipulators who have driven uses of the data. This can cause problems because a technical manipulation may progress through a series of mechanical steps but it is not

77

6  Reading aerial images part of a process of subjective archaeological interpretation which calls on experience and knowledge. This is not a problem that is unique to lidar, or satellite photographs, but can also be identified amongst users of aerial photographs, field surveyors and geophysicists. It can often be identified when practitioners talk of ‘cropmarks’, ‘features’, ‘anomalies’ or ‘lines’, and stop at this basic characterisation, rather than progressing to archaeological interpretations of ditches, field boundaries and tracks, for example. In my opinion, anyone who uses those terms is avoiding the problem of interpreting what has been identified. Remotely sensed datasets, be they aerial photographs or lidar, frequently record traces of ancient ditches, pits and walls which once formed parts of settlements and other features which partitioned the landscape to identify ownership and use, and it is these that are the proper interest of an archaeologist, not simply ‘cropmarks’, ‘anomalies’ and ‘lines’. Whatever processes are applied to analysis of sources of archaeological information, someone at some stage has to look at them and decide what is on them. Knowing how to do this, and how knowledge and experience play their part, is the focus of this contribution. The process of interpretation may be divided into examination of a photograph (or image or visualisation) and archaeological description, though in practice these processes are interwoven. When we examine or interpret a photograph we may initially see ‘cropmarks’ or ‘lines’. This, however, is a preliminary step towards archaeological interpretation when we translate those perceived anomalies into their past forms when it is possible to do so. And it is not always possible to do this, so interpretative maps may include uncertain categories such as ‘possible ditch’ and ‘potential archaeological sites’ (e.g. Hesse this volume). We can rarely be 100% certain that everything we identify through its low relief or as different coloured crop growth is what we claim it to be, but by questioning the evidence, by relating what we see to what has been found on the ground elsewhere (through excavation or by field survey), we can often have a pretty good guess at it. One of the good things about the ‘art’ of interpretation is that we do not expect always to be right. This is not the same as expecting to be wrong, which is not good practice, but making mistakes can be a good way to learn, and is part of what should be an iterative process of archaeological observation and interpretation.

In the rest of this contribution I will discuss the interplay of these issues from my own experience interpreting aerial photographs – issues that have a clear resonance with interpretation of lidar.

Knowledge and its roles We cannot confidently interpret aerial images (or lidar) without first having a level of knowledge and/or expertise about what we are looking at or for. Take the example of pottery with which almost all archaeologists have some familiarity. A beginner may be able to identify that an object is a piece of pottery. If the sherd includes diagnostic characteristics, someone with more learning, or an interest in a particular place or period may be able to state that it is, for example, a piece of Neolithic pottery. At more expert level, an archaeologist may examine the mineral content of the sherd and be able to tell us where the clay came from. So it is with the examination of aerial images that we should expect the most thorough and reliable interpretations to come from people who have had most experience using them. This does not mean only peering at photographs every day, but embracing the knowledge that comes from field investigation (by which I do not mean excavation) and observation of the modern landscape to see how it changes seasonally and through the action of modern farming and developments. A lot can be learned by looking out of the window during journeys. This can also add knowledge of landform in the area of interest and will give the chance to see different soils and locations and how these may affect modern landuse. On recurrent routes these journeys will show seasonal changes to crops and soils. It’s all valuable knowledge that will help our image interpretation regardless of whether we are working in a plough-levelled rural landscape, a less-tamed upland area or in a tree-covered environment. Many of these observations will be specific to particular localities but the principles are the same everywhere: to unravel images, beginning with today and working backwards in time (Palmer 2011); to build up a mental compendium of observations and knowledge; and create a narrative of the main points of interest. Modern and ancient features, natural and cultural must all be considered, as these will have an effect on our interpretation and reconstruction of archaeological landscapes.

78

Rog Palmer

How we see things From experience when teaching, it is apparent that problems of interpretation are often related to the qualities of an image such as tone, texture, pattern, shadow, size, shape and relief, of which size/scale is a principal ‘unknown’ while lighting (shadows) and textures can confuse as well as enlighten. Scale Take scale, for example, and aerial photographs of the countryside – a frequent archaeological subject. Almost all images contain information about scale in the form of buildings, roads or tracks, trees, livestock, tractor lines and so on, and they can be used to approximate the size of other things such as archaeological features. However, if these factors are not explicitly considered by the interpreter, they may fixate on their object of archaeological interest and then ignore potentially helpful and relevant non-archaeological information. Inexperience can lead to amusing errors of scale. As an expert witness in a legal case I was asked how I knew that what I was calling shrubs were not the cabbages that the opposing party claimed to be growing. In this example, the proximity of a caravan on the photographs provided scale and showed clearly that we did not have a new strain of giant cabbages as it was partly overhung by the shrubs. A sense of scale is important if an image interpreter is not going to be fooled into generating a new class of feature that is known to fit within a certain size range. This applies equally well to archaeological categories as it does to cabbages. Scale can be seen in at least two ways in the size of the archaeological features and in their relationship to the landscape. Vision and perception An appreciation of the form of the landscape is helped by sympathetic lighting, and manuals of photo interpretation and books about examination of images often state that vertical aerial photographs should be viewed with shadows falling towards the viewer. “The brain (perhaps quite sensibly) seems to assume that light comes from above and can therefore decode shading and shadows to infer depth relationships.” (Snowden et al. 2006, 217). Thus when we are trying to resolve major and minor topographic forms, and examining detail on aerial photographs, which can be hard enough at the best of times, we

should really reduce factors such as this that may otherwise make it more difficult. We can see the same positioning of the sun above us in the work of hillshading cartographers, who conventionally place the sun in the northwest, and also in many published views of hillshaded lidar products. So, for those in the northern hemisphere, to follow GIS and cartographic conventions of north to the top by putting the source of light below us, hinders our ability to perceive form and can easily cause us to see reverse topography. The impact of lighting on interpretation should be carefully considered, including that resulting from shading techniques applied to lidar products. For example, ambiguous lighting is in a demonstration of lidar and the Stonehenge landscape by Wessex Archaeology (2011). This is lit from the southeast and the web page includes a footnote that with this lighting “…the majority of people will perceive the topography as inverted.” They suggest that it is possible to train the eye-brain to correctly read this – but it seems curious to present a view that initially will confuse most people. Even if a viewer succeeds in correctly visualising such an image, it is possible that if may suddenly flip in the manner of a Necker cube or other such ambiguous images and present again the inverted view. Subjectivity and interpretation As well as learning what tricks are played on us by vision and perception, once we start looking at aerial photographs or lidar images our brain begins to compile a compendium of what it has seen. The scope and breadth of such a compendium increases with time and shows the value of experience in knowledge-based interpretation. Information can be extracted, from photographs, from fieldwork and by looking out of the window on journeys. Skills such as these are not conventionally taught – in fact archaeological field skills tend to focus on excavation, sometimes with artefact collection, but rarely teaching how to read surviving earthworks. Yet understanding earthwork remains on the ground is an important basis for being able to read three-dimensional aerial images, be these stereoscopic views of photographs or lidar images (see Halliday this volume). There is no limit to this knowledge and it will continue to expand and accumulate as new images are examined, as work moves to new places and from observations made during daily life. All this accumulated knowledge adds to the experience we draw on and it will affect what we perceive on aerial images.

6  Reading aerial images

For me, one of the good things about image interpretation is that it is subjective, that there can be several right interpretations of the same set of photographs, and that I expect my own interpretations to change as my experience increases and as my questions change. Take this example from a photo reading exercise. One student jumped directly to what he considered to be the archaeological content of the photograph and said he could see two Bronze Age barrows, adding that those are all he is interested in. Another said that he would prefer to call the barrows ring ditches – because that is what the photograph showed – and that they were set close to a palaeochannel which may have been a river or stream at the time of their construction. And, yes, they may indicate

79

Bronze Age barrows, but he would prefer to be cautious on their date. From this very simple example we have two different answers. Both may be correct, but one shows wider knowledge (i.e. not only the archaeological content) and a greater understanding of the potential and uses of aerial photographs than the other. Similarly, information about the location of the group of enclosures photographed in Figure 6.1 would be reduced in value without the site context shown by the curving palaeochannel that places it on a small promontory. It doesn’t matter if the channel was flowing at the time the enclosures were occupied – it shaped the ground, as is still apparent on the photograph, and probably influenced the location of the enclosures. To interpret a ‘site’ from aerial photographs without

Figure 6.1: A group of small enclosures that palaeochannels show were located on a small promontory. The natural information provides context for the archaeological site. Source: Author: 94.100/15

80

Rog Palmer noting its context is like excavating a pit to find only objects. The nature of ‘rightness’ of interpretations is something that will change – and must be expected to change – as the questions that are asked of the original data differ with time and with the research aims of each interpreter or archaeological project.

Processes of interpretation The book Wessex from the Air (Crawford and Keiller 1928) is rightly regarded as a pioneering use of aerial photographs that were taken by the authors four years before its publication. It is, however, much more than this and it repeats a way of working with aerial photographs that was previously published by Crawford in Air Survey and Archaeology (1924), where he included a number of maps compiled during his photo interpretation and used to explain features on the photographs. Wessex from the Air did the same thing on a much grander scale by including 55 sites illustrated by aerial photographs of which 46 had accompanying maps. By working in this way, Crawford was following the traditions of field archaeology that were active in Britain at the time (i.e. the Committee on Ancient Earthworks and Fortified Enclosures (e.g. Allcroft 1908); publications in the Wessex area by WilliamsFreeman (1915) and Heywood Sumner (1913)). Crawford described his work for the volume thus: “The book took the form of fifty plates, each one an air-photograph of an ancient site which had first to be visited and walked over on the ground. I had a matte print made of each, which I annotated in the field; from this I drew an explanatory diagram, of which a fair copy had to be drawn by a draughtsman.” (1955, 171). The work for the compilation of Wessex from the Air took place in the early days of using aerial photographs for archaeological purposes, and so it is unsurprising that Crawford took the new material (aerial photographs) and compared it with what he knew from his former experiences (field investigation). Working with upstanding sites allowed comparison of one source directly with the other and so enhanced his future photo interpretation skills. This is good practice that is recommended to all aerial photo interpreters. We have seen that field observation also has an important role for people using lidar, especially if features have been recorded that are not fully understood (Doneus et al. 2008), It would

seem desirable, therefore, that lidar users and field investigators have had previous experience of earthwork survey or build it up by working with lidar in the field wherever possible (i.e. see Ainsworth et al. this volume). Wessex from the Air is not just the acclaimed book of early aerial photographs but, more importantly, is a book that used those photographs to illustrate analytical field investigation. Explanations in the book come from the field-prepared drawings and the accompanying text. These may help a reader to understand the photographs as, without such diagrams, there is a limit to what a single aerial image and a description can convey (e.g. Wilson 2000). Any form of remotely sensed image will benefit from ground observation and there are some things that can only be resolved in the field. For example, mounds recorded by airborne lidar in a wood in the Leitha mountains, Austria, included both burial mounds and clearance piles from woodland management cutting. After ground inspection had identified them as such it was possible to use filtering algorithms to remove the woodland clearance piles from full waveform laser data (Doneus et. al. 2008). At a finer level of detail, field visits may be necessary to resolve the relative dates of abutting banks that are parts of a defended earthwork. Observation and interpretation – revisited The second site to be documented in Wessex from the Air is Hambledon Hill in Dorset. The entry on Hambledon is written by Eric Gardner who was a medical doctor with a keen archaeological interest and involvement – which probably explains his contact with Crawford. Illustrations in the book include two aerial photographs from the seven that were taken of the hill by Crawford at about 7pm on 14 July 1924. One shows the Iron Age hillfort, the other the Neolithic causewayed enclosure and the evening light is excellent for showing the earthwork ramparts, hut platforms and the ‘modern’ quarrying within the causewayed enclosure (Figure 6.2: A). Those 1924 photographs have rarely been bettered and would have provided very useful field documents for Gardner to take on the hill. The entry on Hambledon Hill makes it clear that Gardner’s examination of the earthworks was not a rushed job and as a piece of analytical fieldwork it is an epitome of clear and precise thinking. In the 1970s I spent a considerable time on the hill undertaking field survey for Roger Mercer’s Hambledon Hill Project (Mercer and

6  Reading aerial images

Healy 2008) and I worked through Gardner’s report several times. My survey began when I suggested that a low earthwork enclosure on the Stepleton Spur that had been surveyed by RCHME (1970, 104) was not Iron Age as they suggested, but Neolithic. This date was confirmed by trial excavation in 1977 after which the survey became an official part of the project. My reasons for suggesting a Neolithic date for the enclosure provide another example of use of accumulated knowledge. In this case, the knowledge that the main causewayed enclosure had two pairs of cross-dykes that echoed its plan was transferred to Stepleton where similar cross dykes could be seen on some aerial photographs. It was that combination of features, rather than the enclosure itself, that suggested similarity of date. The basis of much of my survey information about Neolithic Hambledon was aerial photo interpretation that was usually carried out between excavation seasons and later backed up by the input of information from the ground (often from excavation) to direct its questions. To assist the survey of the hill, three series of 1:5000 scale vertical photographs, with overlap to be viewed stereoscopically, were taken by CUCAP. All other available photographs of the hill were examined and many were purchased

81

for use in the field. By the 1970s, changes in landuse had masked some features that were clear on the 1924 photographs and, in fact, all of my ‘new’ Neolithic features could be identified on the oldest photographs. In most cases, Gardner had noticed these on the photographs and on the ground but often had interpreted them as parts of the Iron Age hillfort. The Iron Age attribution and Gardner’s explanation fitted well with the knowledge of the 1920s – a time before causewayed enclosures were identified as a class by Curwen (1930). In Wessex from the Air, the Hambledon causewayed enclosure was described as ‘a very old circular camp’ (Gardner 1928, 44) but no date was suggested. We are only able to interpret reliably within the parameters of current knowledge. When I am asked what has been my ‘best’ archaeological find, my answer is an earthwork bank a few centimetres high that became known as the Western Outwork (Figure 6.2: C–D). It had been identified by Gardner on the 1924 aerial photographs and he had checked it on the ground. With clarity of thought he explained that it marked the path of a track leading towards the hillfort’s south-west entrance that, he implied, was contemporary with its use (Gardner 1928, 51–2). He supported this date by identifying that it cut through earthworks related to the ‘small

Figure 6.2: The Neolithic causewayed enclosure (A) at Hambledon Hill and the southern part of the Iron Age hillfort (B) on the right. In 1928, this aerial photograph and fieldwork was thought to show the path of a track (C–D) leading towards the hillfort’s south-west entrance (E) that was implied to be contemporary with its use (Gardner 1928, 51–2). An Iron Age date was supported by his identification that the track cut through earthworks (F) related to the causewayed enclosure (but see Palmer and Oswald 2008, Fig. 2.11). Later examination of this aerial photograph suggested the feature was a ditch and bank as was confirmed by field survey and later by excavation that produced a Neolithic date. Source: Crawford and Keiller 1928, plate III

82

Rog Palmer camp’ (the causewayed enclosure, see Figure 6.2:F) and it was obviously ancient rather than something related to more recent land use. Some fifty years after Gardner, when undertaking my survey of the hill, I was not entirely satisfied with his explanation of this feature. Examination of Crawford’s aerial photograph early in 1982 brought about realisation that the feature was lit by the sun in exactly the same manner as the adjacent part of the main causewayed enclosure. Scanning from south-west to north-east, the photograph shows shadow-highlight-shadow. This was highly suggestive of a ditch and bank and also matched the earthworks of the main causewayed enclosure which was another ditch and bank structure. More recent vertical photographs showed the Western Outwork as little more than a line of differential growth on a minimal scarp or terrace that lay just down slope from a modern fence. The increased stereoscopic resolution of these later photographs also showed the feature to be on steep ground – not the ideal location for a track which, given a free choice of route, would be expected to run along the gentler slope a few metres higher to the east. So why was this my ‘best’ archaeological find? By the date that the Western Outwork was identified (and confirmed as Neolithic by trial trenches a year later) I had spent several months on and around Hambledon Hill over a period of several years and had also worked on aerial photographs of it between the summer seasons. Sometimes when I was wandering over the hill looking for features I felt as if I was ‘thinking Neolithic’ because I knew the hill so well and knew where features had been cut some 5500 years earlier. At the time, well before phenomenological or experiential approaches were popular, this was not a fashionable thing to say – academic archaeology was going through its ‘objective’ phase – but it certainly led to me to look in the right place a few times. The features we had identified on the hill comprised the main causewayed enclosure at the domed junction of three spurs, the Stepleton enclosure on the southern spur and various outworks on the eastern Shroton spur (Figure 6.3: 1977). A series of outworks tended to link the Stepleton enclosure to the main causewayed enclosure (although we could never completely close the gap) and it seemed logical to my ‘Neolithic mind’ that something other than a single long barrow lay within or below the Iron Age defences on the northern spur. The Western Outwork suggested

a link from the main causewayed enclosure on to the hillfort spur and strengthened the suggestion that there should be more Neolithic features there. The so-called ‘phase 1’ of the hillfort was a prime candidate for a Neolithic enclosure on that spur although trial excavation suggested otherwise. But to return to Gardner and his entry in Wessex from the Air, it is no surprise that the Western Outwork was not identified as Neolithic at the time, as such an interpretation required developments in wider Neolithic archaeology (i.e. identification of causewayed camps as a class), and further field investigation over an extended period drawing on a broader base of information – both to ask the questions and, perhaps, to accept the answers. This theme of how accumulation of knowledge can affect our perception is well-demonstrated at Hambledon Hill and can be tracked through Figure 6.3. It came about through a combination of aerial photo interpretation and field investigation of features thus identified. This was followed, in most cases, by excavation to confirm the nature and date of those features. Excavation fed off the survey and future survey was informed by the results of the excavations. In 1974 when our excavations on the hill commenced, our knowledge was based on Gardner’s analysis and its enhancement by Desmond Bonney’s survey for RCHME (1970). The date of the causewayed enclosure had been confirmed as Neolithic by his small-scale trenching. The ‘site’ then comprised the main causewayed enclosure, two pairs of cross-dykes and an outwork on the Shroton (east) spur. There were also two long barrows on the hill, one on the ridge of the hillfort spur, the other between the main enclosure and the southern cross dykes (Figure 6.3: 1974). Examination of aerial photographs for my undergraduate dissertation had suggested a small inner enclosure – but no evidence of this was found during later excavations. Comparison of the four parts of Figure 6.3 shows how features were identified, suggested as Neolithic and tested by excavation. Some features appear in all years, others were removed after excavation showed them to be of a different date. The 1999 map includes the results of Al Oswald’s survey (RCHME 1996) and the slight traces he identified that extend the Neolithic earthworks along the hillfort spur. Further detail is published in the chapter about the field survey in the Hambledon project report (Palmer and Oswald 2008).

6  Reading aerial images

83

Seeing beyond the archaeology Aerial photographs and lidar data may record evidence of millions of years of past activity that can be read to indicate the presence of anything from oil deposits to the grave of a recent murder victim. They record complementary and unique information and have their individual strengths and weaknesses. Both sources can offer information about archaeological and non-archaeological features and, when examined individually or in conjunction, are able to help our understanding of past uses of the land. Lidar can give us the naked form of the ground. We can undress it by removing the trees and other objects and it provides us with the macro and micro relief of the area we are examining. In some cases intensity values may provide information on subsurface features with no surface expression, but much of this may come from interpretation of aerial images. To effectively produce a more complete picture of the land we also need to be able to identify a range of non-archaeological features. An important part of what we need to learn to see are components of the past and present landscape and to assess their relevance to any archaeological objects in the area being examined. As an example of this, consider the form of the landscape as it may have been in parts of temperate Europe before cultivation became widespread, perhaps in the later Bronze Age, maybe later in some places. As people working with aerial images and as field archaeologists we are used to the modern landscape, and its form may exert an undue influence on our interpretations. But what was the landscape like for early settlers? In places, environmental records indicate extensive forested land. But what may lie under the trees? Not the flat fields that we see now but uneven land, hummocked in places by periglacial activity and strewn with stones and boulders where glacial action had cut and eroded the land. Periglacial structures are well documented, but illustrations of these for archaeologists are mostly in modern arable land (e.g. Wilson 2000, 164–75) although geographers may show examples in relief (e.g. Harris 1990). In periglacial areas we may expect there to be remains of pingos (Figure 6.4), stripes and patterned ground and ice wedges all of which would have left humps, depressions and fissures on the ground surface. New occupants were moving into lumpy landscapes. They may

Figure 6.3: Four stages in the identification of Neolithic earthworks on Hambledon Hill through aerial photo interpretation, field survey and excavation. Lines schematically show the courses of bank-ditch constructions and, by 1977, two flint scatters of probable Neolithic date had been identified (cross-hatched). 1982 added the western outwork and a series of what excavation suggested may be small flint mines. These were later discounted. The ?fields that first appear in the 1982 view are very slight traces identified on aerial photographs on alignments that do not respect the earthworks of the spur. Their date is unknown. Source: author. See also Palmer and Oswald 2008, Fig 2.5

have cleared stones and levelled some of these features while others were incorporated into ditch-defined or stone-walled boundaries. Traces of past uses of such natural features can be suggested from interpretation of aerial images, others may be identified on lidar surveys in undisturbed woodland. How do we deal with this information? Is it valid as archaeological or should we ignore it because it is natural?

84

Rog Palmer

Figure 6.4: Walton Common, Norfolk where there are surface remains of pingos. The light green bands in the photograph show ridges that were formed around the ice casts when this landscape was in periglacial conditions. Such ridges would have remained to confront early settlers. Source: Google Earth

On the gravel terraces of Cambridgeshire and Lincolnshire there are examples where I believe that people have used natural fissures (frost cracks) as part of their settlement design. Frost cracks in their natural condition are visible as fissures with raised edges and are routinely recorded in cultivated land as crop-marked features. I have seen frost cracks meandering across the landscape and was content to map them as natural until someone in the past attached a ditched enclosure to them. How do I map it now? Which parts of its length may be ‘archaeological’ (i.e. used in the past, perhaps cleaned out to help drainage) and which parts remain natural? The questions are difficult to answer but such mixtures of natural and anthropogenic show the need to understand and map relevant natural features. In these cases they offer more than context and become part of past structures. This is illustrated by an example from Etton, Cambridgeshire, where a meandering frost crack appears to have become a major spine in a chain of enclosures (Figure 6.5), presumably because it was a visible feature when that landscape was first settled. It is highlighted on the aerial photograph and shown in a different colour in the map below. Its crop-marked characteristics are dissimilar to the definite archaeological ditches and it shows the irregularity and ‘soft’ edges that are common to periglacial features in this part of England. Two enclosures seem to be attached to it, it closes the fourth side of another and it also appears to set an alignment for fields and pit alignments in

territory outside the figure. On this basis, it is an important part of the ditched landscape in this area and needs to be mapped as a component of it. Only excavation can show whether this interpretation is correct and whether, perhaps, it was cleaned and recut along with the man-made ditches once the local occupation was flourishing and the network of ditches served to drain as well as to divide and mark ownership. This example also shows the value of accumulated knowledge of an area I have worked in for several years, mostly on developer-funded projects prior to mineral extraction. During this time, I have examined hundreds of aerial photographs, sometimes the same ones for different projects, made a few field visits, talked to local archaeologists (one of whom is also an aerial photographer) and visited excavations. Problems and ideas have been thrown around at each meeting and the way that I examine aerial photographs has been tempered by this growing and changing knowledge. However, none of this need take years to accumulate and a lot can be transferred by sympathetic teaching. For example, in a recent ArchLand supported workshop in Serbia, students were given photographs of this area as part of an interpretation and mapping example. My teaching had emphasised the differences between archaeological and natural features on similar gravel soils as a result of which most of them had some confidence in distinguishing natural frost cracks from archaeological ditches. Doing

6  Reading aerial images

85 Figure 6.5: The aerial photograph shows part of a complex settlement of ditched features in the Etton area of Cambridgeshire. The yellow feature is a periglacial fissure that appears to have been an important component on the settlement. Enclosures were attached to it and it sets an alignment for the local field system. Below this is part of a map derived from many photographs. This shows in greater detail how a natural feature became part of an archaeological, probably Iron Age, land­ scape. Source: author

the same with pits presented much more of a problem to all of us. Another example that mixes archaeological with natural and recent features near Covington, Cambridgeshire, will be used to assess the evidence that can be derived from each combination of information (Figure 6.6). The first panel, A, shows the archaeological information interpreted from a single vertical photograph on the clay

area of west Cambridgeshire against a schematic modern map. The figure shows part of a medieval landscape of ridge and furrow cultivation that formerly covered most of the clayland. This ridge and furrow was recorded in modern arable fields and has been levelled from its original corrugated form which allows earlier ditched features to affect crop growth and so be recorded from above as cropmarking. Excavation has shown that other

86

Rog Palmer

Figure 6.6: Addition of natural and recent information affects our understanding of archaeological features. A: shows only archaeological features against a schematic modern background. B: adds bands of deeper soil that indicate former streams and so provide a natural context for the archaeological sites. C: includes modern features and shows places where damage is likely to have occurred to buried archaeological contexts. Source: author from original photo interpretation and mapping by Ania Sokołowska based on Aerofilms, AF96C-564/Run 11/2180

pre-medieval sites in the county that were later covered by ridge and furrow have been damaged as a result of medieval ploughing. For example, excavation near Soham, Cambridgeshire, showed a Roman floor had been scored by parallel ridges produced during medieval cultivation (Gdaniec pers. comm.). Because the information in the photograph of Covington is all in crop-marked form we do not know the extent of damage to the early ditches, be this from medieval or recent cultivation. In panel B, areas of deeper soil have been added which, from their shape, show palaeochannels (definitely that on the west as a modern stream flows along its centre) or a dry valley (south) into which soil has slipped

following cultivation of the adjacent higher ground. Regardless of their origin, in the absence of contours, these deeper soil areas show the lowest ground and indicate that the ditched features were located on higher ground. This provides a level of context for the archaeological features. The final panel, C, includes four types of recent (i.e. post-medieval) activity which help inform about the archaeological integrity of the map and, importantly, about the condition of the archaeological layers should any protection or conservation be proposed. The recent field boundaries, like those in current use, help show parcels of land in which no archaeological information was recorded on this particular day. They show that there may have been different regimes of cultivation and so help provide reasons for some of the gaps in the archaeological picture. Examination of other photographs may help fill those gaps. A pipeline cuts through some features in the southern part of panel C and is likely to have damaged or destroyed archaeological contexts on that line and possibly on either side of the pipe trench where a swathe of land was probably cleared to enable access. Similar damage is likely to have occurred in the northern group of features where field drains have been laid and cut through some of the archaeological ditches. Hand-dug quarries are common in many rural areas and result from local exploitation of particular soils. The two small quarries in this figure are close to some of the recorded archaeological ditches. If, as may be possible, the site extended west, the quarries may have damaged earlier features. Damage caused by the pipeline and field drains may help decisions on whether or not to conserve these particular sites. These few examples have shown the importance of context through the addition of selected natural features, and the role of this additional information in research and conservation.

Conclusions This contribution is based on many years of working with aerial photographs, and inevitably focuses on visual interpretation. However, the issues of interpreting aerial photographs apply in equal measure to lidar, where contexted interpretation drawing on a broad base of evidence is key to creating reliable archaeological information. And while visual inspection has been

6  Reading aerial images emphasised, the central point is understanding how we see and how this influences what we see, and this is equally important to the use of algorithms and visualisations to produce images for interpretation. There is no magic to undertaking work of this kind. The information is there and it is up to interpreters to expand knowledge so as to ‘see’ and understand the range of features that may have relevance to archaeological investigations. Earlier in this chapter we saw how experience drives our path through everyday activities and how the accumulation of knowledge helps us perceive and understand more. There is no substitute for experience among the skills of an image interpreter – and it is a skill that needs to be learned, practised, tested and honed. Our experience will provide a level of ‘rightness’ to our interpretations as will the questions through which we approach the aerial data. There can be several ‘right answers’ but never a single ‘only answer’ as the results we get from interpretation are a direct response to the knowledge, experience and interest that affects our perception of them. Every aerial image holds a vast amount of information that we do not need to know about and from this we need to identify and isolate that little amount we will understand and that is of relevance to our archaeological study. Learning about, and from, image interpretation never stops and that is one of the most interesting things about working with these data. To be able to start with an image – be it an old vertical photograph or last week’s lidar data – and end with an analytical map provides a way into the past that few other archaeological activities can match. [Dedicated to the memory of Collin Bowen (1920– 2011), analytical field archaeologist, teacher and inspiration, who died while this was being written.]

References Allcroft, A. H., 1908. Earthworks of England: prehistoric, Roman, Saxon, Danish, Norman and medieval. London: Macmillan. Bradford, J.S.P., 1957. Ancient Landscapes: studies in field archaeology. London: Bell. Crawford, O.G.S., 1924. Air Survey and Archaeology. Ordnance Survey Professional Papers 7. London: HMSO. Crawford, O.G.S., 1955. Said and Done: the

87 autobiography of an archaeologist. London: Weidenfeld and Nicholson. Crawford, O.G.S. and Keiller, A., 1928. Wessex from the Air, Oxford: Clarendon Press. Curwen, E.C., 1930. Neolithic camps. Antiquity 4, 22–54. Doneus, M., Briese, C., Fera, M. and Janner, M., 2008. Archaeological prospection of forested areas using full-waveform airborne laser scanning. Journal of Archaeological Science 35, 882–93. Gardner, E., 1928. Hambledon Hill. In Crawford, O.G.S. and Keiller, A., (eds). Wessex from the Air, Oxford: Clarendon Press. Gojda, M., 1997. Letecká Archeologie v Čechách (Aerial Archaeology in Bohemia). Prague: Czech Academy of Sciences. Harris, C., 1990. Periglacial landforms. In Stevens, N. (ed.). Natural Landscapes of Britain from the Air. Cambridge: University Press. Mercer, R. and Healy, F., 2008. Hambledon Hill, Dorset, England. Excavation and survey of a Neolithic monument complex and its surrounding landscape. Swindon: English Heritage. Palmer, R., 2005. “If they used their own photographs they wouldn’t take them like that”. In Brophy, K. and Cowley, D., (eds). From the air: understanding aerial archaeology, 94–116. Stroud: Tempus. Palmer, R., 2011. Knowledge-based aerial image interpretation. In Cowley, D.C., (ed.). Remote Sensing for Archaeological Heritage Management. EAC Occasional Paper No. 5/Occasional Publication of the Aerial Archaeology Research Group No. 3. Archaeolingua, Hungary, 283–91. Palmer, R. and Oswald, A., 2009. The field survey. In Mercer, R. and Healy, F. (ed.). Hambledon Hill, Dorset, England. Excavation and survey of a Neolithic monument complex and its surrounding landscape. 15–39. Swindon: English Heritage. RCHME, 1970. An inventory of historical monuments in the county of Dorset. Volume three: central Dorset, London: HMSO. RCHME, 1996. Hambledon Hill, Childe Okeford, Handford and Iwerne Courtney or Shroton, Dorset. NMR numbers ST81SW 10 and 17. Request Survey: June–September 1996. Cambridge: RCHME. Parcak, S.H., 2009. Satellite Remote Sensing for Archaeology. Abindgon/New York: Routledge. Snowden, R., Thompson, P and Troscianko, T., 2006. Basic Vision: an introduction to visual perception. University Press, Oxford. Sumner, H., 1913. The Ancient Earthworks of Cranborne Chase. London: Chiswick Press. Wessex Archaeology, 2011. http://www.wessexarch. co.uk/stonehenge/explore-stonehenge-landscape-lidarsurvey, Accessed 18 October 2011 Williams-Freeman, J.P., 1915. An Introduction to Field Archaeology as illustrated by Hampshire. London: Macmillan. Wilson, D.R., 2000. Air Photo Interpretation for Archaeologists. Stroud: Tempus.

7 Messy landscapes: lidar and the practices of landscaping Dimitrij Mlekuž The paper is an attempt to accommodate lidar as a relatively new technology within the practice of landscape archaeology. It develops an argument that lidar allows us to see and understand archaeological landscapes in a new way. Although topographic survey has a long tradition in archaeology, the sheer density and quantity of data that became available with lidar is transformational. In contrast to traditional topographic survey lidar does not map only important places but everything, landscape as a whole. This allows us to see landscape as a mess, a continuum of traces of daily practices and activities materialised in a landscape. By focusing on the practices and their material traces more attention should be given to the questions of time and temporality of landscape. Landscape is not just a palimpsest of traces of activities but a palimpsest of different temporalities. And messy landscapes require a more reflexive approach to the practice of landscape archaeology. Keywords: lidar, landscape archaeology, landscape, practice, temporality

Introduction This paper is a result of my ongoing frustration with lidar. It is about struggles to meaningfully accommodate lidar in my engagement with past landscapes. It is a messy account of resistances and accommodations between my understanding of landscapes and technology. It is also a ‘bricolage’ rather than systematic ‘reflection’, where I am drawing upon disparate theories, practices and technologies and trying to patch them together and accommodate them within the practice of landscape archaeology. Lidar is a technology that allows very precise three-dimensional mapping of the surface of the earth, even where the surface is obscured by forest and vegetation. The level of detail on digital surface and terrain models produced from high-resolution lidar topographic data helps us enormously in identification of past events, which reworked and modified the surface of the earth. But is this all? In this paper I will try to develop an argument that lidar allows us to see and understand the landscape in a new way.

Lidar is basically a topographic survey tool, and topographic survey is one of the oldest field techniques in the landscape archaeologist’s toolbox, but the sheer density and quantity of data that become available with lidar can be transformative: we can literally see landscapes in a new light. And how do they look in a new light? Are they only precise three-dimensional images of archaeological sites in an environmental context? Or is it only a mess of different traces? I am sure of only one thing: if we want to approach the mess of reality at all then we have to teach ourselves to think, to practice, to relate, and to know in new ways.

Landscape It is almost impossible to define what landscape is (see Johnson 2007; Wylie 2007), not only because landscapes are very complex entities, but mainly because they are created through our engagement with them.

7  Messy landscapes: lidar and the practices of landscaping Among archaeologists, one of the most common, and often implicit, definitions is landscape as naturally or humanly created features that exist ‘objectively’ across space. Landscape thus consists of features within their natural context. Landscape archaeology in this sense is simple to define: it is about what lies beyond the site, or the edge of the excavation. The practice of landscape archaeology is to recognize, map and interpret these features (Johnson 2007, 3). Traditionally landscapes have been defined as interactions between ‘natural’ conditions (i.e. weather, terrain, soil type, etc.) and sets of cultural practices (i.e. agriculture, religious or spiritual beliefs, social organization etc.). Landscape is interaction between nature and culture, and processes where nature and culture mutually shape each other (see Sauer 1963). But this is not the only definition. ‘Landscape’ is also a way of seeing, a way of thinking about the physical world (Cosgrove 1984; Daniels and Cosgrove 1988). Then again some, myself included, would add a third concept: landscape as engagement with the world – a process or way of doing things rather than a thing or an idea (Ingold 2000; Thrift 2007). And there might be other definitions. Landscape is therefore a very vague term, used in many different ways. The meaning ranges from landscape as an idea and way of seeing and depicting the world to a purely material space: from something to be looked at from afar to being part of it; from representation to experience. These tensions are productive as they animate the discussions which create new understandings of landscape. But perhaps instead of separating the landscape into categories, tensions and oppositions, nature and culture, it might be more productive to think first about landscaping (Wylie 2007). That is, we should think about practices, habits, actions and events, ongoing processes of relating and unrelating, that come before any separation into categories, definitions and tensions. Instead of landscape being the outcome of interactions of nature and culture, practices of landscaping such as everyday things like walking, looking, building and also the practices of scientific engagement with the landscape, are the cause and origin of our ideas of what is landscape, what is ‘nature’ and what is ‘culture’ (Wylie 2007, 11). Therefore, landscape is a vague and slippery thing, always evading definition. It is much messier than just sets of features and their natural

contexts. We are always caught in these tensions when we are dealing with landscapes, whether we acknowledge it or not, and there is no simple, ‘objective’ way of resolving them. It seems that practice of landscape archaeology is much more complex than objective recognition, mapping and interpretation of ‘features’.

Archaeological practices of landscaping Thus different practices of landscaping tend to produce not only different perspectives, but also different realities. Consider, for example, the phenomenological work on Neolithic landscapes by Chris Tilley (1994) and compare it with the lidar survey of the Stonehenge area (Bewley et al. 2005). Different methods, different practices, produce entirely different landscapes. So, how do we produce knowledge about past landscapes? Development in science, technology and society studies (STS) has convincingly demonstrated that knowledge is produced through a more or less messy set of practical contingencies (Latour and Woolgar 1979; Latour 1987, 1996, 1999; Law 2004). In the same way the work of landscape archaeologists employs a diverse set of practices, tools, technologies, theories, skills, institutions and traditions through which we produce knowledge about landscapes. The work of the landscape archaeologists involves the creation of ‘inscription devices’, or ‘immutable mobiles’. These accounts of external, real, material world (texts, maps, sketches, illustrations, graphs, photographs, point clouds …) keep some types of relations of reality intact – therefore immutable, but can circulate around – thus mobiles, allowing new associations, translations and articulations to take place (Latour 1999). The most common inscriptions produced by landscape archaeologists are maps. Every map is a work, whether realized by hand or by the most advanced technological means, constructed, crafted, using what we have at hand: people, tools, technologies, taken-for-granted practices, theories and skills (Pickles 2004, 88; Wickstead 2009). The practice of landscape archaeology is therefore transformation of material, “real” landscape into inscriptions and translations of these inscriptions, which are processes of transformation, conversion, simplification and combination of inscriptions (Latour 1984, 1999).

89

90

Dimitrij Mlekuž Maps are produced, constructed and crafted, but that does not make them less real. As Bruno Latour observes (2001, 19) the “more instruments, the more mediation, the better the grasp of reality is.......the more human-made images are generated, the more objectivity will be collected.” Mapping or map making is also engagement with landscape, a practice of landscaping, a practice by which we fix and define certain relations in landscape when leaving others fluid. Mapping the land is equally going over, not only in the sense of flying over (as we do with lidar and aerial photographs) but in the sense of becoming acquainted with something in order to become more familiar with it. But maps are not fixed at the moment of production. They are more than just static representations of landscape, as they replace original situations (Latour 1997, 67). Map making and mapping should be seen and practiced as open-ended processes ‘detachable, reversible, susceptible to constant modification‘ (Turnbull 2000; see also Wickstead 2009). As such, maps and mappings are both representations and practices simultaneously. They are open to interpretation, contested and mutable. Mapping is therefore an active, engaged, open-ended process of engagement of landscapes. Mapping is landscaping.

Lidar as topography Topography is the creation of inscriptions about particular places. Topographic survey has a very long tradition in landscape archaeology and requires a pair of good boots, advice from local people and a keen eye for “humps and bumps”. In this way, topography is also about the shape of those places, that can hint about the nature of an archaeological site, and the potential existence of structures buried beneath the soil. It is about recording “humps and bumps” that make up the place and fill it with meaning. The English topographic school, for example, developed topography to the level of art, producing beautiful and rich inscriptions of places using hachure depictions (Johnson 2007). With the development of technology, for example differential GPS and laser theodolite, places can be recorded in a more metrically accurate way, using three-dimensional represent­ ations. This allows us to produce even more inscriptions, using different visualization techniques. Lidar can be understood as an

extension of these technologies. But what makes lidar different from other topographic techniques is its lack of selectiveness. Lidar is not limited to selected places, does not record only important or understood “humps and bumps”, but the whole surface of the earth, all the mess of humps and bumps. Lidar records landscape in an indiscriminate way, every place, every feature, and every square metre is treated with the same attention and resolution (vegetation allowing).

Practices, scrapes, scars and works We are all perplexed by the ubiquity and richness of features captured by lidar. Lime kilns, charcoal burning platforms, fields, hollow-ways, tracks, lynchets, quarry pits, and so on (Figure 7.1) – landscape is never empty, but rather full of these traces. These features overlap, crisscross, are destroyed, reworked or incorporated into other features – they are traces of daily routines, non-discursive practices that scrape the surface of the Earth. They turn Earth inside out and put its material contents on display on the surface. They give the materiality to the practices that occur on it. And this is what we see on the highresolution topographic data, what lidar captures so well and vividly – this is the landscape itself. Practices, activities and tasks, materialized in a landscape. The everyday, the mundane, the routine, humdrum of everyday living consists of a continuous flow of daily tasks. People’s time is consumed by these no matter how insignificant or trivial they might seem. Working in fields, attending other people and animals, transporting stuff and substances, and moving around are routines, manifest as habits, which allow us to cope with the life and go on in the world. We are born into this flow and begin to participate in it from the beginning of our lives, and become fully aware of it only when something goes wrong. Following Bourdieu (1977) we can grasp these taken-for-granted background practices as embodied dispositions. As living in the world is a material practice, these practices leave traces, get worked into, get scraped onto the land, the surface. Of course, not all activities are pre- or nondiscursive, daily and common. Besides scrapes and scars, the traces of daily activities, we also have works, intentional constructions with the idea of durability, control, aesthetic beauty or

7  Messy landscapes: lidar and the practices of landscaping

0

25

0

50

25

0

0

100 m

50

100 m

25

25

50

0

50 m

100

150 m

0

0

25

25

25

0

50

50

91

100 m

100

50

100 m

25

50

0

0

150 m

100 m

25

25

0

25

0

25

50

100 m

0

25

25

50

100

150 m

0

50

100

150 m

0

50

100 m

25

0

50

100 m

50

50

25

100 m

100

150

200 m

50 m

monumentality or symbolic power. These are traces we easily recognize, as castles, ramparts, walls, hillforts, monuments, barrows, parks and gardens, built to change the way people move, access, see and understand the landscape (Figure 7.2).

Figure 7.1 (above)��:� Scrapes, scars and other traces in the landscape: paths, lynchets, fields, extraction pits, charcoal burning platforms, boundaries, furrows.....

Practices, places and landscapes

Figure 7.2������������� :������������ Monumental ramparts of the Poštela hillfort with clusters of Iron Age barrows on the ridges below it

The exploitation of the physical environment for the practicalities of life – materials, energy, shelter and sexual reproduction – is a means through which society is established and propagated. Such practices may be understood as ‘embodied acts of landscaping’ (Wylie 2007, 169) in which self, landscape and culture emerge, reproduce and circulate (Thrift 2007; Cadman 2009; Macpherson 2010). Landscape is therefore full of scrapes, the traces of daily activities, and thus the continuous flow of daily life. These scrapes are not something superficial, added to the landscape, but its main constituent. Practices occur at places that are shaped by scrapes of previous activities; thus

100 m

practices create places and are shaped by places where they are performed. By producing and reproducing places, landscape is produced and reproduced. Places have no boundaries, they are osmotically incorporated in other places and wider landscape (Casey 1996). Lidar records landscape in an indiscriminate way, and does not select important places as ‘sites’. Places are thus embedded in the landscape; their shape, dimension, context and structure are the result of complex and lasting

92

Dimitrij Mlekuž

Figure 7.3������������ :����������� Limekilns in dolines on Slovenian Karst and the network of tracks that connects them

100 m

50 m

Figure 7.4��:� ����������� Network of hollow ways, tracks and paths in the vicinity of Novo Mesto, Slovenia

interactions with the landscape in flux. For example, a karstic doline offers easy access to the limestone, and its sides can become open face quarries (Figure 7.3). At the same time it is a windbreak, and placing a kiln at the base of doline shelters it to help keep temperature constant. But then again it is a place where through bodily interaction with the materiality of landscape (e.g. stone and wind) new scrapes and traces are produced… The limekiln is connected with other places, where other activities occur, and from where bodies and stuff comes here. It is a node which connects charcoal burning platforms, villages and

quarries, but also biographies of people, animals, things and substances that travel through it. As nodes in a network of flows and paths, places are connected and act together (Casey 1996). Things bodies, stuff, and substances, are brought together where they can interact and form or fix new associations and things. Limekilns are not only places where lime is produced from limestone and charcoal, but locales where people, things and plants interact with each other, where environmental knowledge and skills are learned and controlled, and social roles and identities are defined, maintained and contested. For example, the task of lime making enables expressions of identity by specific age or gender groups. Things, people, stuff and substances move around landscape and stop at places. It can be said that movement is an essential and ubiquitous act of landscaping. Practices of movement leave scrapes related to movement such as tracks, hollow ways and paths (Figure 7.4; Hindle 1993; Taylor 1979; Muir 2010, 67–93). Landscape can then be viewed as a product of practices, trajectories, interrelations and flows. And this sense of landscape as a continuous weaving, relating and associating, forever in the making, is in my opinion much more productive than static notions of landscape in terms of territory, boundedness, area, scale and so on (Cadman 2009; Allen 2011). We interact with these places too, and in a

7  Messy landscapes: lidar and the practices of landscaping way enact or trace these connections through our own practices of mapmaking. We are all familiar with constant zooming in on the places, scrapes and features, and interaction with them through different visualizations, drawing information about them from different inscriptions and then zooming out again, connecting with other places and the wider landscape. This is what Rachel Opitz and Laure Nuninger (2010) call ‘contextual topography’ and is way of creating knowledge of landscape through practices of mapmaking, transformations and translations of maps. It is a practice of landscaping, through which we create places and weave connections between them. In this way our own landscaping practices become interwoven with past practices that created scrapes in the landscape. Through our encounter with places, by tracing connections we reiterate landscaping tasks that created and reproduced those places (see Lucas 2001, 202). Perhaps we should be more aware of how knowledge of landscape is produced through our own practices of landscaping, through our own practices of contextual topography (e.g. Halliday this volume). Maybe our reports about landscapes should be accompanied with reflexive accounts – short ‘auto-ethnographies’ about our practices of mapmaking.

Who scrapes? There are agencies other than human that scrape the land as well. Natural processes accumulate and erode layers, subside, uplift, grind, slide apart and collide them..... creating mountain chains, river valleys and plains. These processes are slow, moving in the rhythms of tectonic and geological time. But water can scrape deep scars into the land relatively quickly (Mlekuž 2009). These agencies produce places as well. For example, caves are products of millennia of dissolution of limestone. And they are not just natural places. They can be places where human and non-human activities (think bears) occur. In a sense there is no difference between a ‘cultural’ feature such as house and a ‘natural’ feature such as cave. People may live in a house built by ancestors, or in a cave, where ancestors have dwelt. The fact that a cave is natural may not make it any different from a house for the people who dwell in it. As houses, caves can be seen as the works of ancestors, and their successors may have felt a responsibility to look after those places, to harness their power

or even to transform them into something else (Bradley 1998). Wind can create places too. A tree-throw in a forest leaves a scar in landscape, but also a place with its own specific sociality that can attract deer, become a focal point for Mesolithic hunter-gatherers or a starting point for Neolithic forest clearance (Davies et al. 2005). Landscape is also a habitat for humans and non-human animals. Animals scrape their traces into the land during their daily routines as well. They make their own places: badger sets, boar rootlings, deer digs, and animal trails are scrapes of non-human animal practices. These places, created through their own practices of landscaping have their own animal sociality. For example, think about an anthill, a real ‘work’ of ants and a place of intensive insect sociality. But it can also be reference for other, human and nonhuman, practices with their own socialities. And there are distributed agencies. People and things are mutually constituted. We act with and through material culture and material culture acts through us (Knappett 2005; Latour 1992, 1994; Miller 1987). Tools, equipment and machines change the ways we scrape the earth. For example, Tim Ingold (2004) writes a historical account of how, with the development of the technology of the shoe, people have been afforded differing movements through the landscape. Development of tools and machines such as ploughs and ox carts all change the ways we scrape into the land as we move around on our daily courses and routines. Thus, for example, the development of the plough and ploughing techniques changed the way people turn the land. We can observe these changes in the shape of scrapes, and ridge and furrows fields (Figure 7.5; Beresford 1984; Butlin and Baker 1973; Hall 1998) reflect the practices and bodies involved in the process. Their form is shaped by the way an association of plough, a team of oxen and humans negotiated the field. Thus at the end of the furrow, the leading oxen meets the end first, and is turned left along the headland, while the plough continued as long as possible in the furrow. By the time the plough eventually reached the end, the oxen are lined up facing leftwards along the headland, producing the curved soft S shaped form (Figure 7.5). Even the length of field, which used to be a unit for length (a furlong) usually around 220 m long, is shaped by the materiality of the oxen bodies – it is the length that pair of oxen can plough in one stretch.

93

94

Dimitrij Mlekuž 50 m

Figure 7.5��:� ��������� Remnants of ridge and furrow fields in the Dravinja Valley, Slovenia, cut by a segment of the Vienna-Trieste railway constructed in 1846

And finally there is the agency of landscape and places. As people move around the landscape and perform their practices, they relate to the traces, scrapes and scars. Places act back. The human body takes shape through its interactions with other objects, bodies and places. Thus humans come into being within their local environment and technologies rather than simply being or persisting (Ingold 2000; Latour 2005; Macpherson 2010): “…body and landscape are complementary terms: each implies the other, alternately as figure and ground” (Ingold 2000, 193). Thus land as we see it on high-resolution lidar topographic data offers the potential not to study only the landscape, but practices, bodies, places and their mutual becoming. Even more, to fully capture landscape, we have to take account of technologies, bodies (human and non-human), tools and the material world that are integral parts of the landscape. Lidar data, with its continuum of scrapes, traces and scars, thus offers understandings of life in the dense landscapes that are not only ours, but are shared with other agencies: animal, tool, machine and natural.

Mess of temporalities Landscape is full of scrapes, scars and traces of past practices. They are material traces of dynamic and mobile practices, through which landscape is continuously produced. Thus time is inscribed in its very constitution at multiple levels and scales. Landscape has a temporal dimension.

Life in a landscape consists of a continuous flow of daily tasks. People’s time is consumed by these no matter how insignificant or trivial they might seem. Walking, cooking, caring for children and animals, attending plants, hunting, building and talking are all part of this flow, and are activities which perpetuate life and create histories. We are born into this flow and participate in it from the beginning of our lives. But those tasks are not isolated, discrete events: every task has its own “thickness and temporal spread” (Gell 1992, 223; see also Lucas 2005). They make sense only when related to those that were already performed and those to be. Life is not just a succession of isolated seasonal tasks, but a flow of tasks implicitly or explicitly related to one another. Each task is made possible by past tasks and future tasks give it purpose. This network (frame of reference (Gosden 1994) or taskscape (Ingold 2000)) unfolds over space and time. Therefore time arises from the flow of life – it is created through rhythms of bodily involvement with the material world. As people move around their business they constantly scrape traces into the land and refer to existing traces. In this sense the landscape has its own time depth. Even if those practices collapse into the land and become fixed and materialized as scrapes there is still an inherent temporality of the landscape. As Barbara Bender (2002, 103) observes “Landscape is time materialized. Or, better, landscape is time materializing: landscapes, like time, never stand still. Landscapes are in a constant process of becoming.” The most often used metaphor we use to describe the build up of landscape, or its inherent temporality, is palimpsest. Palimpsest is a parchment on which earlier writing has been erased to make way for new text. In terms of landscapes, it refers to the traces of multiple, overlapping activities over variable periods of time and the variable erasing of earlier traces. This passage from O.G.S. Crawford (1951, 51–2) beautifully captures idea of historical layering in the landscape: “The surface of England is like a palimpsest, a document that has been written on and erased over and over again; and it is the business of the field archaeologist to decipher it. The features concerned are of course the roads and field boundaries, the woods, the farms and other habitations, and all the other products of human labour; these are the letters and words inscribed on the land. But it is

7  Messy landscapes: lidar and the practices of landscaping not easy to read them because, whereas the vellum document was seldom wiped clean more than once or twice, the land has been subjected to continual change throughout the ages.” But landscape is not just a palimpsest, not just a historical layering, in which the present is merely the sum of past episodes. It is also an active, present and futureorientated engagement with the environment through the process of landscaping (Lee 2007). Landscape with its scrapes and works is part of a life world of people who inhabit it. The actions of people refer to these features, these features structure the way people act. People inhabit the world shaped by their predecessors. We are thrown into the world created by our ancestors. People are born into an already settled landscape pockmarked with scrapes and works and by performing their tasks they are alert to the conditions and changes in the environment and adjust their actions accordingly. Thus past scrapes and works can be reworked, incorporated into new ones, respected or just acknowledged. Traces of different periods can exist simultaneously, enduring in a land for different lengths of time because there are variations in change or turnover (Lucas 2005). The landscape is therefore multi-temporal, made up of a series of past durations. The way different temporalities combine into a multi-temporal present is often messy and complex. Consider this passage by Olivier (2001, 62): “The house … was built towards the beginning of this century, in the courtyard of an ancient farm whose structure is still visible... I see an interweaving of houses and constructions, most of them dating back to the 19th century, sometimes including parts of earlier constructions from the 18th or 17th century. The 20th century here looks so localized, so secondary: it is reduced to details, such as windows, doors or, within houses and flats, furniture … Right now, the present here is made up of a series of past durations that makes the present multitemporal.” Past is therefore incorporated and reworked into the present. For example, although the life world around a prehistoric hillfort in coastal Slovenia has changed (Figure 7.6), the massive walls of the hillfort were left untouched. Even more, land division followed the contours of the hillfort incorporating it into a new landscape. It is still there, it still acts back on people, but in a very different way that it did in the Iron Age, and is integrated into the new practices of people in a nearby village. But there might be no reference between new

95

Figure 7.6��:� ��������� Hillfort above Dekani, Littoral Slovenia, incorporated in a later terraced landscape

50 m

and old scrapes, as older scrapes get reworked, modified or are just ignored by new practices. In this way they may become less and less visible due to the scrapes of new activities, which pay no respect to them. They are dominated, reworked, or erased by new scrapes. But older scrapes still come through as visible, and still have potential to reference activities around them. See, for example, the possibly prehistoric clearance activities in a Karst environment with clearance cairns and remains of dry stone walls (Figure 7.7).

Figure 7.��� 7:� ����������� Network of dry stone walls covering the traces of (prehistoric?) clearance and field division in the Slovenian Karst

96

Dimitrij Mlekuž

0

0

0

25

25

50

50

25

100 m

0

100 Meters

50

25

0

100 Meters

Figure 7.8������������ :����������� Different surface textures, from those of the forest floor and floodplains, karstic landscape to the textures of fields, meadows, pastures, ...

0

25

25

50

50 Meters

50

100 Meters

0

100

150 m

25

50

100

150 m

0

25

50

100 m

0

25

50

100

New field divisions, visible as a network of dry stone walls, paid no reference to the older scrapes, which were probably already overgrown and not very noticeable, or not considered relevant, at the time the new layout was established. But they are still here. This is a palimpsest in a narrow sense of the word. But sometimes, new scrapes can completely erase, destroy or rework older traces. There is absolutely no relation to past landscape. The agency of water, for example, can be very powerful, and in speed and scale can dwarf human agency and completely rework whole landscapes. It is important to note that although these scrapes destroyed older ones there is no less land in a sense that something is missing, this is just a new episode in the landscape history, part of the process of landscaping. Ploughing, grazing, clearance – all leave distinctive textures which contrast with the textures of the forest floor, floodplain or mountains. Ploughing as a means of fragmenting and aerating the soil results in a distinctive texture, which has been deliberately created for the properties of the texture itself. It is at

0

25

0

150 m

0

25

25

50 Meters

50

100 m

50 Meters

its surface that we are in constant, immediate and close physical contact with the landscape. People also relate to places through tactile and visual experience of textures beneath their feet. Thus, place acts back through textures (Jones 2003, 41–72). The textures have a temporality of slow but constant change of the very texture of surface. Mundane practices which might have a minimum impact on the surface can in the long term combine to form distinctive textures (Figure 7.8). These temporalities combine in different ways. For example, we can clearly observe how texture changes on a sequence of river terraces. Floodplain, itself a sequence of scrapes created by water, is slowly smoothed out and reworked into the texture of fields (Figure 7.9). Textures have great potential in deciphering the biography of surfaces and practices that occurred on those surfaces but also ways people interacted with the land in close physical contact. All these processes combine the products of different temporalities in different ways; destroy, blur or sharpen their apparent boundaries. These effects are important, for they determine where we see traces of past

7  Messy landscapes: lidar and the practices of landscaping

97

200 m

human practices and what these look like. Some traces of human activities get worked into, or get buried in, the������������������������������������� soil. ������������������������������������ They became part of the Earth itself and disappear as scrapes on the surface. But they can be retrieved by other techniques, for example geophysics. All these processes combine and create messy landscapes. Messy not only because they are full of different scrapes, traces and scars, but because these scars, traces and scrapes combine in complex, often messy ways.

Conclusion What happens when we try to describe things that are complex, diffuse and messy in simple terms? Well, we make a mess of it. This is because simple clear descriptions do not work if what they are describing is not itself very coherent. The very attempt to be clear simply increases the mess (see Law 2004, 1–17). Landscapes are a messy affair.

Through the very practices of landscape archaeology we are also involved in the production of landscape. Our practices of mapmaking become intertwined with the practice that left the scrapes into the land. Thus the practice of landscape archaeology is a necessarily messy job. As we produce knowledge we are all located somewhere: in our practices, our bodies, technologies, and theories that are at hand. We are caught up, as Donna Haraway (1990) puts it, “in a dense material–semiotic network”, sets of relations that simultaneously have to do with meanings and with material. All knowledge is localized within those networks. There is no detached, natural position. Mapmaking is essentially a ‘motley’ grouping of practices (Turnbull 2000, 39), instrumentations, theories and people, more or less successfully brought together. Perhaps we should think more how we employ tools, knowledge and skills when we map the landscape, and offer accounts on

Figure 7.9��������������� :�������������� A palimpsest of temporalities: a��� sequence of Holocene terraces of the River Sava with different surface textures

98

Dimitrij Mlekuž our practices of landscaping. After all, it is how we painstakingly construct past landscapes and imperfectly stitch them together with whatever we have at hand (Halliday this volume). On the other hand, we see past landscapes as mostly empty, dotted with sites such as settlements and cemeteries. But past landscapes were not like this. They were busy with people moving around on their daily routine, herding sheep, working in the field, ploughing, extracting stone and burning charcoal. These activities left their traces and scrapes on a landscape. Archaeologists are also trained to see time as chronology, as a succession of events, phases and periods. But time is much more complex. Time is created by the tasks we perform, and the effect of those tasks can endure after a task was finished. Landscape is charged with time. Time is materialized in the landscape. The present-day landscape is not just a series of fragments of different periods, each surviving to varying degrees, according to their age. The landscape is not just a palimpsest of scrapes and features, but also a palimpsest of multiple temporalities. Landscapes and landscaping are a messy affair. There are no discrete features but the continuum of them, and there is no chronological succession but a mess of temporalities. And by classifying landscape into simple well-ordered features we would just make a more mess out of it. So the only way is to describe things as they are; produce accounts on how our practices of landscaping mingle with those of the past, no matter how messy these accounts might be.

References Allen, C.D., 2001. On Actor-Network Theory and landscape. Area 43.3, 274–80. Bender, B., 2002. Time and landscape. Current Anthropology 43, 103–12. Beresford, M., 1948. Ridge and Furrow and the openfields. Economic History Review 1, 35–45. Bewley, R.H., Crutchley, S. and Shell, C.A., 2005. New light on an ancient landscape: lidar survey in the Stonehenge World Heritage Site. Antiquity 79, 636–47. Bordieu, P., 1977. Outline of a Theory of Practice. Cambridge University Press: Cambridge. Bradley, R., 1998. Ruined buildings, ruined stones: enclosures, tombs and natural places in the Neolithic of south-west England. World Arch­ aeology 30(1), 13–22. Butlin, R.A and Baker, A.H., 1973. Field systems in the British Isles. Cambridge University Press: Cambridge.

Cadman, L., 2009. Nonrepresentational Theory/ Nonrepresentational Geographies. In Kitchin, R. and Thrift, N., (eds). International Encyclopedia of Human Geography Vol. 7, 456–63. Elsevier: London. Casey E.S., 1996. How to get from space to place in a fairly short stretch of time. In Field, S. and Basso, K.H. (eds). Sense of Place, 13–52. School of American Research Press: Santa Fe (NM). Cosgrove, D., 1984. Social Formation and Symbolic Landscape. Croom Helm, London. Crawford, O. G. S., 1953. Archaeology in the Field. Phoenix House, London. Daniels, S. and Cosgrove, D., 1988. Introduction: iconography and landscape. In Cosgrove, D. and Daniels, S. (eds). The Iconography of Landscape, 1–10. Cambridge University Press: Cambridge. Davies, P., Robb, J.G and Ladbrook, D., 2005. Woodland clearance in the Mesolithic: the social aspects. Antiquity 79(304), 280–8. Evans, J.G., 2003. Environmental archaeology and the social order. Routledge: London. Gell, A., 1992. The anthropology of time: cultural constructions of temporal maps and images. Berg: Oxford. Gosden, C., 1994. Social Being and Time. Blackwell: Oxford. Hall, D., 1982. Medieval fields in their many forms. British Archaeology 33, 6–7. Haraway, D.J., 1990. Simians, Cyborgs, and Women: The Reinvention of Nature. Routledge: New York. Hindle, B.P,. 1993. Roads, tracks and their intep­ retation. Batsford: London. Ingold, T., 2000. The Perception of the Environ­ ment: Essays in Livelihood, Dwelling and Skill. Routledge, London. Ingold, T., 2004. Culture on the Ground: The World Perceived Through the Feet. Journal of Material Culture 9(3), 315–40. Johnson, M., 2007. Ideas of landscape. Blackwell: Oxford. Knappett, C., 2005. Thinking through material culture: An interdisciplinary perspective. University of Pennsylvania Press: Philadelphia (Pa). Latour, B., 1986. Visualization and cognition: Drawing things together. Knowledge and Society: Studies in the Sociology of Culture and Present 6, 1–40. Latour, B., 1987. Science in Action: How to follow Scientists and Engineers through Society. Open University Press: Milton Keynes. Latour, B., 1992. Where Are the Missing Masses? The Sociology of a Few Mundane Artifacts. In Bijker, W.E. and Law, J., (eds). Shaping Technology/ Building Society. MIT Press: Cambridge (MA). Latour, B., 1994. Pragmatogonies: A Mythical Account of How Humans and Nonhumans Swap Properties. American Behavioral Scientist 37(6), 791–808. Latour, B., 1996. Aramis, or the love of technology. Harvard University Press: Cambridge (MA). Latour, B., 1999. Pandora’s Hope: Essays on the Reality

7  Messy landscapes: lidar and the practices of landscaping of Science Studies. Harvard University Press: Cambridge, (MA). Latour, B., 2004. How to talk about the body? The normative dimension of science studies. Body & Society 10(2–3), 205–29. Latour, B., 2001. What Is Iconoclash? Or Is There a World beyond the Image Wars? In Latour, B. and Weibel, P., (eds). Iconoclash: Beyond the Image Wars in Science, Religion and Art, 14–38. MIT Press: Cambridge. Latour, B. and Woolgar, S., 1979. Laboratory Life: The Social Construction of Scientific Facts. Sage: Berverly Hills. Law, J., 2004. After Method: Mess in social science research? Routledge: London/New York. Lee, J., 2007. Experiencing landscape: Orkney hill land and farming. Journal of Rural Studies 23(1), 88–100. Lucas, G. 2001. Critical approaches to fieldwork. Routledge: London. Lucas G. 2005. The archaeology of time. Routledge: London. Macperson H., 2010. Non-Representational Approaches to Body–Landscape Relations. Geography Compass 4/1, 1–13. Miller, D., 1987. Material culture and mass con­ sumption. Blackwell, Cambridge. Mlekuž, D., 2009. Floodplains in a new light: LiDAR and the taphonomy of aluvial landscapes. Arheo 26, 7–22. Muir, R., 2010. The New Reading the landscape. University of Exter Press: Exter.

Olivier, L., 2001. Duration, memory and the nature of the archaeological record. In Karlsson, H., (ed.). It’s About Time. The Concept of Time in Archaeology, 61–70. Bricoleur Press: Göteborg. Opitz, L. and Nuninger, L., 2010. Thinking through Topography. Presentation at the Theoretical Roman Archaeology Conference. Oxford, 25–8. March 2010. Pickles, J., 2004. A History of Spaces: Cartographic Reason, Mapping and the Geo-Coded World. Routledge: London. Sauer, C.O., 1963. The morphology of landscape. In Leighly J., (ed.). Land and Life: A Selection from the writings of Carl Ortwin Sauer, 315–50. University of California Press, Berkeley. Taylor, C., 1979. Roads and Tracks of Britain. Dent: London. Thrift, N., 2007. Non-Representational Theory: Space, Politics, Affect. Routledge: London/New York. Tilley, C., 1994. A Phenomenology of Landscape: Places, Paths and Monuments. Berg: Oxford. Turnbull, D., 2000. Mason, Tricksters and Cart­ ographers: Comparative Studies in the Sociology of Scientific and Indigenous Knowledge. Harwood, Amsterdam. Wickstead, H., 2009. The Uber Archaeologist: Art, GIS and the male gaze revisited. Journal of Social Archaeology 9(2), 249–71. Wylie, J., 2007. Landscape. Routledge: London and New York.

99

8 Visualizations of lidar derived relief models Žiga Kokalj, Klemen Zakšek and Krištof Oštir

Archaeological interpretation of lidar derived relief models is strongly dependent on the specific characteristics of different data visualization techniques, especially when not combined with extensive field surveying. Archaeologists dealing with such interpretations are mainly confined to analytical hillshading, the most frequently used technique. It is a well known technique, implemented in most geographic information systems and readily available as a product by most lidar data providers; however, it is inadequate for a thorough investigation of microrelief structures. This chapter addresses different visualization techniques, their specifics, advantages and weaknesses in the context of archaeological interpretation of various types of historical landscape features from high resolution digital elevation models (DEMs). The techniques addressed in detail are analytical hillshading, derivatives of hillshading from different directions (range of hillshadings, mean of hillshadings, PCA of hillshadings), elevation differentiation, trend removal, slope severity, sky-view factor, solar insolation modelling and some others (composite images of a normalised digital surface model (nDSM) and shaded relief, and a greyscale orthophoto image and shaded relief ). Keywords: relief visualization, relief mapping, digital elevation model, historic landscape, airborne laser scanning, lidar

Introduction The potential archaeological information that may be gathered from lidar is highly dependent on the methods applied to view the data. For example, even with a very detailed elevation model of a fairly simple environment (e.g. Hill of Tara lidar survey, Ireland (Corns and Shaw 2009)) produced at considerable expense one has the potential to miss some features simply because of their orientation (Challis et al. 2008). Using just a hillshaded image (‘standard lidar data’ in the average archaeologist’s mind) features can be hidden or visually suppressed due to their aspect rather than their size. To overcome this, many advanced visualization techniques for presenting high resolution

elevation data have been developed specifically for archaeological purposes or have been adapted from other scientific fields. To successfully apply these methods their basic concepts, characteristics, advantages and disadvantages must be understood. This chapter first establishes the need for appropriate use of scientific visualization in archaeological exploration of lidar data. It continues with some reasons why knowledge of processing techniques and visualization methods used for archaeological interpretation is important, sometimes critical, if we want to produce identification maps and interpretations with quantifiable reliability. This is followed by descriptions of a range of (two dimensional)

101

8  Visualizations of lidar derived relief models

visualization techniques that can be used to aid archaeological exploration of lidar data. A discussion on the techniques is included in their descriptions.

Scientific visualization Scientific visualization is an interdisciplinary scientific field primarily concerned with realistic renderings of volumes, surfaces and illumination sources, sometimes with a dynamic (time) component (Friendly 2008). It is the study of the visual representation of information abstracted in some schematic form, with a main goal to communicate information clearly and effectively through graphical means (Friedman 2008). To convey ideas supported by data effectively, both aesthetic form and functionality need to go hand in hand. Such a balance provides an insight into a complex dataset by presenting it in the most intuitive way. Yet scientists often fail to achieve this balance, which usually leads to poor presentation of their results. The balance between aesthetic form and functionality can be achieved using some simple rules. The studied quantities can be displayed using various means: points, lines, numbers, symbols, words, coordinates, shading and colour (Tufte 2001). The question is how to employ these means to appropriately display them? Tufte’s principles (2001) for good graphical display of data are: • • • • • • •

show the data, encourage the viewer to think about the data, avoid distorting what the data have to say, make large data sets coherent, reveal the data at several levels of detail, serve a reasonably clear purpose, and be closely integrated with the statistical and verbal descriptions of data.

Following these principles the visualization should be easier to interpret, since only an accurate interpretation of visual variables makes it possible to understand the displayed data. A rational framework for creating an effective visualization has to focus on the strengths of human perception (Wong 2010b). We have to avoid displays of data that are misleading or difficult to discern – for example using colour hue (Figure 8.1) to display quantitative data (Borland and Russell 2007; Wong 2010a). In 1967 the French cartographer Bertin proposed a theoretical framework to avoid misleading visualizations (Bertin 1984), while Cleveland and McGill (1985) took the first steps in measuring people’s perception ability. They ranked perception tasks according to their accuracy and found that position on a common scale is the easiest to interpret while colour hue is the most difficult (Table 8.1). Advances in computer technology have energized scientific visualization research. Most scientists can now relatively easily manipulate their data, but this is not always done in the most effective or even a meaningful way. The role of data manipulator is crucial in providing not just information that is most reliable in general, but also information that is best fitted for the intended use. Producing a digital elevation Rank 1

Aspect to compare position on a common scale

2

position on the same but non-aligned scale

3

length

4

angle, slope

5

area

6

volume, colour saturation

7

colour hue

Figure 8.1: Colour hue is well suited to display categorical data but it is less appropriate for quantitative data. (A) ‘Rainbow’ colour ramps (of saturated colours) are used frequently with continuous geographical data, but without looking at the legend it is difficult to say which colour means more. (B) The transitions from red to yellow to green and so on are uneven, breaking the correspondence between colour and numerical value. Using colour lightness is more appropriate to display such data (Borland and Russell 2007; Wong 2010a)

Table 8.1: Elementary perceptual tasks – ordered from most to least accurate (Cleveland and McGill 1985)

102

Žiga Kokalj, Klemen Zakšek and Krištof Oštir basic principles of data processing can also benefit archaeologists when ordering new lidar data acquisitions, because they can more easily communicate required scanning parameters and data preparation with the data provider.

Figure 8.2: ‘Fish scales’ are a result of direct point cloud rasterization (A) and wave-like features are a consequence of poor registration of scan lines (B). 0.5 m resolution lidar data of an area around Besancon, France, used with permission of the University of FrancheComte. Charcoal burning platforms can be seen on (B). Black stars typically form around small protruding features on a sky-view factor (SVF) image when calculating it with 8 or 16 directions (C). 0.5 m resolution lidar data of World War I trenches near Kobarid in Slovenia. All images display SVF calculated in 8 directions with a 10 m search radius

The need to know

model (DEM) is not enough – in increasingly specialised environments it has to be specifically tailored to the task at hand. When such a model is obtained for archaeological purposes, it has to be properly displayed to provide maximum benefits. With advances in remote sensing and geographic information systems (GIS) in archaeology, many archaeologists have the knowledge to manipulate spatial data, but now need the skills to present quantitative data appropriately. This is especially important since many processing methods and visualization techniques are increasingly available as free tools or collections of tools (e.g. ZRC SAZU 2010; ESRI Mapping Center Team 2010), and because a large number of quite distinct layers is likely to confuse or overwhelm the untrained user. Knowing the

Information about how raw data has been acquired and processed, and about the method and settings for its presentation, has a great impact on the feature detection and interpretation processes. For example, if the interpreter knows the original scanning density, the method of point cloud filtering and the method of digital elevation model generation, they can judge and take decisions about the various artefacts that can be found in the data. It is extremely useful to know that ‘fish scales’ sometimes found in forest datasets (Figure 8.2A), are a result of a direct point cloud rasterization (i.e. without help of a triangulated irregular network – TIN), and that wave-like features resembling ridge and furrow are a consequence of poor registration of scan lines (Figure 8.2B). Black stars, sometimes seen in a sky-view factor image, can be linked to dedicated processing, where the point-cloud filtering process has been optimized to leave the archaeology as intact as possible. Eight or sixteenpointed black stars may be formed around very small (in area) protruding features that are the remains of conifer trees (Figure 8.2C). This occurs where conifers are too dense for a laser pulse to reach the ground and filtering is set so as not to over smooth the derived elevation model leaving star-shaped ‘bumps’ in the relief generated by the fact that sky-view factor is usually calculated in eight or sixteen directions. This is also true for diffuse solar insolation maps, because skyview factor is used for their calculation (Figure 8.16A). Because visual interpretation is based on contrast, the latter is very often enhanced by histogram manipulation and scale exaggeration, or artificially introduced; a good example is the well-known analytical shading. Consequently, the extent and shape of features being recorded can be, and usually are, altered. It is therefore clearly necessary to know how different visualization techniques work and how to use them to best advantage according to the characteristics of data, general morphology of the terrain, and the scale and preservation of features in question. When a certain technique is chosen

8  Visualizations of lidar derived relief models

103

scanner type platform date swath width [m] flying height [m] pulse repetition rate [kHz] average density of all returns per m2 on a combined dataset GPS error [m] spatial resolution of the final elevation model [m]

for detection or interpretation of features, it is particularly important to know what the different settings do and how to manipulate them. Of course, no perfect technique exists and they are complementary to a great extent. Because the techniques show various features in different ways, emphasizing edges, circular or linear forms differently, a combination of methods is usually required to get most out of the data.

Visualization techniques for lidar derived elevation models Each of the visualization techniques described in this section is illustrated by a figure of the Žerovinšček hillfort to allow direct comparison of the effect of particular techniques. Because not all methods work well on hilly terrain, and to help illustrate some of the peculiarities of the techniques, additional figures are also included. All the visualizations are presented with a minimum-maximum histogram stretch, usually with a saturation of minimum, maximum or both. Using this type of histogram stretch is essential for comparability and for the legends to make sense. Wherever reasonable, the displayed range is presented in a legend. The composite images with hillshading are usually blended by a mathematical procedure based on colour intensity, not by lowering the opacity. The Žerovinšček hillfort, near Bločice in Slovenia is Iron Age in date and lies above an ancient routeway which leads to the famous Amber Route at the Postojna gates. The visible remains comprise stone ramparts, numerous terraces, plateaus and other potential structures but no archaeological excavations have been done on the site (Laharnar 2009). The site is completely covered by forest and the landscape surrounding it is hilly and rugged. Laser scanning and data processing of the Žerovinšček region was not done for

Optech ALTM 3100EA aeroplane 8, 9, 21 and 24 May 2007 350, 1300 1650, 1900, 2200 50, 100 23.2 0.08 1.0

archaeological purposes. Filtering of the lidar point cloud was performed with Repetitive Interpolation algorithm (REIN – Kobler et al. 2007) that works especially well in hilly forested landscapes, but the algorithm settings were adjusted for production of a ‘true’ terrain model. Despite the high density of the final point cloud, the derived terrain is therefore not optimal for archaeological purposes but has been selected for the study because elevation models with this type of processing are usually available to the general lidar data user. Relief shading Relief shading gives the best visual impact of all techniques – it is a representation of the relief in a natural and intuitive manner. Relief shading and hachures used to be the domain of cartographers who manually drew these depictions of slope. Manual shading in the form of a digital brush can be used even today (Jenny 2001; Patterson 1997; Patterson and Hermann n.d.; Price 2001). However, in recent years standard and automatic relief shading tools have become available to users of cartography, GIS and remote sensing software, most commonly in the form of analytical hillshading. This is a computer-based process of generating a shaded relief from a DEM. It is a description of how the relief surface reflects incoming illumination based on physical laws or empirical evidence. There are numerous analytical hillshading techniques (Horn 1981; Phong 1975; Blinn 1977; Batson et al. 1975; Minnaert 1961), although only the method developed by Yoëli (1965) has become a standard feature in most GIS software. Therefore, when analytical hillshading is discussed it is the method developed by Yoëli that is concerned. Standard analytical hillshading is easy to compute and straightforward to interpret even by non-experts and without training. It has a basic assumption that the relief is a Lambertian surface illuminated by direct light from a fictive

Table 8.2: Lidar scanning parameters of the Žerovinšček region (Slovenia)

104 Figure 8.3: An orthophoto image (A), a composite of a greyscale orthophoto image and a normalized digital surface model (nDSM) (B), and an analytically shaded lidar derived terrain model (315° azimuth and 45° Sun elevation; 1 m resolution) (C) of the Žerovinšček Iron Age hillfort near Bločice, Slovenia. The area is completely covered by a thick forest canopy and nothing can be seen of the site from the air. The height of objects recognizable on an orthophoto (trees in this example) can be distinguished from the composite image. The terrain model was created using Repetitive Interpolation (REIN) algorithm (Kobler et al. 2007)

Žiga Kokalj, Klemen Zakšek and Krištof Oštir light source at an infinitive distance (the light beam has a constant azimuth and elevation angle for the entire area). The computed grey value is proportional to the cosine of the illumination incidence angle on the relief surface – the angle between the surface normal and the light beam. Areas perpendicular to the light beam are the most illuminated, while areas with an incidence angle equal or greater than 90° are in a shade (Figure 8.3C). Direct illumination restricts the visualization in dark shades and brightly lit areas, where no or very little detail can be perceived. A single light beam also fails to reveal linear structures that lie parallel to it (Figure 8.4) which can be problematic in some applications, especially in archaeology (Devereux et al. 2008). As a solution Brassel (1974) proposed local adaption of the position of the illumination source in problematic areas. Imhof (1982) suggested local variations in the aspect and the inclination of the global illumination vector by angles of up to a maximum of 30°. This reveals details in ridges and valleys oriented parallel to the primary direction of illumination. Analytical hillshading is useful for exploratory purposes in areas with only slight variations in general topography because under very low light source angles (below 10°) features of extreme subtleness can be made visible (Figure 8.5). Derivatives of hillshading from different directions Producing multiple relief shading outputs by illuminating a surface from multiple directions enhances the visualization of topography.

Figure 8.4: Angle dependence of analytical hillshading: 315° azimuth illumination (A), and 45° azimuth (B), both with 45° Sun elevation. Note the difference of the relief structures that can be observed because of the change in the illumination azimuth. Overgrown cultural terraces near Kobol in Slovenia as evidenced by 1 m spatial resolution terrain model

8  Visualizations of lidar derived relief models

105

Hillshaded images are sometimes used to guide ground surveys, but comparing multiple images in the field is extremely inconvenient. A step towards an improved understanding of the results is combining multiple shadings by considering only the mean (Hobbs 1995), the maximum, or the range of values, for each pixel (Figure 8.6). In order to display the areas with a low range of values more clearly, the result can be square rooted. A common example is also a combination of standard hillshading (315° azimuth) with vertical illumination (Imhof 1982; Hobbs 1995). Hillshades from three different directions, preferably with a consecutive 90° angle difference, can be used to create an RGB image (Devereux et al. 2008; Hobbs 1999). Because images created by illumination from several angles are highly correlated (the same scene is viewed), it is possible to ‘summarize’ information by a mathematical transformation with principal component analysis (PCA) (Devereux et al. 2008). The first three components computed from multiple (e.g. 16) directions usually contain a high percentage (typically over 99%) of the information or variability in the original dataset. They can thus be expected to provide a basis for the visualization of all the shade direction data with minimal loss of archaeological features. The PCA – especially the combination of the first and second principal components, or the false colour composite image of the first three – simplifies the interpretation of the multiple shading data (Figure 8.7). However it does not provide consistent results with different datasets. Elevation differentiation Elevation differentiation, also referred to as colour shading, colour cast, or constrained colour ramps method, limits the range of values that are presented simultaneously. The relative height of archaeological features is usually of a much smaller scale than that of the landscape topography. The height values that archaeologists are interested in presenting are therefore accumulated in only a small part of the whole image histogram. A histogram is a graphic representation of image values distribution (in case of a DEM raster, the values are elevations). It therefore presents the frequency of elevations by vertical rectangles, with widths equal to class interval (e.g. 1 m) and heights equal to the frequency of elevations in that interval. The technique applied is called a histogram stretch,

Figure 8.5: Very low light source angles expose features of extreme subtleness: a standard 45° Sun elevation (A) and (B), and low light 5° Sun elevation (C), all with 45° azimuth. However this only works in areas with very gentle relief morphology, such as this example of the Site A embanked enclosure(s) in Brú na Bóinne World Heritage Site in Ireland.1 m spatial resolution lidar data used with permission of the Discovery Programme. Local histogram saturation is used to present (B) and (C). The first to show the difference this makes when compared with normally presented shaded relief (A), and the second because the image is otherwise too dark to expose any details

whereby the values of interest are stretched to the whole histogram, enhancing the contrast between light and dark areas. There are many techniques and methods to enhance contrast and emphasize details (nonlinear enhancements include: logarithmic stretch, square root enhancement, exponential stretch, and histogram equalization). However, because the preservation of the relative differences between the values (elevations) is important, a basic method known as linear stretch with saturation is used to cut off the extreme values in the upper and lower parts of the histogram. The lowest (Is_min) and highest (Is_max) value of the elevation range to observe are defined. The histogram is then stretched to fill the whole range of values between 0 and Rmax, where Rmax is the

106

Žiga Kokalj, Klemen Zakšek and Krištof Oštir

Figure 8.6: The mean (A), maximum (B) and square rooted range of hillshadings (C) from 16 directions with 45° Sun elevation. An RGB image of hillshadings from three directions (315°, 0°, and 45° azimuth with 45° Sun elevation) (D)

Figure 8.7: A composite of the first two com­ ponents (A) and RGB composite of the first three components of a principal component analysis of analytical hillshading from 16 directions with 45° Sun elevation (B)

maximum radiometric value (e.g. 0 to 255 in an 8-bit display). If I is the original value, than the stretched value Istretched equals: ‫ܫ‬௦௧௥௘௧௖௛௘ௗ ൌ ൞ܴ௠௔௫

Ͳ ‫ ܫ‬൑ ‫ܫ‬௦̴௠௜௡ ‫ ܫ‬െ ‫ܫ‬௦̴௠௜௡ ‫׷‬ ‫ܫ׷‬௦̴௠௜௡ ൏ ‫ ܫ‬൏ ‫ܫ‬௦̴௠௔௫  ‫ܫ‬௦̴௠௔௫ െ ‫ܫ‬௦̴௠௜௡ ‫׷‬ ‫ ܫ‬൒ ‫ܫ‬௠௔௫ ܴ௠௔௫

All the pixels with values lower than the defined Is_min are presented with 0, and pixels with values greater than Is_max are presented with Rmax. This technique is very useful for visualization of features of interest in flat landscapes and is easy to interpret, especially when an appropriate colour ramp is used (Figure 8.8). It is also the only visualization technique presented that retains the original information about relief elevation. It is therefore easy to assess factors

such as the depth of ditches or height of tumuli. However, even with slight variations in the general morphology of terrain, the technique becomes less useful, because archaeological earthworks are obscured by the variation in topography and because intensive manipulation of the histogram is required. For the same reasons the technique completely fails in rugged terrain (Figure 8.9A). A related technique is height coding with modulo distribution (Wood 1996). It dissects the area into equal elevation bands and colours them accordingly. The colour coding interval scheme is repeated in every band interval. This technique reveals small differences within a flat landscape while the bands can be interpreted as contours on a steep and diverse terrain (Figure 8.9B).

8  Visualizations of lidar derived relief models

107

Figure 8.8: Histogram stretch to a narrow range of values. Even a 5 m range of values can be a problem to identify shallow features, such as ridge and furrow on an inclined slope (A). Tame Valley in the English Midlands, 1 m spatial resolution lidar data, collected with an Optech ALTM 2033, © Infoterra Global Ltd, used with permission of the University of Birmingham (A). Past riverbeds of Nadiža 0.5 m spatial resolution lidar data, West of Kobarid in NW Slovenia (B) Figure 8.9 (left): Histogram stretch to the range of data values represented in the figure (A). A local relief model presented with height coding with modulo distribution (B).The techniques do not work well on diverse relief because the range of values is too wide

Trend removal Archaeological features are generally of a much smaller scale than the landforms on which they lie. It is therefore necessary to adjust visualization techniques appropriately, for example defining a small search radius for skyview factor calculation or setting a suitable range for elevation differentiation, although this is not possible with all techniques. A procedure that separates local small-scale features from largescale landscape forms is called trend removal (Figure 8.10). This removes the height variation of large-scale forms and produces a local relief model (LRM) that can be used as input for visualization with other methods. The LRM presents heights (depths) of small-scale features directly (Figures 8.11 and 8.12) and can also be combined with other data sources such as aerial photographs and hyperspectral imagery. The trend can be assessed by generalizing a detailed DEM. Because processing is straightforward, generalization is usually done with a low pass convolution filter, such as average or median, or by resampling a DEM to a lower resolution. A better option is a Gaussian filter that produces a smoother transition between features, but it is computationally more

Figure 8.10: A schematic presentation of trend removal

demanding (Reitberger et al. 2008; Wagner et al. 2008). A LRM (Figure 8.11C) is the difference between the original relief model (Figure 8.11A) and the assessed trend (Figure 8.11B). Hesse (2010) has refined the process by introducing a ‘purged DEM’. The first approximation of large-scale relief forms is assessed with a low pass (average) filter. Because small-scale features are smoothed rather than eliminated by the low-pass filter, the model derived by this approach is biased towards small features, i.e. the local relief elevations are progressively underestimated as the spatial extent of the features increases. Therefore zero contours (contours that match both the trend DEM and the original DEM) are calculated from this first approximation and elevation data is assigned to points along these contours. A new approximation is then interpolated from

108

Figure 8.11: Subtracting an averaged terrain model (B) from the original relief model (A) produces a difference map or a LRM (C)

Žiga Kokalj, Klemen Zakšek and Krištof Oštir

the points. A LRM derived using this approach results in a less biased representation of smallscale topographic features, which reflects more truthfully the elevations of these features relative to the surrounding landscape.

The level of smoothing is defined by the kernel size of the filter, where a smaller kernel exposes smaller features and vice versa. The precise kernel size should reflect the size of the small-scale landforms while a generally safe bet is a kernel size of about 25 m. The method works best on terrain with gradual slopes, while it produces artefacts such as artificial banks and ditches where relief is diverse and/or changing abruptly. Slope severity Slope severity (gradient) is the first derivative of a DEM and is aspect independent. It represents the maximum rate of change between each cell and its neighbours and can be calculated either as percentage of slope or degree of slope. Its use for detection of archaeological features has been reported by Doneus and Briese (2006, 104). Challis et al. (2011) found slope severity the best visualization technique for most circumstances among the methods they analysed. If presented in an inverted greyscale (steep slopes are darker), slope severity retains a very plastic representation of morphology. However, additional information is needed to distinguish between positive/convex (e.g. banks) and negative/concave (e.g. ditches) features since slopes of the same gradient (regardless of rising or falling) are presented with the same colour.

Figure 8.12: A hillshaded image (A) and a histogram stretch (B) of a local relief model, and the LRM presented with a purposely designed colour ramp overlaid on a hillshaded image (Hesse 2010) (C). The LRM was calculated with a 25 m Gaussian filter

Sky-view factor Sky-view factor (SVF) can be used as an alternative method of relief mapping in order to overcome the directional problems of hillshading (Kokalj et al. 2011; Zakšek et al. 2011). SVF is a geophysical parameter (if we do not manipulate elevation data by vertical exaggeration) that measures the portion of the sky visible from a certain point. The portion of the sky visible above the surface is especially relevant in energy balance studies (Bewley et al. 2005; Bourbia and Awbi 2004) and computation of diffuse solar insolation (Yard et al. 2005; Robinson 2006). Diffuse solar insolation rasters can be used to

8  Visualizations of lidar derived relief models visualize archaeological features as well (Challis et al. 2011), but require additional calculations and the results are more generalized. The SVF computation is based on diffuse illumination. An imaginary light source illuminates the relief from the celestial hemisphere, which is centred at the point being illuminated. Additionally it assumes that: • the hemisphere is equally bright across its entire area; • there is no additional directional illumination source; • the Earth’s curvature is ignored over short (not more than 10 km) distances.

In such a case, the relief illumination is correlated to a portion of the visible sky that is limited by the relief horizon (either terrain or surface, depending on the application) corresponds to the relief illumination; a ridge is more illuminated than the bottom of a steep valley because both are illuminated from the bright sky and more sky can be seen from the ridge than from the valley. The most convenient measure for expressing the portion of the visible sky is the solid angle, Ω (Figure 8.14), which is proportional to the surface area, S, of the projection of the object onto the sphere centred at the observation point, divided by the square of the sphere’s radius R: Ω = k • S/R2. In order to normalize the SVF between 0 and 1 the proportionality constant k is set to the value of 1/2π (Marks et al. 1979). Values close to 1 indicate that almost the entire hemisphere is visible, which is the case in exposed features (planes and peaks), while values close to 0 are present in deep sinks and lower parts of deep valleys where almost no sky is visible. The light that falls from the sky onto a certain part of the surface is reduced by the obstacles that form the horizon. These obstacles can be described in all directions by the vertical elevation angle above the horizontal plane. A good SVF approximation can therefore be performed with the estimate of this angle in several directions. After the vertical elevation angle is determined in the chosen number of directions n, the SVF is determined as a sum of all portions of the sky within each direction: ∑ (1 – sin γi)/n, where γi is the vertical angle of the horizon in the direction i. The computation of the horizon in multiple directions is time consuming so simplified methods have been developed (Duffie and Beckman 1991; Tian et al. 2001), that can

109 Figure 8.13: Slope severity (A). The elevation profile (B) refers to the CD line in (A). Note that the structure the profile line crosses can be quite easily – and wrongly – interpreted as convex instead of concave, because the image looks similar to shaded relief

Figure 8.14: Sky-view factor is defined as the proportion of visible sky (Ω) above a certain observation point as seen from a two-dimensional representation (A). The algorithm computes the horizon angle γ in n (presented are eight) directions to the specified radius R (B)

Figure 8.15: Sky-view factor image (10 m search radius in 16 directions) of Žerovinšček hillfort. Many details can be perceived in this image despite the variable relief morphology

110

Žiga Kokalj, Klemen Zakšek and Krištof Oštir be considered as a manipulation of the slope severity histogram. SVF can be used with raster elevation data of different scales, from very high resolution lidar relief models (applicable in archaeology) to coarser resolution global datasets (such as SRTM DEM and ASTER DEM). The computation of SVF is influenced by the search radius of the horizon and the larger the search radius the more generalized the results. In contrast, a small search radius can be used to visualize and classify local morphological forms. For example, a 10 km search radius can be used in meteorological studies, while a 10 m search radius is suitable for discrimination of archaeological features. Locally flat terrain, ridges and earthworks (e.g. building walls, cultivation ridges, burial mounds) which receive more illumination are highlighted and appear in light to white colours on a SVF image, while depressions (e.g. trenches, moats, ploughing furrows, mining pits) are dark because they receive less illumination. To display features more clearly a histogram stretch is usually required. A minimum-maximum stretch with saturation of minimum is recommended, because it preserves the relative distribution of values. This renders narrow valley floors, very steep slopes, and the vicinity of high objects (such as buildings) black, which can be useful when delineating concave features on flat or undulating terrain. If this is undesirable because features of interest can be ‘hidden’ in such areas, a standard deviation stretch or a standard minimum-maximum stretch provides a visualization with no or minimal saturation, but less contrast. We find SVF the most appropriate general visualization technique, because it enhances visibility of both simple and complex small-scale features whatever their orientation and shape, on most types of terrain. The SVF calculation code can be modified to omit closest neighbour pixels in the calculation. This greatly improves the visibility of archaeological features where the data are noisy due to strip misalignment or overambitious resolution setting. The SVF based relief visualization as described above is implemented in a remote sensing software ENVI+IDL (Exelis Visual Information Solutions 2012). The code is publicly available and can be downloaded from ZRC SAZU (2010). The difference between results this code gives and some other implementations (e.g. SVF in SAGA GIS) are enormous, especially from the point of visualizations, because the latter usually

give very saturated areas with low SVF. This means no details can be perceived in valleys. Solar radiation Another sophisticated visualization technique that can also be used in archaeology is solar insolation mapping (Challis et al. 2011). Meteorological stations used to collect data on solar insolation duration, but today global irradiance [Wm-2] is usually measured – that is solar energy flux received by a horizontal surface. It consists of direct and diffuse parts and it is possible to measure the latter separately. Because the global irradiance changes over time we can map either its average for a certain period or the global radiance [Jm-2 or Whm-2] – energy density that is received by a horizontal surface. Global radiance varies slowly in space and the obvious changes only occur on the borders between climatic areas and elevation changes. Maps of global radiance are therefore almost unusable for visualizing terrain features. For visualization of smaller areas it is more appropriate to map quasi-global radiation exposure, which is the global irradiation transformed from the horizontal to the actual terrain surface. Many different methodologies have been developed for this in recent years (e.g. Corripio 2003; Zakšek et al. 2005). These stress that the direct part of quasi-global radiance depends mostly on the angle of incidence, which is defined by astronomical and terrain parameters (slope and aspect), as well as the horizon of the surrounding terrain that influences the effective possible duration of solar insolation (the period of sunshine in clear weather subtracted by the period of shade due to terrain obstacles). The terrain is also significant for the diffuse part of radiation since diffuse radiation is reduced by the sky-view factor. A part of the diffuse radiation does not come from the sky but from the surface and its amount depends again on the sky-view factor and the surface albedo. Sunny and shady inclines are immediately evident in the quasi-global radiance maps. The radiance is higher on sunny slopes in comparison with the global radiance of horizontal surfaces and the shady slopes receive significantly less radiance. The output looks similar to hillshading. However, in contrast to classical hillshading the greatest values are usually on the south facing slopes (in the northern hemisphere), which may confuse the interpreter – the concave terrain forms appear convex and vice versa. In addition,

8  Visualizations of lidar derived relief models

111 Figure 8.16: Diffuse (A) and solar insolation map (B). The latter gives more cues on the general terrain morphology because direct solar insolation is also considered

classical hillshading ignores cast shadows. The computation of solar radiance is also significantly longer, because the cosine of the incidence angle has to be computed for each moment considered in the analysis (e.g. every hour of every tenth day). This is a numerical solution and a faster analytical solution is also possible but in this case terrain obstacles are not considered (Allen et al. 2006). Such an analytical solution should provide results similar to hillshading from multiple directions. However, the changes are not only in the azimuth but in the zenith angle as well. This has the consequence that the irradiation is greater at noon than at sunrise or sunset, and also that the influence of winter months in the whole year result is smaller than the influence of summer months. As can be seen in Figure 8.16, solar insolation mapping is a powerful tool for terrain visualization. To enhance the visualization we can manipulate its parameters to create one of the visualizations described in previous sections (hillshading, hillshading from multiple directions, sky-view factor) or an arbitrary combination of them. Other techniques There are many other methods that can be used to aid recognition of small scale landforms and interpretation of the environment of archaeological features in general. For example, logical and arithmetical operations, classification, visibility analysis, overlaying procedures and moving window operations can be used to enhance the edges of features or otherwise improve their recognition. An interesting approach uses combinations of normalised digital surface model (nDSM) and shaded relief (Figure 8.17A) or greyscale orthophoto image (Figure 8.3C) that help evaluate the environment of the artefacts, especially when covered by

Figure 8.17: A composite of a normalised digital surface model and shaded relief (A), a composite image of relief curvature in the direction of slope and shaded relief (B), and a composite image of total relief curvature and shaded relief (C)

forest. An approach that is especially effective in visualizing hydrological networks was proposed by Kennelly (2008) who combined relief

112

Žiga Kokalj, Klemen Zakšek and Krištof Oštir hillshading and curvature (Figure 8.17B and C). All the visualizations are usually combined with analytical hillshading for interpretation purposes.

Conclusions National heritage records usually contain vector data of recorded features. With the help of remote sensing and lidar especially, a vast number of new discoveries is made each year, even in areas that have been intensively studied for decades. To minimize ambiguity it would be wise to include some information on how the features have been recorded. With this we do not have in mind only the name of the technology, but also the specific details about base data acquisition, processing and visualization, as well as about the interpretation process. Despite the many metadata standards, not even the minimum required is usually supplied to or used by the interpreters; hence there is a clear need for development of good practice. The following metadata can assist the interpretation process at several stages: • data scanning: scanner type, scanning density, density of a combined dataset, scanning date; • data processing: method(s) used, parameter settings, description of the processing goal (e.g. producing a terrain model, removing just the vegetation), elevation model resolution; • visualization: method(s) used, parameter settings (e.g. hillshading (Sun elevation and azimuth), LRM (method, distance), SVF (distance, directions); • interpretation process: reliability of the results (qualitative if quantitative evaluation is not possible, e.g. low to high, description of each class is recommended).

In this chapter we have described a number of visualization methods. We have evaluated relief shading and derivatives of hillshading from different directions, elevation differentiation, trend removal, slope severity, sky-view factor, solar insolation and mentioned several other techniques. The visualization methods can be related to physical quantities or derived only for relief representation. They range from simple to computationally very extensive. Most of them are already included in standard GIS software, are available as add-ons or as independent products.

When choosing the most suitable method one has to differentiate two main goals of visualization of the relief: representation of elevation/topography and detection of features. For elevation any standard method, e.g. the most commonly known hillshading, is adequate. For automatic or manual feature detection, the choice is more complex. Usually the interpretation can be done based on a limited number of visualizations at the same moment; and in fact the human interpreter can observe only one view at a time. It is therefore crucial to limit the visualisations, but can we select just one? As we have shown, different artefacts are better seen on different representations. Nevertheless, we believe that elevation differentiation, trend removal and sky-view factor are most effective, especially if combined with hillshading. Crucially, we suggest that at least two methods should be used in the interpretation.

References Allen, R.G., Trezza, R. and Tasumi, M., 2006. Analytical integrated functions for daily solar insolation on slopes. Agricultural and Forest Meteorology 139 (1–2), 55–73. Batson, R.M., Edwards, E. and Eliason, E.M., 1975. Computer Generated Shaded Relief Images. Journal of research of the U.S. Geological Survey 3(4), 401–408. Bertin, J., 1984. Semiology of Graphics: Diagrams, Networks, Maps 1st ed. University of Wisconsin Press. Bewley, R.H., Crutchley, S.P. and Shell, C.A., 2005. New light on an ancient landscape: lidar survey in the Stonehenge World Heritage Site. Antiquity 79(305), 636–47. Blinn, J.F., 1977. Models of light reflection for computer synthesized pictures. SIGGRAPH Comput. Graph. 11(2), 192–198. Borland, D. and Russell, M.T.I., 2007. Rainbow Color Map (Still) Considered Harmful. IEEE Comput. Graph. Appl. 27(2), 14–17. Bourbia, F. and Awbi, H.B., 2004. Building cluster and shading in urban canyon for hot dry climate: Part 1: Air and surface temperature measurements. Renewable Energy 29(2), 249–62. Brassel, K., 1974. A Model for Automatic HillShading. Cartography and Geographic Information Science 1, 15–27. Challis, K., Forlin, P. and Kincey, M., 2011. A generic toolkit for the visualization of archaeological features on airborne lidar elevation data. Archaeological Prospection 18(4), 279–89. Challis, K., Kokalj, Ž., Kincey, M., Moscrop, D. and Howard, A.J., 2008. Airborne lidar and

8  Visualizations of lidar derived relief models historic environment records. Antiquity 82(318), 1055–64. Cleveland, W.S. and McGill, R., 1985. Graphical perception and graphical methods for analyzing scientific data. Science (New York, N.Y.) 229(4716), 828–33. Corns, A. and Shaw, R., 2009. High resolution 3-dimensional documentation of archaeological monuments & landscapes using airborne LiDAR. Journal of Cultural Heritage 10, Supplement 1(0), e72–e77. Corripio, J.G., 2003. Vectorial algebra algorithms for calculating terrain parameters from DEMs and solar insolation modelling in mountainous terrain. International Journal of Geographical Information Science 17(1), 1–23. Devereux, B.J., Amable, G.S. and Crow, P., 2008. Visualisation of LiDAR terrain models for archaeological feature detection. Antiquity 82(316), 470–9. Doneus, M. and Briese, C., 2006. Full-waveform airborne laser scanning as a tool for archaeological reconnaissance. In Campana, S. and Forte, M., (eds). From space to place: 2nd International Conference on Remote Sensing in Archaeology: proceedings of the 2nd international workshop, CNR, Rome, Italy, December 4–7, 2006 (BAR international series, 1568): 99–105. Oxford: Archaeopress. Duffie, J.A. and Beckman, W.A., 1991. Solar Engineering of Thermal Processes. 2nd ed. WileyInterscience. ESRI Mapping Center Team. 2010. Hillshade Tools. Available at: http://mappingcenter.esri.com/index. cfm?fa=arcgisResources.modelsScripts. Exelis Visual Information Solutions. 2012. ENVI Software – Image Processing & Analysis Solutions. Available at: http://www.exelisvis.com/language/ en-us/productservices.aspx [Accessed September 28, 2012]. Friedman, V., 2008. Data Visualization and Infographics. Graphics. Available at: http://www. smashingmagazine.com/2008/01/14/mondayinspiration-data-visualization-and-infographics/ [Accessed May 12, 2011]. Friendly, M., 2008. Milestones in the history of thematic cartography, statistical graphics, and data visualization, in Seeing Science: Today. Boston. Available at: http://www.math.yorku.ca/SCS/ Gallery/milestone/milestone.pdf [Accessed May 12, 2011]. Hesse, R., 2010. LiDAR-derived Local Relief Models – a new tool for archaeological prospection. Archaeological Prospection 17(2), 67–72. Hobbs, K.F., 1995. The rendering of relief images from digital contour data. The Cartographic Journal 32(2), 111–6. Hobbs, K.F., 1999. An investigation of RGB multi-band shading for relief visualisation. International Journal of Applied Earth Observation and Geoinformation 1(3–4), 181–6.

113 Horn, B.K.P., 1981. Hill shading and the reflectance map. Proceedings of the IEEE 69(1), 14–47. Imhof, E., 1982. Cartographic relief presentation. Berlin, New York: Walter de Gruyter. Jenny, B., 2001. An Interactive Approach to Analytical Relief Shading. Cartographica: The International Journal for Geographic Information and Geovisualization 38(1), 67–75. Kennelly, P.J., 2008. Terrain maps displaying hillshading with curvature. Geomorphology 102(3–4), 567–77. Kobler, A., Pfeifer, N., Ogrinc, P., Todorovski, L., Oštir, K. and Džeroski, S., 2007. Repetitive interpolation: a robust algorithm for DTM generation from Aerial Laser Scanner Data in forested terrain. Remote sensing of environment 108(1), 9–23. Kokalj, Ž., Zakšek, K. and Oštir, K., 2011. Application of sky-view factor for the visualization of historic landscape features in lidar-derived relief models. Antiquity 85(327), 263–73. Laharnar, B., 2009. The Žerovnišček Iron Age hillfort near Bločice in the Notranjska region. Arheološki vestnik (60), 97–157. Marks, D., Dozier, J. and Davis, R.E., 1979. Clear-Sky Longwave Radiation Model for Remote Alpine Areas. Archiv für Meteorologie Geophysik und Bioklimatologie Serie B-Klimatologie Umweltmeteorologie Strahlungsforschung 27(2–3), 159–87. Minnaert, M., 1961. Photometry of the Moon. In Kuiper, G.P. and Middlehurst, B.M., (eds). Planets and Satellites, 213–48. Chicago: The University of Chicago Press. Patterson, T., 1997. A Desktop Approach to Shaded Relief Production. Cartographic Perspectives 36, 38–40. Patterson, T. and Hermann, M. Creating valueenhanced shaded relief in Photoshop. Available at: http://www.shadedrelief.com/value/value.html. Phong, B.T., 1975. Illumination for Computer Generated Pictures. Communications of the ACM 18(6), 311–7. Price, W., 2001. Relief Presentation: Manual Airbrushing Combined with Computer Technology. The Cartographic Journal 38(1), 107–12. Reitberger, J., Krzystek, P. and Stilla, U., 2008. Analysis of full waveform LIDAR data for the classification of deciduous and coniferous trees. International Journal of Remote Sensing 29(5), 1407–31. Robinson, D., 2006. Urban morphology and indicators of radiation availability. Solar Energy 80(12), 1643–1648. Tian, Y.Q., Davies-Colley, R.J., Gong, P. and Thorrold, B.W., 2001. Estimating solar insolation on slopes of arbitrary aspect. Agricultural and Forest Meteorology 109(1), 67–74. Tufte, E.R., 2001. The Visual Display of Quantitative Information. 2nd ed. ������������ Graphics Pr. Wagner, W., Hollaus, M., Briese, C. and Ducic,

114

Žiga Kokalj, Klemen Zakšek and Krištof Oštir V., ���������������������������������������� 2008. 3D vegetation mapping using smallfootprint full-waveform airborne laser scanners. International Journal of Remote Sensing 29(5), 1433–52. Wong, B., 2010a. Points of view: Color coding. Nature Methods 7(8), 573. Wong, B., 2010b. Points of view: Design of data figures. Nature Methods 7(9), 665. Wood, J., 1996. The geomorphological characterization of digital elevation models. PhD thesis. University of Leicester, UK. Available at: http://www.soi.city. ac.uk/~jwo/phd [Accessed December 7, 2009]. Yard, M.D., Bennett, G.E., Mietz, S.N., Coggins Jr., L.G., Stevens, L.E., Hueftle, S. and Blinn, D.W., 2005. Influence of topographic complexity on solar insolation estimates for the Colorado

River, Grand Canyon, AZ. Ecological Modelling 183(2–3), 157–72. Yoëli, P., 1965. Analytische Schattierung. Ein kartographischer Entwurf. Kartographische Nachrichten 15(5), 141–8. Zakšek, K., Oštir, K. and Kokalj, Ž., 2011. Sky-view factor as a relief visualization technique. Remote Sensing 3, 398–415. Zakšek, K., Podobnikar, T. and Oštir, K., 2005. Solar insolation modelling. Computers & Geosciences 31(2), 233–40. ZRC SAZU. 2010. IAPS ZRC SAZU | Institute of Anthropological and Spatial Studies ZRC SAZU. Available at: http://iaps.zrc-sazu.si/?q=en/svf#v [Accessed July 9, 2010].

9 Worth a thousand words – Photogrammetry for archaeological 3D surveying Fabio Remondino After a period during which laser scanners have dominated surveying applications, photogrammetry is slowly coming back, both at research and commercial level. Indeed recent developments and improvements of hardware and software are drawing attention back to the image-based approach for many surveying and modelling purposes. This article reports the state-of-the art of photogrammetry for archaeological 3D mapping applications with examples at different scales. Keywords: photogrammetry, image-based modelling, automation, lidar

“It is quite evident that even with our past progress, we have only scratched the surface of the possibilities in the use of photogrammetry” (George D. Hardy, 1973)

Introduction A decade’s experience of recording and creating digital 3D datasets such as point clouds, digital surfaces or solid models has demonstrated that these data can be a truly valuable resource for many heritage applications and activities like interpretation, analyses, computer-aided restoration, cross-comparison, monitoring of shape and colour, the creation of 3D repositories and catalogues, and multimedia exhibitions; all this is in addition to the more traditional activities

of historical documentation, digital conservation and interactive visualization. Unfortunately examples of research applications which move beyond visualizations and metrics, however sophisticated, remain limited (e.g. Figure 9.1). Indeed 3D models often do not move beyond serving as very detailed documentation of an object, site or landscape and their aesthetic value is sometimes treated as an end in itself. In too many cases archaeologists do not succeed in extracting new archaeological information from 3D data which may have a real impact on knowledge and allow us to address larger research questions. This gap between the creation of a model and its interpretation and use in upstream research can be traced, at least in part, to both a

Figure 9.1: Examples of 3D modelling from terrestrial images found on the Internet or acquired during holiday trips

116

Fabio Remondino communication gap between the archaeological and surveying communities and also to the absence of proper tools (mainly software) for non-experts that are able to produce detailed information appropriate for archaeological needs. This gap in communication and expertise sharing and the lack of effective tools is present at each stage in the digital recording and 3D modelling process. The complete process is multi-faceted, involving data collection, processing, visualization, interpretation, reprocessing, manipulation and cleaning, and more visualization and interpretation, with deliberate choices made at every stage. The manner of data acquisition and processing, the nature of the surveyed subject, and the intended purpose of the 3D recording all have significant impacts on the overall process and the final result. The steps of digital data capture, data processing, archiving and long-term data management, visualization and data reproduction form a chain of procedures and data transformations performed to derive new products, losing and altering elements of the original data along the way. To create high-fidelity digital models, the operator should make choices and employ algorithms that ensure that any processes where large amounts of information are lost occur as late as possible in the processing chain. This paper focuses specifically on the choice between laser scanning and photogrammetric approaches made at the initial data collection step of the 3D digital modelling process, emphasizing the potential and value of photogrammetric approaches, and comments on the impact of this choice on later stages. Examples throughout the text illustrate the potential of photogrammetry and the entire image-based modelling procedure. Photogrammetry (Mikhail et al. 2001; Luhmann et al. 2007) is a robust technique for processing image data to create 3D models. Starting from corresponding features in images (tie points), metric and detailed 3D information with statistical estimates of unknown parameters are derived. Photogrammetry developed from its origin as an analogue process to a fully digital modelling technique in the 1980s. While satellite imagery is now of a suitable resolution, aerial imagery is still the most common source of data for large scale mapping. In recent years developments in photogrammetry have been complemented and strengthened by those in the fields of computer vision, image processing and 3D reconstruction-

most importantly those developing ‘structure from motion’ procedures. Computer vision researchers have focused on developing a fully automated processing pipeline based on projective geometry. These fully automated approaches may reduce the accuracy of the results, but can be useful for visualization, object-based navigation, location based services, robot control, shape recognition, augmented reality, annotation transfer or image browsing purposes – and hence are of interest for some archaeological applications. The greatest benefit of advances in computer vision to photogrammetry is the continuous development of new image analysis algorithms and 3D reconstruction methods. These have been adopted by the photogrammetric community in order to automate most of the steps of the 3D modelling pipeline

Collecting 3D data The creation of a 3D model begins with data collection. Reality-based 3D surveying and modelling of heritage sites and objects may be achieved using passive sensors and image data (Remondino and El-Hakim 2006), active sensors and range data (Vosselman and Maas 2010), classical surveying (e.g. total stations or GNSS), extrusion functions from existing 2D drawings and maps (Yin et al. 2009) or an integration of the aforementioned techniques, as is common in cases of large and complex sites (Stamos et al. 2008; Guidi et al. 2009; Remondino et al. 2009). In general 3D surveying involves the acquisition of unstructured 3D data (e.g. point clouds) using a chosen sensor and technique, while 3D modelling denotes the procedure of converting those unstructured data into structured 3D data (e.g. polygonal meshes). The surveying technique (e.g. 3D scanning or imaging) is selected according to the object’s dimensions, location constraints, instrument portability and usability, surface characteristics, working team experience, budget, final goal of the survey and so on. An image-based survey is defined as a survey where photographic images, together with an element of scale or some ground control points, are processed (e.g. with photogrammetry) to produce 3D reconstructions and line drawings. Passive, image-based sensors (e.g. digital cameras) deliver image data which are processed using mathematical formulae and photogrammetric techniques to infer 3D

9  Worth a thousand words – Photogrammetry for archaeological 3D surveying information from 2D measurements. A rangebased survey is defined as a survey where range sensors (e.g. terrestrial or airborne laser scanning) are employed to directly record an unstructured 3D point cloud of the surveyed scene. To date, many projects have elected to use (airborne or terrestrial) laser scanning due to the high reliability and ease of use of such instruments, in spite of challenges posed by time consuming post-processing and comparatively bulky and expensive equipment. Recent (i.e. post-2008) developments in automated and dense 3D reconstruction from images (Hirschmueller 2008; Remondino et al. 2008; Hiep et al. 2009; Furukawa and Ponce 2010; Verhoeven 2010; Pierrot-Deseilligny et al. 2011), based on photogrammetry and computer vision methods have shown very promising results. This development, coupled with the release of web-based, automated processing tools such as ARC3D and Photosynth (Vergauwen and Van Gool 2006; Snavely et al. 2008), for example, and open-source image processing algorithms (Bundler, Apero, MicMac, PMVS, etc.) have renewed attention on image-based 3D modelling as an inexpensive, robust and practical alternative to 3D scanning. A 3D surveying and modelling process, in addition to delivering metric and possibly georeferenced results, should quantify accuracy (the closeness between measured values and reference values) and precision (the consistency with which a measurement or a set of measurements can be repeated), and provide good value in performance, completeness, portability, flexibility and, ideally for archaeologists, should also be low cost. While it is not easy to achieve all this in a single technique – hence sensor and data integration, in particular for large and complex sites – thanks to recent developments in processing algorithms and software, images can be used as a unique data source to produce very satisfactory 3D models for many applications. To achieve good results experience is still required for image acquisition and processing, but increasing levels of automation are allowing non-expert surveyors to achieve aesthetically pleasing 3D models. Increases in automation and consequent ease of use should not, in theory, come at the price of the geometric quality of the digital model. The quantitative comparison of the results of automated processes with manual, expert processing methods is ongoing and a topic beyond the scope of this paper, rather the

usability (for archaeologists) of the results of automated processes is the focus here.

Recording optical sensors and imaging platforms A variety of image-based sensors are available which may be used for 3D photogrammetric documentation (Remondino 2011). Images can be acquired with terrestrial, aerial or satellite sensors. Terrestrial digital cameras include CCD/CMOS sensors, frame, linear, multiple head, SLR-type, industrial, off-the-shelf, highspeed, panoramic head cameras, and still-video. Common terrestrial cameras have at least 10–12 Megapixels at very low price while highend digital cameras feature more than 40–50 Megapixel imaging sensors. Airborne cameras for photogrammetric surveying are generally classified as small, medium and large format cameras (Sandau 2009), they feature frame or linear array sensors and they are usually coupled with GNSS/IMU (Global Navigation Satellite System/Inertial Measurement Unit) systems. In addition to available aerial acquisition platforms, particular attention has been devoted to UAVs (Unmanned Autonomous Vehicles) like lowaltitude model helicopters or drones. These platforms can fly in an autonomous mode, using an integrated GNSS/IMU which facilitates many procedures in the data acquisition and processing chain, a stabilizer platform to achieve precise, high-quality image overlap and digital cameras. UAVs can be used to collect image data from otherwise difficult to access areas (Figure 9.2) or integrated with terrestrial surveying, and applications are increasing rapidly due to their flexibility, the ability to fly in restricted areas, ease of manoeuvrability and transport, flexible working heights and the possibility to shoot in all directions (vertically down, but also in convergent and horizontal modes). The images acquired can be transmitted to a ground station and processed in situ to ensure complete coverage and provide some quality control for the final models (Figure 9.3). While much of the recent development in photogrammetric documentation has used closerange ground based techniques or UAVs, satellite imagery has the potential in the near future to provide high quality models for large landscape areas (Dowman et al. 2011). Many optical satellite sensors have a return-to-location time

117

118

Fabio Remondino

Figure 9.2: A set of images acquired with an UAV platform over a temple (left), the image orientation results (centre) and the surface model produced (right) for orthoimage and site map generation

Figure 9.3: A mosaic of UAV images (left) acquired above the excavation area of Pava (Siena, Italy). The DSM, derived for exaction documentation, monitoring, GIS and archiving purposes, is shown in shaded (centre) and textured mode (right)

Figure 9.4: Dense 3D reconstruction of a temple in Paestum from about 150 images for documentation, conservation and communication purposes

Figure 9.5: Archaeological artefact surveyed and modelled in 3D using images. To derive metric and scaled results, a known distance (here on a scaled bar) is necessary. Results achieved with interactive measurements (c. 260 polygons) and automated 3D processing (1.9 Mil. polygons)

Figure 9.6: Image-based surveying of a Buddha statue and generation of a 3D point cloud, shown in colour-code and textured mode respectively

9  Worth a thousand words – Photogrammetry for archaeological 3D surveying of less than three days, providing close temporal resolution. Large archives of imagery are available, often as stereo-pairs, providing historical depth unavailable from other sensors. In addition to time-depth, the spatial resolutions achieved are comparable to those from laser scanning. The smallest geometric element which is visible in the images, generally called the Ground Sample Distance (GSD) or footprint can reach 45 cm for civil applications using high resolution satellites, while airborne sensors can deliver images with a 10–20 cm GSD and UAV flights can produce images with very high GSD of about 1–5 cm (Figure 9.4) and terrestrial close-range imagery can produce a GSD of less than 1 mm.

Data processing and point cloud generation The typical image processing pipeline consists of the following steps: • Sensor calibration to determine the internal deformities and parameters of the employed sensor; • Sensor orientation (or image triangulation) to retrieve the six DoF camera poses (i.e. positions where the images have been acquired), normally using a bundle adjustment method; • Surface measurement to derive a dense or sparse point cloud of the surveyed scene using multiimage matching algorithms in the former case, or manual measurement in the latter; • Feature extraction to create vector data or layers for GIS applications; • Digital Surface Model (DSM) and Digital Terrain Model (DTM) generation for site docu­ mentation, volume computation, orthophoto creation, etc. for other applications; • Orthophoto production (e.g. precisely rectified images where scale error and perspective distortion have been removed) and texture mapping.

All these steps can be performed in most commercial photogrammetric software, using both manual and automated procedures. The automated processing of multiple images is the key feature that brings photogrammetry away from the (inefficient) old days of manual point identification and stereo-pair creation and makes it a pragmatic choice for an archaeological project. User interaction is still required for scaling and geo-referencing, and unless the image processing is performed simply to derive a nice-looking 3D model without any metric requirement this step will require the most interaction and effort (Figure 9.5). At each step in the processing chain the choice between manual and automated approaches, and between technically correct and approximating techniques, affects the quality of the final result and the time needed to achieve it. In the case of terrestrial acquisitions, digital cameras must be accurately calibrated, preferably in a controlled laboratory environment, with a 3D testfield and a bundle adjustment solution with additional parameters to fully compensate for systematic errors (Remondino and Fraser 2006). Generally a target-based approach is used in order to automatically identify and precisely measure the match points in the images. When archive images are used (Gruen et al. 2004), camera calibration becomes tricky, requiring assumptions and hypotheses to correctly process the ��������������������������������������� images and derive an accurate 3D model. ������ Once a basic model has been created from the image data, further reconstruction of 3D scenes and model refinement will depend on the project requirements. In the case of complex architectural scenes, man-made objects, detailed city modelling and cartographic applications at large scale, manual or semi-automated approaches are generally the preferred tools for surface measurement and feature extraction, as they remain more precise and reliable (Figures 9.6 and 9.7). On the other hand, small free-form

119

Figure 9.7: Aerial oblique survey over a medieval village in Tuscany to produce a dense and detailed 3D point cloud

120

Fabio Remondino objects, detailed ornaments or small/medium scale DSM can be automatically reconstructed using advanced matching techniques in order to produce dense 3D point clouds and textured polygonal models (Hirschmueller 2008; Remondino et al. 2008; Haala, 2009; Hiep et al. 2009; Gehrke et al. 2010). Methods based on structure-from-motion procedures combine the sensor orientation and 3D reconstruction in a single phase, leaving the scaling and geo-referencing, if required, to the end of the procedure. The scaling and geo-referencing is accomplished with a simple Helmert transformation. While structure-from-motion techniques make for very efficient processing pipelines, this aspect of the technique is less than ideal; the technically correct approach requires the introduction of the datum information (e.g. Ground Control Points) in the orientation phase.

The lidar alternative

Figure 9.8: A burial excavation (left) reconstructed in 3D (centre, right) from a set of terrestrial images for metric documentation and physical replica purposes

Lidar is probably the most significant measure­ ment technology introduced in the topographic mapping field in the last decade (Shan and Toth 2008). The term lidar (Light Detection And Ranging) is a general term used to indicate laser scanning surveying. Laser scanners can record the intensity of the surveyed scene, and are effective even on textureless areas (in contrast to photogrammetry). Since 2000 Airborne Laser Scanning (ALS) and Terrestrial Laser Scanning (TLS) have been used in various applications, with continuous scientific investigations and improvements in

both hardware and software (Opitz this volume). Commercial ALS systems are based on the Time of Flight (ToF) measurement principle with near infrared (NIR) laser light and commercial TLS are based on ToF and triangulation measurement principles with laser light in the range 450–1500 nm. The main advantages of laser scanning is the direct creation of a dense 3D point cloud and, in some cases, the possibility to map structures hidden by vegetation thanks to the multiple return or full waveform digitization. Penetration through the vegetation canopy is a key factor in the wide acceptance of lidar as a tool for landscape archaeology. Ten years ago, as laser scanning was growing in popularity as a means to produce dense point clouds for 3D documentation, mapping and visualization purposes at various scales, photogrammetry could not efficiently deliver results similar to those achieved with lidar instruments. Consequently lidar became the dominant technology in 3D recording in archaeology, and replaced photogrammetry in many application areas. Further, many photogrammetric researchers shifted their research interests to lidar, resulting in further decline in advancements in image-based photogrammetric techniques. Over the past four years improvements in hardware and software, primarily from the computer vision community, have improved photogrammetry-based tools and algorithms to the point that lidar and photogrammetry now can deliver comparable geometrical 3D results for many terrestrial applications (Figure 9.8; Remondino et al. 2008; Pierrot-Deseilligny et al. 2011)

9  Worth a thousand words – Photogrammetry for archaeological 3D surveying

Lidar vs. Photogrammetry As terrestrial scanning and photogrammetry can produce comparable results, the choice between the two depends primarily on project constraints and requirements. Terrestrial laser scanning instruments are still relatively cumbersome and expensive compared to terrestrial digital cameras (off-the-shelf or SLR-type) and their bulkiness might be problematic in some field campaigns or research projects. The point clouds recorded with lidar instruments are already metrically correct, but they are not based on redundant measurements which may be problematic for projects concerned with absolute accuracy. Typical photogrammetric measures derived in the adjustment procedures (variance estimations and other statistical matrices) are not available to evaluate the quality of the lidar point clouds produced; laser scanners provide just a few statistics to describe errors for the entire dataset (normally these numbers are provided by the vendors). TLS errors are normally treated as ‘black boxes’ as they lack well-defined procedures to assess per-project quality, and because internal procedures are hidden, raw data are normally not accessible and calibration or correction procedures are not clear. Photogrammetric processing, while more labour intensive in this case, can be carried out so that calibration procedures, corrections, and error metrics are explicitly stated. Photogrammetric processing algorithms can suffer from problems with the initial image quality (noise, low radiometric quality, shadows, etc.) or certain surface materials (shiny or texture-less objects), resulting in a noisy DSM or difficulties in feature extraction. Furthermore, in order to derive metric 3D results from images a known distance or ground control points are required. On the other hand, for aerial applications photogrammetric approaches can be advantageous in open areas. The typical point density of lidar datasets is around 1–20 points/m2 while an aerial photogrammetric image typically

121

has a GSD on the order of 10 cm and could theoretically be used to produce a DSM with 100 points/m2.

Conclusions The article has described the state-of-the art of photogrammetry for archaeological 3D surveying and modelling. The richness of image content information cannot be matched by any other data acquisition device and examples demonstrate the potential of an image-based methodology. Image-based 3D modelling is a cheap but powerful alternative to 3D scanning. Although experience in image acquisition and 3D modelling is still needed, various open-source and web-based tools are available to create 3D results from images with little expertise required (Figure 9.9). Although these automated processes are rapidly being adopted by archaeological projects, the user should always keep in mind the goal of the 3D reconstruction and be able to distinguish between nice-looking 3D models and accurate metric results, and which is important for any given application. From a research point of view, the great scientific and technological challenges today are no longer related to sensor and system development, but rather to algorithm design and improvement. Indeed, having a large variety of imaging sensors (from satellite to underwater) at our disposal, the obvious question is not the source of data or the processing technique, but the final accuracy of the 3D data that can be derived. So there is a clear future for photogrammetry at least in geo-spatial and precision mapping applications. Photogrammetry is definitively out of the shadow of lidar and is once again an active research area. Of course, the two techniques should not be considered as competing, but as complementary given all the technical differences between them. Despite all the potential of photogrammetry

Figure 9.9: 3D modelling of the lower part of a column for documentation and virtual reconstruction purposes. (left) images of the acquired sequence; (centre) produced shaded 3D model; (right) final textured 3D model

122

Fabio Remondino (and lidar) and the constant pressure of inter­ national heritage organizations, a systematic and targeted use of 3D surveying and modelling in the Cultural Heritage field is still not yet employed as a default approach. Furthermore when a 3D model is produced, it is often subsampled or reduced to a 2D drawing due to a lack of software or knowledge for the proper handling of 3D data by non-experts. It should be clear that the availability and use of 3D metric data opens a wide spectrum of further applications and allows new analyses, studies, interpretations, conservation policies or digital restoration. Thus 3D virtual heritage should be used more frequently, taking advantage of all that remote sensing technologies and the third dimension offer to the heritage world.

References Agarwal, S. Snavely, N., Simon, I., Seitz, S.M., Szeliski, R., 2009. Building Rome in a day. Proc. Int. Conference on Computer Vision. Kyoto: Japan. Dowman, I., Jacobsen, K., Konecny, G., Sandau, R., 2011. High Resolution Optical Satellite Imagery. Whittles Publishing. Furukawa, Y. and Ponce, J., 2010. Accurate, dense and robust multiview stereopsis. IEEE Transactions on Pattern Analysis and Machine Intelligence 32(8), 1362–76. Gehrke, S., Morin, K., Downey, M., Boehrer, N. and Fuchs, T., 2010. Semi-global matching: an alternative to LiDAR for DSM generation? International Archives of Photogrammetry, Remote Sensing and Spatial Information Sciences 38(1), Calgary, Canada (on CD ROM). Gruen, A., Remondino, F. and Zhang, L., 2004. Photo­ grammetric Reconstruction of the Great Buddha of Bamiyan, Afghanistan. The Photogrammetric Record 19(107), 177–99. Guidi, G., Remondino, F., Russo, M., Menna, F., Rizzi, A. and Ercoli, S., 2009. ������������������� A Multi-Resolution methodology for the 3D modeling of large and complex archaeological areas. International Journal of Architectural Computing 7(1), 39–55. Haala, N., 2009. Comeback of digital image matching. Photogrammetric Week, 289–301. Hirschmueller, H., 2008. Stereo processing by semiglobal matching and mutual information. IEEE Transactions on Pattern Analysis and Machine Intelligence 30(2), 328–41. Hiep, V.H., Keriven, R., Labatut, P. and Pons, J.P., 2009. Towards high-resolution large-scale multiview stereo. Proc. Computer Vision and Pattern Recognition. Kyoto: Japan Luhmann, T., Robson, S., Kyle, S. and Hartley, I., 2007. Close Range Photogrammetry: Principles,

Techniques and Applications. Whittles: Dunbeath, UK. Mikhail, E.M., Bethel, J.S. and McGlone, J.C., 2001. Introduction to modern photogrammetry. Wiley. Pierrot-Deseilligny, M., De Luca, L. and Remondino, F., 2011. Automated image-based procedures for accurate artifacts 3D modeling and orthoimage generation. 23rd Int. CIPA (International Scientific Committee for Documentation of Cultural Heritage) Symposium, Prague, Czech Republic (on CDROM). Remondino, F., 2011. Heritage Recording and 3D Modeling with Photogrammetry and 3D Scanning. Remote Sensing 3(6), 1104–38. Remondino, F. and El-Hakim, S., 2006. Image-based 3d modelling: a review. The Photogrammetric Record 21(115), 269–91. Remondino, F. and Fraser, C., 2006. Digital camera calibration methods: considerations and com­ parisons. International Archives of Photogrammetry, Remote Sensing and Spatial Information Sciences 36(5), 266–72. Remondino, F., El-Hakim, S., Gruen, A. and Zhang, L., 2008. Development and performance analysis of image matching for detailed surface reconstruction of heritage objects. IEEE Signal Processing Magazine 25(4), 55–65. Remondino, F., El-Hakim, S., Girardi, S., Rizzi, A., Benedetti, S. and Gonzo, L., 2009. 3D ����������� Virtual reconstruction and visualization of complex architectures – The 3D-ARCH project. International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences 38(5/W1). Sandau, R., 2009. Digital Airborne Camera: Intro­ duction and Technology. Springer. Shan, J. and Toth, C., 2008. Topographic Laser Ranging and Scanning: principles and processing. CRC Press. Snavely, N., Seitz, S. M. and Szeliski, R., 2008. Modeling the world from internet photo collec­ tions. International Journal of Computer Vision 80(2), 189–210. Stamos I., Liu L., Chen C., Woldberg G., Yu, G. and Zokai S., 2008. Integrating ���������������������� automated range registration with multiview geometry for photorealistic modelling of large-scale scenes. International Journal of Computer Vision 78(2–3), 237–60. Vergauwen, M. and Van Gool, L., 2006. Web-Based 3D Reconstruction Service. Machine Vision Applications 17, 411–26. Verhoeven, G., 2010. Taking computer vision aloft – archaeological three-dimensional reconstruction from aerial photographs with PhotoScan. Archaeo­ logical Prospection 18(1), 67–73. DOI: 10.1002/ arp.399. Vosselman, G. and Maas, H.-G., 2010. Airborne and terrestrial laser scanning. CRC Press. Yin, X., Wonka, P. and Razdan, A., 2009. Generating 3d building models from architectural drawings: A survey. IEEE Computer Graphics and Applications 29(1), 20–30.

10 From lidar to LSCM: micro-topographies of archaeological finds Adrian A. Evans, Mhairi L. Maxwell and Gemma L. Cruickshanks Laser Scanning Confocal Microscopy (LSCM) uses similar principles as lidar to reconstruct 3D manipulable topographies of artefact surfaces. This is a relatively new technology which, when employed in archaeological material culture studies, opens up new avenues of enquiry and interpretations. Three case studies are presented to illustrate the interpretative potential of LSCM for understandings of artefact materiality. Importantly this method allows us to delve into incisions, reconstruct texture of wear and uncover residues on an object’s surface. These are traces of past performances of making, use and re-use. The benefits of this type of analysis are not just visual, but experiential, allowing us to build narratives of peoples’ engagement with objects. The first case study explores the use of the technique to study Upper Palaeolithic stone tool function; a second case study explores the biography of an Iron Age antler object from Broxmouth Hillfort, East Lothian, Scotland; and the third case study reviews the study of tool marks on Iron Age worked osseous materials. It is hoped that the implications of employing LSCM to understand materialities discussed in this paper can be translated for applications of lidar to the wider landscape. Issues of scale, biography and identity are highlighted in these case studies. Keywords: Laser Scanning Confocal Microscopy, artefact Biographies, antler objects, bone objects, lithic tools, Quantitative Analysis of Surface Features, tool profiles, use-wear, topographies, materiality, scale of analysis

Introduction While the use of lidar to study sites and landscapes is now routine, this chapter looks at the use of similar technology on a much smaller scale, and consequently to a much higher resolution. Aerial lidar may achieve 60 points per metre and tripod based/bench top scanners record from around 1000 points per metre (1 cm resolution) up to 5000 points per metre (0.2 mm resolution). Laser scanning confocal microscopy (LSCM), described here, can scan 8 million points per metre (120 nm resolution). Thus, this technology allows the investigation of the micro-topography of object surfaces in the way that lidar allows

the investigation of landscape surfaces, opening up a new area of the application of 3D laser scanning technologies for the investigation for object biographies. LSCM allows one to present a visual representation of surfaces that could not normally be resolved and makes the invisible visible in creating representations of the artefact surface in order to reconstruct immediate human experiences with the object over time. Thus, LSCM can provide an immediacy to physical and sensual human engagement with studied objects and has from the beginning been employed to answer questions of the immaterial. At the scale of the artefact surface LSCM opens up the ability to study resource investment and

124

Adrian A. Evans, Mhairi L. Maxwell and Gemma L. Cruickshanks

Figure 10.1: The Olympus LEXT 4000 instrument, an example of a laser scanning confocal microscope (LSCM) designed for materials science (Photograph D. Macdonald)

tool use in manufacture. It can characterize tool and wear marks, and enables an understanding of use-life through wear analysis. These capabilities come with the same benefits found in lidar: it is completely non-destructive, non-contact technology, the results can be used for display in manipulable 3D environments to support analysis or for virtual museum output, or as snapshots at intuitive orientations such as viewing along a cut mark to observe profile and any detail in profile changes (impossible with conventional and electron microscopy due to limited working distance). The capabilities of this new technology for archaeological materials and the interpretative implications of working at this immediate scale are highlighted in three case studies. The first focuses on stone tool function and the advantages of texture measurement with a case study from Wey Manor Farm; a second explores the Iron Age biography of an antler object from Broxmouth Hillfort, East Lothian, Scotland; and the third case study case study explores the study of bone and antler objects from two additional sites. These case studies highlight past human engagements with artefacts and also the relationship of the researcher with the material under study.

The Laser Scanning Confocal Microscope Laser scanning confocal microscopy is a technique that has been in use in biological sciences for imaging for quite some time (see Pawley 1995). Recently versions of this type of microscope have been developed that use similar principles for imaging surfaces for materials science, including the LEXT 4000 (Figure 10.1) used in this study. This microscope has a highly accurate vertical scanning head and horizontal scanning stage, with a highly refined optical system of panchromatic lenses at high numerical aperture which provides resolutions at the limit of optical systems and colour accuracy. The imaging system comprises the common white-light optical system for normal viewing purposes and a laser illumination system in a confocal pinhole arrangement for microscopic 3D modelling. It is this ability to produce 3D modelling of surfaces which has parallels with lidar; it provides essentially the same capability under the microscope and the opportunity to perform optical remote sensing or non-contact surface metrology (Figure 10.2). The confocal microscope forms images by collecting reflected light from a discrete focal plane, in contrast to lidar where all laser light reflected from a surface by Rayleigh scattering is detected. To improve the resolution under the microscope a confocal system is required. This

10  From lidar to LSCM: micro-topographies of archaeological finds discards most light that is not immediately in focus at any given point across the sample surface by using a pinhole aperture that is optically conjugate to the focal plane. The pinhole does not cut out all light that is out of focus and close to the focal plane some out of focus light does return to the detector, but with decreasing luminosity. In this case luminosity is inversely proportional to surface conjugation and this diminishing return of light by depth allows the system to estimate surface location even when it falls between vertical slices. The incident laser light used by the LEXT 4000 system is a laser at 405 nm. The short wavelength light in combination with a rapid response photodiode allows high resolution imaging. The laser light is scanned across the surface using a mirror linked to a servo system. The mirror is an electroformed (galvano) mirror which resonates in the beams path using micro-electromechanical servos. The objective lens is displaced through the vertical axis and slices of optically focussed sections are produced. As the vertical position of each slice recorded is known, the slices can be processed together to create 3D representations of the object or surface data (cloud data). The planar resolution that can be achieved with this system is 120 nm and up to 10 nm vertical resolution. Magnification depends on the objective lens used and, as standard, the highest magnification objective is 100× (0.95 NA), which allows magnifications over 2000× to be achieved. In practice there are a few apparent limitations to the LEXT 4000 confocal system. Most notable is that it does not react well to areas of darkness in the field of view caused by a surface or areas that do not reflect light. This is a common problem in laser imaging systems and the LEXT 4000 attempts to negate this by using dual pinholes with detectors of different sensitivities. This effectively increases the dynamic range and enables the imaging of complicated surfaces of uneven reflectivity. LSCM is highly suitable for the imaging of bone, lithic and metal surfaces. Usual observation of worn tool surfaces is by optical microscopy or scanning electron microscopy (SEM), but these systems have some imaging issues. Optical microscopy suffers from very limited depth of focus which limits appreciation of complex textures, while SEM is cumbersome and requires mounting and often coating of samples. Neither SEM nor optical microscopy can produce 3D models without complicated additional software.

For the SEM, software is available that uses multiple images taken using a tilting eucentric stage to produce useable surface models (e.g. MEX, Alicona), while for the optical microscope a vertical stage motor can be added to collect images taken at different focal depths that can be reconstructed using focus region detection during image analysis. There are commercial systems that provide this, one of which has been used in the imaging of cut-marks and fractures on archaeological bones (Bello et al. 2011). Free systems are also available that can do this type of reconstruction using any microscope (e.g. Helicon Focus, ImageJ plugins) but the quality is dependent on the quality of lenses and accuracy in stage height adjustment (motorised or manual) – which usually is not as high as purpose build precision engineered equipment.

125

Figure 10.2: A schematic illustration showing the operating principles of a LSCM

126

Figure 10.3: Laser intensity images, captured using the LSCM, of worn tool edges produced by different processes. Each pair (upper and lower) represents wear types that are traditionally hard to distinguish. Visualisation of microtopography using confocal microscopy makes this possible

Adrian A. Evans, Mhairi L. Maxwell and Gemma L. Cruickshanks

Case studies The biographical study of archaeological artefacts is important for understanding their social, political and ideological roles in particular contexts over time, since the role and value of objects is not stable (Appadurai 1986; Gosden and Marshall 1999) and are crucial in forming and maintaining social relationships (e.g. Chapman and Gaydarska 2006; Gell 1998; Hurcombe 2007; Thomas 1996; Henare et al. 2007). In the main these studies have looked to ethnography, experimental archaeology, microwear studies and the depositional contexts in which archaeological artefacts are found. As demonstrated here, LSCM can contribute to this endeavour to unpick palimpsests of wear and reconstruct the tools which create the micro-

topographies of artefact surfaces. Furthermore LSCM working at the immediate scale has the potential to reveal the networks of materials (including organics) and people in which the object under study was involved. Stone tools Understanding of function and means of production, principally through wear analysis, are key aspects of stone tool research (Evans and Macdonald 2011; Verges and Olle 2011). LSCM offers the analysis of stone tools the power to move beyond the production of images and surfaces for morphological analysis, into the realm of surface texture investigation. This is important because it represents a crucial step towards improving the currently established

10  From lidar to LSCM: micro-topographies of archaeological finds

127 Figure 10.4: A) Experimental texture data (rms roughness) collected from a set of tools used to work a set of tools against a range of different common materials. B) Overlay of data from the working edge of the Wey Manor Farm burin

approach of functional analysis in lithics, which has relied on subjective comparative microscopy of modern experimentally used specimens with archaeological counterparts (i.e. traceology, lithic use-wear analysis or microwear analysis (Tringham et al. 1974; Keeley 1980)). In these traditional approaches analysis is conducted at a range of magnifications, from stereomicroscopy through to SEM, and can focus on macroscopic edge fracturing, scratches, fine stria, edge rounding and polished surfaces. Features are recorded as the observer interprets them and from this conclusions of use are made through comparison of features on modern experimental tools that have been used in different processes. These approaches have become a serious point of contention amongst analysts, with experimental tests designed to question how effective the traditional approaches are in identifying tool function (Newcomer and Keeley 1979; Odell and Odell-Vereecken 1980; Gendel and Pirnay 1982; Newcomer et al. 1986; Unrath et al. 1986; Shea 1987; van den Dries 1998, 99–112; Rots et al. 2006). These blind tests were experiments where analysts were presented with modern experimental tools that had been used to perform tasks without the analyst’s knowledge in order to assess how accurately the method could be used to identify tool function. On average the method has been able to identify contact material to an accuracy of under 50% and the type of motion tools are applied in is identified correctly 73% of the time (Evans 2009, 101–5). Despite these results, this method is still applied (Hardy and Moncel 2011; Nunziante Cesaro and Lemorini

2012; Donahue and Evans in press); this in itself being a good indicator of how important understanding stone tool use is viewed. However, the application of LSCM has the potential to put these studies on a much more secure footing, relying less on a highly subjective interpretation of a range of wear patterns and markings. The imaging capabilities of LSCM in visualizing surface texture at a high resolution (Figure 10.3) makes visual distinction of wear types far easier than it has been previously using traditional microscopy techniques. The top two images in this figure, for example, represent wear from working antler and wood. These are very similar but fine stria can be seen on the polished surface on the antler working tool, while the polished surface of the wood working tool appears to be flatter and without micro-texture. Identification of such differences increases confidence in making interpretative decisions on the use of archaeological tools. In addition, the metrology capabilities of the system can be harnessed to differentiate worn surfaces with using a quantitative approach. This is enabled through the use of texture analysis to study the 3D surfaces and identify differences in surface polish quantitatively rather than visually. Experimental work using LSCM has illustrated its potential for crystalline silicates such as flint and chert through imaging of surfaces and subsequent study using texture analysis methods derived from engineering (Figure 10.4A; Evans and Donahue 2008). The analytical technique used is the measurement of roughness (as defined by ISO standards) over areas of the surface. An

128

Adrian A. Evans, Mhairi L. Maxwell and Gemma L. Cruickshanks

Figure 10.5: The burin tool from Wey Manor Farm

Figure 10.6: The antler drum with incised ring and dot motifs from Broxmouth Hillfort (finds code GDC). The image on the left shows how it was sawn off from a larger object and secondly the internal wear in the perforation, indication of secondary use as a bead or suspension on a thong (Photograph M. Maxwell)

individual blind-test of this method confirmed the potential capability (Evans and Macdonald 2011) and preliminary research on its application to amorphous silicates such as obsidian may also be useful (Stemp and Chung 2011). The example presented here is from the Upper Palaeolithic site of Wey Manor Farm (Surrey), one of only a few open air Upper Paleolithic sites found in situ in the UK and all the more remarkable for its undisturbed contexts and good state of preservation. The tool in question is typologically a ‘burin’, that is a blade modified to have a robust chisel-like edge (Figure 10.5). ‘Traditional’ analysis suggested that this tool was used to work a hard material such as wood or bone/antler, an interpretation common when the analyst is not confident about the specific type of hard material involved. However, metrology has the most potential when differentiating broad classes of wear identification made using traditional approaches. The data produced from the analysis of the tool’s surface using the LEXT, when presented with the experimental data, clusters in an area that overlaps with four of the six processes tested (see Figure 10.4B).

However, the majority of data corresponds with antler scraping, suggesting that the tool was used to work antler (or bone, which produces a similar texture) rather than wood. The additional interpretative data derived from this analysis is of considerable use. In this case it takes a tool that has an interpretation as being used for a craft activity – working a hard material such as wood or bone/antler – and furthers this to the identification of working osseous material. Wood working and bone working, while both craft activities, represent different types of time investment and produce objects that were treated in different ways, bone objects are far more likely to have been curated owing to the additional time investment required in their production. This enables hypothesis testing/building related to retooling strategies and landscape use. In this particular case there is only one tool discussed here but, for example, this suggests the production of objects out of bone was occurring at this location. This enables reflection on material selection strategies and brings one a step closer to a good understanding of the activities at the location of discard.

10  From lidar to LSCM: micro-topographies of archaeological finds

129 Figure 10.7: LSCM microscopy images of the eight ring and dot motifs incised on the antler drum (numbered 1 to 8 from the far left). A: motif numbers 7 and 8. B: motif numbers 1, 5 and 4. C: motif numbers 2, 3 and 6. These three groups of motifs have been incised using different points

Antler drum, Broxmouth Hillfort The high resolution 3D mapping of visual and textural surface properties of an antler drum bearing eight incised ring and dot motifs (Figure 10.6) from Broxmouth Hillfort, East Lothian (705 cal BC – cal AD 260) (Armit and McKenzie forthcoming) allowed us to examine its specific biography over time. This artefact was an off-cut from a larger handle and subsequent wear within the perforation shows that it was re-used as a bead or suspended on a thong before deposition in association with structures 400–210 cal BC. LSCM allowed for a more detailed understanding of this object’s biography than would have been possible using other techniques by indicating that three different tool profiles had been used to create the ring and dot motifs. It also allowed us to explore the efforts of the craftspeople who made and decorated this object, since scratches and other marks represent evidence of the failure and frustration of attempts to incise. By taking 3D measurements of the incisions it became evident that multiple points, perhaps multiple compass tools, were used to incise the ring and dot motifs of the antler drum (Figure 10.7). Compass tools with centre scribing bits were used to decorate this object, because slight sporadic variations in the width between the central dot and the outer ring, along with the varying depth of the incised outer rings (average range of difference between two measurements was 129 ųm–139 ųm) represent differences in applied pressure. On the other hand, a cylindrical boring tool would be expected to give uniform or relative depth measurements, or a fixed double-pronged trepanning tool or bar would be expected to give relative width measurements

between the central dot and outer ring, taking into account the pressure applied by the human hand. Neither of these are the case here and cannot therefore have been used to decorate the drum. A lack of precision and variability was exactly the reason why scholars have thought compass tools would have been an unlikely choice to create motifs which look concentric and precise to the eye (MacGregor 1985, 75; Tuohy 1995, 57). However, LSCM has shown the adaptability and variability of the tool used to incise the motifs. The width, depth and shape of the central dots of each motif were examined and 3-D reconstructed using LSCM, and the results show that three different tool profiles were used: a rounded bowl point, a steep-sided V-profile and a bevel bottomed point (Figure 10.7). Therefore, it could be that three different compass tools or points attached to the compass were used for decoration (Figure 10.8). Alternatively we are seeing the sharpening of a bowl point into a steep V-profile, and then blunting into a bevel point. Variations in the height and width show blunting or adaptations to inconsistencies of the antler surface whilst incising each motif, but the three tool profiles remain recognisable and form

Figure 10.8: LSCM profiles of central dots, which show three groups of tool profiles (A, B and C, also represented in Figure 10.7) used to incise the ring and dot motifs. This may be the result of sharpening a bowl point (C) into a Vshaped point (B), which blunted to a bevel (A), or alternatively the use of three different points

130

Adrian A. Evans, Mhairi L. Maxwell and Gemma L. Cruickshanks

Figure 10.9: (left) Height coloured model of ring and dot motif number six; (right) LSCM image of central dot 6, showing a secondary dot to the bottom, and two other entries of the compass point

three groups (shown in Figure 10.7). Therefore we are getting an insight into the three-stage planning of the decoration of this artefact, where the three groups of motifs are spaced at regular intervals, and between each stage either a different point/compass tool is used or the same point is sharpened. The detailed mapping of the surfaces of the drum also brought to attention failures in, and re-attempts at incision. In particular, motif six appeared to have been particularly difficult to incise; three points of entry into the central dot are evident (Figure 10.9) corresponding with episodes of slippage during incision of the outer ring. Slippages correspond with natural variations in the antler surface, from very smooth to pock-marked and uneven. There is no evidence for smoothing of the antler tine prior to incision which may have made it easier to work. As such, a considerable investment of skill is manifested in the decoration of this object; the craftsperson, feeling through their compass, had to adapt to changes in the natural antler surface, a task not always easy to achieve (Figures 10.6 and 10.7). The variation in tool profiles enables the consideration of a detailed understanding of artefact history. Figure 10.7 shows the three different combinations of motifs incised in three groups, either indicative of sharpening or of multiple points. It might be tempting to suggest that multiple individuals, with their own personal compass tools, may have been involved in the decoration of this artefact. The ring and dot motif may have been an owner’s ‘stamp’, similar to a trophy won at a football match with the teams name incised. Another possibility is that if we presume a time gap between each

group it may be that the ring and dot motifs were incised during transitional moments over generations or in an individual’s life associated with their coming of age or change in status. Similar interpretations have been argued for Bronze Age spacer necklaces, for example one found in a cist associated with a young woman in Mount Stewart, Bute (Clarke et al. 1985, 289; Shepherd 1985, 213). The Mount Stewart necklace has a replacement toggle which is unworn (Shepherd 1985, 213), many of the beads show varying degrees of wear and some beads are argued to be made from originally larger beads (Clarke et al. 1985, 289). This antler object from Broxmouth was originally part of a larger object and was sawn down to size and subsequently worn. Therefore, it too has evidence for curation and re-use. It was argued that necklaces like this Mount Stewart example were passed down and circulated via group or family members and embellished and repaired over the course of these transitions (ibid.). The motifs incised into the Broxmouth antler object can be thought of in a similar way, as embodiments of a network of enchainment (of human relations or individual achievements). The testing of some of these hypotheses is not possible but previously such things were hard to discuss without the good understanding of variations in engravings that this research method allows, whilst repeated incision over time re-enacted and re-interpreted the materialities of the antler drum and kept it in use. If soaked for working, cast antler ‘bleeds’ a smelly red substance. Red appears to have been a powerful colour in the British Iron Age, linked to ideas of fertility, death and violence (Giles 2007) and it is therefore possible that repeated

10  From lidar to LSCM: micro-topographies of archaeological finds

131 Figure 10.10: (left) Underside of pin head or gaming piece from Uamh an Eich Bhric showing cut marks; (right) Fragment of antler tine from Uamh an Ard Achaidh with cut mark (Photograph G. Cruickshanks)

incising of this object was an important act of purification, bringing it back to life. The employment of LSCM in this case has greatly added to our capability to discuss and understand the planned manufacture and biography of this antler drum. It allows for the consideration that ring and dot motifs were not haphazardly decorative, but were planned and executed with skill. It can be suggested that what has often been considered to be a mundane artefact was, in fact, maintained and embellished by more than one person through incising ring and dot motifs, thus indicating its social value, each re-working performed to mark moments of change. Furthermore the discovery of a compass tool is important as there is only one example of an actual compass tool of Iron Age date in the archaeological record from Fairy Knowe, Stirlingshire (Main 1998). Objects from the Uamh an Eich Bhric and Uamh an Ard Achaidh LSCM of worked antler has highlighted some of the potentials and limitations of the technique for the high resolution examination of toolmarks. This case study examines artefacts from two Iron Age sites on the Isle of Skye. Uamh an Eich Bhric (Cave of the Speckled Horse) is a rock shelter with extensive eroding middens and evidence of craft activities dated to between 50 BC and 250 AD (Wildgoose and Birch 2010, 8) and Uamh an Ard Achaidh (High Pasture Cave) is a late Bronze Age/Iron Age votive and feasting site (Birch and Wildgroose in prep). The well-preserved tool-marks on bone and antler provide a proxy-record for the tools being used by Iron Age people, where the actual tools

are often poorly preserved or entirely absent from the archaeological record (Cruickshanks in prep.). In order to distinguish different toolmarks examination under high magnification is required, and to date, attempts to identify different tool types and materials have primarily used SEM of a silicone cast of the tool-mark (i.e. creating a positive impression of a negative feature (e.g. Olsen 1988; Greenfields 2002; Cristiani and Alhaique 2005). It is not ideal to make casts of osseous surfaces, especially fragile objects where this process can destroy the surface including any tool marks. Such a procedure is relatively crude and LSCM allows this to be avoided. As a result it is also quicker, offering much more powerful means for visualising the surface topography and opening up new possibilities for analysis. A pin head or pegged gaming piece from Uamh an Eich Bhric (Figure 10.10 left) displays a variety of tool-marks from manufacturing and modifications through its life. Analysis focussed on two cuts across the base. Their purpose is unclear and they may represent later modification of the object. Initial visual analysis suggested the cut marks were likely to have been made by the same blade due to their similar width (c.1.5 mm), although one was considerably deeper. However, the LSCM images showed different profiles for the two cut marks (Figure 10.11). The deeper cut (cut B) has a square section with striations along the edges, which is typical of a metal saw (MacGregor 1985, 55), while the shallower cut has an asymmetric profile with slightly sloped sides and base. The shallower cut (cut A) is more difficult to classify, being somewhere between a saw profile and the typical V-shaped section of

132 Figure 10.11: LSCM images of cut marks from the pin head/gaming piece, showing flat base and steep sides with striations typical of a cut made with a saw

Figure 10.12: Symmetrical, steep sided V-shaped cut mark on the antler tine, typical of a very sharp metal blade

Adrian A. Evans, Mhairi L. Maxwell and Gemma L. Cruickshanks

10  From lidar to LSCM: micro-topographies of archaeological finds a knife or axe cut. It was possibly created by the same saw as the deeper cut, but slightly angled which would cause the asymmetry; perhaps a reason for abandoning the shallower cut in favour of the second, straighter and deeper attempt. The piece examined from Uamh an Ard Achaidh is an antler tine which has been hollowed out, probably for use as a handle, and has a deep cut parallel to the end (Figure 10.10 right). It was recovered from a context dating to around 600–500 cal. BC (Birch and Wildgroose in prep). The LSCM image showed the profile to be symmetrical and V-shaped with very steep, smooth sides (Figure 10.12). This is typical of a sharp metal blade, the smooth sides suggest a chopping action from a heavy blade such as an axe or cleaver, rather than a sawing or slicing action from a slighter blade which would create striations along the edges (Walker and Long 1977, 608). The base of the cut has a hump in the middle indicating that the blade edge may have been concave. This can occur from sharpening a blade and poses the intriguing possibility that this technique could be used to match distinctive cut marks to a specific blade. Although the LSCM image clearly shows that this was made by a metal tool, more comparative work is required to conclude whether this was a copper alloy or iron tool, a particularly interesting question given the early Iron Age date. A limitation of LSCM for the examination of tool-marks is the scale of the topography. LSCM in the case of the Olympus LEXT 4000 is limited to the study of surfaces where topography varies less than 1 cm in depth. As a result, only a section through part of the cut mark on the antler tine from Uamh an Ard Achaidh could be scanned as the depth of the cut along with the curve of the bone proved too deep to scan the entire surface. It should however be noted that this can be negated by the use of two scans at different depths and post-production manipulation.

Discussion and conclusion These three case studies have introduced some potential applications for LSCM as a technique for addressing questions of materiality. LSCM has great potential for the analysis of tool-marks on ivory, antler and other materials (potentially organics) and seems particularly suited to shallow, micro-topography. Analysis of the burin from Wey Manor Farm enabled an understanding

of the tool’s use and helps to build a picture of the activities which took place at this site in the Mesolithic. The examination of the Broxmouth drum revealed use of a specific tool set and the extent of investment for the decoration of the object, highlighting the multiple hands that may have been involved. High-resolution examination of Iron Age worked bone and antler artefacts from the Isle of Skye clearly showed the use of metal blades (a saw and a heavy chopping blade) providing rare glimpses of tools which are rarely found in the archaeological record. Examining object surfaces at high resolution adds a different dimension to the biography of objects, shedding light on objects that no longer exist and illuminating networks of social relations and individual engagements over time. In the case of the Broxmouth drum, it demonstrates the existence of tool technology that previously might have been considered rare in the British Isles. There is only one example of an Iron Age compass tool in the UK (Main 1998) but these results indicate a much wider circulation of such specialist items. It shows the planning in manufacture and the potential curation of an everyday antler object with ubiquitous ring and dot motif (this motif is found on many types of objects, combs, torcs, shields and mounts of bone and metal throughout the Iron Age in Europe). On the Isle of Skye further tool types are revealed that are incredibly rare in the archaeological record, perhaps through envaluation the metal tools were of high value and therefore recycled, rarely making it into the archaeological record (Taylor 1999). The discovery of these tools situates the artefacts under study within a complex assemblage of materials indicative of enchained social networks (Hurcombe 2007). The Wey Manor Farm burin analysis brings to light the past use of the tool and, with the analysis of other tools at the site, places it within a social context of landscape use and personal organization (the spatial patterning of activities within living spaces). These results are a direct challenge to traditional methods of interpretation that have been the subject of heated methodological debate in material culture studies, offering an approach where qualitative and quantitative approaches can be applied with a degree of consistency at a finegrained immediate scale to contribute to understandings of society in the past. For example, further work comparing LSCM images of tool-marks from copper alloy and iron tools could provide a new

133

134

Adrian A. Evans, Mhairi L. Maxwell and Gemma L. Cruickshanks insight into the introduction and development of iron tools, enabling the identification of individual workshops or tooling based on the identification of characteristic wear marks. In conclusion, by working at the immediate scale to analyse textures, tool-marks and residues, LSCM has opened up new avenues of exploration into the materiality and social lives of artefacts. The immaterial becomes material at this scale of analysis.

Acknowledgements We would like to thank Professor Ian Armit and Dr Jo McKenzie, with whose help the study of this antler drum has emerged as a distracting side project! Adrian Evans was funded by the AHRC (Grant No. AH/J007935/1). The AHRC (Grant No. ���������������������������������������� AH/I505741/1) �������������������������� and Historic Scotland are funding M Maxwell’s PhD research�������� . Gemma ������ Cruickshanks’ PhD is funded by the AHRC (Grant No. AH/I021418/1). Thanks to Steven Birch and Martin Wildgoose for permission to use objects from the High Pastures Cave and Environs and Fiskavaig Projects, and National Museums Scotland for access to the objects, and to Rob Poulton, Surrey County Council for permission to use objects from the Wey Manor Farm assemblage. Dan Bashford and Lindsey Büster must also be thanked for providing helpful comments on an earlier draft of the Broxmouth antler drum case study.

References Appadurai, A., 1986. The Social Life of Things: Commodities in Cultural Perspective. Cambridge: Cambridge University Press. Bello, S.M., Parfitt, S.A. and Stringer, C.B., 2011. Earliest Directly-Dated Human Skull-Cups. Plos One 6(3): e17026. Birch, S. and Wildgroose, M., in prep. High Pastures Cave and Environs Project, Srath, Isle of Skye 2010. Broadford Environmental Group, Data Structure Report. Archaeological and Ancient Landscapes Survey. Chapman, J. and Gaydarska, B., 2006. Parts and Wholes: Fragmentation in Prehistoric Context. Oxford: Oxbow Books. Clarke, D.V., Foxon, A. and Cowie, T.G., 1985. Symbols of Power: At the Time of Stonehenge. Edinburgh, Her Majesty’s Stationery Office. Cristiani, E. and Alhaique, F., 2005. Flint Vs. Metal: The Manufacture of Bone Tools at the Eneolithic

Site of Conelle Di Arcevia (Central Italy). In Luik, H., Choyke, A., Batey, C. and Lõugas, L., (eds). From Hooves to Horns, from Molluscs to Mammoth: Manufacture and Use of Bone Artefacts from Prehistoric Times to the Present. Tallinn: Trükitud Tallinna Taamatutrükikojas. 397–403. Cruickshanks, G., in prep. Iron in Iron Age Scotland: A Regional Study of Production and Use of Iron from c.800 BC to AD 800. University of Edinburgh/ National Museums Scotland, PhD Thesis. Donahue, R.E. and Evans, A.A., in press. Microwear Analysis. In Boismier, W.A., Gamble, C. and Coward, F., (eds). Neanderthals among Mammoths: Excavations at Lynford Quarry, Norfolk. London: English Heritage. Evans, A.A., 2009. Microwear Analysis and the Role of the Microlith in Mesolithic Britain. PhD, University of Bradford. Evans, A.A. and Donahue, R.E., 2008. Laser Scanning Confocal Microscopy: A Potential Technique for the Study of Lithic Microwear. Journal of Archaeological Science 35, 2223–30. Evans, A.A. and Macdonald, D., 2011. Using Metrology in Early Prehistoric Stone Tool Research: Further Work and a Brief Instrument Comparison. Scanning 33(5), 294–303. Gell, A., 1998. Art and Agency: An Anthropological Theory. Oxford: Clarendon Press. Gendel, P.A. and Pirnay, L., 1982. Microwear Analysis of Experimental Stone Tools : Further Test Results. Studia Praehistorica Belgica 2: 251–65. Gosden, C. and Marshall, Y., 1999. Cultural Biography of Objects. World Archaeology 31(2). Greenfields, H., 2002. Distinguishing Metal (Steel and Low-Tin Bronze) from Stone (Flint and Obsidian) Tool Cut Marks on Bone: An Experimental Approach. In Mathieu, J. (ed.). Experimental Archaeology: Replicating Past Objects, Behaviours and Processes. Oxford, BAR International Series 1035, 35–54. Hardy, B.L. and Moncel, M.-H., 2011. ������������ Neanderthal Use of Fish, Mammals, Birds, Starchy Plants and Wood 125–250,000 Years Ago. Plos One 6(8), e23768. Henare, A., Holbraad, M. and Wastell, S., (eds) 2006. Thinking through things: theorising artefacts ethnographically. London: Routledge. Hurcombe, L., 2007. A Sense of Materials and Sensory Perception in Concepts of Materiality. World Archaeology 39(4), 532–45. Keeley, L.H., 1980. Experimental Determination of Stone Tool Uses. Chicago, Chicago University Press. Macgregor, A., 1985. Bone, Antler, Ivory and Horn: The Technology of Skeletal Materials since the Roman Period. London: Croom Helm. Main, L., 1998. Excavation of a Timber RoundHouse and Broch at the Fairy Knowe, Buchlyvie, Stirlingshire, 1975–8. Proceedings of the Society of the Antiquaries of Scotland 128, 293–417. Newcomer, M., Grace, R. and Unger-Hamilton, R., 1986. Investigating Microwear Polishes with

10  From lidar to LSCM: micro-topographies of archaeological finds Blind Tests. Journal of Archaeological Science 13(3): 203–17. Newcomer, M.H. and Keeley, L.H., 1979. Testing a Method of Microwear Analysis with Experimental Flint Tools. In Hayden, B., (ed.). Lithic Use-Wear Analysis. New York: Academic Press. Nunziante Cesaro, S. and Lemorini, C., 2012. The Function of Prehistoric Lithic Tools: A Combined Study of Use-Wear Analysis and Ftir Microspectroscopy. Spectrochimica acta. Part A, Molecular and biomolecular spectroscopy 86, 299–304. Odell, G.H. and Odell-Vereecken, F., 1980. ���������� Verifying the Reliability of Lithic Use-Wear ��������������� Assessments by “Blind Tests”: The Low Power Approach. Journal of Field Archaeology 7, 87–120. Olsen, S., 1988. The Identification of Stone and Metal Tool Marks on Bone Artifacts. In Olsen, S., (ed.). Scanning Electron Microscopy in Archaeology. Oxford, BAR International Series 452, 337–60. Pawley, J., (ed.) 1995. Handbook of Biological Confocal Microscopy. New York: Springer. Rots, V., Pirnay, L., Pirson, P. and Baudoux, O., 2006. Blind Tests Shed Light on Possibilities and Limitations for Identifying Stone Tool Prehension and Hafting. Journal of Archaeological Science 33(7), 935–52. Shea, J.J. 1987. On Accuracy and Relevance in Lithic Use-Wear Analysis. Lithic Technology 16, 44–50. Shepherd, I.A.G. 1985. Jet and Amber. In Clarke, D.V., Foxon, A. and Cowie, T.G., (eds). Symbols of Power: At the Time of Stonehenge. Edinburgh: HMSO. 204–28. Stemp, W.J. and Chung, S., 2011. Discrimination of Surface Wear on Obsidian Tools Using Lscm

and Rela: Pilot Study Results (Area-Scale Analysis of Obsidian Tool Surfaces). Scanning 33(5), 279–93. Taylor, T., 1999. Envaluing Metal: theorizing the Eneolithic ‘hiatus’. In Pollard, M., Budd, S.Y. and Rixer, P., (eds). Metals in Antiquity. Oxford: Archaeopress. Thomas, J. 1996. Time, Culture and Identity. An Interpretative Archaeology. London: Routledge. Tringham, R., Cooper, G., Odell, G., Voytek, B. and Whitman, A., 1974. Experimentation in the Formation of Edge Damage: A New Approach to Lithic Analysis. Journal of Field Archaeology 1: 171–96. Tuohy, C., 1995. Prehistoric Combs of Antler and Bone. University of Exeter, PhD thesis. Unrath, G., Owen, L., Van Gijn, A., Moss, E.H., Plisson, H. and Vaughan, P., 1986. An Evaluation of Use-Wear Studies: A Multi-Analyst Approach. Early Man News 9/10/11, 117–76. Van Den Dries, M.H., 1998. Archaeology and the Application of Artificial Intelligence. Leiden: Leiden ������� University. Verges, J.M. and Olle, A., 2011. Technical Microwear and Residues in Identifying Bipolar Knapping on an Anvil: Experimental Data. Journal of Archaeological Science 38(5): 1016–25. Walker, P.L. and Long, J.C., 1977. An Experimental Study of the Morphological Characteristics of Tool Marks. American Antiquity 42, 605–16. Wildgoose, M. and Birch, S. 2010. Uamh and Eich Bhric (Fiskavaig 1): Excavation of an Eroding Midden Site. West Coast Archaeological Services, Data Structure Report. Archaeological and Ancient Landscapes Survey.

135

11 Using lidar data – drawing on 10 year’s experience at English Heritage Simon Crutchley This chapter looks at the different ways that lidar data can be visualised and used by different user groups depending on the level of technology available to them and the intended purpose. The wide variety of formats available are discussed along with how these impact on the ease of recognition and interpretation of features. The majority of the examples illustrate the ways that English Heritage has used the data either directly or with partner organisations in desktop and field based work, ranging from using processed data in a pseudo 3D environment to paper printouts and including standard hillshaded jpegs, ASCII data and raster surfaces. The use of data by a number of volunteer organisations and other groups in England in desktop analysis is also discussed. Keywords: ��������������� visualisation, lidar �������������������� applications, lidar ����������� data formats, ��������� community ������������������ groups, desk-based ����������� assessment, field survey

Introduction Since its introduction to the archaeological community the use of lidar data has grown exponentially, and it is now seen as a standard survey tool that is often considered at the outset of project proposals. Within English Heritage (EH) in the United Kingdom (UK) the number of projects containing a lidar element has grown considerably since the first lidar related project on Stonehenge was commissioned in 2001. However, examination of project proposals received by EH, and external discussions, suggest that use of lidar is often proposed in inappropriate situations or with no clear understanding of how the data can be visualised and the necessary time and skills required. This is often due to unrealistic expectations of what lidar can reveal. As the technique has proliferated the variety of imagery and data products available has increased (Devereux et al. 2005, 2008) and the number and variety of both amateur and professional users has grown enormously. While development of visualisation and maximising information return is largely driven by professional and academic

users, there has also been an increase in the number of end users, including volunteer groups, who see lidar as an excellent means for getting to grips with archaeology in their area, especially in wooded locations. Whilst there are undoubted advantages in the growth of visualisation techniques to assist interpretation (Crutchley and Crow 2009; Kokalj et al. this volume), the abundance of approaches may confuse some end-users and raises the question of cost/benefit for those without access to expensive software or hardware, or who have little experience with GIS. Simply put, if readily available products generate 90% of the information produced by other techniques, is it worth spending a lot of time and energy to get that extra 10%? Although there has been some assessment of the range of lidar sources and visualisation techniques (Kokalj et al. 2010, 2011), no one has attempted to look at the full range of lidar data and how it is being used by different users. Specifically this concerns how to best match applications to the requirements of the end user and the division between desk based assessment

11  Using lidar data – drawing on 10 year’s experience at English Heritage Disadvantages

Advantages Image file

x x x x x

Data

x Flexibility x Potential to process data in a variety of ways and extract more information

Simple to view Easy to distribute Small file size Print out for use in the field Less restrictive copyright arrangements

(DBA) and field checking. This is central to appropriate use of lidar – if users do not have the proper tools for viewing or manipulating the data, then they will not get the full benefit, and misuse may lead to misleading and erroneously derived archaeological information.

Using images or data files The first key decision for users is whether to use image files created by someone else or to directly use the data to create a variable visualisation on a computer screen. Often the choice will be dependent on availability and cost, and how the data will be used as a field source or in a DBA. It is also dictated in part by how confident those intending to use the data are with data and image manipulation. There are a growing number of software packages, including many that are free or open-source, available to download from the internet, but unless the end-user has a relatively good grasp of GIS and the concepts of image and data analysis they may be unwilling to use them. Furthermore, in many cases the endusers are groups who want to be able to use the information they derive from the data in the field and so need hard copy. It is worth noting that the line between images and data is not always as clear as it may at first seem and both have advantages and disadvantages (Table 11.1). While to most people who do not have experience with GIS the word raster denotes an image file, rather than a file containing data, this is not strictly the case. A raster is technically a grid of data. For a raster image the value of each cell is represented by a colour, for a raster surface the value is a figure

x Inflexibility x An image is fixed and shows a specific element

x x x x x

Not easy to distribute Requires specialist viewing software Time consuming analysis? Generally large file sizes More restrictive copyright arrangements

that relates to something measurable in the real world e.g. a height value or a degree of slope. The issue is further complicated by certain types of geo-tiffs that appear to be image files, but which retain the height value in a form that is simply represented by a colour. The image The simplest and most readily accessible form of lidar data is the jpeg image (or similar format). These depict the data in various ways, but unlike the ‘data’ they do not contain any intrinsic information about height (or intensity or slope etc.). Each pixel merely contains the RGB code that makes it a certain colour. However, they are very easily viewed on any PC, may be loaded on many handheld devices such as GPS loggers and can be imported into most GIS packages. If they are to be used for mapping in a digital environment accurate georeferencing is important. Any image can also be provided as a paper print for those situations where use of computers is less feasible. Image files are created using a variety of methods that translate the data into greyscale shades or colour tones and so potentially contain much more information than is immediately apparent. The main proprietary image editing packages (e.g. Adobe Photoshop or Corel Draw) provide image editing tools such as level equalisation or colour replacement, which can be used to manipulate the data at a relatively basic level and enhance subtle details and changes. Similar tools are available in some free software (e.g. Gimp). Data Much more useful and flexible than images

137 Table 11.1: Advantages and disadvantages of image and data compared

138

Simon Crutchley are the data, containing height (and in some cases intensity) values that can be manipulated and analysed given the appropriate software. End-users can create multiple images to help pick out specific details, or they can view the data interactively in a pseudo-3D environment. Whilst this can improve confidence in any interpretations made it can also be time consuming, and this raises the question of how much more information can be gained over working with simple images, and whether it is worth the extra time and effort?

The English Heritage experience When EH staff started using lidar data they were in the position that many prospective users find themselves in; they did not have access to hardware or software to effectively process the data themselves and therefore had to rely on imagery provided by a third party. Gradually over time, working through a number of projects, they have acquired the hardware and software to process the lidar data themselves, and have also gained an improved understanding of how it can be used to best effect. There is little doubt that the way in which the data were originally provided and the problems these caused has influenced the way that EH have continued to work with the data. The nature of the organisation has probably also had an effect. The Aerial Survey & Investigation team at EH (and its predecessor the Royal Commission on the Historical Monuments of England – RCHME) has over 50 years’ experience of mapping archaeological features and landscapes from aerial photographs and other aerial imagery. Since about 1998 they have worked in a largely digital environment using successive versions of the AutoDesk CAD drawing programme as their primary means of mapping. This allows rectified imagery to be imported from which relevant archaeological features can be transcribed with various data attached to allow the creation of a rudimentary GIS. The fact that large numbers of staff need to produce a consistent product has meant that the team have developed practices to utilise different aspects of lidar data as the data available has changed, but the fundamental aim has been to integrate the lidar data into a structured flowline with minimal impact on existing working practices. Undoubtedly, this in part reflects the needs of a large organisation

trying to maintain common standards and whilst it may seem bizarre to users used to working in bespoke GIS and mapping packages, it suits the context in which the lidar is used. Today, in many cases, there is ready access to the actual height data, and from this it is possible to produce different visualisations as discussed in further detail below; however, in the beginning staff had to rely very much on the data that was provided, which originally was imagery. Using simple imagery The lidar derived image files can be coded according to height, as greyscale (e.g. low = dark to high = light), or using a colour range such as blue to red. However, such height profiling/ contour profiling does not provide much that is particularly useful for the archaeologist on its own. The use of hillshading, using a virtual light source to produce shadows and highlights similar to the effects of low level (oblique) sunlight, is much more effective at revealing features of interest. To maintain consistency, a common standard is to illuminate from the northwest, the standard light source for many illustrations. A combination of hillshade with underlying colour coded height shading is a common product. Such is the case for the UK Environment Agency lidar derived images, which were used by EH at an early stage (Holden et al. 2002). The Environment Agency has flown nearly 70% of the UK over the 15 years (at least 2 m resolution, with a large proportion covered at 1m or higher), concentrating on low-lying land and especially that which is liable to flooding. Within the UK their image tiles are available through the Geomatics Group for a small cost, but in some areas copies are held by the local government authorities who may make them available for viewing. The digital tiles are preferred as these can be enhanced in Photoshop or similar packages to allow the extraction of considerably more data than might otherwise seem to be contained in the image (Figure 11.1). If the data and appropriate software are available it is possible to use differing light sources set up to highlight the varying features according to the optimum location and produce images for staff in the field or in DBA. This could be done at the outset of a project by one specialist to provide data for others to use. In these cases the image can be viewed in the same way as a vertical aerial photograph with very low oblique lighting and features can be interpreted based on

11  Using lidar data – drawing on 10 year’s experience at English Heritage shadows and highlights. The interpreter must understand the direction of the light source, since this information is vital to differentiate between features built up above the surrounding ground surface (i.e. mounds and banks) and those that are cut into it (i.e. pits and ditches). The first project where EH used lidar data (at Stonehenge) used single hillshaded images of this sort. The project was carried out using data captured in 2001, but only had access to image files created by Colin Shell at Cambridge University (Bewley et al. 2005) since staff did not have the capacity at the time to deal with the data. Initially five images, each covering the entire WHS, were created using lighting from different directions. Because each image was only lit from a single direction it ran the risk, inherent in any single lit image, of missing features that lie parallel to the light source. This is a problem familiar to archaeologists using conventional aerial photographs lit by oblique sunlight as features that run perpendicular to the sun can be identified easily by their shadows, but those that lie along the line of illumination cast no shadow and are rendered virtually invisible. It is thus necessary to view a number of different images from different angles to be sure that no features are missed. A similar situation prevailed for the next project in the Witham Valley (Lincolnshire). Here EH had access to the raw data and had the hardware and software to process it and create imagery, but they were still inexperienced in what could be learned from the data. Furthermore, the data could only be provided to those undertaking the mapping as imagery. So, although the project suffered many of the same problems as at Stonehenge, because they had the data it was possible to produce hillshaded imagery (albeit single illumination images), and also colour profiles, which proved useful in the visualisation of some of the heavily ploughdamaged barrows that had almost no surface expression. Having access to the appropriate software also enabled more detailed analysis of the underlying topography that helped with the understanding of the landscape (Crutchley 2006). Where the data has been available, later projects have used a combination of image files and active data as discussed below, but EH have continued to use the Environment Agency lidar derived imagery as a standard resource wherever it is available.

These are colour coded according to height on a standard scale and are hillshaded with illumination from the northwest. While these do not use height exaggeration and the colour scale is biased to the lower end of the scale because it is designed primarily for low-lying land liable to flooding, basic image manipulation (i.e. level equalisation or colour replacement) brings out features not immediately evident. Because they are georeferenced it is simple to import tiles into AutoDesk and where available they are now used as a standard resource in any National Mapping Programme project, alongside conventional photographs (Horne 2011; e.g. Marden Henge Environs in Wiltshire (Carpenter and Winton 2011) and the Hoo Peninsula in Kent). The key problem with single illumination images is the risk of missing features that lie parallel to the light source, but there are various ways to get around this, the easiest of which is to create images that combine more than one light source. The simplest are created as composite images by using overlays of different colours with different degrees of opacity and more complex composite images can be created using algorithms such as Principal Component Analysis (PCA; Devereux et al. 2008). While there are clearly advantages with such techniques, primarily the reduction of the risk of missing features that lie parallel to the light source, the use of multiple lighting angles can lead to uncertainty over whether features are positive or negative since two adjacent images may present different forms of shadow and highlight due to different illumination. This can be offset by using contour coloured imagery at the same time to show which features are higher than others. Whilst EH have the capacity to create PCAs they have not used them in a major mapping project because they have preferred to work with the active data. However, they have been involved in projects where PCA imagery has been used, but this has primarily been in paper format. In the field Paper copies have proved very useful in the Forest of Dean project (Gloucestershire). Here the lidar data was initially used as a single hillshaded image, but in later stages of the project a composite image illuminated from four different directions was produced by the Forest Research branch of the Forestry Commission. These were taken into the field and used alongside a proforma recording sheet to identify features on

139

140

Simon Crutchley the ground and interpret them. Various types of printout (e.g. at different scales) were used and their advantages and disadvantages recorded as part of the survey report (Hoyle 2005, Appendix L). This preliminary analysis identified over 1000 potentially significant features, none of which had been identified by earlier archaeological research. Within woodland these included extensive areas of charcoal burning platforms and open cast mineral workings, many of which are likely to date to the medieval or early post-medieval periods (Hoyle 2005). In the next phase of the project a sample of the features were examined in detail on the ground and the multiple hillshades were replaced by single hillshaded images of the feature illuminated from both the northwest and northeast, generally produced on A4 sheets. They also used these single illumination images as georeferenced jpeg images on a data logger with a small screen. Monochrome images lit from one direction only were preferred to the polychrome images lit from between four and eight directions as these proved difficult to comprehend on the small (7.8 × 5.9 cm) screen. The paper proforma was also replaced by direct digital entry to a database on the logger (Hoyle 2011, 2.4.1–2). A similar methodology was used by the Weald Forest Ridge Historic Environment Awareness (HEA) Project in East Sussex. Here survey groups were provided with printouts, usually at 1:3000 scale which provided a balance between being able to recognise quite small features, and a pixelated image. At this scale 1 km² easily fits on an A3 sheet and so gives a good overview of an area. The lidar images used were generally four colour composite hillshaded images created by Peter Crow at Forest Research. While there was a pre-survey DBA which had identified possible new features as points, lines or polygons, the results of this were not given to the community groups. This was done to avoid the risk that groups would just target the features shown on the lidar transcription instead of looking at the whole landscape and recording everything they found. This was of particular concern since previous experience in the Forest of Dean had shown that small features, such as sawpits, do not always shows clearly on the lidar imagery and the survey design wished to avoid this bias. Lidar images have also been used in the field as paper or film copy, or on handheld tablets or GPS tablets within EH. Initial tests found some issues with screen resolution, but otherwise it was found

to be a very useful tool to quickly and accurately locate features, which in some cases were difficult to see on the ground. More recently, lidar has been a key element in the Miner Farmer project in the North Pennines, both in office and in the field (Oakey et al. 2011; Ainsworth et al. this volume). This project has used the data in a variety of ways, further developing it as a powerful means of landscape interpretation and mapping. Desk based assessments Where possible, EH has preferred to take advantage of the capacity to use active data, exploited for the first time in the Mendip Hills AONB in 2007 2009. geo-tiffs were initially created with a standard light source, but it soon became obvious that different topographic areas required different sources in order to highlight features to best effect (e.g. north and south facing slopes require different light sources as a northwest light source will leave the majority of a south-facing slope in shadow). Therefore, further geo-tiffs were created according to where they showed most features to their best advantage, but because they were single illumination images they risked missing features. To resolve this, interactive data files were created in QT Modeler at the same time. This program is designed for viewing and analysis of lidar data and the creation of interactive files that can be viewed via a freely downloadable viewer. This allowed all staff to view and map from the image files imported into CAD, whilst at the same time having access to active data and the ability to manipulate the light source in real time. When they saw features in the interactive data for which they did not have a relevantly lit tile from which to map, they requested another image with lighting from that angle. The project recorded a large number of previously unknown sites and added additional data to known sites in a significant enhancement to what was visible on traditional aerial photographs (Truscoe 2008; Firth and Truscoe 2011). While the combination of imagery and data had proven to be quite successful in Mendip it was recognised that greater efficiency might be gained if it was possible to map from the actual data. This highlights the difference between visualisation on the one hand and the processes of mapping and interpretation on the other. Most programs available for viewing lidar and other surface datasets are designed predominantly to create impressive visual experiences and models

11  Using lidar data – drawing on 10 year’s experience at English Heritage (e.g. fly-through) and not to facilitate accurate mapping and recording of visible features. Thus, whilst there has been an increase in the number of programs that do allow mapping direct from the data, the emphasis on visualisation remains and EH has maintained a pragmatic approach that has changed as improved software has become available. The use of lidar data in EH is essentially in plan form, and although there are certain limitations inherent in this approach it works very well with established flowlines developed using conventional aerial photographs as the primary mapping resource within a CAD environment. Advances in the AutoDesk suite have seen the introduction of the capacity to read raster surfaces directly into the mapping environment (i.e. AutoDesk Map), so in the most recent projects the raster surface can be manipulated to highlight features, whilst at the same time remaining correctly georeferenced to allow accurate mapping (Crutchley 2008; Crutchley et al. 2010; Oakey et al. 2011). This ability to assimilate lidar in an existing flowline is important to a large organisation, as it has eased effective integration with minimum disruption and maximum benefit. The latest versions of AutoDesk allow the creation not only of hillshades based on the surface, which EH makes extensive use of, but also slope and aspect models (which have yet to be applied). A combined approach has been adopted using QT reader together with AutoDesk. Whereas previously QT was used to view the data interactively, but mapping was carried out against a fixed image, in the new versions of AutoDesk hillshading can be applied directly to the surface that is used for mapping and interpretation. The rapid response of QT, operating in real-time, is useful to aid analysis and so runs alongside. A similar approach could presumably be used in any other software that allows vector mapping (i.e. GlobalMapper or ArcGIS). Whilst there may seem to be disadvantages to using two programmes EH have found the benefits of the complementary approach outweigh any drawbacks. Other techniques as yet unutilised within EH The discussion above has dwelt on EH experience, as an example of a large national organisation dealing with progressively more available lidar data for a variety of mapping projects. The developing approach has created a ‘best fit’ to

the requirements of the organisation that will continue to grow. It is recognised, of course, that there are a range of possibilities beyond that realised in practice at EH and aspects of these are discussed here, in particular visualisation and analytical techniques. The simplest of the visualisation techniques have a long history of use in GIS and exploit different properties of the surface model (i.e. slope or aspect analysis). These, and other visualisation techniques such as Sky View Factor (Kokalj et al. 2010, 2011) or Local Relief Models (Hesse 2010), have their strengths and weaknesses, and while some are discussed in greater detail in other papers in this volume, a brief summary of potential benefits and pitfalls is presented here (Table 11.2). These techniques are essentially now standard approaches to visualisation, but there are two other innovative techniques that may be beneficial. The first is controlled animation, which exploits the abilities of the human eye and brain at picking out change rather than extracting detail from still images. Thus a series of different hillshaded images are cycled in a loop and as the sun apparently moves around the landscape casting shadows various features appear and disappear. This ‘animation’ presents interesting possibilities to interpretation, but a problem in mapping, because mapping on a moving image presents technical and probably user challenges. It is possible to run the animation on a loop in ArcGIS, but it is still necessary to pause the loop to map. Finally, and probably the most-underused way of visualising is to view the data stereoscopically taking advantage of the mind’s ability to see three dimensions as a spatial, not just a tonal, feature. This technique has been used since the earliest days of aerial photography and should be the basis of interpretation from aerial imagery for archaeological purposes, but has thus far had little application with lidar data. 3D viewers are available, but most are expensive, though the increasing use of 3D viewing in the entertainment market, however, means that these can be expected to drop in price. In the short term, however, options such as the creation of anaglyphs are more readily achievable and it is also possible to produce physical prints of the stereo pairs that can be viewed in the same way as aerial photographs using manual stereoscopic viewers. With suitable software it should be possible to map truly in 3D whilst interpreting the stereo images on the screen as

141

142

Simon Crutchley Technique

Advantages

Single hillshaded image

Easy to produce; no need for expensive software

Multiple hillshade composite image Principal Components Analysis (PCA) hillshade Slope and aspect

Easy to produce; no need for expensive software

Combination composites (Slope and hillshade) Sky View Factor (SVF)

Local Relief Model (LRM) Animation Stereoscopic viewing

Table 11.2: Summary of potential benefits and pitfalls of visualisation techniques

A single image that captures most of the key variations in the data that may be of interest Relatively easy to produce; no need for expensive software; Detects some features missed by hillshaded imagery Relatively easy to produce; no need for expensive software; Detects some features missed by hillshaded imagery Relatively easy to produce; no need for expensive software Detects some features missed by hillshaded imagery Detects some features missed by hillshaded imagery Easy to produce; no need for expensive software Detects some features missed by hillshaded imagery Relatively easy to produce; no need for expensive software if using anaglyphs Stereoscopic viewing is a readily known technique for those used to viewing aerial photos

has been the norm for digital photogrammetry for many years. The different techniques described here ultimately rely on the use of a raster surface as this is the only format from which surfaces (e.g. with hillshading, slope) can easily be generated (Figure 11.1). However, there are some viewers that allow the original point clouds to be visualised. In most cases this is very much inferior to the use of surfaces, but there are some situations where it is beneficial to view the points themselves, especially in exceptionally dense undergrowth. There are several free viewers available (e.g. FugroView, Fusion) and there will no doubt be more viewers coming out as a large section of the lidar community, outside the heritage industry, is recognising the importance of viewing point data as its primary source (Opitz this volume; Paul Burrows pers. comm.). This is particularly true where colourised clouds can be produced either using simultaneous image capture or linking to digital imagery.

Disadvantages Risk of missing features parallel to the light source Areas with variable topography will require a number of different hillshading to maximize benefit for each facet Multiple images with degrees of transparency mean subtle features can be masked Not as easy to interpret as a single image; more complex and time consuming to create; Difficult to interpret without experience because features do not appear as expected Areas of variable topography require different combinations of hillshades to maximise benefit Difficult to interpret as features do not appear as expected Appears to work better with “hard” surfaces than subtle earthworks Complex processing required to create it Difficult for those unskilled in GIS to understand the effects of parameters Because it is moving it is impossible to map from Only available for Desk Based Applications Only currently available for Desk Based Applications Full 3D visualisation using polarising glasses and/or screens are expensive

Conclusions The large amounts of lidar data that are becoming available, often at a national coverage, can benefit archaeologists greatly for rapid identification survey, where a standard visualisation technique using image files is likely to be appropriate, and for detailed assessment and management survey for specific sites, where the advantages of using interactive data for in depth interpretation are apparent. Lidar is proving accessible to a variety of users. For those with neither technical expertise nor appropriate hardware or software, the data can be accessed through standard image files. If these can be provided as a bespoke service, in a format that brings out the specific user requirement, they can be extremely useful. Case studies from the UK show that using image files in the field is an efficient and practical tool for archaeological field survey, whether undertaken by inexperienced volunteers or experienced field workers. Where users are confident in data manipulation there are a growing number of programs and packages certainly providing a wide

11  Using lidar data – drawing on 10 year’s experience at English Heritage

variety of options for visualising and analysing the data. The EH experience thus far suggests that hillshades are most easily understood and provide a high proportion of the ‘total’ information that might be extracted. However detailed comparison of multi-sensor survey results illustrates the importance of complementary data sources (see Bennett et al. this volume), and further analysis of the relative benefits of different visualisation techniques for lidar over a range of landscapes to discover what percentage of information each approach generates will be useful in establishing standard approaches (e.g. for comparison of ‘information return’ from aerial photographs and lidar see Crutchley et al. 2010). Such analysis is vital to quantifying costbenefit for varying approaches across different landscapes, as there are good reasons to believe that certain techniques will be more efficient in generally flat areas, whereas others will be of particular benefit in rough terrain. For example,

features such as charcoal burning platforms that are cut into the slope may not appear readily in hillshaded images, while slope analysis and local relief models have proven very successful, though the former has the great advantage of being easy to perform. Fitting specific project aims, landscape type and the character of the expected archaeological features to appropriate techniques is absolutely vital. And for most archaeologists, professional and amateur, the key to successful use of lidar is easy access to straightforward tools for visualisation and the skills and experience to interpret the archaeological features, rather than having to learn unfamiliar computer techniques to process the data. As with other sources, such as aerial photographs, the quality of the archaeological products (e.g. mapping, database records, and detailed interpretation) from lidar will depend on the experience of the user, the availability of appropriate tools and some understanding

143

Figure 11.1: Comparison of some key viewing techniques at West Woods near Avebury, Wiltshire. Clockwise from top left the visualisations are colour height shaded with single light source; PCA; LRM and slope. Lidar source Environment Agency (May 2006)

144

Simon Crutchley of the source data. The EH experience has demonstrated that this can be managed across a variety of workers, from full-time highly experienced fieldworkers (see Ainsworth et al. this volume) to part time volunteers with variable experience. Such approaches produce archaeological information of varying types, from monuments that are reliably characterised and mapped, to those that require other information, such as that derived from ground-observation. A final point for consideration concerns what constitutes ‘good quality lidar data’, a potentially complex issue influenced by resolution (i.e. number of points) and multiple processing and recording specifications (e.g. Doneus and Briese 2006, 2011). For the relatively simple matter of resolution comparison of the data from the Witham Valley with that for Mendip suggests that 1m data collection can be expected to record the vast majority of features that characterise the archaeology of the UK. Depending on the platform used, aerial lidar data can be captured at much higher resolutions than this (theoretically up to 100 points per metre) and the greatest density so far actually used for an archaeological project is 60 points per metre at the Hill of Tara, Ireland (Corns and Shaw 2009; Corns et al. 2008). These examples are specialist applications, and EH’s experience suggests that 1m data is sufficient to record most features. The clear exception, however, is in woodland where greater density is required to improve canopy penetration. This concurs with the experience in Baden-Württemberg, Germany where DTM data at a grid width of 1m and elevation accuracy of ± 0.15 m was judged sufficient in most circumstances for the identification and visualisation of archaeological structures (Bofinger and Hesse 2011). Such perspectives are vital to establish cost-effective applications for lidar in the archaeological community and require explicit judgements on cost-benefit based on understandings of what lidar offers, how it may be analysed and visualised and the objectives of the projects to which this data source is being applied.

References Bewley, R.H., Crutchley, S.P., and Shell, C., 2005. New light on an ancient landscape: lidar survey in the Stonehenge World Heritage Site. Antiquity 79, 636–47. Bofinger, J. and Hesse, R., 2011. As far as the laser can reach…: Laminar analysis of LiDAR

detected structures as a powerful instrument for archaeological heritage management in Baden-Württemberg, Germany. In Cowley, D. (ed.). Remote Sensing for Archaeological Heritage Management. EAC Occasional Paper No. 5 Brussels 2011 Carpenter, E. and Winton, H., 2011 Marden Henge and Environs, Vale of Pewsey, Wiltshire: A report for the National Mapping Programme. English Heritage Research Department Report Series No. 76-2011 Corns, A., Fenwick, J. and Shaw, R. 2008 ‘More than meets the eye’. Archaeology Ireland 22, 3 (85), 34–8 Corns, A. and Shaw, R. 2009. High resolution 3dimensional documentation of archaeological monuments & landscapes using airborne LiDAR. In Lasponara, R. and Masini, N., (eds). ICT and Remote sensing for Cultural Resource Management and Documentation. Journal of Cultural Heritage, Volume 10, Supplement 1, December 2009, 72–7, DOI: 10.1016/j.culher.2009.09.003 Crutchley, S.P., 2006. Lidar in the Witham Valley, Lincolnshire: an assessment of new remote sensing techniques. Archaeological Prospection 13, 251–7. Crutchley, S., 2008. Ancient and modern: combining different remote sensing techniques to interpret historic landscapes. In Lasaponara, R., and Masini, N., (eds). Remote Sensing for Archaeology and Cultural Heritage Management; Proceedings of the 1st International EARSeL Workshop on ‘Remote Sensing for Archaeology and Cultural Heritage Management’. Rome: ARACNE, 103–6. Crutchley, S., Small, F., and Bowden, M., 2010. Savernake Forest: Report for the National Mapping Programme. English Heritage Research Department Report Series No. 29–2009. Crutchley, S., and Crow, P., 2009. The Light Fantastic: Using airborne lidar in archaeological survey. Swindon: English Heritage. Devereux, B.J., Amable, G.S., Crow, P., and Cliff, A.D., 2005. The potential of airborne lidar for the detection of archaeological features under woodland canopies. Antiquity 79, 648–60. Devereux, B.J., Amable, G.S., and Crow, P., 2008. Visualisation of lidar terrain models for archaeological feature detection. Antiquity 82(316), 470–9. Doneus, M. and Briese, C., 2006. Full-waveform airborne laser scanning as a tool for archaeological reconnaissance. In Campana, S. and Forte, M. (eds). From Space to Place: 2nd International Conference on Remote Sensing in Archaeology. BAR International Ser 1568, 99–105. Doneus, M. and Briese, C., 2011. Airborne Laser Scanning in Forested Areas – Potential and Limitations of an Archaeological Prospection Technique. In Cowley, D. (ed.). Remote Sensing for Archaeological Heritage Management. EAC Occasional Paper No. 5 Brussels 2011, 59–76. Firth, H., and Truscoe, K., 2011. The Aggregate Landscape of Somerset: Predicting the Archaeo­

11  Using lidar data – drawing on 10 year’s experience at English Heritage logical Resource. In Lewis, J. (ed.). Archaeology of Mendip: 500,000 Years of Change and Continuity. Heritage Marketing & Publishing: King’s Lynn. 387–420. Hesse, R., 2010. LiDAR-derived Local Relief Models (LRM) – a new tool for archaeological prospection. Archaeological Prospection 17, 67–72. doi: 10.1002/ arp.374. Holden, N., Horne, P. and Bewley, R.H., 2002. High-resolution digital airborne mapping and archaeology. In Bewley, R. and Raczkowski, W. (eds). Aerial Archaeology: Developing Future Practice. 173–80. Horne, P., 2011. The English Heritage National Mapping Programme In Cowley, D. (ed.). Remote Sensing for Archaeological Heritage Management. EAC Occasional Paper No. 5 Brussels 2011, 143–52. Hoyle, J.P., 2005. The Forest of Dean, Gloucestershire, Archaeological Survey, English Heritage Project Number 2727, Stage 2: Pilot Field Survey, Unpublished circulation draft for English Heritage, Gloucestershire County Council Archaeology Service April 2005 Hoyle, J.P., 2007. The Forest of Dean, Gloucestershire, Lidar survey of selected areas of woodland and the Aggregates Resource Area. The Forest of Dean Archaeological Survey Stage 3A. Unpublished draft report for English Heritage. Hoyle, J.P., 2011. Forest of Dean Archaeological Survey Stage 3B: Survey for management of lidar-detected

earthworks in Forestry Commission woodland (Project Number 5291 SURV). Phase 1: Rapid field validation and scoping analysis for characterisation of archaeology in woodland – Project Report Volume 1: Methodology, Results, and Discussion. Unpublished report for English Heritage. Kokalj, Z., Zaksek, K. and Ostir, K., 2011. Application of sky-view factor for the visualisation of historic landscape features in lidar-derived relief models. Antiquity 85(327), 263–73. Kokalj, Z., Zaksek, K. and Ostir, K., 2010. Archaeological Application of an Advanced Visualisation Technique Based on Diffuse Illumination. In Reuter, R. (ed.) Proceedings of the 30th EARSeL Symposium: Remote Sensing for Science, Education Natural and Cultural Heritage, (http://www.earsel.org/?target=publications/ proceedings/symposium-2010),113–9. Oakey, M., Radford, S. and Knight, D., 2011. Alston Moor, North Pennines Area of Outstanding Natural Beauty: Aerial Investigation and Mapping Summary Report. English Heritage Research Department Report Series No. 4/2012 Truscoe, K., 2008. Archaeological aerial survey in the Central Mendip Hills. The Aggregate landscape of Somerset: predicting the archaeological resource. Aggregates Levy Sustainability Fund project 3994, unpublished report: Somerset County Council/ English Heritage. http://www.english-heritage.org. uk/publications/aggregate-landscape-of-somersetcentral-mendip/

145

12 Lidar and World Heritage Sites in Ireland: why was such a rich data source gathered, how is it being utilised, and what lessons have been learned? Anthony Corns and Robert Shaw This paper explores the capture and use of lidar data for a selection of inscribed and tentative World Heritage Sites in Ireland. The detailed specifications for these projects are presented and the impact of these variables on the subsequent use of the data is discussed. The paper examines how different commissioning agencies have managed their lidar resources and to what extent they have enabled access to others and encouraged the wider use of the data. Some unexpected outcomes from the lidar surveys, both positive and negative, are explained, and the extent to which lidar projects have fulfilled expectations assessed. The paper also looks at how the experiences gained from working on these initial projects will be used to help define future projects. Finally alternative approaches to generating high resolution relief shaded models that may challenge the use of lidar in the future are considered. Keywords: Ireland, landscape, lidar, World Heritage Site, DSM, DTM, relief shaded model

Introduction Archaeological and built heritage applications of lidar surveying began to emerge in Europe a little over ten years ago, following its widespread use in other scientific fields such as environmental monitoring and topographic survey. One of the earliest examples was at Stonehenge, the World Heritage Site (WHS) in Wiltshire, England (Bewley et al. 2005; Crutchley this volume). Here English Heritage, already aware of the enormous benefits that high resolution digital surface models could deliver to management and research issues, had the foresight to liaise with the Environment Agency (UK) and commission a lidar survey of the Stonehenge landscape. This survey, approximately 40km 2 centred on Stonehenge itself, proved an invaluable resource in addressing complex and controversial management issues (such as access roads and visitor facilities) and became a core resource in the development of a management plan for the WHS. Following a detailed examination of the

resulting lidar relief shade models the survey also lead to the discovery of several new archaeological features and provided further insight and detail into those monuments already present in the record. The first archaeological application of lidar in Ireland was at Loughcrew, Co. Meath, a spectacular landscape with the largest concentration of megalithic monuments in central Ireland. An area of 30 km2 was surveyed using the Cambridge University Unit for Landscape Modelling’s airborne lidar. This was a Heritage Council funded project conducted by University of Cambridge and University College Cork (Shell and Roughley 2004). The results were, again, a spectacular endorsement of lidar as a landscape modelling technique, with the generated relief shade models revealing significant new finds, enhanced detail on known monuments and resolving significant positional issues within the Register of Monuments and Places.

147

12  Lidar and World Heritage Sites in Ireland The survey generated significant publicity within Irish archaeology through publications and seminars, highlighting the potential archaeological value of lidar for prospection and management purposes. Loughcrew was quickly followed by a number of lidar projects, largely supported by the Heritage Council, which aimed to explore further issues such as optimum resolution, processing and visualisation techniques, all based on specifically commissioned data sets. Whether by design or good fortune the result of this proliferation is that Ireland now has lidar coverage for some of the most important cultural heritage sites in the country. This presents the opportunity to examine some of the issues about why the surveys were undertaken, how the data is being utilized and how the experiences gained from these projects may influence future lidar ventures.

UNESCO World Heritage List, Ireland The UNESCO World Heritage List (whc. unesco.org/en/list) includes 936 cultural and natural heritage properties which are considered by the World Heritage Committee as having outstanding universal value. Currently only two sites are listed for the Republic of Ireland: the archaeological complex of the Bend of the Boyne (Brú na Boinne, see Davis et al. this volume) and Skellig Michael, both on the basis of cultural heritage. In addition to this, a Tentative List of sites for nomination to the World Heritage List was completed and presented to UNESCO in 2010 (Figure 12.1), including the Hill of Tara and Dún Ailinne – both royal sites of ancient Ireland. Taking the two listed sites and two tentative sites together presents an opportunity to consider the application of lidar in very different environments, with different management, conservation and research challenges. They also present an opportunity to consider the different roles state agencies, local authorities, universities and the wider research community play in the commissioning, processing and utilisation of the lidar resource.

Lidar project specification The lidar specifications for the four projects concerned vary in their detail (Table 12.1). At Brú na Boinne the primary objective was

to capture high resolution 3D survey data for the WHS, and in the process investigate the capability of lidar to define known archaeological monuments and reveal previously unknown monuments. The specification was informed by the Stonehenge lidar project with further advice from the Discovery Programme. Tiled data was supplied as a Digital Surface Model (DSM), and a filtered Digital Terrain Model (DTM), with orthorectified imagery – geo-referenced and coregistered with the lidar tiles – with a ground resolution of 25 cm. The Discovery Programme was given this data and using some basic ArcGIS functions generated relief shaded models of both the DSM (Figure 12.2) and DTM. These, along with the orthoimages, were distributed to Meath County Council as both GIS datasets and publication images and posters which provided the basis for the interrogation and analysis of the landscape.

Figure 12.1: World Heritage Sites and tentative sites in Ireland. Underlined names identify where lidar has been specifically captured for heritage applications

148 Table 12.1: Project information and specification for the four lidar projects

Anthony Corns and Robert Shaw Status

Brú na Boinne World Heritage Site

Dún Ailinne WHS Tentative List

Hill of Tara WHS Tentative List

Skellig Michael World Heritage Site

Funding

Heritage Council

Heritage Council

Heritage Council

NMS

i

Client

Meath County Council

Discovery Programme

Discovery Programme

NMS

i

Data Capture

Environment Agency UK

BKS Fugro

BKS Fugro

BKS Fugro

Sensor

Optech ALTM 3100

Flimap 400

Flimap 400

Flimap 400

Platform

Fixed wing aircraft

Helicopter

Helicopter

Helicopter

Acquisition

Nov 2007

Feb 2007

Dec 2007

2007

1m

25cm (18cm hilltop)

12.5cm

12.5cm

GSD Area

ii

2

96km

2

2.4km

2

2.38km

2

0.3km

i

National Monuments Service (NMS) of the Dept. Of Environment, Heritage and Local Government, (Heritage was move to the Dept of Arts, Heritage and Gaeltacht (DAHG) in 2011).

ii

Ground sampling distance – spacing between data points on the ground.

Figure 12.2: Brú na Boinne complete reliefshaded DSM, with an enlargement around Newgrange passage tomb

The survey of the large hilltop enclosure Dún Ailinne was undertaken to enable an assessment of the value of a higher resolution lidar system. The effectiveness of a variable Ground Sampling Distance (GSD) was evaluated while also gathering valuable data to support the research

being undertaken on the site by the University of Washington (Johnson 2009). The absence of a highly accurate and high resolution survey plan, and the problems associated with mapping features obscured by dense gorse vegetation suggested lidar as a potential solution. GIS

12  Lidar and World Heritage Sites in Ireland

149 Figure 12.3: Comparison between the relief-shaded DSM and DTM models for Dún Ailinne hilltop enclosure

Figure 12.4: Reliefshaded DSM of Hill of Tara with enlargement to illustrate the increased model resolution than can be achieved with the Flimap system

modelling was undertaken by the Discovery Programme, and the remarkable resolution of the Flimap system, and the quality of DTM after vegetation filtration was immediately apparent (Figure 12.3). Following the success of the Flimap survey at Dún Ailinne a similar specification lidar survey was commissioned for one of the iconic sites of Ireland, the Hill of Tara. This complex archaeological landscape, one of Ireland’s principle royal sites, includes the earthwork remains of many features including a hilltop enclosure, cursus monument and many ring barrows. In some cases these are reflected as micro topographic features difficult to discern with the naked eye, or have only been discovered through geophysical survey. The lidar project was commissioned for an area encompassing the land in the care of the state (the Office of Public Works – OPW) to add to the research resource and examine the benefits of very high resolution terrain data. GIS modelling by the Discovery Programme resulted in a DSM and DTM of extremely high resolution (Figure 12.4). In the case of Skellig Michael the archaeology consists primarily of upstanding stone-built monastic structures in contrast to the earthwork features of the previous sites. The remote nature of the island and the extreme topographic conditions present many challenges to conventional survey. There is also an archaeological challenge, with

150

Anthony Corns and Robert Shaw

Figure 12.5: Clockwise (from top left): Aerial view of the dramatic landscape of Skellig Michael (Photo: Con Brogan); The monastic settlement at the north of the island, with the dramatic south peak in the distance (Photo: Con Brogan); Enlarged plan view of the skyview shaded DSM of the monastery; Enlarged plan view of the skyview shaded DSM of the south peak; Plan view of the Sky View Factor DSM of the whole island

most features being stone built structures, many still standing to a significant extent (Figure 12.5). The primary objective was to provide an accurate baseline survey for the whole island as a resource for ongoing conservation and research. Only a DSM data set was provided as the output due to the absence of significant vegetation and the fact that upstanding walls and structures

constituted the core archaeological monuments. The Discovery Programme was commissioned to generate the GIS layers and map outputs to facilitate the data interrogation. To maximise the visibility of archaeological features a number of different processing algorithms were applied (Kokalj et al. 2010, this volume) with sky-view factor proving most effective (Figure 12.5).

12  Lidar and World Heritage Sites in Ireland The derived GIS layers, maps and plans from all four of these lidar projects have been available to the commissioning organisations for at least two years now and have been integrated into research frameworks and management plans.

Working with the lidar data Having established the reasons for investing in lidar data capture for the four projects it is appropriate to look now at how the data has been used in practice and consider whether and to what extent it has met expectations. Brú na Boinne This is perhaps the most successful and far reaching lidar project undertaken in Ireland to date. The primary stakeholder of this project, Meath County Council, points to a number of anticipated and unexpected benefits and applications from the investment in lidar data. In most cases this has been in collaboration with other state agencies and research programmes, a result of a positive attitude to data sharing which will be discussed later. Examples include:• The data set was made available to the Archaeological Survey of Ireland where a systematic examination of the lidar data has found evidence of new monuments. To date 27 new monuments have been confirmed and added to the Record of Monuments and Places, and a further 71 have been identified as requiring further investigation to confirm their archaeological potential. This work alone confirmed one of the research objectives demonstrating that lidar is indeed a powerful tool in the study of historic landscapes. • The Boyne Valley Landscape Project funded by Irish National Strategic Archaeological Research (INSTAR) Programme from 2008–10 also made extensive use of the lidar data (Lewis et al. 2008, 2009). The aim of this project was to produce an integrated, comprehensive landscape archaeological model of the evolution of the Boyne catchment and so develop an environmentally contextualised understanding of a key element of Ireland’s archaeological heritage. The lidar data not only fulfilled a primary data function but was also integral to significant components of environment research, and as the final report on the project describes ‘lidar…together with other existing geomorphological and archaeological datasets,

151 allow a robust, integrated archaeological and geomorphological GIS to be developed, which has the potential to significantly advance alluvial geoarchaeology in Ireland.’   In a further phase of the Boyne Valley Landscape Project the lidar data was again systematically analysed to identify potential new sites, with over 100 possible sites noted for further investigation (Davis et al. 2010, this volume). • The Heritage Council used the lidar data extensively in the Brú na Boinne Research Framework Project. The publication of a research framework is considered best practice by UNESCO and it is significant that the lidar survey was embedded in the final publication as a key resource (Smyth 2009). • Finally, and significantly, the lidar data forms part of Meath County Council’s GIS dataset expanding the user base beyond archaeology and heritage into everyday issues of management and planning.

Dún Ailinne This lidar survey was the first archaeological application of a high resolution helicopter system in Europe, being flown in 2007. The primary purpose was to consider the value of generating surface models at such high resolutions, and examine whether the micro-topographic archaeological components of the landscape are present in the models. New discoveries such as access routes to the hill and details of the quarry pits were some of the immediate observations. The landscape also presented an excellent environment to test vegetation filtration processes (Figure 12.3). The results achieved by the Discovery Programme were reported to the Heritage Council and published in a number of conference papers and journals (Corns and Shaw 2009) and went on to inform future project planning and specification. The data has been made available to academic researchers, notably the ongoing research project of the University of Washington (Johnson 2009), presenting the opportunity to combine high resolution surface models with intensive geophysical survey data. Unfortunately the lidar relief models were not received in time to be included in the published research of this phase of research, but it remains a spectacular resource awaiting further interpretation and application which will advance the understanding of this important landscape.

152

Anthony Corns and Robert Shaw

Figure 12.6: Top, an area of the Hill of Tara relief shaded DTM identified as having archaeological potential. Bottom, the result of the magnetic gradiometry survey that followed, revealing a possible medieval settlement

Hill of Tara The lidar data for the Hill of Tara has been extensively used in research since it was processed into relief shaded surface models. Core to these activities has been the Discovery Programme which has an on-going research interest at the site and it has actively sought to maximise the potential of the data. The Department of Archaeology, NUI Galway, have been a long term partner of the Discovery Programme and they undertook a detailed study of the lidar

data, comparing it with previous topographic surveys and geophysical evidence and revealed a number of significant new finds (Corns et al. 2008). Beyond archaeological prospection the lidar models were used in 2010 by the Discovery Programme and NUI Galway to identify zones of potential for a further campaign of geophysical survey. This resulted in a major new discovery with a possible medieval component to the settlement evident (Figure 12.6). The Office of Public Works (OPW) is

12  Lidar and World Heritage Sites in Ireland responsible for the management of the Hill of Tara and has accessed the GIS data and high resolution prints of the relief models. An important function of the OPW is to liaise with the local population, who lease grazing rights on the state owned lands and others who farm in the surrounding landscape. When discussing management issues they have found the lidar images to be far more engaging to non-technical people than plans or maps. The lidar data sets are considered a central resource for the Consultative Group on day to day Management issues at the Hill of Tara, hosted by the National Monuments Service (Department of Arts, Heritage and Gaeltacht – DAHG). The Discovery Programme has also been commissioned to prepare a Conservation Plan for the Hill of Tara, and it is envisaged that the lidar relief models will be a core mapping resource in this document. Skellig Michael Skellig Michael is owned by the Minister for the Arts, Heritage and the Gaeltacht on behalf of the Irish people, with the OPW having responsibility for management of the site. It has been the subject of long-term conservation and stabilisation programmes since it was taken into state guardianship in 1880, processes which have intensified and been subject to greater scrutiny since 1996 when the site was inscribed on the World Heritage list. Insufficient detail in base mapping, even when supplemented by a large scale (1:1000) photogrammetry project, was the driving force behind the decision to commission the lidar survey (see Skellig Michael World Heritage Site: Management Plan 2008–18). The lidar survey has been used by the archaeologists of the National Monuments Service (NMS) of the DAHG to further both the research and management agenda. Examples include:• The extraction of detailed and accurate plans of the main archaeological complexes. These have been used as a base map to accurately position old survey plans and drawings. • Detailed sections generated from the DSM ‘on demand’ have proved an extremely valuable additional resource, identifying structural components of the buildings and associated terraces, tracks and steps (Figure 12.7). • The DSM has been scrutinised in great detail and a number of new discoveries made, such as previously unidentified continuations of steps, and improved definition of many existing archaeological features. The ability to scrutinise

153 the model in the GIS environment cannot be over-emphasised as in many cases access on the ground is impossible or highly dangerous. • Management of the site has benefited from access to the lidar survey, with the models being used to identify and map areas requiring routine maintenance. This information, stored as GIS data, will establish those areas most at threat from erosion caused by tourists and enable action to be taken to reduce their impact.

Managing, accessing and sharing data The Brú na Bóinne project has been by far the most successful at enabling access to a wider range of users. The lidar data is administered by the commissioning agency, Meath County Council, with the Discovery Programme responsible for GIS preparation and management. Meath County Council have an open attitude to sharing data and facilitating access for other organisations, as can be seen from the variety of organisations that have accessed and worked with the lidar survey to date (Table 12.2). This policy of open access has been informed by the County Meath Heritage Plan 2007–11 adopted in 2007. The following three actions extracted from the heritage plan illustrate the principles which have shaped this enlightened policy:• Establish mutual links with third-level institutes to encourage heritage research and activity in Meath. • Utilise Information Communication Technology (ICT) to record, preserve and present heritage data. • To work in partnership with the relevant agencies/organisations and the public to promote, understand, conserve and manage the Brú na Bóinne UNESCO World Heritage Site.

The practical impact of this policy is that the council will facilitate requests for data on the understanding that full acknowledgement is made and that any further reports or papers produced are made available to Meath County Council. For the other projects significantly less access has been made by third parties. In the case of Dún Ailinne the data is managed by the Discovery Programme on behalf of the Heritage Council. Few requests have been received for access to GIS data, but a number of requests to include generated images in publications have been

154

Anthony Corns and Robert Shaw

Figure 12.7: Two crosssections extracted from the DSM, with a skyview shaded DSM showing their location

agreed, including a forthcoming study of the Barrow Valley by Dr Annaba Kilfeather. The Discovery Programme also manages the data for the Hill of Tara. As a high profile heritage site with major issues of conservation and management there has been an active encouragement of accessing the data. The OPW, NMS, and Meath County Council have all been given access to GIS layers, and high resolution large format prints produced on request. Beyond these management agencies the Department of Archaeology, NUI Galway, has been given GIS data and prints for teaching and advanced research. The GIS data has also been supplied to M.A. students for use in diverse applications such as developing digital humanities research tools. The lidar data for Skellig Michael is managed by the National Monuments Service (NMS) with extensive support from the Discovery Programme, made necessary due to the limited GIS resources at the NMS. The Discovery Programme has supplied printed maps and plans and applied a range of processing algorithms to generate relief shaded DSM products. These are now being used by archaeologists at the NMS in GIS. Currently the NMS have not opened up access to this lidar data set due to some issues which will be discussed in the next section. However, in principle they believe that the data should be shared with other institutions. All of the above is essentially done on an ad hoc basis. The awareness of data is often the result of informal contact and conversations, requests are made by email or phone calls, and data distributed by DVD or hard drives. Spatial Data Infrastructure (SDI), an area of development being actively pursued in Irish archaeology

(Corns and Shaw 2010), and through EU initiatives such as the ArchaeoLandscapes Europe project, should provide the solution. The establishment of well archived data with detailed metadata made available through geoportals would radically improve the ability to discovery, access and use lidar data sets.

Unforeseen outcomes Although the outcomes of the lidar surveys (Table 12.3) are somewhat diverse, they are largely positive with some common themes. Brú na Boinne and Hill of Tara demonstrate the ease with which the public can engage with and understand maps and plans based on the relief shaded DSM. This was particularly useful in public consultation, planning and outreach events. This ease of interpretation also resulted in larger than anticipated requests for the use of lidar models as a base map for illustration in publications (e.g. Stout 2010; Meehan 2011). The archaeological application of the high resolution (flimap) lidar generated significant publicity for the technique itself, particularly at Hill of Tara. The data provider (BKS Fugro) was so impressed by the exceptional quality of the data and modelling that they used the models in a number of articles, presentation and even advertising material. This highlighted the potential they saw in the future of lidar as an archaeological and heritage management tool and resulted in a proliferation of similar surveys in Northern Ireland. From an archaeological perspective perhaps the most significant remark was from the Brú

12  Lidar and World Heritage Sites in Ireland Heritage Council Meath County Council National Monument Service Office of Public Works The Discovery Programme University College Dublin Dundalk Institute of Technology Individual Researchers General Public

Project Brú na Boinne

Dún Ailinne Hill of Tara Skellig Michael

155

State Body – role is to propose policies and priorities for the identification, protection, preservation and enhancement of the national heritage Local Authority Government –Dept of Arts, Heritage and Gaeltacht Government – management of state properties Archaeological research company (state funded) Third level education Third level education Public Local landowners

Positive Quality of research Volume of research Public engagement Publication illustration Remote modelling Flimap publicity Public engagement Flimap publicity Scientific: other applications Scientific: positional accuracy

na Boinne where the volume and quality of research based on lidar had not been anticipated. In particular the number of new discoveries has exceeded expectations given the past intensive studies of the area. An inherent advantage of all aerial approaches to landscape mapping was noted at Dún Ailinne where private ownership of the site severely limits public access. The detailed terrain model generated remotely from the lidar survey will reduce the need for researchers to spend time on the ground in the future. Other general scientific benefits arose in diverse ways with data used in disciplines beyond the core archaeological applications. At the Hill of Tara it has been used in the management of hedgerows and natural habitat studies as part of the Consultative Groups activities, whilst at Skellig Michael the underlying geodetic information measured has been used to resolve gross positional errors in the base mapping. Some negative experiences were reported, notably problems modelling the data. Meath County Council initially attempted to use MapInfo Vertical Mapper to process the Brú na Boinne data but struggled to generate grid files due to the size of the data set. This was resolved by tiling the data into more manageable units and building raster files in ArcGIS, but exposed

Negative Processing difficulties

Minimal use

Processing difficulties Quality issues related to terrain

a skills deficit which was also encountered by other clients e.g. NMS when working with the Skellig Michael data. Only the exceptionally steep and rugged terrain of Skellig Michael presented more survey problems than had been anticipated. The survey had to be flown twice as the first attempt contained a large data gap at the South Peak – one of the major zones of archaeological interest but also one of the most inaccessible parts of the island. Interpolation errors also occurred during the initial processing of the raw lidar data that only became apparent in some areas of the model with the creation of the DEM surface. These appeared as an interference pattern in the model, but when viewed in profile looked as if they were associated with extreme slope angles or perhaps overhanging cliffs. In other areas of the model some unusual issues occurred on some of the modern features which were accounted for as reflectance from standing water. Issues of this nature had not been encountered in previous projects where the remarkably clean data was a notable feature. These issues are still being investigated but have not had a significant impact on the usefulness of the lidar survey. Finally, at Dún Ailinne a surprising outcome was noted given the overwhelmingly positive

Table 12.2: Organizations that have accessed the Brú na Bóinne lidar survey data

Table 12.3: Summary of the unexpected outcomes from lidar surveys

156

Anthony Corns and Robert Shaw experiences with other projects. Only minimal use has so far been made of the data for advanced archaeological research. This can largely be attributed to the fact that the lidar survey was primarily an exercise to prove the quality of the technique, rather than in support of a wideranging management or research agenda.

Future directions, lessons learned To the commissioning organisations and funders these projects are the first step in using lidar in archaeological research and heritage management, each project providing interesting lessons which will influence how the data may be used in the future. Brú na Boinne Meath County Council has identified some further ambitions for the Brú na Bóinne lidar data. Developing current research themes they hope to see comprehensive archaeological investigation of the newly identified potential sites, and hope that the lidar data can be used to inform land management within the WHS. Building on the evidence that the public engage well with lidar models it is hoped to create interpretative display panels and also to make the data available to the public online. The lidar project is considered an unqualified success by Meath County Council. It has exceeded expectations in the range and volume of research supported and there is a belief that more applications will become apparent in the future. Undoubtedly, the large extent (96 km2) of the survey in such an archaeologically rich area has been a factor. However, core to the success and the sense that the project achieved value for money was the concept of partnership – sharing resources and skills in a joint venture that included Meath County Council, the Heritage Council and the Discovery Programme. Dún Ailinne The specification for this high resolution lidar project helped to define the specification for the projects that followed. In practice, the variable GSD applied at Dún Ailinne failed to provide the significant savings in overall project cost that had been anticipated. The resulting inconsistency in the model resolution was found to be potentially prejudicial when using the data for prospecting or when examining the detailed

ground morphology. The boundary between the two resolutions is not visible in the model and it is therefore possible that micro-topographic features present on the ground could be missing from the model, distorting distributions and site interpretation. As a result this practice was not repeated in future surveys. Hill of Tara The OPW have expressed an interest in repeating the lidar survey at some stage in the future to assess potential erosion changes in the landscape, including the impact of visitor numbers on the site, and land management and development in the area. The high resolution and positional accuracy of the 2007 survey would provide an excellent baseline for any comparison. One issue which may be addressed in the near future is the extent of the survey, which was largely determined by affordability rather than archaeological reasons. Landscape planning and management is increasingly looking at the Hill of Tara in a much wider context, issues that are being addressed in documents such as the Draft Tara Skryne Landscape Conservation Area 2010 report. However, it may be that the surveyed area can most economically be extended using the latest photogrammetric techniques rather than commissioning additional lidar projects. Recent tests carried out on an existing set of digital vertical aerial images (1:7,500 scale) using the Erdas LSP eATE processing module have generated models at a remarkably high resolution, (Figure 12.8, see also Remondino this volume). However, it is worth remembering that photogrammetry can only generate a DSM, it cannot be filtered to generate a DTM (i.e. remove trees and hedgerows etc.) in the way that lidar can. There also appears to be a greater noise level in the model with some issues with seam edges between models, but these may be resolved by further refining the processing options. Skellig Michael Working with this lidar data set has provided great benefits to the archaeologists researching Skellig Michael, taking the mapping and plans to another level of precision and accuracy. However, it is only by viewing the relief models in conjunction with other resources such as photographs and physically visiting the site that the most complete interpretation can be achieved (i.e. Figure 12.9). Since many of the archaeological features are stone built structures and the definition of these was

12  Lidar and World Heritage Sites in Ireland

157 Figure 12.8: The potential of a new photogrammetric processing technique tested at Hill of Tara. Top left, relief-shaded DSM generated by conventional digital photogrammetry. Top right, relief-shaded DSM produced from same vertical images but using ERDAS LPS eATE multi-ray matching algorithm. For comparison bottom left is the relief-shaded DSM from high resolution lidar, with the filtered lidar relief-shaded DTM bottom right

a primary objective of the scanning project, the lidar-based relief model GSD of 12.5 cm may not be adequate in certain areas. Here the application of terrestrial based scanning systems with a

resolution of better than 5 mm may be necessary. Indeed, to complete the 3D documentation of Skellig Michael a terrestrial laser scanning project is planned with the intention of creating a

158

Anthony Corns and Robert Shaw

Figure 12.9: Identifying archaeology on Skellig Michael. Top is an oblique photograph taken from a helicopter, bottom in the skyview shaded DSM the corresponding area. Location 1, a preexcavated pathway was not visible on the lidar or photograph, only located by site visit. Location 2, rock cut steps are visible on the photograph but not on the lidar. Location 3, the lower traverse is visible on both photograph and lidar. Location 4, the Needle’s Eye, is neither visible on the lidar nor photograph, only located by site visit. Finally, location 5, the Prayer Station is visible on both photograph and lidar

seamless integration with the lidar data. Looking to the future, the fragile nature and inaccessibility of Skellig Michael point to the need to control and manage the number of visitors and offer alternative engagements with the site through

the development of virtual tours of the island in a digital environment. The level of detail in the relief model generated from the lidar survey would provide an excellent basis for such a development (see Challis and Kincey this volume).

12  Lidar and World Heritage Sites in Ireland

Conclusions The four case studies demonstrate how lidar has become firmly embedded as a landscape mapping technique for archaeology and cultural heritage in Ireland. For each project the decision to ‘speculate’ on the value of lidar has been vindicated; its significance for both management and research clearly demonstrated. This is particularly so for the two World Heritage Sites, Brú na Boinne and Skellig Michael. High resolution relief models derived from lidar data have been important resources when producing management plans and research framework documents – standard working practice for World Heritage Sites. The most successful and positive experiences of lidar have come from partnership and collaboration, where positive attitudes to sharing resources exist. This is best illustrated by the Brú na Boinne project. However, even in this case technological limitations were identified as a significant obstacle, the solution to which undoubtedly lies in the development of spatial data infrastructure. This would be of enormous benefit: formalising access arrangements, widen­ ing the number and range of uses, ensuring long term secure archiving and ultimately maximising the return on investment in data. Lidar technology is constantly evolving and the options available expanding. High resol­ ution systems are now being flown from fixed wing aircraft adding a further option when commissioning projects. In Ireland commercially available ‘off the shelf ’ lidar has emerged since the projects in this discussion were commissioned, notably from the Ordnance Survey Ireland (OSi). The latter is actively pursuing the heritage and archaeological market in their marketing and promotional material. In parallel other sources of high resolution surface models are likely to emerge through the use of photogrammetric methods (Figure 12.8, Remondino this volume). These factors may well reduce the requirement for specifically commissioned lidar projects, but as the case studies have shown it is access to the relief models themselves, not lidar data per se that is the important resource. At the time these projects were commissioned lidar happened to be the most appropriate technology but access to alternatives can only have a positive influence, increasing applications and lowering cost. For those working with lidar data in Ireland there is great satisfaction in seeing the technique embedded in the research

159 and management agenda – elevating lidar data from being merely spectacular images into a valued and fully utilised resource.

Acknowledgements The authors wish to thank Edward Bourke, Ana Dolan, Ian Doyle and Loreto Guinan for their valuable comment on and insight into the use of lidar in Ireland.

References Bewley, R.H., Crutchley, S. and Shell, C.A., 2005. New light on an ancient landscape: lidar survey in the Stonehenge World Heritage Site. Antiquity 79:305, 636–47. Corns, A., Fenwick, J. and Shaw, R., 2008. More than meets the eye. Archaeology Ireland 22, no. 3, Issue no. 85, 34–8. Corns, A. and Shaw, R., 2009. High resolution 3-dimensional documentation of archaeological monuments & landscapes using airborne LiDAR, Journal of Cultural Heritage 10, Supplement 1, 72–77. Corns, A. and Shaw, R., 2010. Cultural Heritage Spatial Data Infrastructures (SDI) – Unlocking the Potential of our Cultural Landscape Data. In Rainer, R. (ed.). Proceedings of EARSeL Symposium 2010, 1–8. Davis, S., Megarry, W., Brady, C., Barton, K., Lewis, H., Mulrooney, G., Cummins, T., Guinan, L., Turner, J., Gallagher, C. and Meehan, R, 2010. Boyne Valley Landscapes Project. Phase III Final Report. Unpublished Report, The Heritage Council, Kilkenny. Johnston, S., Campana, D., and Crabtree, P., 2009. A geophysical survey at Dún Ailinne, County Kildare, Ireland. Journal of Field Archaeology 34(4): 385–40. Kokalj. Ž., Zakšek, K., and Oštir, K., 2010. Archaeological Application of an Advanced Visualisation Technique Based on Diffuse Illumin­ ation. In Rainer, R. (ed.) Proceedings of EARSeL Symposium 2010, 113–20. Lewis, H., Gallagher, C., van Breda, W., Mulrooney, G., Davis, S., Meehan, R., Turner J., Brown, A., Guinan, and Brady, L. 2008. An integrated comprehensive GIS model of landscape evolution and land use history in the River Boyne Valley and its catchment. Unpublished Report, The Heritage Council, Kilkenny. Lewis, H., Gallagher, C., Davis, S., Turner, J. and Foster, G., 2009. An integrated comprehensive GIS model of landscape evolution and land use history in the River Boyne Valley. Unpublished Report, The Heritage Council, Kilkenny.

160

Anthony Corns and Robert Shaw Meehan, R., 2011. The Making of Meath: The County’s Natural Landscape History. Meath County Council. Shell, C. and Roughley, C., 2004. Exploring the Loughcrew Landscape: a New Approach with Airborne Lidar. Archaeology Ireland, 18, no. 2, Issue no. 68, 20–23. Smyth, J. (ed.), 2009. Brú na Bóinne World Heritage

Site: Research Framework. The Heritage Council, Kilkenny. Stout, G., 2010. Monumentality and inclusion in the Boyne valley, County Meath, Ireland. In Leary, J., Darvill, T. and Field, D., (eds). Round Mounds and Monumentality in the British Neolithic and Beyond, 2010 Neolithic Studies Group Seminar Papers 10, Oxford: Oxbow, 197–210.

13 The role of lidar intensity data in interpreting environmental and cultural archaeological landscapes Keith Challis and Andy J. Howard Laser intensity images, collected through airborne survey and created by visualising the amplitude of each returned laser pulse, usually as a greyscale image, are often ignored when working with the results of lidar surveys. Fortunately a growing body of research demonstrates the value and uses of intensity data and imagery. In this paper we summarise the problems inherent in collecting and using intensity data, review examples of its use in a variety of environmental disciplines and discuss experiments that may help to determine its potential applications in archaeology. Keywords: intensity data, alluvial landscapes, Trent Valley, knowledge-based mapping

Introduction The archaeological applications of airborne lidar topographic data are now well known. As well as geoarchaeological mapping and prospection (Brunning and Far-Cox 2005; Carey et al. 2006; Challis 2005; Challis 2006; Challis et al. 2006; Jones et al. 2007), published applications include over-arching landscape studies (Barnes 2003; Bewley et al. 2005; Bofinger et al. 2006; Harmon et al. 2006; Powlesland et al. 2006), investigation of the potential for lidar to detect upstanding archaeological remains beneath the vegetation canopy (Crow et al. 2007; Deveraux et al. 2005; Doneus and Briese 2006; Risbøl et al. 2006; Sittler and Schellberg 2006) and contributions to the compilation and refinement of records of the historic environment (Holden et al. 2002; ������������������������������������������� Bewley 2003; Crutchley 2006; Challis et al. 2008). Rather less well explored by archaeologists are the potential applications of lidar intensity data, a secondary output of topographic measurements that records a laser ‘image’ of the land surface derived from measurements of the amplitude of each reflected laser pulse (Figure 13.1). A �� comprehensive summary of applications of

airborne lidar intensity is provided by Höfle and Pfeifer (2007), but thus far, ���������������������� its applications have been restricted to a few specialist fields including characterising species diversity in forest plantations (��������� Donoghue et al. 2007), the determination of the age of lava flows in active volcanoes (���������� Mazzarini� et al. 2007,��������������� 2009) and the classification of glacial surfaces (Figure 13.2; Arnold et al. 2006; ����� Lutz et al. 2003).

Figure 13.1: The lidar principle. Intensity values are an 8 or 10 bit digital number recording the amplitude of each reflected pulse

162

Keith Challis and Andy J. Howard

Figure 13.2: Examples of lidar intensity usage; (left) Raw and range normalised intensity image of a sector of the southern flank of Mount Etna, where scoria cones, lava flows, lobes, and channels are clearly identified.(from Mazzarini et al. 2007, fig. 3; (right) Rangenormalized intensity image of the midre Lovenbreen glacier, Svalbard. Snow-covered areas appear as white/very light grey, the glacier surface as mid/dark grey and the surrounding rock or moraine as mid/light grey (from Arnold et al. 2006, fig. 11)

A number of studies, such as those by Song et al. (2002), Chust et al. (2008) and Yoon et al. (2008) suggest that lidar intensity may be more generally useful for determination ����������������� of land cover and could serve as a useful addition to lidar elevation data when interpreting lidar survey results. Work by the present authors suggest that lidar intensity data and imagery also have significant potential for elucidating cultural and environmental archaeological remains, particularly in river valleys where extensive lidar data has been routinely collected and where contrasting biological, geochemical and sedimentological conditions exist (Challis et al. 2011a and b).

The intensity record Lidar intensity measurements �������������� represent the reflected energy from a highly focused beam of

near infrared radiation (NIR), which provides a concentrated measurement of an object’s reflectivity. The physical principle underlying lidar intensity is summarized by Höfle and Pfeifer (2007). Since ����������������������� NIR reflectance varies in response to a number of earth-surface characteristics, intensity data has the potential to provide a qualitative descriptor of earth surface materials. Boyd and Hill (2007) have demonstrated that lidar intensity records capture a reliable and accurate analogue of NIR solar reflectance, while recognising that for any given scenario reflected solar radiation captured by a broad spectrum multispectral or hyperspectral instrument is likely to provide a better and more discriminating record of soil and vegetation reflectance properties. Lidar intensity data is routinely collected as part of surveys aimed primarily at recording topography but rarely examined. This suggests that there is a significant body of data suitable

163

13  The role of lidar intensity data

Figure 13.3: examples of intensity output from Optech, Leica and FliMap instruments with spectral reflectance graph showing the different wavelengths of the three instruments

for analysis for archaeologically interesting ends that is largely ignored. Lidar instruments store the intensity return of each pulse (digitised wave amplitude) as a digital number, most frequently on an 8-bit (values 0255) or 10-bit (values 0-1096) scale. Post-flight processed data usually provide intensity data as a component of the point-cloud (so-called xyzi datum, where i is the intensity value for each pulse) or less frequently as an interpolated greyscale image; although this latter format is less desirable, as interpolation and scaling of intensity values from flight data to finished product can have a significant impact on values. It is important to recognise that there are significant variations in �������������������������� intensity measurements between individual lidar systems largely due to differences in receiver properties and type and wavelength of the laser used. The most frequently used lidar instruments in the British Isles comprise a series of sensors produced by a Canadian manufacturer, Optech, and

instruments manufactured by Fli-Map and Leica. Optech instruments use ���������������������� a NIR laser operating between 1047–1068 nm (varying by system) and are in common use in the UK by commercial contractors and government agencies for surveys. The Leica ALS50-I and allied instruments uses a similar laser wavelength (1064 nm) while that produced by Fli-Map uses a middle infra-red laser operating at 1500 nm (Lemmens ��������������� 2007)�. Variations in laser wavelength are significant as the laser is, in effect, the illumination source for the recorded image and reflectance of earth surface material varies considerably in the infra-red (Figure 13.3). One consequence of this is that while the Optech and Leica instruments maximise the potential difference in reflection between green and parched vegetation (the fundamental physical property of anthropogenically produced vegetation marks) the Fli-Map instrument, operating at a longer wavelength, offers less discrimination between green and dry vegetation and, because of

164

Keith Challis and Andy J. Howard variations in reflectance properties in the middle infra-red, a reversal of the reflectance properties of the Optech and Leica instruments. Therefore, it is essentially to know the make and model of instrument that has collected the data.

Working with intensity data Lidar intensity records are highly susceptible to influence by a range of equipment and environmental variables, divided broadly into: 1) system variables; 2) target variables; and 3) processing procedures. System variables include factors such as: the distance between the lidar system and the target (the range – controlled largely by the altitude of the survey aircraft, but also influenced by both natural and anthropogenic topographic variation and the scan angle of individual lidar pulses); the peak pulse power of the laser system, which may vary as a factor of pulse frequency, beam divergence and laser footprint size (a product of range and beam divergence); and angle of incidence. Target variables include the cross sectional area of target within the laser footprint, target reflectivity and surface roughness. Processing procedures include variables introduced to the intensity data by factors such as the interpolation technique applied to convert the point cloud data into a regular grid. It is important to recognise that in order to maximise potential results from analysis of intensity data, access to unprocessed point cloud data as well as flight and instrument meta-data are usually required. It may often prove necessary to post-process intensity data using appropriate methodologies to counter some or all of the flight and instrument effects on data, although many system variables are not routinely possible to correct for. Luzum et al. (2004) have explored normalisation of intensity measurements from regions of high relief for path length by calculating survey aircraft altitude and laser scan angle. While it is relatively easy to correct for range based on survey aircraft altitude, correction for scan angle is more problematic. ������������������������� Coren and Sterzai (2006) report that in Optech instruments angles below c.15° from nadir have little effect on intensity values, although since Optech instruments may scan up to 25° off-nadir (Optech 2003), some data can be affected by such instrument induced variations. Significantly, Yoon ����� et al.

(2008) found that there was little variation in intensity due to path length on variant reflectors such as vegetation and concluded that this was due to the heterogeneous character of returns from such materials compared to those for invariant reflectors such as man-made surfaces. Kaasalainen et al. (2009) have attempted radiometric calibration of intensity data against targets of known reflectance in order to allow comparison and mathematical modelling of multiple survey flights. Finally, the impact of ��� atmospheric attenuation on the transmission and reflection of the laser pulse may also affect results, but is difficult to account for, requiring detailed observations of atmospheric conditions at the time of survey and complex computational modelling and consequently this correction is not routinely performed.

Archaeological applications of intensity data: a case study from the Trent Valley, UK For over 30 years, the alluvial valley floor of the River Trent in central England has been one of the most intensely investigated geoarchaeological landscapes in the world containing a range of both buried and upstanding multi-period archaeological remains and significant environmental deposits preserved in palaeochannels (e.g. Whimster 1986; Baker 2003; Brown 1998; Brown et al.. 2001; Howard 2005; Howard et al. 2007; Knight and Howard 2004; Salisbury 1992). In recent years, a significant body of this research has been focused on the application of remote sensing technologies, including lidar (Challis 2005, 2006; Howard et al. 2008) and the following example illustrates work undertaken in the middle and lower part of the valley where the terrain comprises a mixture of gravel terraces and islands of late Pleistocene age (the Holme Pierrepont Sand and Gravel) as well as Holocene age overbank alluvium and palaeochannels. Lidar data for the study area were acquired by Infoterra Global Ltd on 26th July 2007 using an Optech ALTM 2033 Lidar flying at an average altitude of 914 m and recording two returns (first and last) at a maximum scan angle of 20°��������������������������������������������� ������������������������������������������������ . Data was supplied as a x,y,z,i point cloud in ASCII format for first and last pulse returns processed to WGS84 datum with ellipsoidal elevation values. The point clouds were processed

13  The role of lidar intensity data

165 Figure 13.4: Comparison of lidar elevation and intensity data with field collected environmental variables. Top, lidar elevation, FP and LP intensity data for the study area. Bottom, scatter plots showing comparison of LP elevation and intensity with soil moisture and soil organic content

a

c

using Applied Imagery’s Quick Terrain Modeller software to generate first and last pulse digital surface models at 1 m spatial resolution and 8bit greyscale images derived from first and last pulse intensity values. Aircraft recorded intensity values were ‘histogram-stretched’ to the full 8-bit dynamic range (values 0-255), but were otherwise unaltered. The DSM and intensity images were re-projected to British National Grid and ellipsoidal elevation values converted to Ordnance Datum using Erdas Imagine 9.1. In the analysis below image based visual analysis of intensity makes use of data re-projected to British National Grid. Since re-projection alters original values through a process of interpolation, statistical analysis made use of data in the original WGS84 datum with no additional processing steps employed on the point cloud.

b

d

Comparison of intensity and field data In addition to lidar, this study involved simultaneous ground based sampling to collect data relating to soil organic content, moisture, earth resistance and sediment stratigraphy. Soil organic content displays a strong negative linear relationship to elevation, defined by the lidar DSM (Figure 13.4a; linear R square 0.79). As elevation increases, the organic content of the soil decreases, highlighting the difference between the topographically lower palaeochannel and adjacent higher Pleistocene terrace. Much weaker relationships are again observed between the lidar intensity data and soil organic content, with both graphs displaying moderate negative linear relationships (Figure 13.4b; linear R square values of 0.35 and 0.32 respectively). Soil moisture has a negative linear relationship

166

Keith Challis and Andy J. Howard

Figure 13.5: Representative areas of terrace (A) and floodplain (B). In each case lidar topography and range normalized intensity are shown. Figure annotations indicate the principle features discussed in the text

with elevation (Figure 13.4d; linear R square 0.57), lower than the corresponding value for elevation and soil organic content (linear R square 0.79). Like soil organics, soil moisture has weaker relationships with lidar first pulse (FP) and last pulse (LP) intensity data, both moderate negative linear relationships with intensity (Figure 13.4e; FP linear R square 0.345). All of the measured variables are highly correlated. However, this does not mean that lidar intensity can be used in a predictive sense to identify areas of differential sediment character and in particular the preservation potential of organic remains, as the relationships of both FP and LP intensity to soil moisture content, soil

organic content and earth resistance survey are non linear. Indeed, the graphical analysis of the variables showed that soil moisture and soil organic content were most strongly related to elevation, this being a function of the geomorphological variation within the study area.

Knowledge-based mapping using intensity In this final section last pulse ground lidar elevation and normalised intensity data for two locales in the lower Trent (Figure 13.5), which are representative of the entire study

13  The role of lidar intensity data area, are examined in order to assess the utility of these data to inform a series of empirical statements about the character of floodplain and terrace topography, the sediments and their archaeological potential. Such information might be typical of a first level assessment of valley floor landscapes and could provide a framework for the rapid assessment of large areas. In order to reduce the impact of topographic effects on intensity, our data were normalized to range by applying a procedure adapted from Luzum et al. (2004; see Challis et al. 2011a for details). Figure 13.5A shows an area of predominately Holme Pierrpont Sand and Gravel river terrace at South Muskham. The terrace is bisected from north to south by a sinuous palaeochannel, part of a major former channel of the River Trent. Mapping of superficial geology (BGS sheet 113, Ollerton) indicates the channel is infilled with fine grained alluvium. However, close examination of the lidar elevation data shows that the channel is both laterally more extensive and morphologically more complex than BGS mapping suggests. Shallow sinuous depressions within the terrace surface are also evident in the lidar elevation data, which might indicate areas where superficial colluvial and alluvial deposits might enhance preservation of buried cultural and environmental archaeological material. Whilst lidar intensity data for this area also show some difference in crop colour and growth on the terrace to the east of the palaeochannels, suggesting variations in the composition of the underlying terrace, they otherwise contribute little to the qualitative understanding of the burial environment. Figure 13.5B shows an area of predominantly fine grained (overbank) alluvium adjacent to the present channel of the River Trent between Kelham and Averham. Lidar elevation data clearly distinguishes the lower lying areas of floodplain alluvium and marginally higher islands of the Holme Pierrepont Sand and Gravel. A number of palaeochannels are indicated by sinuous depressions, and in all probability represent former channels of the Trent, which in this area may have had an anastomosing character (Knight and Howard 2004). The level of detail recorded by lidar elevation data is substantially greater than that present on the existing mapping of superficial geology (BGS sheets 126, Nottingham and 127, Grantham) and this additional imagery provides an appropriate base map for further geomorphological refinement and identification

167 of areas of potential differential preservation prior to archaeological field investigation. For example areas of terrace might be identified as being of greater potential for settlement related activities; along terrace margins overbank flooding may have buried cultural deposits and palaeochannels may provide prime locations for the recovery of organic sediments capable of preserving proxy records environmental change (Howard and Macklin 1999). The intensity data for this area adds considerably to qualitative understanding of the burial environment. Variations within intensity values on the floodplain indicate cropmarks mirroring the orientation of several of the sinuous palaeochannels. To the east of the central island of terrace material, itself indicated by an area of differential colour showing as lower intensity return, probably indicating parched crop, several areas of very low intensity returns probably indicate highly saturated ground and perhaps in several cases shallow standing water. These low-lying areas correspond to several very shallow topographic depressions indicated in elevation data and suggest channel features or perhaps simply depressions that may contain waterlogged, well preserved, organic sediments. Other linear cropmarks, probably buried field drains, are also apparent on the intensity data.

Intensity: a way forward in archaeology We have found little evidence in the data analysed of variation in intensity values due to instrument factors. Furthermore, normalisation of intensity values for range based on a variety of parameters made no significant difference to the visual quality of data and we conclude that in areas of low relief variation where the survey aircraft has flown at a uniform altitude, range normalisation is of no benefit for archaeological interpretation. This is probably largely due to the relatively low relief variation within our study area, and the heterogeneous land cover, largely arable land and other vegetation covered surfaces. In areas of higher relief variation and/or more homogenous land cover we would expect normalisation to be of greater benefit. Examination of lidar intensity imagery from a variety of archaeological and geomorphological settings indicates that these data do contain information not present in the corresponding elevation data. Lidar intensity data informs on the character of soils, sediments and vegetation

168

Keith Challis and Andy J. Howard in specific topographic conditions, a correlation which allows empirical analysis of the relationships between these variables and adds qualitative information to the interpretation of the landscape. In effect the intensity image is subject to the same knowledge-based interpretation as might be used to extract information from conventional aerial photographs. Since intensity data is (or can be) routinely recorded during a lidar flight aimed primarily at gathering topographic information, it is suggested that the examination of these data are routinely incorporated in the archaeological interpretation of existing lidar data, and that their collection always form part of the parameters of any airborne lidar survey commissioned for archaeological purposes. Whilst we recommend the collection of intensity data, there are a number of key areas where further research is needed to exploit the full potential of this data: • At present, a very limited number of topographic settings and geographic areas have been studied and there is a need for further systematic investigation. • The majority of data is collected in the winter months and there is a need for data collection during non-traditional seasons (e.g. summer), which would allow comparison with periods of the year when cropmarks are most easily observable. • For anticipated use of intensity to inform survey strategy (e.g. flight lines, altitude, single epoch, etc.). • Further work with respect to the calibration of intensity data to targets of known reflectance. • Further work is required on normalisation procedures for intensity data. • Procedures need to be developed to fuse intensity data and other RS data (e.g. for object driven classification of land cover). • In UK, examination of intensity data from the new generation of Optech instruments is needed. • Future work on full waveform intensity data as it becomes available.

Acknowledgements This research was supported by DEFRA through the Aggregates Levy Sustainability Fund administered by English Heritage. Fieldwork

associated with the lidar studies outlined conducted by Dr Chris Carey, Mark Kincey and Paul Breeze.

References Arnold, N.S., Rees, W.G., Devereux, B.J. and Amable, G.S., 2006. Evaluating the potential of highresolution airborne LiDAR data in glaciology, International Journal of Remote Sensing, 27(6), 1233–51. Baker, S., 2003. The palaeochannel record in the Trent Valley UK: contributions towards cultural heritage management. Internet Archaeology 20. http://intarch.ac.uk/journal/issue20/baker_toc. html. Barnes, I., 2003. Aerial remote-sensing techniques used in the management of archaeological monuments on the British army’s Salisbury Plain training area, Wiltshire, UK. Archaeological Prospection 10, 83–90. Bewley, R.H. Crutchley, S.P and Shell, C.A., 2005. New light on an ancient landscape: lidar survey in the Stonehenge World Heritage Site. Antiquity 79(305), 636–47. Bewley, R., 2003. Aerial survey for archaeology. The Photogrammetric Record 18(104), 273–92. Brown, A.G., 1998. Fluvial evidence of the Medieval Warm Period and the Late Medieval Climatic Deterioration in Europe. In Benito, G., Baker, V.R. and Gregory, K.J., (eds). Palaeohydrology and Environmental Change. 43–52. Chichester: Wiley. Brown, A.G., Salisbury, C.R., and Smith, D.N., 2001. Late Holocene channel changes of the Middle Trent: channel response to a thousand year flood record. Geomorphology 39, 69–82. Brunning, R. And Farr-Cox, F., 2005. The River Siger rediscovered: lidar survey and relict landscape on the Somerset Claylands. Archaeology and the Severn Estuary 16, 7–15. Bofinger, J., Kurz, S. and Schmidt, S. 2006. �������� Ancient Maps – modern data sets: different investigative techniques in the landscapes of the Early Iron Ahe princely hill fort Heuneburg, Baden-Wurttemberg. In Campana, S. and Forte, M. (eds). From Space to Place: Proceedings of the 2nd International Workshop on Remote Sensing in Archaeolo;gy, CNR, Rome, Italy, Dec 4–7, 2006. British Archaeological Reports, International series 1568, 87–92. Boyd, D. S. and Hill, R.A., 2007. Validation of Airborne LiDAR Intensity Values from a Forested Landscape Using HyMap Data: Preliminary Analyses. In Rönnholm, P., Hyyppä, H., and Hyyppä, J., (eds). Proceedings of the ISPRS Workshop: Laser Scanning 2007 and SilviLaser 2007, September 12–14, 2007, Espoo, Finland. 71–6. Carey, C., Brown, T., Challis, K. Howard, A.J. and Cooper, L., 2006. Predictive modelling of multiperiod geoarchaeological resources at a river

13  The role of lidar intensity data confluence: a case study from the Trent-Soar, UK. Archaeological Prospection 13(4), 241–50. Challis, K., 2005. Airborne LiDAR: A Tool for Geoarchaeological Prospection in Riverine Landscapes. In Stoepker, H. (ed.). Archaeological Heritage Management in Riverine Landscapes. Rapporten Archeologische Monumentenzorg, 126, 11–24. Challis, K., 2006. Airborne laser altimetry in alluviated landscapes. Archaeological Prospection 13(2), 103–27. Challis, K., Kincey, M., Carey, C. and Howard, A.J., 2011a. Airborne Lidar Intensity and geo­ archaeological prospection in river valley floors. Archaeological Prospection 18, 1–13. Challis, K., Kincey, M., Carey, C. and Howard, A.J., 2011b. Assessing the preservation potential of temperate, lowland alluvial sediments using airborne lidar intensity. Journal of Archaeological Science 38, 301–11. Challis, K., Kokalj, Z., Kincey, M., Moscrop, D. and Howard, A.J., 2008 Airborne Lidar and Historic Environment Records. Antiquity 82 (318), 1055–64. Challis, K., Howard, A.J., Smith, D.N., Gearey, B.R., Moscrop, D., Carey, C.J. and Thompson, A., 2006. Using Airborne Lidar Intensity to Predict the Organic Preservation of Waterlogged Deposits. In Campana, S. and Forte, M., (eds). From Space to Place: Proceedings of the 2nd International Workshop on Remote Sensing in Archaeology, CNR, Rome, Italy, Dec 4–7, 2006. British Archaeological Reports, International series 1568, 93–8. Chust, G., Galparsoro, I., Borja, A., Franco, J. and Uriarte, A., 2008. Coastal and estuarine habitat mapping, using LIDAR height and intensity and multi-spectral imagery. Estuarine, Coastal and Shelf Science 78,633–43. Coren, F. and Sterzai, P., 2006. Radiometric correction of laser scanning. International Journal of Remote Sensing 27, 3097–104. Crow, P., Benham, S., Devereux, B.J. and Amable, G.S., 2007. Woodland vegetation and its implications for archaeological survey using Lidar. Forestry 80(3), 241–52. Crutchley, S., 2006. Light detection and ranging (lidar) in the Witham Valley, Lincolnshire: an assessment of new remote sensing techniques. Archaeological Prospection 13(4), 251–7. Devereux, B.J., Amable, G.S. Crow, P. and Cliff, A.D., 2005. The potential of airborne lidar for detection of archaeological features under woodland canopies. Antiquity 79(305), 648–60. Doneus, M. and Briese, C., 2006. Digital terrain modelling for archaeological interpretation within forested areas using full-waveform laserscanning. In Ioannides, M., Arnold, D., Niccolucci, F., and Mania, K., (eds). Proceedings of the 7th International Symposium on Virtual Reality, Archaeology and Cultural Heritage (VAST 2006). 155–62. Donoghue, D., Daniel, N.M., Watt, P.J., Cox, N.J. and Wilson, J., 2007. Remote sensing of species

169 mixtures in conifer plantations using LiDAR height and intensity data. Remote Sensing of Environment 110, 509–22. Harmon, J.M., Leone, M.P., Prince, S.D. and Snyder, M., 2006. Lidar for archaeological landscape analysis: a case study of two eighteenthcentury Maryland plantation sites. Antiquity 71, 649–70. Holden, N., Horne, P, and Bewley, R.H., 2002. High resolution digital airborne mapping and Archaeology. In Bewley R. and Raczkowski W., (eds). Aerial Archaeology: Developing Future Practice. NATO Series 1, Vol. 337, IOS Press, Amsterdam, 173–80. Höfle, B. and Pfeifer, N., 2007. Correction of laser scanning intensity data: Data and model-driven approaches. ISPRS Journal of Photogrammetry & Remote Sensing 62, 415–33. Howard, A.J., 2005. The contribution of geoarch­ aeology to understanding the environmental history of the Trent Valley, UK. Geoarchaeology 20(2), 93–107. Howard, A.J., Bridgland, D.R., Knight, D., McNabb, J., Rose, J., Schreve, D., Westaway, R., White, M.J. and White, T.S., 2007. The British Pleistocene fluvial archive: East Midlands drainage evolution and human occupation in the context of the British and NW European record. Quaternary Science Reviews 26, 2724–37. Howard, A.J., Brown, A.G., Carey, C.J., Challis, K., Cooper, L.P., Kincey, M. and Toms, P., 2008. ������ Archaeological resource modelling in temperate river valleys: a case study from the Trent Valley, UK. Antiquity 82(318), 1040–54. Howard, A.J. and Macklin, M.G., 1999. A generic geomorphological approach to archaeological interpretation and prospection in British river valleys: a guide for archaeologists investigating Holocene landscapes. Antiquity 73, 527–41. Jones, A.F., Brewer, P.A., Johnstone, E. and Macklin, M.G., 2007. High-resolution interpretative geomorphological mapping of river environments using airborne LiDAR data. Earth Surface Processes and Landforms 31, 1574–92. Kaasalainen, S., ����������������������� Hyyppä, H., Kukko, A., ������������ Litkey, P., Ahokas, E., Hyyppä, J., Lehner, H., Jaakkola, A., Suomalainen, J., Akujärvi, A., Kaasalainen, A. and Pyysalo, A��������� ., 2009. Radiometric calibration of LIDAR intensity with commercially available reference targets. IEEE Transactions on Geoscience and remote sensing 47, 588. Knight, D. and Howard, A.J., 2004. Trent Valley Landscapes. Heritage Marketing and Publications Ltd: Kings Lynn. Lemmens, M., 2007. Product Survey on Airborne Lidar Sensors. GIM International 21(2). Lutz, E., Geist, Th. and Stötter, J., 2003. Investigations of Airborne Laser Scanning Signal Intensity on Glacial Surfaces – Utilizing Comprehensive Laser Geometry Modelling and Orthophoto Surface Modelling (A Case Study: Svartisheibreen, Norway). ��� In ������������������������������� Maas, H.-G., Vosselman, G. and

170

Keith Challis and Andy J. Howard Streilein, A. (eds). Proceedings of the ISPRS Workshop on 3-D reconstruction from airborne laserscanner and INSAR data. 143–48. Dresden. Luzum, B., Starek, M. and Slatton, K.C., 2004. Normalizing ALSM intensities. GEM center report No. Rep_2004-07-001. U.S.A.: University of Florida. Mazzarini, F., Colini, L., Neri, M., Behncke, B., Salvatori, R., Buongiorno, M. and Pareschi, M., 2009. ����������������������������������������������� Spectral properties of volcanic materials from hyperspectral field and satellite data compared with LiDAR data at Mt. Etna. International Journal of Applied Earth Observation and Geoinformation, Volume 11(2), 142–55. Mazzarini, F., Pareschi, M. T., Favalli, M., Isola, I., Tarquini, S. and Boschi, E., 2007. Lava flow identification and aging by means of LiDAR intensity: Mount Etna case. Journal of Geophysical Research, 112, B02201. Optech, 2003. ALTM 2033 Airborne Laser Terrain Mapper. Optech: Toronto. Powlesland, D., Lyall, J., Hopkinson, G,. Donoghue, D., Beck, M., Harte, A., and Stott, D., 2006. Beneath the sand – remote sensing, archaeology, aggregates and sustainability: a case study from Heslerton, the Vale of Pickering, North Yorkshire, UK. Archaeological Prospection 13(4), 291–9. Risbøl, O., Kristian Gjertsen, A. and Skare, K., 2006. Airborne laser scanning of cultural remains in forests: some preliminary results from a Norwegian project. In Campana, S. and Forte, M., (eds). From Space to Place: Proceedings of the 2nd International

Workshop on Remote Sensing in Archaeology, CNR, Rome, Italy, Dec 4–7, 2006. British Archaeological Reports, International series 1568, 107–12. Salisbury, C.R., 1992. The archaeological evidence for palaeochannels in the Trent Valley. In Needham, S. and Macklin, M.G. (eds). Alluvial Archaeology in Britain. 155–62. Oxford: Oxbow Monograph 27. Shell, C.A. and Roughley, C.F., 2004. Exploring the Loughcrew Landscape: a New Approach with Airborne Lidar, Archaeology Ireland 18(2), Issue no. 68, 20–3. Sittler, B. and Schellberg, S., 2006. The potential of lidar is assessing elements of cultural heritage ��������� hidden under forest canopies or overgrown vegetation: Possibilities and limits in detecting microrelief structures for archaeological surveys. In Campana, S. and Forte, M., (eds). From Space to Place: Proceedings of the 2nd International Workshop on Remote Sensing in Archaeology, CNR, Rome, Italy, Dec 4–7, 2006. British Archaeological Reports, International series 1568, 117–22. Song, J.H., Han, S.H., Yu, K., Kim, Y., 2002. Assessing the Possibility of Land-cover Classification Using Lidar Intensity Data. IAPRS, 9–13 September, Graz, 34, 4. Whimster, R., 1986. The Emerging Past. Air Photography and the Buried Landscape. RCHME: London. Yoon, J.S., Shin, J. and Lee, K-S., 2008. Land cover characteristics of airborne LiDAR intensity data: a case study, IEEE Geoscience and Remote Sensing Letters 5, 801–5.

14 The changing picture of archaeological landscapes: lidar prospection over very large areas as part of a cultural heritage strategy Ralf Hesse Airborne lidar allows archaeological prospection over very large areas with a single, consistent method, largely irrespective of present-day land cover. Lidar-based prospection thus has the potential to capture spatial patterns of archaeological sites and other elements of the archaeological landscape over large areas. When targeting very large areas, such as the 35,751 km2 of the German federal state Baden-Württemberg, the efficient management, processing and interpretation of large amounts of data is a central issue. Large-scale lidar-based prospection projects have the potential to greatly increase the numbers of mapped features, but may suffer from reduced confidence of interpretation and the results largely lack detailed temporal information. This reflects the reality of working with extensive archaeological landscapes with overlapping remnants of multiple phases of human impact, where spatial patterns and statistical analysis increase in importance over individual features. Lidar-based archaeological prospection is becoming an integral part of cultural heritage strategies. In Baden-Württemberg, this includes in particular (a) the revision of location and extent of sites already registered in the archaeological site database and (b) the expansion of the database by mapping all potential archaeological sites and other elements of the archaeological landscape which are detectable in the lidar data. The results will be used to provide guidance for local and regional planning authorities to better protect archaeological sites and landscapes. Keywords: lidar-based prospection, Baden-Württemberg, land use, data management, potential sites, heritage management, archaeological landscape

Introduction Recent decades have witnessed an increasing attention in archaeology to a landscape-scale perspective as aerial photographs, geophysical survey and, in recent years, airborne lidar have widened the scope from individual sites to site distribution patterns and landscapes. Airborne lidar allows visualisation and enhancement of subtle relief features and detection of traces of former human activities, facilitated by new data acquisition, processing and visualisation techniques (e.g. Devereux et al. 2008; Doneus

et al. 2008; Doneus and Briese 2011; Hesse 2010; Zakšek et al. 2011). In particular the ability of lidar to penetrate vegetation makes it the first remote sensing technology that allows the detection of archaeological remains under forest canopy (e.g. Gallagher and Josephs 2008). Very large areas can now be prospected and mapped, largely irrespective of presentday vegetation cover. This has stimulated the increasing use of lidar in archaeological prospection, both for archaeological research and heritage management. This chapter describes

172

Ralf Hesse

Figure 14.1: Map of Baden-Württemberg showing the progress of the lidar-based archaeological prospection to September 2011 (completed areas outlined in blue). Numbered dots mark the location of sites shown in further figures; red lines show the courses of the Roman Limes of Germania Superior and Raetia and the Odenwald Limes. Inset: location of Baden-Württemberg in Germany

one of the spatially most extensive projects in which airborne lidar is used for archaeological prospection in a heritage management and protection context carried out in the German federal state Baden-Württemberg.

Archaeological sites in the topography of Baden-Württemberg With an area of 35,751 km2, Baden-Württemberg is Germany’s third-largest federal state (Figure 14.1). It is rich in archaeological sites spanning all periods since the Palaeolithic. Since the Neolithic, human activities have left traces in the relief of the landscape, some of which still persist as topographic features. These include

Neolithic and Bronze Age hilltop settlements like the Kirchberg near Ammerbuch-Reusten (Kimmig 1966; Biel 1987) and Bronze Age burial mounds like those near Nehren (Pirling 1980, 80). Of particular interest in this region are the Early Celtic Iron Age sites including the Heuneburg, possibly the earliest urban-like settlement north of the Alps (Kurz 2010). From the late pre-Roman Iron Age, several oppida (e.g. Heidengraben), rectangular bank-and ditch enclosures (Viereckschanzen) as well as numerous burial mounds are still preserved (Wieland 1996; Bittel et al. 1990). Several sections of the Limes, today part of the UNESCO multinational World Heritage Site ‘Frontiers of the Roman Empire’, ran through Baden-Württemberg, demarcating the Roman provinces of Germania

173

14  The changing picture of archaeological landscapes Superior and Raetia from the Germanic tribes to the northeast (Deutsche Limeskommission 2009). Several Roman villas as well as parts of the road network (e.g. near Pforzheim) are still recognisable as relief features (Kortüm 1995, 27ff.). Some Merovingian burial mounds and hill forts (e.g. the Runde Berg near Urach and the Zähringer Burgberg near Freiburg (Steuer 1997)) also survive as relief features. Besides the extant buildings and settlement layouts, there are numerous relict medieval and postmedieval remains like agricultural terraces, ridge and furrow and hollow-ways as well as abundant mining, quarrying and charcoal production sites (Simms 1976; Kempa 2003; Ludemann 2010). Military trenches and bomb craters from World War Two can be considered to be among the most recent archaeological features in the region. All these remains, and others, have the potential to be detected and recorded in the lidar data, and form part of the ‘reference collection’ that informs the archaeological interpretation of the lidar data.

Archaeology in the context of past and present land use Land use has an important role in the differential preservation of the archaeological heritage. Present-day land use in Baden-Württemberg is dominated by forestry (37.5%), agriculture (25.6%) and grassland (17.4%). While 25.4% of the forested area is coniferous and 5.8% is deciduous forest, 68.9% are mixed coniferous and deciduous forests. The high percentage of forest cover in particular has several consequences for archaeological research and heritage management. On the one hand, archaeological sites may have been preserved in areas where forest cover has been temporally continuous for centuries or millennia, while on the other hand, forest cover largely precludes the use of aerial reconnaissance for archaeological prospection. Thus, although aerial photography has a long tradition in BadenWürttemberg and the State Office for Cultural Heritage holds more than 200,000 oblique aerial photographs taken by O. Braasch, more than one third of the state is forest and therefore, until the advent of lidar, a closed book to aerial reconnaissance. In contrast, most archaeological sites on arable land and grassland (which in many cases have been ploughed in previous decades or centuries) have been levelled intentionally or

unintentionally. Written sources from the 19th century document the intentional removal of earthworks and burial mounds, for example in the vicinity of the Iron Age site Heuneburg (Kurz 2008, 194ff.). Long-term agricultural activities, in particular ploughing, have a very destructive effect on archaeological remains, levelling, disturbing and finally removing them (e.g. Oxford Archaeology 2002). Soil and cropmarks as well as surface finds show not only where a site is located but are evidence that it is being destroyed (Powlesland 2011). While sites on arable land and grassland are more accessible by aerial photography or surface survey, they often survive only as very low topographic features.

Known sites in the archaeological database Most of the known archaeological sites and find spots in Baden-Württemberg and thousands of oblique aerial photographs are recorded in a database. This contains entries derived from field survey and excavation (from research by the State Office for Cultural Heritage and from published research), aerial photographs and historical documents. However, in addition to the land use biases discussed above, there are a number of issues associated with the database that impact on its reliability. The addition of sites to the database is an ongoing task, and some publications may be overlooked. Therefore, some known sites are not (yet) incorporated in the database. For example, extensive traces of medieval iron ore mining have been investigated by fieldwork (Kempa, 2003), but are not yet included in the database (Figure 14.2). They are easily recognisable in the lidar Sky-View Factor visualisation (see below), which also allowed the identification of additional mining traces both within and beyond the research area of Kempa (2003). Accuracy of coordinates for the database entries is very variable (Figure 14.3). This is particularly true for entries based on historical documents or early research, but even for sites recorded by GPS, coordinates can be off-target by more than 30 m where satellite signal reception is poor (e.g. in dense forest). In addition, many spatially extensive sites such as groups of burial mounds are represented as points in the database which does not convey the spatial extent of the site (Figure 14.4). As the database is a tool for management and protection of sites in a

174 a

Ralf Hesse b

Figure 14.2: Mining traces on the northern slope of the Swabian Alb. (a) results of ground-based mapping by Kempa (2003, 24, Abb. 9), (b) SVF visualisation of the lidar DEM Figure 14.3: A group of burial mounds near Heidenheim (visualisation: colourcoded LRM draped over shaded relief ). Crosses indicate the positions of burial mounds as recorded in the archaeological database

Site-specific applications of lidar

Figure 14.4: A group of burial mounds near Ilshofen (visualisation: colour-coded LRM draped over shaded relief ). A single point represents the entire group of burial mounds in the archaeological database

planning and development context, these issues can become very problematic. These issues are much in mind for both site-specific applications of lidar and the state-wide lidar-based mapping project reported on here.

The Department of Archaeology of the State Office for Cultural Heritage Management in Baden-Württemberg has a relatively long history of using airborne lidar to detect and visualize archaeological sites (Bofinger et al. 2006). As early as 2003, a 23 km2 segment of the Danube valley around the Heuneburg early Iron Age hill fort in Upper Swabia (Figure 14.5) was scanned by a specialised company in preparation for extended field investigations (Bofinger et al. 2006, 2007). A second, partially overlapping, 9 km2 survey flown in 2008 extended the area to the west and north to include several burial mounds and an Iron Age rectangular bank-andditch enclosure. Both surveys had a point density of approximately four points per square metre. These two surveys played an important role in assessing the suitability of coarser resolution, but spatially more extensive and cost-efficient, lidar data. Comparison of the two Heuneburg data sets with coarser-resolution (approximately one point per square metre) state-wide lidar data available from the State Topography Authority (see below) indicated that the statewide data would normally be sufficient for the identification and visualization of archaeological structures (Bofinger and Hesse 2011). The largest site-specific lidar surveys in BadenWürttemberg target the Limes of Germania Superior and Raetia which runs through BadenWürttemberg for approximately 160 km and the Odenwald Limes with a length of approximately 40 km. These areas were surveyed in 2008

14  The changing picture of archaeological landscapes a

b

and 2010 respectively, both at a resolution of four points per square metre. The lidar scans with widths of approximately 500 m serve to document the preservation of the Limes and ancillary sites (e.g. watchtowers) as relief features and to revise and improve the historical mapping of the Limes. Until recently, the use of these data was based on single direction hill shade images (e.g. Bender 2009); however, in the context of the management of the UNESCO World Heritage Site “Frontiers of the Roman Empire” (Deutsche Limeskommission 2010), new visualisation techniques will be applied in the future.

The state-wide prospection project As outlined above the archaeological database is biased by differential preservation and visibility due to past and present land use and datacollection issues. It is thus unclear how well the database represents the known sites in terms of their location, extent and preservation as relief features, or how many otherwise unknown sites may survive in the landscape. Thus, while residential and industrial sprawl, construction of roads, railway lines and pipelines, mechanised agriculture and forestry practices, as well as looting, pose serious threats to known and unknown archaeological sites, sites cannot be protected without at very least the knowledge that they exist. This provides the background for the archaeological prospection based on airborne lidar data launched by the State Office for Cultural Heritage in May 2009. The two goals of this state-wide project are (i) the revision of location and extent of known archaeological sites and the determination of their preservation

175 c

as relief features and (ii) the full spatial coverage archaeological prospection. Lidar data covering the entire federal state are made available to the State Office for Cultural Heritage by the State Topography Authority under a General Agreement on Geodata. The data were thus produced primarily for topographic purposes, and vegetation filtering was therefore not specifically adapted for prospection of archaeological features. The data was supplied as a vegetation-filtered point cloud with an average point-to-point spacing of approximately one metre. While this is not an optimal data set for archaeological purposes, it has the advantages of being available for this project and covering the entire state. Thus, it provides for the first time the possibility of full-coverage archaeological prospection of the state using a single method – especially important as it is the first remotesensing method that allows investigation of the large areas of forest. Working with such a large, state-wide data set presents new possibilities and challenges in comparison with smaller, site-specific approaches. While time constraints are usually not a major issue in small, site-specific studies, the progress and success of a state-wide project requires an efficient workflow from the raw (vegetationfiltered point cloud) data through intelligible visualisations to time-efficient recording of the interpretation results. Furthermore, the level of detail aimed at in site-specific studies to fully utilize the potential of an expensive lidar survey can probably not be achieved in a state-wide prospection project.

Figure 14.5: Comparison of lidar shaded relief images of the Heuneburg from three different surveys. (a) survey 2003, four points per square metre, (b) survey 2008, four points per square metre, (c) data provided by the State Topography Authority, one point per square metre

176

Ralf Hesse

Figure 14.6: User interface 1

Figure 14.7: User interface 2

Data management, processing and visualisation Efficient data management, processing and interpretation are key issues in the feasibility of this project. The dataset is massive and the project has an ambitious timescale and spatial coverage. The lidar point cloud data was delivered as about 160,000 files with a total volume of more than one Terabyte. The revision of approximately

64,000 database entries and the full coverage prospection of the 35,751 km2 by a single researcher and within a tight temporal frame (the ArchaeoLandscapes Europe project which partially funds this work ends in September 2015) demands an efficient workflow. As handling each of the 160,000 files individually was not feasible, the data had to be combined at the point cloud level by an automatic process. By trial and error it was found that the available hardware and software allowed the efficient processing of files up to about 0.5 Gigabyte, which at a resolution of one metre is equivalent to 25 to 100 km2 per file. This still results in between 400 and 1600 files per raster data layer. However, as well as the lidar-based DEM visualisations and other derived data such as Local Relief Model and Sky-View Factor (see below), lidar point density, data gap maps, topographic and geological maps and orthophotos are used. Therefore, there are between eight and nine raster data layers for each 100 km2 segment. In addition to this, there are up to 22 vector data layers containing the known archaeological sites and find spots from the archaeological database, the present-day land use, built-up areas, road and railway networks and the administrative and land parcel boundaries. The large number of data layers required an efficient data management setup, achieved by the development and implementation of two graphical user interfaces. The main user interface (Figure 14.6) serves to (i) execute a number of data processing steps, (ii) document the current status of data processing and interpretation on the basis of 100 km2 data segments, (iii) display metadata such as topographic maps, average point density and number of known sites for each data segment and (iv) automatically create Global Mapper workspace files to load all required data layers for a given data segment. Data processing includes the creation of raster DEM from point cloud data, data quality analyses (point density, data gaps) and the creation of visualisations and is carried out using ENVI as well as tools implemented using VBA under MS Excel. A second user interface (Figure 14.7) serves to document the progress of interpretation on a screen-by-screen basis (using a resolution of one metre per pixel) and to interactively modify DEM illumination. It also provides templates for the quick creation of common point, polyline and polygon objects. To overcome the limitations of the con­

14  The changing picture of archaeological landscapes ventional hill-shading representation of DEM data (see ������� Kokalj� et al. this volume) and to make the interpretation of the lidar data as timeefficient as possible, two novel visualisation techniques are used in this project. The Local Relief Model (LRM; Hesse 2010) is based on the extraction of local topographic anomalies which can then be used to create greyscale or colour-coded images and to measure the heights and volumes of these anomalies. For colour-coding, using blue for negative and red to yellow for positive relief anomalies eases the intuitive reading of the images. It was found that draping the colour-coded LRM over a shaded relief DEM (cf. Figures 14.3 and 14.4) allows a strong visual accentuation of local topographic anomalies while at the same time retaining the possibility of further enhancement by changing illumination azimuth and elevation. The second novel visualisation used in this project is SkyView Factor (SVF; Zakšek et al. 2011). It is based on the modelling of diffuse ambient illumination which results in bright pixels for exposed and dark pixels for topographically enclosed areas (see Kokalj� et al. this volume). The SVF visualisation is very suitable in particular for concave features (e.g. mining traces, hollow ways; cf. Figure 14.2) and still retains a good overall impression of the topography.

Progress of the project The development and implementation of data management and processing tools as well as data processing itself were largely finished by January 2010. By September 2011, 11,800 km2 (33% of the area of Baden-Württemberg) had been analysed, both reviewing known sites and identifying previously unknown sites. Some 15,400 database entries for known sites have been reviewed, and their state of preservation as relief features recorded. Of these 15,400 sites only 6060 can reasonably be expected to be visible in the lidar data. For example, 28.0% of the entries are located in built-up areas, 8.6% are find spots rather than monuments, 3.7% are burials with no surface expression and 14.6% are former excavations, geological structures, buildings etc. Finally, for 4.9% of the entries insufficient information was available to properly locate or identify the feature. Only 0.8% of the entries are located in areas where lidar data gaps made identification impossible. This should not be

177 misunderstood as stating that data quality is not a serious issue. Rather, in many areas with low lidar ground point density (i.e., under dense forest) better data quality would be desirable in order to move beyond the mere assertion that a known site is there. Looking at some of details of the resulting data, of the 6060 revised known sites which can be expected to be visible in the lidar data, 36.0% can be classified as fortifications, of which only 13.5% are recognisable as relief features. Generally similar results are obtained for settlements (17.8% recognisable), burial mounds (13.0% recognisable), roads (23.8% recognisable), mining traces (5.8% recognisable), agricultural structures (15.8% recognisable) and charcoal burning platforms (18.5% recognisable). These low detection rates are due to different factors. While many database entries are based on excavations, historical documents or soil and cropmarks, lidar generally records only surface topography (though see Challis 2011 and this volume for discussion of intensity values) and so will not record sites with no surface expression. The available lidar ground point density of approximately one point per square metre (and less under dense vegetation) is almost certainly a factor, and an increase in ground point density can be expected to increase the detection rate. Lidar interpretation thus complements rather than replaces other prospection methods. In the same period to September 2011, 215,000 previously unknown potential archaeo­ logical features have been mapped. The use of this advisory term rather than ‘archaeological site’ reflects a number of issues, but it should be emphasised that most of these features are indeed archaeological and clearly of anthro­pogenic origin. One issue arises from the lack of chronological control and the question of how old a feature has to be to be classified as archaeological. Another issue is related to scale: while discrete features (e.g. individual mounds, field boundaries, ramparts) are mapped as individual points, lines or polygons, the term ‘site’ is more commonly understood as comprising an interrelated set of such features (e.g. a burial mound cemetery, a field system or a fortified settlement). Furthermore, some types of remains are commonly considered of no or low heritage value (e.g. thousands of charcoal burning sites or field boundaries), which many regard as ‘elements of the cultural landscape’ rather than ‘archaeological features’, let alone ‘archaeological sites’. As lidar-based elevation is almost the exclusive data source for this project,

178

Ralf Hesse

Figure 14.8: Percentages of mapped potential archaeological features classified into broad categories

Figure 14.9: Former field boundaries. (a) mapped from lidar data (blue lines; visualisation: colour-coded LRM draped over shaded relief; present-day parcel boundaries shown in red), (b) comparison of the mapped field boundaries (blue) with a historic cadastre map

a

further research methods and data sources have the potential to greatly enhance the validity of the results. Extrapolation to the entire state area suggests that more than 600,000 such sites can be expected, and taking into account that only a fraction of the existing sites are detectable in the lidar data, it is likely that an even larger number of sites will remain undetected. Figure 14.8 shows the percentages of the mapped sites by category. Most are related to agricultural activities (33.0%: former field boundaries, agricultural terraces/lynchets, ridge and furrow), followed by mounds (14.7%), mining and quarrying traces (10.2%), roads (9.3%) and charcoal burning

b

platforms (5.1%). Only 0.6% of the sites are classified as fortifications; not surprising given that these are easily recognised as monuments by ground field workers. By far the largest number of newly identified sites are traces of former resource exploitation such as agriculture (Figure 14.9), charcoal production (Figure 14.10a) and mining (Figure 14.2). Field boundaries are often characterised by the accumulation of soil along the margins of ploughed fields which can form banks between a few centimetres and more than a metre in height and these can be mapped on the basis of the LRM, while a vector data layer of present-day field parcel boundaries allows the identification of relict features (Figure 14.9a). Comparison with 18th and 19th century historical cadastre maps shows that the former field boundaries in many cases conform to historic land parcels (Figure 14.9b). However, it also becomes clear that only some of the historic land parcels are recorded as relief features. Fortifications and other enclosures are less common and include rectangular bank-andditch enclosures (Figure 14.10b, Viereckschanzen – typical for the late Iron Age in southern Germany, but in some cases possibly related to post-medieval military or forestry), promontory fortifications and fortified hilltops (Figure 14.10c, mostly prehistoric or medieval) and smaller earthworks. Traces of former road networks can appear as negative relief features as in the case of swarms of hollow ways (Figure 14.10d) or

14  The changing picture of archaeological landscapes as positive relief features as in the case of some Roman roads. In many cases hollow ways are associated with mining traces. Mounds are very numerous features and often occur in clusters. However, these raise an issue with the lidar derived results – that it is not always possible from the lidar data alone to determine whether they are burial mounds (Figures 14.3 and 14.4), other anthropogenic features (e.g. related to mining) or natural features. This is particularly true for mounds which appear to be severely eroded by agriculture.

179

The results of the lidar-based archaeological

prospection are added as potential archaeological features to the archaeological database. This reflects their identification from almost exclusively desk-based work with remotely sensed data and consequently the limited knowledge about them. In this respect, the lidar-based mapping results are similar to sites known exclusively from aerial photographs. Crucially, by adding them to the archaeological database, they can be considered in planning and development, such as recent cases of a pipeline and the new Stuttgart-Ulm railway line, as well as research. In a very few cases, particularly conspicuous or interesting features, such as fortifications, are followed up by literature searches, field visits and, in selected cases, geophysical investigations. Given the large number of mapped features, this can only be achieved for very few features.

a

b

c

d

Application of the results in heritage management

Figure 14.10: Examples for potential archaeological features. (a) charcoal burning platforms in the southern Black Forest, (b) possible Iron Age rectangular bank-and-ditch enclosure near Tuttlingen, (c) likely medieval castle site near Lake Constance, (d) hollow ways and related mining traces near Dettingen

180

Ralf Hesse

a

b Figure 14.11: Spatial distribution analyses: density of (a) charcoal burning platforms and (b) mounds in the southern Black Forest (number of features per square kilometre)

Lidar and the changing picture of archaeological landscapes The use of lidar is changing, if not revolutionising, our picture of archaeological landscapes. Not only does it contribute to the dramatic increase of recorded traces of former human activities, but also to the growing recognition that sites cannot be viewed in isolation and the rapidly expanding possibilities of analyses related to (inter-)visibility, spatial relationships, least-cost paths or resource potential (e.g. Posluschny 2008; Steiniger and Hay 2009). Context matters, and can be provided by combining all available archaeological data with

geomorphological analyses of the DEM. Blank areas on archaeological maps are disappearing, and it becomes undeniable that the present landscape is full of traces of earlier human activities. With the progress of lidar-based archaeological prospection projects like the one in Baden-Württemberg, archaeological landscapes finally can and must be seen as diachronic and coherent rather than as assorted sites separated by areas without relevance. In Ireland, for example, this was prominently shown by the controversy around the construction of a motorway in the vicinity of the Hill of Tara (Ronayne 2008). In the research domain, the sheer numbers

14  The changing picture of archaeological landscapes of potential sites may very likely lead to a stronger focus on spatial distribution patterns and statistical analyses (Figure 14.11). Despite the large number of features identified on the basis of a consistent data set and methodology, this requires an awareness of the biases and limitations of lidar data analysis, recognising the provisional nature of some of the results. Total field observation would require considerable manpower presently far beyond the capacity of the State Office for Cultural Heritage or other institutions, while wholesale excavation is neither conceivable nor desirable. Therefore, learning to deal with uncertainty may be one of the necessary consequences of the extensive application of lidar in archaeology. This is particularly true for temporal uncertainty in the often intense overlapping of traces from multiple periods and the usually unknown relationships between past landscapes and their human inhabitants, which all lead to an apparent messiness of archaeological landscapes (Mlekuž this volume). For heritage management and protection lidar data, used extensively in the manner described above, has two main impacts. On the one hand, the revision of known sites in many cases provides improved information on their location and size. This can enhance the efficiency with which heritage concerns are taken into account in planning processes. On the other hand, the large number of newly identified sites and the recognition of diachronic, coherent archaeological landscapes highlights the complex issue of setting priorities for the strongest protection and for those sites which may be sacrificed for economic development. In Germany, while there is recognition of cultural landscapes and the need for their preservation, the extent to which this includes features such as mining traces, relict agricultural remains or charcoal kiln sites, however, is limited. Urban and industrial sprawl and the construction of highways, railways and pipelines are common threats to archaeological sites and landscapes. Land consolidation, agriculture and increasingly mechanised forestry threaten and destroy sites in many other areas. In some cases, it can be argued that features which are so numerous as to be characteristic of a certain area or region should be protected so as to maintain the unique character of a certain archaeological landscape. In other cases, rare features may be the focus of attention for the simple reason that they are rare. And while this paper has demonstrated the huge

181 impact of the lidar mapping project in BadenWürttemberg on the archaeological database, the concomitant focus on topography entails the risk that sites without a surface expression as relief features remain overlooked.

Conclusions Airborne lidar is becoming increasing important to archaeological prospection, both in research and for heritage management. While dedicated lidar surveys with high point density and optimised vegetation filtering are desirable, lidar data produced for other purposes – in particular topographic survey – are often the only choice for financial and spatial coverage reasons. An ambitious project aimed at the fullcoverage lidar-based archaeological prospection of Germany’s third-largest federal state of BadenWürttemberg shows that such data can be used effectively for archaeological purposes. Given the very large area to be prospected, efficient data management, processing and interpretation are required. In the case of the presented project, this is achieved by the implementation of user interfaces, data processing workflows and the use of new visualisation techniques such as LRM and SVF. In the course of the project, the location and extent of all known archaeological sites are reviewed and previously unknown sites are mapped. This will improve the quality of the existing archaeological database and adds several hundred thousand otherwise unknown archaeological features which can be taken into consideration in archaeological research, heritage protection, planning and development. The lidar data emphasises the need to develop awareness of how to deal with sites whose interpretation or chronology is uncertain, to address the need to prioritise protection efforts and to raise new concerns regarding the protection and protectability of archaeological landscapes.

References Bender, S., 2009. Spuren von Kleinkastellen beim Gleichener See? Der Limes 3, 10–1. Biel, J., 1987. Vorgeschichtliche Höhensiedlungen in Südwürttemberg-Hohenzollern. Forschungen und Berichte zur Vor- und Frühgeschichte in BadenWürttemberg 24. Theiss: Stuttgart. Bittel, K., Müller, D. and Schiek, S., 1990. ���� Die keltischen Viereckschanzen. Atlas archäologischer

182

Ralf Hesse Geländedenkmäler in Baden-Württemberg. Theiss: Stuttgart. Bofinger, J. and Hesse, R., 2011. As �������������� far as the laser can reach… Laminar analysis of LiDAR detected structures as a powerful instrument for archaeological heritage management in BadenWürttemberg, Germany. In Cowley, D.C., (ed.). Remote sensing for archaeological heritage management. Proceedings of the 11th EAC Heritage Management Symposium, Reykjavík, Iceland, 25–27 March 2010. 161–71. Bofinger, J., Kurz, S. and Schmidt, S., 2006. �������� Ancient maps – modern data sets: different investigative techniques in the landscape of the Early Iron Age princely hill fort Heuneburg, Baden-Württemberg. In Campana, S. and Forte, M., (eds). From Space to Place. 2nd International Conference on Remote Sensing in Archaeology. BAR International Series 1568. Archaeopress, Oxford. 87–92. Bofinger, J., Kurz, S. And Schmidt, S., 2007. ������ LiDAR – High Resolution Raster Data as a survey tool. In Figueiredo, A. and Velho, G.L., (eds). The world is in your eyes. Proceedings of the XXXIII Computer Applications and Quantitative Methods in Archaeology Conference, Tomar, Portugal 2005, Lisbon 2007. 255–60. Deutsche Limeskommission, 2009. Der Limes. Nachrichtenblatt der Deutschen Limeskommission 3(2). Deutsche Limeskommission, 2010. ManagementPlan 2010–2015. UNESCO-Welterbe “Grenzen des Römischen Reiches: Obergermanisch-Raetischer Limes”/UNESCO-World Heritage Site “Frontiers of the Roman Empire: Upper German-Raetian Limes”. Beiträge zum Welterbe Limes, Sonderband 1 (bilingual document available for download at: http://www.deutsche-limeskommission.de/ fileadmin/dlk/images/dlk/pdfs/ManagementPlan-2010-2015.pdf ). Devereux, B.J., Amable, G.S., Crow, P. and Cliff, A.D., 2005. The potential of airborne lidar for detection of archaeological features under woodland canopy. Antiquity 79, 648–60. Doneus, M. and Briese, C., 2011. Airborne Laser Scanning in forested areas – potential and limitations of an archaeological prospection technique. In Cowley, D.C., (ed.). Remote sensing for archaeological heritage management. Proceedings of the 11th EAC Heritage Management Symposium, Reykjavík, Iceland, 25–27 March 2010. 17–32. Doneus, M., Briese, C., Fera, M. and Janner, M., 2008. Archaeological prospection of forested areas using full-waveform airborne laser scanning. Journal of Archaeological Science 35, 882–93. Gallagher, J.M. and Josephs, R.L., 2008. Using LiDAR to detect cultural resources in a forested environment: an example from Isle Royale National Park, Michigan, USA. Archaeological Prospection 15, 187–206. Hartke, W., 1954. Über die “Ackerberge” und ihre Bedeutung als Index für das Alter agrarlandwirtschaftlicher Grenzen. Bemerkungen

zu einem Buch von E. Juillard. Zeitschrift für Agrargeschichte und Agrarsoziologie 2, 173–7. Hesse, R., 2010. LiDAR-derived Local Relief Models – a new tool for archaeological prospection. Archaeological Prospection 17, 67–72. Kempa, M., 2003. Archäologische Untersuchungen an früh- und hochmittelalterlichen Verhüttungs­ plätzen. Abbau und Verhüttung von Eisenerzen im Vorland der mittleren Schwäbischen Alb. Forschungen und Berichte zur Vor- und Frühgeschichte in BadenWürttemberg 86, 9–116. ������������������ Theiss: Stuttgart. Kimmig, W., 1966. Der Kirchberg von Reusten. Urkunden zur Vor- und Frühgeschichte aus Südwürttemberg-Hohenzollern 2. Kortüm, K., 1995. Portus Pforzheim. Untersuchungen zur Archäologie und Geschichte in Römischer Zeit. Thorbecke: Sigmaringen. Kurz, G., 2008. Ein Stadttor und Siedlungen bei der Heuneburg (Gemeinde HerbertingenHundersingen, Kreis Sigmaringen). Zu den Grabungen in deer Vorburg von 2000 bis 2006. In Krausse, D. (ed.). Frühe Zentralisierungsund Urbanisierungsprozesse. Zur Genese und Entwicklung frühkeltischer Fürstensitze und ihres territrorialen Umlandes. Kolloquium des DFGSchwerpunktprogram������������������������������ ms 1171 in Blaubeuren, 9.–11. Oktober 2006. Forschungen ���������������������������������� und Berichte zur Vorund Frühgeschichte in Baden-Württemberg 101, 185–208. ������������������ Theiss: Stuttgart. Kurz, S., 2010. Zur Genese und Entwicklung der Heuneburg in der späten Hallstattzeit. In ������������ Krausse, D. and Beilharz, D. (eds). “Fürstensitze” und Zentralorte der frühen kelten. Abschlusskolloquium des DFG-Schwerpunktprogramms 1171 in Stuttgart, 12.–15. Oktober 2009. Forschungen und Berichte zur Vor- und Frühgeschichte in Baden-Württemberg 120, 239–56. Theiss. ����������� Stuttgart. Ludemann, T., 2010. Past fuel wood exploitation and natural forest vegetation in the Black Forest, the Vosges and neighbouring regions in western Central Europe. Palaeogeography, Palaeoclimatology, Palaeoecology 291, 154–65. Oxford Archaeology, 2002. The management of archaeological sites in arable landscapes. BD1701, CSG15. Final Project Report (available upon request from the UK Department for Environment, Food and Rural Affairs, www.defra.gov.uk). Pirling, R., 1980. Die Mittlere Bronzezeit auf der Schwäbischen Alb. PBF XX 3: Munich. Posluschny, A., 2008. Sehen und gesehen werden – Sichtbarkeitsanalysen als Werkzeug archäologischer Forschungen. In Krausse, D. and Steffen, C. (eds). Frühe Zentralisierungs- und Urbanisierungsprozesse – Zur Genese und Entwicklung frühkeltischer Fürstensitze und ihres territorialen Umlandes. Kolloquium des DFG-Schwerpunktprogramms 1171 in Blaubeuren, 9.–11. Oktober 2006. 367–80. Theiss: Stuttgart. Powlesland, ��������������������������������������� D., 2011. Identifying the unimaginable – managing the unmanageable. In Cowley, D.C., (ed.). Remote sensing for archaeological heritage management. Proceedings of the 11th EAC

14  The changing picture of archaeological landscapes Heritage Management Symposium, Reykjavík, Iceland, 25–27 March 2010. 17–32. Ronayne, M., 2008. Commitment, objectivity and accountability to communities: priorities for 21stcentury archaeology. Conservation and Management of Archaeological Sites 10(4), 367–81. Simms, A., 1976. Deserted medieval villages and fields in Germany, a survey of the literature with a select bibliography. Journal of Historical Geography 2(3), 223–38. Steiniger, S. and Hey, G.J., 2009. �������������� Free and open source geographic information tools for landscape ecology. Ecological Informatics 4, 183–95.

183 Steuer, H., 1997. Herrschaft von der Höhe. Vom mobilen Söldnertrupp zur Residenz auf repräsentativen Bergkuppen. Die Alamannen. Ausstellungskatalog. 149–62. Theiss: Stuttgart. Wieland, G., 1996. Die Spätlatènezeit in Württemberg: Forschungen zur jüngeren Latènekultur zwischen ��������� Schwarzwald und Nördlinger Ries. Forschungen und Berichte zur Vor- und Frühgeschichte in BadenWürttemberg 63. ������������������� Theiss: Stuttgart. Zakšek, K., Oštir, K. and Kokalj, Z., 2011. Sky-view factor as a relief visualization technique. Remote Sensing 3, 398–415.

15 Lidar in Mediterranean agricultural landscapes: reassessing land use in the Mauguio Nicolas Poirier, Rachel Opitz, Laure Nuninger and Krištof Ostir The value of lidar data for the study of forested areas has been repeatedly demonstrated, recording large numbers of previously unknown sites and features where most survey methods are ineffective. However, the question of the value of lidar in cultivated areas already investigated through historical mapping and archaeological studies such as field walking survey, aerial photographic survey and excavation remains open. This paper summarizes the results of recent research in the open and heavily cultivated Mauguio region of southern France and reflects on the challenges of integrating information from lidar survey into an existing body of knowledge on the development of an agricultural landscape. Keywords: lidar, Mediterranean, agricultural landscapes, visualization

Introduction This paper presents an assessment of the added value of lidar data for the study of a Mediter­ ranean landscape in the Mauguio area of the French Languedoc which has been intensively studied by archaeologists since the 1980s, primarily through the substantial field walking surveys led by F. Favory and C. Raynaud (Favory et al. 1994). This area was further studied through a number of thematic projects at national and European level, enhancing the information about land use and settlement patterns in this region area through regional and inter-regional comparanda. In focusing on an agricultural Mediterranean landscape this paper highlights a regional divide in what might be termed the fundamental building blocks of knowledge about landscape history. Earthworks, ranging from ridge and furrow field systems to massive Bronze and Iron Age hillforts have long been essential to the character and archaeological description of landscapes throughout northern and central Europe. Landscape archaeology and the practice of survey in the Mediterranean, on the other hand, developed within a different research

tradition – one heavily dependent on the ceramic and stone remains found through field walking. Given a tradition of studying rural landscapes that does not take as much account of topography, what kind of role can lidar data play and how should its evidence be balanced with that from traditional data sources? While many lidar surveys are commissioned, or existing data acquired, with the primary aim of identifying new sites (in the traditional sense of the term) or documenting existing ones, for this project a lidar survey in 2007, covering a block of about 40 km², was undertaken with the specific aim of creating an accurate DTM to better understand changes in fluvial patterns, drainage systems and the evolution of the lagoon coastline. Following the completion of an initial palaeo-hydrological study, a second project (ModAgSpace, Marie Curie PIEF-GA-2009236681) was established to explore the relationship between archaeological features visible in the lidar, many of which were interpreted as related to agriculture or drainage, and land use patterns as understood from previous archaeological research. The relationship between areas characterized by extensive agrarian manuring spreads, indicative

15  Lidar in Mediterranean agricultural landscapes: reassessing land use in the Mauguio

of intensive or long term agricultural use, and the visibility of features in the lidar data was of particular interest. Both projects focused on the agricultural landscape and the evolution of land use, rather than the settlement pattern (the usual focus of survey in Mediterranean archaeology), setting lidar up to play a complementary rather than augmentary role in landscape scale research

in the region. In establishing the lidar survey as a source of complementary knowledge, technical challenges were posed by visual noise in the data caused by deep plough marks, and interpretive challenges arose from the process of integrating the detected features into existing datasets and assessing the value of such information in a well studied area.

185

Figure 15.1: Location of the study area

186

Figure 15.2: Detection of micro-relief

Nicolas Poirier, Rachel Opitz, Laure Nuninger and Krištof Ostir

Geographic context The Mauguio study area is located in the Languedoc region of southern France, part of a plain bordering the Mediterranean Sea. It is very flat, lying just a few metres above sea level, surrounding a coastal lagoon. This marshy area was heavily managed during the past and is currently occupied by scrubland, vineyards and agricultural land (Figure 15.1).

History of research Knowledge of local settlement history and land use is primarily based on the work of Favory and Raynaud (Favory et al. 1994) and the Archaeomedes Project undertaken in the 1990s (Van der Leeuw et al. 2003). The area was also studied through limited excavation, and satellite and aerial imagery were used to identify field systems and roads (Favory and Raynaud 1992). In the 2000s this region was studied using remote

15  Lidar in Mediterranean agricultural landscapes: reassessing land use in the Mauguio

187 Figure 15.3: Testing the Automatic Periodic Noise Removal – an earlier pattern of cultivation is revealed (right)

Figure 15.4: Testing two hillshading methods: Swiss Hillshade Model (left) and MDOW (right)

sensing data to detect palaeo-channels and develop a reconstructed palaeo-DTM (Nuninger and Oštir 2005). The results of all these projects allow us to trace the general trends of long-term land use. While the earliest remains date from the Palaeolithic period, it is in the early Iron Age that settlement and activity along the banks of the coastal lagoon increases significantly. The Roman period saw further agricultural intensification on the plain, with the introduction of a cadastral land division system, an extensive settlement network and widespread drainage infrastructure installed. These developments totally changed the environmental dynamics of the area. The Roman drainage systems, roads and land divisions established the framework of the agricultural landscape for subsequent periods,

and they remain apparent in the contemporary landscape. DTM preparation in areas of intensive agriculture The high resolution of the lidar DTM (0.5 m) is appropriate for the detection of microtopographic features such as field boundaries, drainage ditches or ancient roads. However, this precision can be problematic in fields under cultivation because plough furrows, vineyards, and other artefacts of modern land use which are not easily separated from the ‘interesting’ ground surface (Figure 15.2) are also recorded in great detail. And where the agricultural noise is greatest in areas of heavy ploughing archaeological relief will be at its most subtle due to heavy erosion. Identifying subtle traces of past activity in the

188

Nicolas Poirier, Rachel Opitz, Laure Nuninger and Krištof Ostir present-day agricultural landscape therefore required careful filtering and visualization to remove noise and improve visibility. To facilitate the visual recognition of archaeological features in the plough zone, the removal of linear ‘plough noise’ using established image processing techniques was attempted. An initial effort (Nuninger et al. 2008, 36–9) demonstrated that Sobel edge detection, low-pass filters and directional filters, as implemented in Erdas Imagine, did not significantly improve the visibility of archaeological features. However, an approach based on Fourier analysis was more successful in removing the traces of ploughing (Figure 15.3). Two approaches were taken. The application of an ‘Automatic Periodic Noise Removal’ algorithm in Erdas Imagine, requiring a minimum of user input, was applied in the first instance. In cases where this did not improve the appearance of the image, manual editing in Fourier space to remove the periodic plough noise was performed. Removing the linear pattern created by the most recent phase of ploughing sometimes revealed evidence of earlier phases of cultivation and differences in land use in the form of changes in field size, row spacing and orientation of cultivation.

Visualisation in the plough zone The hillshading model applied always influences the visibility of features in a lidar DTM, and in heavily cultivated landscapes this effect is magnified by the shadows created by plough marks. Because fields may be ploughed in a variety of directions, any directional hillshading model applied over a reasonably large area will cause serious shadowing effects in some fields where it is near-perpendicular to the direction of the ploughing. Consequently, while simple directional hillshades are often most effective, in this situation we tested two complex hillshading models: the Swiss Hillshade Model and the MDOW (Multi-Directional Oblique Weighting) Hillshade Model (Figure 15.4). The Swiss Hillshade Model is designed to, “emphasize the major geographic features [and] minimize the minor features, smooth irregularities on the slopes, but maintain the rugged characteristics of ridge tops and canyon bottoms… You can then simulate an aerial perspective that makes the higher elevations lighter and the lower

elevations darker” (Barnes 2002). On first assessment, this model may not seem suitable for processing digital terrain models to detect archaeological entities, since it is supposed to highlight major geographic features, minimize the margins and smooth out slope irregularities. However, in an area dominated by ploughing it ameliorates the problems introduced by a directional hillshade. Further, by lightening shading on the highest altitudes, and darkening shading on lower elevations, absolute elevation differences are highlighted. In an area with a large elevation range such shading will not show up archaeological features, but it can be effective on very flat terrain where relief is very low (a few centimetres elevation change across a feature of several metres extent) and the total elevation variation is very small. In contrast, the Multi-Directional Oblique Weighting (MDOW) Hillshade Model is theoretically an answer to the problem of, “traditional computer-generated shaded-relief maps [which] emphasize structures that happen to be obliquely illuminated, but wash out structures that are illuminated along the structural grain. This […] technique, which emphasizes oblique illumination on all surfaces, provides more detail in areas of an image that would otherwise be illuminated by direct light or left in darkness by a single source illumination.” (Robert Mark, USGS, http:// pubs.usgs.gov/of/1992/of92-422/of92-422.pdf ) Testing this model revealed that rather than improve visibility in areas with multi-directional ploughing, noise associated with the plough lines in the image increased because more fields were affected by shading perpendicular to the direction of their ploughing. While the Swiss Hillshades were judged to be useful, initial assessment did not indicate that any single visualization of the terrain consistently provided the best results. Following the practice of many projects a combined approach was taken, including using directional hillshades and other visualizations of the terrain in the course of interpretation. Successfully working with a variety of visualizations relies on experience interpreting topography and archaeological knowledge (see Halliday this volume) more than it depends on technical prowess or excessive computing power. Hillshades and other terrain models can be easily produced using batch scripts and storage space for the files is increasingly inexpensive and readily obtained. The challenge is in knowing which visualization is likely to be

15  Lidar in Mediterranean agricultural landscapes: reassessing land use in the Mauguio

most useful for any area of landscape, in being able to look at a visualization and, based on its utility, select the next one to apply to the same area, avoiding having to systematically look through every rendering.

Integrating lidar with existing knowledge After processing to remove noise and create good visualizations as described above, a large number of features associated with agriculture, drainage and land division were identified. An essential first step is discriminating between features that document past activity and those which

189

Figure 15.5: Palaeofeatures detected and orthophoto coverage

190

Figure 15.6: Palaeofeatures detected overlain on the Napoleonic cadastre

Nicolas Poirier, Rachel Opitz, Laure Nuninger and Krištof Ostir

are modern. Establishing chronology is always problematic when working with aerially acquired data. In this case, orthophotographic coverage (Figure 15.5) and the Napoleonic cadastre were used to assess relatively recent changes in the landscape and attempt to identify features associated with these changes before undertaking a comparison with data from field walking survey. On the basis of this exercise 85 features were identified as being archaeological and previously

undocumented. These were dominated by linear features such as field boundaries, drainage ditches and paths. Comparison with the Napoleonic cadastre and orthophotographs The Napoleonic cadastre was compiled for France and many other countries under the Napoleonic Empire during the course of the 19th century, continuing even after the end of the Empire. In

15  Lidar in Mediterranean agricultural landscapes: reassessing land use in the Mauguio

this part of southern France these maps were made in 1811 (AD34, 3P3424) and for part of the study area the maps are available in digital form. These were georeferenced and all the field boundaries were digitized. Within the area covered by the digital maps, six of the 24 lidar features (25%) are depicted on the cadastral map, corresponding to established field boundaries. One of these is also visible on the orthophoto coverage (Figure 15.6).

Evidence for early field systems Field walking campaigns between 1986 and 1992 recorded both site (settlements) and off-site (manuring scatters) materials. In the eastern Languedoc almost 300 archaeological features were identified, the majority of which are characterized as settlements. The manuring scatters, while fewer in number, occupy a greater area and consequently play an important role in characterizing the land use history of the

191

Figure 15.7: Field walking data

192 Figure 15.8: Roman cadastre, Napoleonic cadastre and palaeofeatures

Figure 15.9: Comparing the orientations of the Roman cadastres, Napoleonic cadastre and features identified through the lidar survey

Nicolas Poirier, Rachel Opitz, Laure Nuninger and Krištof Ostir

15  Lidar in Mediterranean agricultural landscapes: reassessing land use in the Mauguio

193 Figure 15.10: Distance analysis between lidar features, field walking features and random points

Figure 15.11: Analysis of the durability index and density of lidar features

194

Nicolas Poirier, Rachel Opitz, Laure Nuninger and Krištof Ostir

Figure 15.12: Dispersion graph showing durability index values and lidar features densities

region. These scatters cover over 900 hectares and date from the Iron Age to the Modern Era (Figure 15.7). In the study area defined by the lidar survey, two field systems dating from the Roman period were identified from aerial imagery, historic maps and excavations. These systems, known as Sextantio-Ambrussum and Nîmes A, strongly influenced the development of subsequent field systems. The earlier arrangement, SextantioAmbrussum, is oriented at 22°30’W while the Nîmes A is oriented at 30°30’W. Global orientation Since the features detected from the lidar data were almost all linear features interpreted as field boundaries, their individual and global orientations were studied. The directional spectrum of the features identified in the lidar data was compared with those of the field systems shown on the 19th century cadastral map and the two Roman cadastres discussed above (Figure 15.8). The directionality of these field systems was assessed by identifying the orientation of each linear feature in the field systems (i.e. individual field boundaries) and the results for each system were compared graphically (Figure 15.9). Based on the global similarities in orientation we conclude that many of the features identified in the lidar models are related to the Roman systems. For the lidar-derived features the most frequent orientation (66°;70°) matches that of the Sextantio-Ambrussum (67.5°) cadastre and

the second most common orientation (59.5°)� �������� corresponds to the primary direction of the Nîmes A cadastre ���������������������������� (Ouriachi 2009, 319; Favory et al. 1985; Favory 1997)������������������� . While it appears that the majority of these features respect the orientations of the Roman cadastres, there are also important differences. For example, features oriented between 1° and 10° appear frequently in the lidar data, but not in the Napoleonic cadastre, the Roman cadastres or contemporary field systems. Based on the graphical analysis, we have noted a global correlation between the Napoleonic cadastre and the models proposed for two Roman cadastres. Based on this similarity we suggest that elements of the system of land division established in the Roman period persisted into the 19th century, and its global orientation and structure continue to influence the current landscape. Alternatively, it can be proposed that the various landscape organizations are all influenced by similar environmental and topographic factors, leading to broadly similar solutions for landscape organization. Proximity to sites identified through field walking Following assessment of the orientation of the field systems, the locations of individual features identified in the lidar were compared with data from field walking. Here we assume that consistent spatial proximity, compared to a random distribution, may suggest a link between the presence of features in the lidar data and sites

15  Lidar in Mediterranean agricultural landscapes: reassessing land use in the Mauguio found through field walking. That is not to say that individual features identified in the lidar should be specifically associated with nearby remains from field walking, but that the overlap between the overall patterns from the two data sources allows us to identify more and less active areas of the landscape. Thirty sets of random points with the same count as the lidar-derived feature set were generated within the spatial extent of the lidar coverage and features identified in the lidar data were assigned centroids. The minimum distance between each centroid and the closest point indicating a feature identified through field walking was calculated (Figure 15.10). The analysis was then repeated measuring the minimum distance between each point feature identified through field walking and the nearest point in each random set. The results of this analysis show a difference between the mean minimum distances observed for features seen in the lidar and those for random point sets. The smaller average minimum distance for the lidar features indicates a potential link between the location of these features and field walking sites. Based on this analysis it was suggested that concentrations of features from the two datasets, representing increased activity in the landscape, should be expected to appear together. Areas with evidence of long term agricultural activity To further investigate long term land use dynamics a second analysis was conducted comparing the average duration of agrarian occupation, based on evidence from manuring scatters (as defined by the Archaedyn project, Poirier and Tolle 2008) and the density of lidar features (Figures 15.11 and 15.12). Based on a global comparison there is no demonstrable correlation between the average durability of agricultural activity as represented by manuring scatters and the density of lidar features detected (Figure 15.12). However, further analysis based on refined chronologies of both the lidar features and manuring areas, or analysis based on another metric for concentrations of lidar features, may provide different results.

Conclusions and future directions The project described here sought to improve

knowledge of past agricultural and land management systems in the contemporary agricultural landscape of the Mauguio, combining lidar data, aerial photographs, field walking data and historic maps. Based on the history of research in the Mauguio area, and no doubt influenced by the general habits of researchers who work on Mediterranean landscapes, the approach taken here was one of making the lidar data ‘fit’ with existing evidence from field walking and historical map sources on which rural Mediterranean studies often depend, and noting where it differed. The testing of ‘fit’ between lidar and other data sources was carried out primarily through formal and graphical spatial analysis. This approach, while contributing some new knowledge in this case, presents two notable problems. First, through spatial analyses concentrations of undated features identified in the lidar data were (loosely) related to concentrations of dated archaeological or historical features. This kind of proximity based analysis is inherently problematic. It is well known that individual features that appear together are not necessarily related, and it may be that even general trends in concentrations of features should not be correlated. Even where spatial analysis legitimately suggests how features identified through lidar survey might fit with patterns identified from historical maps and field walking data, it is usually suggested that further study in the field is needed to understand the character, function and chronology of the lidar features. While recognizing the problems involved in interpreting lidar data partly on the basis of spatial analyses, precluding spatial analysis entirely does not seem the way forward, as doing so would seriously limit the possibilities for formally integrating evidence from lidar surveys with other datasets and the creation of new knowledge. Equally problematic is the implicit treatment of evidence from the lidar survey as secondary to field walking and historic mapping data, and the underlying expectation that it will augment, rather than contradict, existing datasets. In wooded areas where little previous archaeological research has been undertaken, evidence from field walking, historic maps and lidar may easily find themselves on equal footing as new knowledge is developed. Landscapes with long research histories start us off with a certain set of prejudices and ideas. By sheer quantity field walking data dominates the available information

195

196

Nicolas Poirier, Rachel Opitz, Laure Nuninger and Krištof Ostir for the Mauguio study area, and as it has been the subject of extended study across several projects it is unsurprising that as interpreters we are more confident in our assessment of this material. But does this mean that we should necessarily be surprised if the evidence derived from the lidar survey points to different, or just partially divergent, conclusions than those drawn by previous studies? Clearly not, and the need to be willing to revise our interpretations over time is well argued by Halliday and Palmer (this volume). At the same time, it is important to consider whether it is our original interpretation or our understanding of the new evidence that should be questioned. The challenge of integrating new evidence and reassessing our confidence in different data sources occurs whenever an area or subject is restudied because new information has come to light. Meeting this challenge is essential if we wish to use lidar data to study multifaceted topics like long term land use and if we plan to include lidar in ongoing research on the rural Mediterranean, where multiple sources of information and complex histories of research abound.

References Barnes, D., 2002. Using ArcMap to Enhance Topographic Presentation. Cartographic ������������ Perspectives 42, 5–11. Favory, F. and Raynaud, C., 1992. La production du paysage en Langudoc oriental dans l’Antiquité et au Moyen Age: Etude de Mauguio (Hérault). Mappemonde, no. 1, 12–6.

Favory, F., Girardot, J. and Raynaud, C., 1994. L’habitat gallo-romain autour de l’étang de l’Or (Hérault). Hiérarchie, dynamique et réseaux du 2è s. av. au 5è s. ap. J.-C. Mélanges Pierre Lévêque 8, Paris, 123–215. Nuninger, L. and Oštir, K., 2005. Contribution à la modélisation des paléo-reliefs de la plaine littorale de l’Etang de Mauguio (Languedoc, France): premières approches par télédétection. In Berger, J., Bertoncello, F., Braemer, F., Davtian, G. and Gazenbeek, M., (eds). Temps et espaces de l’homme en société, analyses et modèles spatiaux en archéologie, actes des rencontres 21–23 octobre 2004, APDCA, Antibes, 123–34. Ouriachi, M.J., 2009. Habitat, terroirs et territoire en Languedoc oriental durant l’Antiquité : approche spatio-temporelle d’un système de peuplement, PhD thesis, University of Franche-Comté, available at: http://tel.archives-ouvertes.fr/tel-00429724/fr/. Poirier, N. and Tolle, F., 2008. Measurements of Diachronic Stability of Agrarian Exploitation. In Posluschny, A., Karsten, L. and Herzog. I., (eds). Layers of Perception. Proceedings of the 35th Computer Applications and Quantitative Methods in Archaeology Conference, Berlin, Germany, April 2–6, 2007 (Kolloquien zur Vor- und Frühgeschichte, vol. 10), Habelt, Bonn. Poirier, N., Opitz, R., Nuninger, L. and Oštir, K., 2010. The ModAgSpace Project: Lidar Data and Landscape Archaeology in Southern France. In Proceedings of CAA 2010, Granada. Sittler, B. 2004. Revealing historical landscapes by using airborne laser scanning. A 3-D model of ridge and furrow in forests near Rastatt (Germany). International Archives of Photogrammetry (ISPRS), 26, 258–61. Van der Leeuw, S., Favory, F. and Fiches, J., 2003. Archéologie et systèmes socio-environnementaux. Etudes multiscalaires sur la vallée du Rhône dans le programme ARCHAEOMEDES, CNRS, Paris.

16 Using lidar as part of a multi-sensor approach to archaeological survey and interpretation Rebecca Bennett, Kate Welham, Ross A. Hill and Andrew Ford The use of airborne remote sensing techniques such as lidar and spectral imaging has found increasing popularity in the historic environment sector over the past decade. Many landscape projects across Europe are incorporating archive digital airborne survey and increasing numbers are commissioning bespoke survey. This is particularly true of lidar data, but despite a number of promising applications, digital spectral surveys (referred to as multi or hyperspectral imaging) have been less frequently utilised and our understanding of the full potential of these rich data sources and how they might best be combined is still in its infancy. Although often compared favourably with traditional aerial photographic survey, each sensor by itself captures only a portion of what can be recognised as being of archaeological significance (through soil and crop marks and earthworks). As no single airborne sensor records all of the indicators we understand to represent archaeological remains, the strength of these techniques has to be in their complementarity to each other. This chapter will explore the developments in the field of airborne multisensor survey in recent years, illustrated by a case study from a grassland environment in the UK as part of a project established at Bournemouth University in 2009. The aim of the project is to develop an approach to multisensor airborne archaeological survey that makes the full information content of topographic and spectral survey accessible to heritage professionals. The contribution of both archive and bespoke lidar to multi-sensor survey will be examined, as well as the quality of the information derived from airborne digital techniques by comparison to the existing archaeological record from previous fieldwalking, excavation and extensive aerial photographic survey. Finally, some of the issues encountered will be discussed along with some potential ways forward for using combined airborne remote sensing data in landscapes that are not dominated by arable production. Keywords: Lidar, multispectral, hyperspectral, airborne remote sensing, grassland

Introduction In the field of archaeological prospection, applications of airborne data, specifically that gathered by high resolution topographic and spectral sensors, have increased exponentially in the last decade. These sensors, once the preserve of environmental and geographical scientists, have become highly desirable tools to assess the historic environment at landscape scale. While

our understanding of their application moves forward with each archaeological case study into which they are incorporated, the full information content of these data remains unexploited. The authors argue that one reason for this is the limited number of multi-sensor surveys that have been undertaken using airborne data. Assessing the localised changes in soil, vegetation and topographical properties that we classify as

198

Figure 16.1: Stacking

of airborne and ground based data for landscape survey

Rebecca Bennett, Kate Welham, Ross A. Hill and Andrew Ford

archaeological features using a sensor which can only adequately capture one of these parameters will no doubt detect some features yet leave others undetected. This applies equally to lidar survey undertaken without aerial photographic analysis, as to hyperspectral data assessed without lidar. In all cases the complementarities of the airborne sensors means that the sum of understanding derived from application of multiple data sets is far greater than that from a single data set. In the following discussion the theory and background of multi-sensor airborne surveys for archaeological prospection is examined, including the perceived benefits, work in the field to date and the factors which inhibit multisensor survey. The authors then present a case study of the application of multi-sensor survey for the management of a temperate grassland environment on the Salisbury Plain, UK, concentrating specifically the need for multisensor survey in challenging environments and on the contribution of lidar data to the overall results of the research.

Airborne multi-sensor survey Airborne remote sensing techniques such as lidar and spectral imaging have found increasing popularity in the historic environment sector over the past decade. Many landscape projects across the world are incorporating archive digital airborne survey and increasing numbers are commissioning bespoke survey. This is particularly true of lidar data, but despite a number of promising applications, digital spectral surveys (referred to as multi or hyperspectral imaging) have been less frequently utilised and our understanding of the full potential of these

rich airborne data sources and how they might best be combined is still relatively poor. Multi-sensor survey is in many respects a natural progression of the multi-method investigation of historic landscapes to encompass the large amount of high resolution digital data that can now be recorded thanks to developments in sensor technology. However where once a single type of airborne data, typically aerial photographs, would have been used in combination with ground-based techniques such as fieldwalking or targeted geophysical survey, now archaeological projects are increasingly looking to use a variety of airborne and ground-based digital data, and require ways to compare the data from a variety of sensors (Figure 16.1). The main benefit of a multi-sensor approach to archaeological survey lies in the complexity and variability that is characteristic of past human interaction with the landscape. Although a useful shorthand, the term ‘archaeological features’ does not convey the complexity or variety of relict remains of the historic environment. In reality these features vary hugely in topology, topography and structure and they can be apparent as direct changes to the surface of the land or as proxy changes to soil and vegetation caused by sub or near surface features. They may not be visible at all, masked by soil, vegetation or other environmental conditions. In all likelihood they will also have been altered, or be in a state of alteration, by taphonomic processes. It is clear that no single sensor could detect such a range of characteristics. The strength of multi-sensor survey therefore is the complementarity of the data that can be collected by deploying multiple sensors, thereby allowing different characteristics of the archaeological features to be detected. This complementarity can lead not just to improved rates of detection but to a better understanding and interpretation of the features detected and their surroundings. Archaeological prospection in a landscape is a selective process (Halliday this volume); picking out by eye or aided by automated processes, the features that have attributes that we believe represent archaeological remains. As with any decision-making process, the more information that can be gathered about these areas the better informed their interpretation can be. Although recent advances in sensor technology have removed some of the barriers to simultaneous collection of airborne spectral and lidar data, the specification and application of multiple

16  Using lidar as part of a multi-sensor approach to archaeological survey and interpretation airborne datasets are not without challenges. The choice of sensor and its calibration to detect archaeological features in a given landscape is determined predominantly by the nature of the features that are anticipated, and in reality our current understanding of how best to apply the technology is limited. Reasons for this are discussed in detail elsewhere (Bennett et al. 2011), but many of the problems stem from the use of archive airborne data that has been collected, processed and visualised for other purposes, such as environmental and hydrological survey, without assessment of the impact of the decision making processes behind the final product. Despite increasingly widespread use there has been very little quantitative study to assess the impact of visualisation techniques on the accuracy of feature mapping and interpretation from remote sensing data. The situation is improving with recent publications by Challis et al. (2011), Bennett et al. (2012) and Štular et al. (2012) providing significant contributions to the impact of processing on lidar visualisations along with the study of the impact of processing techniques for airborne spectral data in grassland environments (Bennett 2011; Bennett et al. 2012b; Bennett et al. forthcoming). Add to these technical considerations the impact of geology, soils, season and rainfall and the variety of factors affecting the detectability of a feature by a particular sensor becomes very complex. Use of multiple sensors can help to pick apart some of these factors, especially so if the surveys are contemporary. Some of the barriers to the use of multisensor survey for any area are clear. Firstly, data of the quality, timeframe and resolution may not be available from archive sources and is often prohibitively expensive to commission. Secondly, obtaining contemporary datasets for ground to airborne data comparison is logistically challenging yet essential for certain datasets like earth resistance where results are highly condition dependent. Thirdly, the large quantities of data produced by this type of comparative analysis are difficult to manage without specialist software and data storage capacity. Finally and crucially, our understanding of how to utilise the data from some airborne platforms is in its infancy and this is especially true of digital spectral data. This can make efficient extraction of useful information from this wealth of data impossible for nonspecialist users. For these reasons, published studies comparing airborne sensors have tended to be limited to

just two datasets; most often the comparison of lidar data with archive aerial photographs (e.g. Bewley et al. 2005; Challis et al. 2008), although there has also been some investigation of the correlation between different hyperspectral sensors (Powlesland et al. 2006; Traviglia 2006; Challis et al. 2009). There have been far fewer opportunities for detailed examination of correlation between sensors of different types. Where both spectral and topographic data have been available, such as in a study of the Salisbury Plain Training Area (Barnes 2003), the combined analysis has focused on ascertaining land cover categories rather than the archaeological information content. Where airborne data have been compared with geophysical data the greatest challenge has been obtaining datasets that are contemporary to ensure comparability (Challis et al. 2011). Consequently gaps in our understanding remain, specifically in the complementarity of digital spectral data and lidar data and correlations between airborne sensors of all types and ground based geophysical techniques. While multi-sensor techniques in airborne remote sensing archaeology are far from commonplace, the select number of instances where they have been applied have shown great potential. The only study to date to explore fully the complementarity of two digital spectral sensors and lidar survey was undertaken in Crete by Rowlands and Sarris (2007). Automated classification of pixels was used to define archaeological features in the multispectral data, and was successful in extracting upstanding stone remains from the surrounding bare earth and also appeared to correlate with some features known from geophysical survey (Rowlands et al. 2007, 798). Experimentation with data fusion has also formed a key part of the investigation of the ruined Roman town of Aquileia in north-west Italy (Sterazi et al. 2008; Traviglia and Cottica 2011). In this project lidar and hyperspectral data were combined using low level (e.g. GIS overlay) and high level (e.g. digital combination) processing, illustrating the importance of lidar topographic and intensity data for improving the classification of features in the spectral data (Sterazi et al. 2008, 371). By integrating the lidar digital elevation model (DEM) and spectral data this project was able to map spatial correlation of mineral deposits, distribution and drainage of deleterious materials on the surface and vegetation cover maps that were sensitive to terrain slope and/or elevation.

199

200

Rebecca Bennett, Kate Welham, Ross A. Hill and Andrew Ford

Figure 16.2: View of a calcareous grassland environment, Salisbury Plain, Wiltshire, UK

Management of unimproved grassland environments Existing studies in the UK demonstrate the potential for airborne sensors to complement or surpass traditional aerial photographic survey techniques in environments dominated by arable production, while those on the continent focus on Mediterranean examples. However, little is known of how these data might aid prospection in other temperate environmental zones such as grasslands or uplands. In the UK, improved and unimproved grasslands accounted for 48% of landcover in 2008 (Food and Agriculture Organisation of the United Nations 2010) (Figure 16.2). The majority of these areas classified as semi-natural, being created and maintained by human intervention. Globally, this land cover accounts for 40% of the world’s landmass (Suttie et al. 2005) and poses significant challenges for traditional archaeological prospection techniques. In the UK, these regions are on the margins of presently sustainable agriculture, and are more readily affected by changes in climate than lower altitudes. They contain a wealth of information about changing subsistence strategies through the prehistoric and historic periods. In addition many of the areas are rich in natural resources, such as metal ores, with evidence of more than 4000 years of exploitation and trade. Many of these areas fall within National Parks and are designated for their natural beauty and the value of their geology, ecology and historic environment. This designation is reflected in their modern land use patterns which are dominated by pasture and recreational use. While the growth of archaeological remote sensing in the UK, especially the application of lidar survey, is undeniable, grassland environments have tended to be overlooked, due in part to the greater availability of archive remote sensing data for other environments. However, the need for both quantifying and characterising the historic

environment in these areas is pressing. While not subject to the industrial-scale agriculture of lower lying regions, many of these areas are under pressure from changes in land management, tourism and environmental and ecological change. Additionally protection of the historic environment must be balanced with that of the natural environment and local economy. In order to represent the historic environment properly in any management strategy it is first necessary to characterise it, identifying as much as possible of the archaeological resource and attempting to understand what is still unknown, and what threatens its survival. Over the past century aerial photographs have revolutionised the understanding of non-arable areas, but this medium is not without limitations. Specifically the hardy nature of ground cover in grassdominated or upland areas when compared with arable crops makes the detection of vegetation change caused by archaeological features almost impossible in photographs that capture only the visible wavelengths. In addition, the identification of features with only ephemeral topography on aerial photographs is highly dependent on lighting, and even in well-studied areas the interrogation of high resolution digital terrain models can play a significant role in feature identification and mapping accuracy. These factors bear directly on the application of multisensor survey to grass dominated landscapes and have driven the assessment of the full information content of airborne remotely sensed data by the Bournemouth-based project. Beyond the picturesque – analysing the full information content of airborne remotely sensed images From the examples above it is clear that the integration of multiple airborne remote sensing techniques can enable the extraction of a range of contextual information in addition to the detection of archaeological features. This makes multi-sensor survey a powerful tool for understanding and managing historic landscapes. This is particularly true for unimproved grassland environments that lie between areas dominated by arable farming and the dense moorland vegetation typical of higher altitudes to identify in which regions of the electromagnetic spectrum archaeological features are most clearly detected and relate these to the biophysical properties of the vegetation known from the environmental sciences. For the lidar data, analysis focuses on

16  Using lidar as part of a multi-sensor approach to archaeological survey and interpretation the derivation of accurate micro-topography to complement and characterise the feature data extracted from the spectral imagery. In undertaking these targeted analyses it is hoped to improve both our understanding of the historic environment but also the application of airborne remote sensing data to archaeological prospection. Beyond the Picturesque project therefore aims to move beyond visualisation to techniques that allow the synthesis of technical details about the archaeological features detected in multiple datasets. For the hyperspectral data this analysis has two levels. Firstly, ancillary data on land use and vegetation type is derived from the imagery to inform archaeological feature detection and interpretation. Secondly, spectral sensitivity is analysed to identify in which regions of the electromagnetic spectrum archaeological features are most clearly detected and relate these to the biophysical properties of the vegetation known from the environmental sciences. For the lidar data, analysis focuses on the derivation of accurate micro-topography to complement and characterise the feature data extracted from the spectral imagery. In undertaking these targeted analyses it is hoped to improve both our understanding of the historic environment but also our application of airborne remote sensing data to archaeological prospection. The test areas for the project were selected for their environment, land use, availability of archive airborne data and the depth of previous archaeological investigation. Two sites on the East Range of Salisbury Plain, Everleigh and Upavon, were selected as the main targets for the study (Figure 16.3). Almost 39,000 hectares of Area

Everleigh

Upavon

Data Type

the rolling chalk outcrops of the Plain is owned and managed by the Ministry of Defence as the largest training area in the UK. The archaeological landscapes of this area are remarkable both for their location between the World Heritage Sites of Stonehenge and Avebury and for the outstanding preservation of their archaeological features. Purchased by the War Office following the agricultural depression of the late 19th century (McOmish et al. 2002, 6) the Plain is the largest area of chalk grassland in Britain that remains unaffected by agricultural intensification. Archive and bespoke digital spectral data and Resolution

Date Flown

201

Figure 16.3: Location of the study areas,Salisbury Plain, Wiltshire, UK

Table 16.1: Airborne Digital Data Sources for the Salisbury Plain Study Areas

Source

Digital Spectral (CASI)

1.5 m

January 2001

Environment Agency

1.5 m

May 2001

Environment Agency

Lidar

1m

2005

Environment Agency

Aerial Photography (Oblique)

0.15 m

Archive (c.1950–2002) Wiltshire Historic Environment Record

4-Band NIR Aerial Photography (Vertical)

2006

2006

2006

2007

Digital Spectral Eagle Hawk

1m 2m

Lidar

0.25 m

March 2010

March 2010

Defence Estates

NERC ARSF NERC ARSF

202

Rebecca Bennett, Kate Welham, Ross A. Hill and Andrew Ford

Figure 16.4: Examples of lidar and spectral visualisation techniques over a 1 km2 area

lidar survey (detailed in Table 16.1) were analysed and compared with the existing transcription of information from aerial photographs by the English Heritage National Mapping Programme (NMP) and the local Historic Environment Record (HER). The process of feature identification followed a top-down approach where every anomaly of potentially anthropogenic origin was initially recorded, including modern routeways, tank tracks and agricultural features. Modern features were then categorised and removed from the subset following field survey and map regression. The remaining features were then categorised by form and topography before being interpreted archaeologically with reference to the existing understanding of features in this landscape from the NMP and HER data. A large number of visualisation techniques were employed including true and false colour

composites, vegetation indices and Principle Components Analysis (PCA) for the spectral data (Bennett et al. 2012b). For the lidar data slope, aspect and PCA of shaded relief images were combined with newer approaches (Figure 16.4) such as local relief modelling (LRM) (Hesse 2010) and sky-view factor (Kokalj et al. 2011, this volume). This allowed the impact and bias of visualisation techniques for each data type to be thoroughly assessed before the data from different sensors were combined (Bennett et al. 2012a). Comparisons of detectability in the airborne remote sensing data were made using both binary visibility and average percentage feature length (APFL) recovery. The APFL is a measure of partial visibility and was calculated using the largest mapped extent of each feature from any data to compare with the extent mapped in each dataset. This allowed the evaluation of the

16  Using lidar as part of a multi-sensor approach to archaeological survey and interpretation

203

contribution of ‘new’ survey techniques to the archaeological record.

Figure 16.5 (above): Relative recovery rates from each of the sources examined for the Salisbury Plain Study

Figure 16.6 (left and below): Assessing the impact of plough damage to a field system using the Local Relief Model, Everleigh

Lidar Local Relief Model showing the location of profiles across a lynchet 0.20

A –Scheduled B –Heavily ploughed

0.15 Height in metres

The contribution of lidar data to multi-sensor survey The analysis of airborne remote sensing data for the Everleigh area allowed the mapping of 170 potential archaeological features of which 68 were known from the HER. Figure 16.5 shows the relative recovery rates of features from each data source examined. In this environment where there has been less damage to upstanding features through intensive ploughing, the lidar data proved to be the key dataset for recovering archaeological feature information. It is clear that the lidar data outperformed the other data sets but the significance of its incorporation to the project runs much deeper. Due to its increased geospatial accuracy (±8 cm in all planes) over the other data it was possible to use the lidar data to standardise the locations of features recorded in the HER and spectral data, which were otherwise of variable metrical accuracy. For a 1 km study area in Upavon, 140 features from the HER could be mapped in the lidar LRM and shaded relief models; of these 117 or 82% of these features had incorrect spatial locations, generally in the range of 8–15 m. Geospatial accuracy is extremely important to any multi-sensor analysis, as location is the only constant factor when comparing datasets that are recording different parameters. Without knowing the location (or at least being able to assess the accuracy to which the location is known) it is impossible to compare different methods of detection with any confidence. The lidar data also provided other benefits. Results of statistical analysis of the impact of land use on feature visibility across the Everleigh area showed that for any of the lidar visualisation techniques trialled there was no significant association of land use and feature visibility across the three categories in the study area (cultivation to a depth >0.25 m, minimal cultivation and disturbed grassland). This was not the case for the digital spectral data and HER data where visibility of features proved to be significantly affected by land use type (Bennett et al. 2011). Lidar data provided a way to analyse the output of the spectral data for feature type. From spectral data alone, interpretation of the surface relief (i.e. topography) of features can be difficult, but by profiling the lidar models, in particular the LRM, it was possible to assess

0.10 0.05 0.00 -0.05 -0.10 -0.15

0

5

10

15

20

25

30

35

Distance along profile in metres

Comparison of lynchet profiled in scheduled and ploughed areas

40

204

Rebecca Bennett, Kate Welham, Ross A. Hill and Andrew Ford the elevation component of the features. In addition, the texture of features as represented in high resolution lidar models is also useful when interpreting the lower resolution digital spectral data. The resulting interpretations could be used to broadly categorise features seen in the spectral data by their topography: positive (e.g. bank or lynchet); negative (e.g. pit or ditch); or neutral (no detectable topography). The detectability of feature types in the spectral data could then be assessed statistically against models of expected feature type distribution for the area and against the detection rates from other data such as aerial photographs. As our understanding of how and why features are detected in spectral data is still very limited, ancillary data such as elevation and texture proved extremely helpful when trying to ascertain the factors that contribute to feature visibility. Although ground-based geophysical techniques and soil survey have also proved valuable to this task, the advantage of lidar is that its coverage can be co-terminous to the spectral data, where other techniques are limited to smaller areas, particularly if they are collected simultaneously with the airborne data. Assessing and monitoring the preservation and degradation of features is an important aspect of historic landscape management. This can be hard to achieve through any single technique although archive aerial photographs can have the benefit of providing a time-depth perspective. By using profiles of the lidar data it was possible to explore the impact of plough degradation on a field system that ran across different land use types, including an area of scheduled protection (Figure 16.6). As historic and modern land use is well recorded in the study area, this provided an opportunity to quantify the changes in the form of a feature (such as the lynchet illustrated in Figure 16.6) that occur as a result of modern agricultural activity. These can be observed in contrast to the longer-term changes to the form of the feature that result from other erosion processes, which are typified in the portion of the feature preserved in the scheduled area. While changes relating to damage and degradation can also be seen in visual data, they can be hard to quantify. High accuracy lidar models provide the opportunity

to assess quantitatively factors such as the rate of erosion. As the depth of the lidar archive increases it should be possible to monitor topographic change, providing repeat acquisitions can be processed identically. Combining this with digital spectral data to examine movement across the landscape or hotspots of current activity can aid the assessment of impact of human and faunal activities on archaeological features.

Conclusions The project has shown the critical role that lidar data can play in the airborne multi-sensor survey of grassland environments. In these landscapes lidar is well suited to feature detection but also provides a wealth of contextual information which significantly aids interpretation of features in other data sources. Detection in lidar data has proved to be less biased by variations in current land use than for digital spectral and archive aerial photographic data. Most importantly, due to its high spatial accuracy, lidar data can be used to assess, and in some cases correct, geospatial errors in other datasets. As this is a critical issue for multi-sensor analysis, it is highly advisable that lidar be incorporated as a primary dataset to improve quality and comparability of results.

Acknowledgements Archive CASI and ALS data were supplied by the Environment Agency and new data acquisition for the project was supported by the NERC Airborne Research and Survey Facility (GB10-07), Field Spectroscopy Facility and Geophysical Survey Facility. The Bournemouth University team would like to thank the Ministry of Defence and Defence Estates for facilitating access and especially Richard Osgood, Senior Historic Environment Advisor. Access to the archaeological archive was facilitated by the staff of the Wiltshire Historic Environment Record and we are especially grateful for the expertise and encouragement of Roy Canham. The research is supported by a Bournemouth University Doctoral Research Bursary.

16  Using lidar as part of a multi-sensor approach to archaeological survey and interpretation

References Barnes, I., 2003. Aerial remote-sensing techniques used in the management of archaeological monuments on the British Army’s Salisbury Plain Training Area, Wiltshire, UK. Archaeological Prospection 10(2), 83–90. Bennett, R.A. 2011. Archaeological Remote Sensing: Visualisation and analysis of grass-dominated environments using airborne laser scanning and digital spectra data. Unpublished PhD, UK: Bournemouth University. Available at http://www. pushingthesensors.com/thesis/. Bennett, R., Welham, K., Hill, R.A., and Ford, A. 2012a. A Comparison of Visualization Techniques for Models Created from Airborne Laser Scanned Data. Archaeological Prospection 19(1), 41–8. Bennett, R., Welham, K., Hill, R.A., and Ford, A.L.J. 2012b. The Application of Vegetation Indices for the Prospection of Archaeological Features in Grass-dominated Environments. Archaeological Prospection 19(3), 209–18. Bennett, R., Welham, K., Hill, R.A., and Ford, A., 2011. Making the most of airborne remote sensing techniques for archaeological survey and interpretation. In Cowley, D.C., (ed.). Remote Sensing for Archaeological Heritage Management. EAC Occasional Paper, 99–107. Hungary: Archaeolingua. Bennett, R., Welham, K., Hill, R.A., and Ford, A.L.J. forthcoming. Using airborne spectral imagery for archaeological prospection in grassland environments – an evaluation of potential. Antiquity. Bewley, R.H., Crutchley, S.P., and Shell, C.A., 2005. New light on an ancient landscape: Lidar survey in the Stonehenge World Heritage Site. Antiquity 79(305), 636–47. Challis, K., Carey, C., Kincey, M., and Howard, A.J., 2011. Assessing the preservation potential of temperate, lowland alluvial sediments using airborne lidar intensity. Journal of Archaeological Science 38(2), 301–11. Challis, K., Forlin, P., and Kincey, M., 2011. A Generic Toolkit for the Visualization of Archaeological Features on Airborne LiDAR Elevation Data. Archaeological Prospection 18(4), 279–82. Challis, K., Kincey, M., and Howard, A.J., 2009. Airborne remote sensing of valley floor geoarchaeology using Daedalus ATM and CASI. Archaeological Prospection 16(1), 17–33. Challis, K., Kokalj, Z., Kincey, M., Moscrop, D., and Howard, A.J., 2008. Airborne lidar and historic environment records. Antiquity 82(318), 1055–64.

Food and Agriculture Organization of the United Nations, 2010. FAO statistical yearbook. Rome: F.A.O. Hesse, R,. 2010. LiDAR-derived Local Relief Models – a new tool for archaeological prospection. Archaeological Prospection 17(2), 67–73. Kokalj, Z., Zaksek, K., and Ostir, K., 2011. Application of sky-view factor for the visualisation of historic landscape features in lidar-derived relief models. Antiquity 85(327), 263–73. McOmish, D., Field, D., and Brown, G., 2002. The field Archaeology of the Salisbury Plain Training Area. Swindon: English Heritage. Powlesland, D., Lyall, J., Hopkinson, G., Donoghue, D., Beck, M., Harte, A., and Stott D., 2006. Beneath the sand: remote sensing, archaeology, aggregates and sustainability: a case study from Heslerton, the Vale of Pickering, North Yorkshire, UK. Archaeological Prospection 13(4), 291–9. Rowlands, A., and Sarris, A., 2007. Detection of exposed and subsurface archaeological remains using multi-sensor remote sensing. Journal of Archaeological Science 34(5), 795–803. Sterazi, P., Coren, F., Creati, N., Vellico, M., and Pietrapertosa, C., 2008. Hyperspectral and LiDAR data fusion applied to archaeological studies: the Aquileia site. In Lasaponara, R. and Masini, N., (eds). 1st International Workshop on Advances in Remote Sensing for Archaeology and Cultural Heritage Management, Rome. Štular, B., Kokalj, Ž., Oštir, K., and Nuninger, L. 2012. Visualization of lidar-derived relief models for detection of archaeological features. Journal of Archaeological Science 39(11), 3354–60. Suttie, J.M., Reynolds, S.G., Batello, C., and Food and Agriculture Organization of the United Nations, 2005. Grasslands of the world. Rome: Food and Agricultural Organization of the United Nations. Traviglia, A. 2006. Archaeological usability of Hyperspectral images: successes and failures of image processing techniques. In Campana, S. and Forte, M., (eds). From space to place: 2nd international conference on remote sensing in archaeology: proceedings of the 2nd international workshop, CNR, Rome, Italy, December 4–7, 2006. BAR International Series, 123–30. Oxford: Archaeopress. Traviglia, A., and Cottica, D., 2011. Remote sensing applications and archaeological research in the Northern Lagoon of Venice: the case of the lost settlement of Constanciacus. Journal of Archaeological Science 38(9), 2040–50.

205

17 Remotely acquired, not remotely sensed: using lidar as a field survey tool Stewart Ainsworth, Al Oswald and Dave Went

This paper describes the evolution of ‘ortholidar’ survey as a ground-based method for rapid survey of earthworks and upstanding remains, especially those in lead-mining landscapes. It illustrates the methods used on three landscape recording projects undertaken by English Heritage where aerially acquired data (geo-referenced, vertical photographs and lidar) has been used as a primary recording tool in the field, and focuses on explaining how the ground-based methods incorporate the data. It presents some of the benefits to be gained, and also highlights some of the weaknesses of such data as a field tool in the mapping and interpretation process. Keywords: lidar, survey, orthophotography, ‘ortholidar’, lead-mining, earthworks

Introduction Although lidar is increasingly employed in archaeological projects, its use as the primary tool for ground-based field survey of historic sites and landscapes has been little explored, while, unsurprisingly, the potential of lidar in wooded environments has attracted considerable attention. There has also been some emphasis on purely desk-based research and the potential contributions of fieldwork to the processes of site discovery and landscape understanding has received less attention. Where fieldwork has taken place, it has usually been designed to ‘validate’ the results of remote-sensing and/or to answer specific, localised questions arising from that stage of work, an approach often referred to by the dubious term ‘ground-truthing’. This methodology derives directly from common practice in traditional desk-based mapping and interpretation of aerial photographs where little fieldwork is undertaken in what is often a purely desk-based process (though see Palmer this volume). A traditional two-stage approach of deskbased assessment followed by limited field observation will clearly remain appropriate for

many archaeological projects. However, this paper argues that for open landscapes with extensive survival of earthworks, such stages can be costeffectively replaced by using high-resolution aerial imagery and data in the field as the basis for rapid survey on the ground. This method, discussed in more detail below, has been developed through a series of conservation-oriented survey projects undertaken by English Heritage (EH), firstly at Scordale in Cumbria, then at Grassington Moor in North Yorkshire, and most recently within an on-going study of Alston Moor in the North Pennines. All these projects deal with extensive, upland lead-mining landscapes (without being completely restricted to the industrial remains) and all have used a combination of field survey and rectified aerial photographic imagery, with the additional benefit of high-resolution lidar in the North Pennines, to map and interpret heritage assets. The emerging fusion of survey techniques is more iterative and less strictly sequential in application and intellectual process than the traditional two-stage approach. It allows a single archaeologist to combine most of the remote observations available to a desk-based researcher with the full suite of additional data

17  Remotely acquired, not remotely sensed: using lidar as a field survey tool that will only ever be accessible to the person who encounters sites, buildings and landscapes on the ground, literally close enough to touch them. The lidar-based approach outlined below is certainly more time-consuming than remote mapping and interpretation without any follow-up in the field, and may often be more time-consuming than the traditional two-stage approach. However, the results of the projects are demonstrating that especially where conservation and management are priorities, the use of lidar as the basis for rapid field survey on the ground can be shown to be more cost-effective if the time invested is compared on the basis of number of assets recognised, subtlety of interpretation, and depth of understanding attained.

The origins of an appropriate survey method While the damage done to historic industrial sites and landscapes by one-off or repeated flood events has long been recognised, the more subtle harm done by prolonged hydrological erosion, caused by rainfall, surface run-off and streams, is only now becoming adequately understood. The scheduled lead-mining landscapes of Scordale in Cumbria and Grassington Moor in Yorkshire are two such cases in point. Both landscapes have extensive and (in places) dense archaeological remains, including water management features which were created to drain mines, power waterwheels and wash ore, all embedded in and entwined with complex natural fluvial systems. These cases brought into sharp focus two pressing needs. The first was to understand more about the processes of natural water erosion and how these affect the survival of important mining and associated building remains over large areas. The second, to chart how complex man-made water-management systems have affected and, following their abandonment, continue to affect the surrounding environment, whether through the dispersal of heavy mineral contaminants into farmland and river systems or, as in the case of Scordale, through the redistribution of eroded material downstream causing structural damage to roads and bridges. This interaction between historic water management and natural processes lies at the very heart of the conservation threats experienced most acutely by mine complexes in upland settings. The fundamental challenge here, and applicable elsewhere, was

to design a recording method that would provide the relevant conservation agencies with detailed evidence-based mapping for the nature, complexity and extent of the historic remains, and for the types and locations of hydrological systems that represent threats to and from the natural environment. There was no ready-made solution to the problem. The investigations at Scordale and Grassington, described below, were important stepping-stones in the development of a method which allowed archaeologists to use remotelysensed (or, perhaps more accurately, aeriallyacquired) imagery and data as the basis for their recording on the ground (see Ainsworth and Hunt 2006, 2009; Hunt and Ainsworth 2010; Ainsworth and Burn 2009). Each project threw up its own unique considerations and methodological challenges, and played a role in the eventual decision to use lidar as a field survey tool for Alston Moor. Scordale, Cumbria Scordale is a deeply incised valley that cuts into the western escarpment of the North Pennines. The surface mining remains of the Murton and Hilton Lead Mines, scheduled in 1999, occupy both the valley floor and the precipitous slopes along its upper reaches. By the time scheduling took place catastrophic erosion was already occurring along the course of the Scordale Beck, particularly after heavy rain and snowmelt, impacting on the archaeological remains within the valley, roads and bridges downstream and the water quality of the wider river system (Lane and Dugdale 2006). The archaeological priorities were to identify, map and ascribe significance to the historic remains along a 4 km stretch, and to record those most at risk in greater detail as the basis for a conservation strategy to be implemented by the landowners, the Ministry of Defence (MoD). A report commissioned by the MoD from Durham University to investigate the causes of the worsening erosion along the wider river system proposed that past land-use, particularly medieval and later lead mining but also prehistoric cultivation might have played a role in sensitising the valley to fluvial erosion (Lane and Dugdale 2006). This astute suggestion was, in due course, supported and amplified by English Heritage’s archaeological survey work, which revealed that vast quantities of spoil had been dumped and ‘hushed’ into the Scordale Beck, and that the valley’s natural

207

208

Figure 17.1: Scordale, extract from the orthophotographs. The desk-top, photogrammetric transcription of the complex outcrops, scree and mining waste failed to extract the subtlety of the multi-period archaeological activities, particularly the mining remains, which extend along the steep valley slopes despite the high resolution of the imagery. However, the scaled hard-copy image provided a ready-made survey base for the unravelling and recording of an extremely high level of detail directly in the field. Direct observation of the remains on the ground allowed features to be drawn directly onto the orthophotographic prints. © Crown Copyright

Stewart Ainsworth, Al Oswald and Dave Went

fluvial pathways had been dramatically altered both by the extraction itself and by the various water management schemes associated with the mining. Ultimately, it was concluded that the cessation of active management when the mines finally closed in the 1920s was a key factor in the problems that emerged decades later. In 2002 Scordale was surveyed as part of a desk-based aerial photographic interpretation of the Warcop Army Training Estate to National Mapping Programme (NMP) standards (Winton and Horne 2010; Horne 2011). However, the scope and small scale (1:10,000) of the recording of the complex industrial remains in the valley only addressed basic research and conservation needs. In the view of both the English Heritage Inspector of Ancient Monuments and the MoD Archaeologist, a more comprehensive and detailed survey was required; yet because of the area’s active use as a military range, with access limited to a small number of non-firing days each year, detailed (i.e. time-consuming) field survey over such a large and relatively inaccessible area of numerous and (in places) complex remains was clearly impractical. Initially, a solution seemed

to be offered by the commissioning of vertical digital colour photography at 20 cm resolution by the MoD in 2005 to underpin their broader environmental management needs (Figure 17.1). It was anticipated that this imagery, once computer-rectified to OSGB36 National Grid and scaled at 1:2,500, would allow a fairly detailed traditional desk-based photogrammetric transcription of the archaeological remains. This could then be followed up by a conventional ground observation stage, designed to follow so hard on the heels of the transcription that the two stages would overlap, thus making efficient use of the few, irregular non-firing days. With this intention, the field team began by using orthophotographic prints as an underlay to the transcribed plot to aid location of archaeological features on the ground. It was quickly discovered, however, that the transcription was neither as informative nor as timesaving as had been anticipated, as many mining features apparent on the ground had not been recognised. For example, it had evidently proved difficult, from the deskbased perspective, to distinguish between natural cliffs and artificial quarry faces, between natural scree and mine spoil, and between natural stream channels and artificially induced ‘hushes’, all of which presented their own distinctive characteristics when viewed in the field. In other cases the form, details and function of structures and complex earthwork remains were disguised by the vertical nature of the photographs, so that the remotely recorded depictions were too generalised and simplistic, or, as in the case of some structures of rubble built within larger expanses of rubble, entirely overlooked. In short, it was apparent that to complete the ground observation stage thoroughly, consistently and to the required standard, would require a much greater investment than anticipated. In fact, a high proportion of features were discernible on the 1:2,500 orthophotographs once their character had been determined through fieldwork, and consequently the field team began to use the orthophotograph itself, rather than the remotely transcribed plot, as the medium for recording archaeological remains encountered on the ground. Points, lines and polygons were drawn directly onto the orthophotograph (or more usually onto a waterproof overlay), while notes and sketches were used to flesh out the depictions. With the resolution of the orthophotograph allowing features as small as about 1 m in diameter to be pin-pointed, there were very few

17  Remotely acquired, not remotely sensed: using lidar as a field survey tool instances (mostly in areas of deep shadow on the orthophotographs) when it was necessary to carry out traditional measured survey; but when required this was easily achieved using mappinggrade GPS sets, which offered accuracy well within the tolerance of the 1:2,500 survey scale (Figure 17.2). The fieldworkers were surprised to discover that recording in this mode was often significantly easier and quicker than amending the remotely transcribed plot, a process which, given the separation of tasks between field and office based staff, sometimes required prolonged consideration to establish exactly what the transcriber had depicted. Ultimately, it was recognised that the project had, more by necessity than judgement, developed a rapid and efficient approach which used aerially-acquired imagery to record features observed on the ground. Most importantly, the experience questioned the appropriateness of desk-based remote transcription and field survey as separate tasks, and pointed the way towards a true hybrid that would be more efficient than the traditional two-stage EH approach. Grassington, North Yorkshire The survey of the lead mines on Grassington Moor in 2008 was prompted by reports of damage to the scheduled remains by modern attempts to improve drainage along the road network that once served the mines; although on closer examination the damage was shown to stem from serious underlying problems of hydrological erosion. The lead mining landscape here is smaller (c.2.5 sq km), more accessible and benign than Scordale, but the density of surface remains is significantly greater. Indeed, in many areas, extensive mining has created a ‘lunar’ landscape of overlapping shafts, pits and waste heaps, whose pattern is made even more difficult to discern by secondary re-working for the reclamation of barytes. Interwoven with these remains are complex multi-period leats and reservoirs, as well as natural watercourses, many disturbed or truncated by later mining-related activity. Grassington, like Scordale, had previously been the subject of desk-based aerial photographic interpretation in the Yorkshire Dales Mapping Project (Horne and MacLeod, 1995; Horne and MacLeod 2004) – a pilot project for the NMP completed in 1992. Here too, while the manual (i.e. ‘sketch-mapped’) depiction at 1:10,560 scale necessarily presented a schematic, broad-brush overview of the industrial activity achieved at the rate of at most one day per km square, the Dales

National Park Authority and English Heritage required a higher resolution depiction of specific individual features and threats in order to prepare an effective management plan. In this case, digital vertical colour aerial photographic imagery at 25 cm resolution, rectified and corrected to OSGB36 geodetic datum, was available to EH from Next Perspectives via a Pan-Government Agreement (PGA). Based on the experience of Scordale, it was decided to dispense altogether with the traditional approach of desk-based transcription followed by ground observation. Instead, orthophotographs were prepared as laminated basemaps at 1:2,500 scale for the archaeologists’ direct use in the field. Even though this was a much more complex landscape than Scordale, and the photographic imagery was of slightly lower resolution, by enhancing the colour, contrast, tones and saturation levels, a remarkable degree of detail was visible (Figure 17.3). In the field it proved possible to accurately locate most features against this background, tracing their extents directly onto the laminated sheet as points, lines, or polygons, while supporting observations were recorded in a notebook and added to the project GIS and database in the office (Figure 17.4). The 2.5 square kilometres of the complex were covered by a team of two in eight days, capturing accurate geospatial information (totalling just over 550 individual records) and the compilation of the GIS and Access database in the office took 15 days. As at Scordale, a small minority of features, including individual ‘meerstones’ (boundary stones defining areas of mining rights), could not be located on the orthophotographs, and these were plotted using mapping-grade GPS. As an experiment, the orthophotographs were

209

Figure 17.2: Scordale, the limestone outcrops in the background have all been mined for lead along surface seams resulting in their buttress-like appearance and spreads of unproductive waste material (deads) deposited in the mined-out gaps in between. These features, and differences in colour of the waste materials on the ground surface (which indicate ore-dressing areas), were visible on the vertical aerial photographs but had not been recorded during the desk-top transcription; they were however easily and speedily added to the orthophotographs on the ground. Features not easily identifiable on the orthophotographs were surveyed using mapping-grade GPS. In this location, loose planks, iron posts and different grades of waste provided the evidence for the line of an ore-shoot and site of an ore-crusher. Alun Bull © English Heritage

210

Stewart Ainsworth, Al Oswald and Dave Went

Figure 17.3: Grassington, orthophotograph of the survey area. The original ‘as photographed’ colour saturation and contrast settings have been refined to produce colour levels where lead-mining waste and re-worked areas can be easily identified (grey and white).The part panel at the top-centre shows the original unaltered image tones. Licensed to English Heritage for PGA, through Next PerspectivesTM

also uploaded onto the GPS sets themselves, along with historic Ordnance Survey (OS) maps, so that it was immediately possible for the archaeologists to equate their position on the ground (or the position of a specific feature) against the orthophotograph and/or the maps. The result was a highly successful, cost-effective survey methodology (Ainsworth and Burn 2009). which allowed detailed analysis of the archaeological remains and the forces which threatened their survival.

Alston Moor, North Pennines and the use ‘ortholidar’ for rapid field survey In 2008, EH launched a multi-disciplinary landscape investigation called ‘Miner-Farmer

Landscapes of the North Pennines Area of Outstanding Natural Beauty (AONB)’, hereafter referred to as the Miner-Farmer project which is due to be completed in 2013. The AONB, the second largest in England, spans parts of Cumbria, Northumberland and County Durham and was designated as Britain’s first UNESCO European Geopark in 2003 in recognition of its complex mineral-rich geology. All the geological resources, but particularly the silverrich lead, have been intensively exploited for hundreds, if not thousands, of years, leaving an extensive and diverse legacy of industrial remains. However, neither the industrial sites nor the other components of the historic environment had ever been comprehensively mapped. Only the most obvious mines and quarries, mostly gleaned from historic OS maps, were identified

17  Remotely acquired, not remotely sensed: using lidar as a field survey tool

211 Figure 17.4: Top: Grassington, feature 311. An area of re-working of 19th century lead-waste for barytes in the 1950s. These complex surface remains are hard to differentiate using desk-top survey methods but the variety of colours and gradation of stone sizes of the waste material resulting from different processes and methods of extraction can be seen clearly on the ground and separated from nonarchaeological activity. In this instance, the remains have been later disturbed by modern, illicit digging for roadbuilding material which has resulted in loss of archaeological stratigraphy, exposure of artefacts and damage to surrounding fragile dressing floors. Stewart Ainsworth © English Heritage. Centre: Extract from the project GIS based on the colourcurved orthophotograph. The enhancement helped identify the heavily disturbed areas (lighter tones) although only the ground stratigraphy could confirm whether this was secondary re-working for minerals, what period, or whether related to non-archaeological activities such as extraction for hardcore. Scaled prints similar to this formed the basis of the field survey and were used as a basemap to define areas of archaeological activity, erosion and threat issues as points, lines and polygons. This survey and textual information were later added to the GIS in the office. Licensed to English Heritage for PGA, through Next PerspectivesTM. Bottom: Extract from the Microsoft Access database. This has been designed to allow a high level of management and threat-level data to be added to the archaeological record

in the local Historic Environment Records (HERs) and the National Monuments Record (NMR), and only a handful of the most prominent lead-mining and processing sites (some quite extensive) had been singled out for public presentation and statutory protection. Against this background, the Miner-Farmer project aimed to investigate, in a holistic manner, the interwoven influences of industry (particularly lead mining) and farming on the development of a representative sample of the region’s landscape to inform the conservation, protection and management of both the historic and natural environments (Ainsworth 2008, 2009, 2010; Ainsworth and Frodsham 2011). Undertaken in partnership with the North Pennines AONB Staff Unit, the Environment Agency, Natural England and the North Pennines Heritage Trust, the project includes contributions from several specialist teams within EH and from commercial contractors, including Infoterra, the Birmingham University Vista Spatial and Technology Unit, and North Pennines Archaeology Ltd. The research concentrates on a 300 square km sample area in and around Alston Moor, a massif mostly over 300 m above sea level, lying between the confluence of the rivers Nent and South Tyne (Figure 17.5). The uppermost parts of the survey area are, like the majority of the AONB, bleak, open moorland and peat bog. Below this, enclosed moorland on the valley sides gives way to hay meadows and pasture on the lower slopes, and it is mostly within this

212

Figure 17.5: North Pennines, the MinerFarmer project area. Based on Ordnance Survey mapping © Crown Copyright and database right 2011. All rights reserved. Ordnance Survey Licence number 100019088

Stewart Ainsworth, Al Oswald and Dave Went

farmland that lead mining remains occur, often side-by-side with remains of medieval, Roman and prehistoric settlement. Furthering the methodological research One of the project’s explicit aims was to build on the methodological research outlined above and, by combining ground survey and remote sensing, systematically test new approaches to the research and management of upland mining landscapes. This was of particular relevance given that several extensive lead-mining complexes on Alston Moor, including some Scheduled Monuments, are experiencing serious fluvial erosion similar to those at Scordale and Grassington, and here too these problems pose a threat both to the historic remains themselves and to the wider landscape. Indeed, the Environment Agency recognises the Nent as the most contaminated river in Britain, but is currently unable to isolate more than 30% of the sources of this contamination, in part because understanding of historic water management associated with the mining is poor.

Experience gained from the earlier projects indicated that neither traditional ground survey alone (due to the time it would take to fully map large areas from scratch), nor desk-based aerial photographic transcription alone (due to inadequate detail and lack of condition and threat assessment) was likely to meet the objectives of understanding the historic environment, the natural environment and the threats faced by each. Thorough, systematic examination of method had not been an explicit objective before, but a ‘nested set’ of techniques has now been developed to allow different combinations of survey to be properly compared and evaluated, deliberately building-in both overlapping and ‘double-blind’ samples. Deskbased interpretation and mapping from new and archived conventional aerial photographs has been applied to the overall project area of 300 square kms. Within this, to test the added value that lidar and less conventional forms of aerial photographs could bring to the process, a series of remote-sensing (that is aerially-acquired) datasets were commissioned from Infoterra covering a core area of 50 square kms focussed on the confluence of the rivers Nent and South Tyne. These comprised 50 cm lidar (and 1 m lidar interpolated from the 50 cm dataset) and intensity data, as a well as digital, rectified, vertical full-spectrum colour, infra-red and hyper-spectral bandwidth aerial photographs. Within the core area all visible detail is recorded by desk-based aerial photograph interpreters at a scale of 1:2,500 and in parallel with this rapid field survey is also being undertaken at 1:2,500, developing the approach trialled at Scordale and Grassington, but using lidar as the primary basemap for fieldwork. Moving from the ‘orthophotograph’ to ‘ortholidar’ In reviewing the method developed at Grassington it was anticipated that high-resolution lidar, even studied in isolation without the benefit of other forms of aerial imagery, might prove more effective than an orthophotograph for identifying, recording and understanding hydrological flowpaths and their impacts on heritage assets (Ainsworth and Burn 2009, 107–15, 120–2). However, while orthophotographs had proved adequate for the mapping and interpretation of most features in the field, it was not certain that the ‘pseudo 3D’ representation of ground features by lidar would offer an equally practical and

17  Remotely acquired, not remotely sensed: using lidar as a field survey tool straightforward medium. A particular question concerned the limitations of the ‘fixed angle’ lidar image taken into the field, compared to the flexibility of lidar viewed on a computer screen, where both the angle of the dataset and the light source could be manipulated to explore the appearance of features. Parallel questions faced those responsible for the desk-based NMPstyle research: how much would lidar reveal in comparison to the use of orthophotographs or the normal range of aerial photographs; and could the findings be justified against the extra time required to explore the lidar dataset? In short, questions of practicality and added value were central to the formative stages of the project. Specially commissioned remote sensing datasets were supplied to the project by Infoterra as georeferenced 1 km square tiles, for ease of use and workable file size. The lidar tiles were provided in Digital Terrain Model (DTM) and Digital Surface Model (DSM) formats, and this was subsequently hillshaded in ESRI ArcGIS. A comparison of 1 m and 50 cm datasets for the North Pennines, and a limited field trial, rapidly confirmed the 50 cm lidar basemap as a practical tool for this detailed survey. The NMP element of the project has been completed (Oakey, Radford and Knight 2012) but the wider project is still in progress so the final results of the methodological comparisons are not yet available, but the development of the process is described and some provisional outcomes are presented below. The field survey method, in practice The field survey of the core area is being undertaken by two separate field teams, one from English Heritage, the other from North Pennine Archaeology Ltd (NPA), who successfully tendered to carry out part of the work, fulfilling one of the project’s key capacity-building objectives. The same lidar and other remotesensing datasets are being used by both teams. The only difference in method is that NPA employ survey-grade GPS rather than mappinggrade GPS to record features not readily apparent on the aerial imagery, due to the ready availability of that equipment within their field unit. Each kilometre square within the core area is thoroughly investigated by an individual archaeologist who maintains regular contact with colleagues in nearby squares to minimise inconsistencies in interpretation along tile boundaries (and also to ensure each others’ safety in the field). Although a methodological blue-

print was agreed in advance, in practice each archaeologist has adopted a slightly different approach, according to individual preference. Most prepare by studying the readily available sources, including the remote sensing imagery, historic OS maps and previous NMR and HER records. These are consulted frequently as the fieldwork is in progress, and usually checked again at the end of each day in the field, when hindsight can make a useful contribution. Regardless of the number and quality of the pre-existing records, the fieldwork is always carried out in the manner of a comprehensive Level 2 or ‘walkover’ survey of virgin territory (as defined in EH guidelines – see Ainsworth et al. 2007). In the course of this, a rapid analysis and succinct record is made of every historic asset: archaeological features, standing buildings, ruins and other structures (including field walls), natural watercourses and key botanical features, such as managed woodland and individual trees, and occasionally individual artefacts. Larger complexes, such as mines, farmsteads and other settlements, are broken down into their component parts to whatever degree is judged necessary to achieve the level of resolution required for effective interpretation. These entities are recorded onto the ‘ortholidar’ basemap (see below and Figure 17.6) as points (if smaller than 5 m in diameter), or lines (mostly used for tracks, leats and drains), or polygons (for other features). Following the initial trials two methods of field recording were adopted. Firstly, hillshaded lidar and orthophotographic imagery, along with historic and modern mapping, were loaded onto mapping-grade GPS receivers (Trimble GeoXT). These light-weight, user-friendly instruments receive differential corrections through the EGNOS (European Geostationary Navigation Overlap Service) satellite, enabling real-time corrections to OSTN02/OSGM02 transformations. Tests have demonstrated that accuracy is acceptable within the normal tolerances of the project mapping scale of 1:2,500 within the OS National Grid. Points, lines and polygons, reflecting archaeological features, can be recorded either by logging positions with the GPS unit itself, or by digitising on the screen with a stylus. The device also holds a specially designed database for inputting key information (including record number, monument type, dimensions, land-use background, condition and threat-related information), and a free-text description for each mapped entity. Mapped

213

214

Stewart Ainsworth, Al Oswald and Dave Went Figure 17.6: North Pennines, km square NY6948. The top image shows the km square hillshaded 50cm lidar basemap used for recording points, lines and polygons in the field. The bottom image shows the resulting field survey layer of the GIS for the same km square. Specific sites were selected during the rapid survey process for more detailed survey at Level 3 to provide a higher level of understanding and inform future conservation and management, for example the diamond-shaped Roman fort near the centre of the lidar image known as Whitley Castle (see also Figures 17.10 and 17.11). The DSM provided the most useful dataset as it displays surface structures which form part of the record such as buildings and walls, with details of roof lines often aiding phasing of building complexes ©English Heritage

features and their associated information can then be exported from the GeoXT directly into the office-based GIS. Early trials soon demonstrated that only three of the datasets (the lidar DTM, DSM and the vertical colour photographs) were routinely used, whilst the ability to compare historic maps on screen against the lidar imagery was also regularly found to be a helpful facility. Despite the attractiveness of an entirely electronic process of data capture and transfer, the hand-held GPS equipment did not, however, prove entirely problem-free. In practical terms, the lidar and orthophotographic images have to be reduced in file size (and therefore resolution) to allow their effective storage and manipulation, and image clarity is further compromised by the GPS set’s small screen (54 × 72 mm) which can be particularly difficult to read in harsh or very low light or when covered in raindrops. Digitising onto this screen with a stylus is more time-consuming than digitising with a mouse while sitting comfortably at a desk and maintaining the image at a size suitable for accurate recording requires regular scrolling across the image, which hinders an overview. The second method was to provide a lowtech alternative to the GeoXT in the form of a printed, hillshaded lidar image at 1:2,500 scale, upon which features could be simply outlined or otherwise marked in pencil or indelible pen. The DSM was the chosen format, as it depicts buildings, field walls and trees, allowing the archaeologists to locate themselves at a glance on the ground, and also modelled the relative heights and angles of roofs, which was sometimes helpful when defining various phases of development within farm complexes

17  Remotely acquired, not remotely sensed: using lidar as a field survey tool (see Figure 17.6). To rain-proof this basemap, the lidar tile was printed as a mirror image on polyester film and sealed, print side down, against a lightweight field drawing board. This simple piece of kit has proved reliable, practical and resilient to the extreme North Pennines weather. Indeed, it has become a mainstay of the recording method for both the EH and the NPA field teams. There is a slight loss of print density when printing the ‘ortholidar’ basemap onto plastic film, therefore a glossy, photographicquality print on paper was a helpful supplement. Glossy prints of the other aerial images, together with historic maps made up the ‘field-pack’ of reference material created for each 1 km square, usually at a reduced scale of 1:4,000 (250 mm square) for more convenient handling. The colour orthophotograph was particularly useful in the open moorland areas where identification of vegetation differences helped both navigation and identification of features. The biggest single benefit of the false-colour infra-red print arose in areas of lead-mining where de-vegetated areas of lead waste were very easy to identify (as green tones) at a glance, and helpful in refining areas of field investigation (see Figures 17.7 and 17.8). Ultimately then, the mature field kit comprised a single drawing board with the ‘ortholidar’ 1:2,500 basemap, a waterproof field notebook for textual recording of features marked on the basemap and a file of weatherproofed maps, records and photos. The hand-held GPS remained useful for recording features not visible on the aerial images, and its real-time positioning facility was extremely valuable when pin-pointing very low features visible in the remote imagery but obscured by high heather and moorland sedges on the ground. In the main, however, the ‘low-tech’ method was favoured for its ease of use. Indeed, NPA found that they used the GPS so rarely that it was more efficient to initially leave their heavier equipment behind and to subsequently fix all the inaccurately located features within the kilometre square in a single operation. Adapting to the ‘ortholidar’ method The method described above has been applied, with minor variations, by all the field staff and the approach of taking lidar into the field to act as a basis for rapid survey has proved to be fundamentally efficient, effective and simple. Different versions and treatments of the lidar data were explored. DTM versions of the lidar proved particularly helpful for areas masked by

215 Figure 17.7: North Pennines, colour-curved, CIR imagery of km square NY 7443. The intense green areas are areas which are vegetation free due to lead and zinc contamination from dressing waste, Having prints of this data available in the field enables rapid identification and verification of dressing waste and allows fluvial dispersal routes to be quickly identified and recorded on the ortholidar basemap. Compare Figure 17.8. © English Heritage

deciduous woodland, for example where the ditches alongside the Maiden Way Roman Road were confirmed in an area which was otherwise impenetrable due to the density of the scrub. In dense coniferous plantations the DTM imagery was of less value. Although it could reflect massive linear features in a fragmentary and coarse-grained form, other features, even quite prominent and distinct earthworks such as infilled mine shafts, remained completely undetectable by any means other than ground survey. The archaeologists’ preference for the drawing board, rather than the GPS unit, does not stem from technophobia, nor from the purely practical difficulties of the GPS screen mentioned above. Rather, the small screen made it conceptually difficult to simultaneously take into account

Figure 17.8: North Pennine, hillshaded DSM of km square NY 7443 using 50 cm lidar (the same area as Figure 17.7). A hard-copy polyester print is made at 1:2,500 scale and mounted on a drawing board to use as a survey document: the equivalent digital tile is loaded into the GPS. This image illustrates well how the process of interpretation and recording of the complexities of the mining at Fletcheras Mine and the associated water management systems is enhanced with this data as a pseudo-3D survey base document for use in the field. Coal, zinc, lead and silver have all been mined here (the last two since at least the 14th century). Differences in the morphology of mining types and the intricacies of complex fluvial channels can be rapidly analysed whilst at the same time undertaking investigation of the threat and erosion issues. © English Heritage

216

Stewart Ainsworth, Al Oswald and Dave Went both detail and big picture. Furthermore, the static, ‘head-down’ process of digitising onto the GPS screen, which is relatively time-consuming (compared to simply outlining a feature with a pen or pencil), itself distracted from the primary tasks of observation and analysis, which require constant movement and a ’head-up’ stance to assimilate the maximum information. Routine use of the GPS set in the field could save time through the reduction of subsequent digitising in the office, but this mode of fieldwork itself was taking three to four times longer to complete. Furthermore, the GPS touch screen was also found to be constraining when it came to data entry, and a considerable gulf soon emerged between electronic records and those compiled in field notebooks: the latter being more detailed, analytical and discursive, and perfectly acceptable as long as they encompassed all the data required to complete the multiple fields in the main project GIS and database. The value of the notebook approach was also evident as the fieldwork progressed through the day and first thoughts concerning features and processes required amendment in light of further knowledge: changes that were easily accommodated on paper, but more time-consuming via the GeoXT. The initial supposition that total electronic survey in the field would provide the most complete and efficient record has not been born out through experience. This conclusion, however, is based on a snapshot of currently available technology. It seems inevitable that within a short space of time a genuinely lightweight ruggedised laptop with a practical screen size and built-in mapping-grade GPS will overcome at least some of the problems encountered. The advantages of the ‘ortholidar’ approach The accuracy of the rectification process has proved to be well within the mapping tolerance for 1:2,500 scale mapping using GPS and OS Mastermap data, and thus could be used as an accurate basemap. The key benefit of the lidar was to have all the natural and artificial components of the landscape visible on a single document in pseudo 3D, which could then be used as the basis for an analytical record. This was pivotal to the rapid speed of the fieldwork as well as being an aid to interpretation. It was found that cross-referencing between the lidar basemap and the features on the ground became a seamless physical and intellectual activity, consideration of each refining the interpretation of the other

in a constant iterative process. As well as adding immense value to both the interpretation and recording in the field, all the archaeologists took very quickly to the method – in most cases becoming entirely comfortable with the approach within a couple of days. ‘What does it look like on the lidar?’ became a frequently asked question whilst actually standing in front of the entity in question. Being able to see the whole km square in hillshaded pseudo-3D (using the low-tech printed basemap rather than the GeoXT screen) helped refine thoughts about the context of those features near to hand, and what other features or areas should be inspected next in the interpretative process. This is a facility which archaeologists in the field have never had in such a comprehensive manner before. The main methodological challenge was to test whether the hillshaded display of pseudo 3D data was a sufficiently accurate portrayal from which to record representative points, lines and polygons. In practice it was very quickly established that the brain translates the direction of hillshading onto the reality of the actual feature with ease, and it becomes very easy to recognise where it is accurate or where it is unrepresentative, allowing the archaeologist to quickly draw the outline of the feature to an accuracy of 100 m in diameter) were targeted through geophysical survey, corroborating and enhancing the lidar-based picture. The use of viewshed analysis is explored, highlighting the importance of tomb visibility against views from the tombs, demonstrating that a Neolithic boat traveller’s first view of the ritual landscape is likely to have been of Newgrange, while all three great tombs would only have been visible from the very apex of the Bend. Finally, while lidar data have provided a wealth of new information regarding the landscape of Brú na Bóinne attention is also drawn to significant recent discoveries which exhibit no topographic expression. Keywords: Ireland, Neolithic, UNESCO World Heritage Site, landscape, viewshed, geophysical survey, visualisation.

Introduction The ‘Archaeological Ensemble of the Bend in the Boyne’ (known colloquially as Brú na Bóinne) is arguably the most widely known and best studied archaeological landscape in Ireland. Brú na Bóinne was inscribed as a UNESCO World Heritage Site (WHS) in 1993 owing to its concentration of megalithic artwork, the scale of Neolithic passage tomb construction, and the long record of continuous archaeological activity within the area (Smyth et al. 2009, 7). The main concentration of monuments lies

to the west of the town of Drogheda, within a rock-cut bend in the River Boyne as it flows eastwards to the Irish Sea. The area is defined to the north by the River Mattock, and to the south by the course of the Boyne itself. The WHS core area comprises about 780 hectares and is surrounded by a ‘buffer zone’ which comprises a further 2,500 hectares. At the time of the last WHS management plan there were 93 recorded monuments listed within the area (Dúchas 2002). The WHS is dominated by three massive passage tomb cemeteries – Newgrange, Knowth and

224

Stephen Davis, Conor Brady, William Megarry and Kevin Barton

Figure 18.1: The Brú na Bóinne World Heritage Site, indicating core and buffer zones. Sites mentioned in text are labelled by letter after Coffey (1912) and O’Kelly (1978). Numbered boxes indicate areas illustrated in more detail in corresponding figures

Dowth, each of which is located at a prominent point upon an east-west shale ridge (Figure 18.1). Many sites within the WHS are known by single (occasionally double) letter codes. These are mostly derived from the early work of Coffey (1912), with addenda provided by Ó Ríordáin and Daniel (1964) and O’Kelly (1982). Lidar data collected for the WHS has become a key part of its management strategy and is being utilized in several research projects (Corns and Shaw this volume). The use and further application of remotely sensed data at Brú na Bóinne is particularly valuable because only a small proportion of the WHS is under state ownership. This paper discusses the use of lidar data for research which, in addition to supporting prospection and the identification of new sites, moves toward building landscape-scale narratives. One of the challenges in building overarching narratives for the WHS is incorporating the lidar data into the dense existing archaeological record developed through excavations, survey, geophysics and aerial photography in a way that goes beyond simply adding to the corpus of sites. An overview of the archaeology of the WHS is presented here, followed by a first look at new research perspectives being explored through the combined sources of information, pursuing the goal of ‘doing more’ with lidar.

Chronology of the WHS The following comprises a brief overview of the chronological development of the WHS; however, more detailed chronological overviews of the archaeology of the area appear elsewhere (e.g. Stout 2002; Smyth et al. 2009). Mesolithic – Early Neolithic The earliest evidence for human activity in the area is represented by a series of lithic artefacts characteristic of the Later Mesolithic period (c.5,500–4,000 BC), recovered during excavations adjacent to Newgrange (O’Kelly et al. 1983). Structural and artefactual evidence associated with the adoption of agriculture in the Early Neolithic period (c.4,000–3,600 BC) is known from a number of sites within Brú na Bóinne as well as from excavations in advance of the construction of the M1 motorway immediately to the east (O’Kelly et al. 1978; Eogan 1986; Eogan and Roche 1997; Moore 2003; Ó Drisceoil 2003, 2007). Middle Neolithic The passage tomb cemetery, which the area is best known for, was constructed mainly during the Middle Neolithic period (c.3600–3100 BC). There are at least forty possible passage tombs

18  Lidar survey in the Brú na Bóinne World Heritage Site within the area (Eogan and Doyle 2010), with many mounds remaining unclassified owing to lack of visible diagnostic features. Newgrange, Knowth and Dowth are the three largest, each measuring in excess of 80 m in diameter and dominating the Brú na Bóinne landscape Newgrange is the best known of these three, in part owing to the deliberate alignment of passage and chamber towards the rising sun at the winter solstice (O’Kelly 1982; Stout and Stout 2008a). The mound is about 85 m in diameter and stands 11 m high. Newgrange was extensively excavated between 1962 and 1975 by Professor M.J. O’Kelly, revealing details of the construction of the passage tomb, in addition to evidence of later settlement activity dating to the Late Neolithic/Early Bronze Age (O’Kelly 1982; O’Kelly et al. 1983; Stout and Stout 2008a). Several other excavations have occurred at Newgrange, including those of Sweetman in the 1980s (Sweetman 1985, 1987 – see below). Knowth lies at the western end of the WHS. It is similarly massive, with a diameter of about 90 m and a height of 15 m. Knowth has been extensively excavated over many years by Professor George Eogan, revealing multiple phases of activity and occupation from the Early Neolithic through to the post-medieval period. Six separate phases of development during the Neolithic alone were identified (Eogan 1986; Eogan and Roche 1997). The third of the great mounds, Dowth, was subject to limited excavation in the mid-19th century (cf. Harbison 2007; O’Kelly and O’Kelly 1983) and again in 1989 (Lynch 1990), but has never been extensively excavated. Its dimensions are similar to those of Newgrange and Knowth, with a diameter of about 85 m and a height of 15 m. Many of the smaller passage tombs and mounds are arranged into sub-cemeteries focused on the larger mounds (Cooney 2000, 153–8; see Smyth et al. 2009, 28–9). At Knowth there is a clustering of up to 18 smaller sites, some of which have been reconstructed since the excavations. At Newgrange, a linear arrangement of mounds lies along the same ridge as the main tomb; Sites K and L lie to the west, with Site Z and Sites E-H to the east (cf. O’Kelly et al. 1978). Three other possible passage tombs, Sites A, B and U, lie between Newgrange and the river. At Dowth, there are further mounds and tombs (Sites I and J) while slightly farther afield, Site T is at Townleyhall and Site S at Monknewtown.

Late Neolithic An intensification in settlement activity in the area in the Late Neolithic (c. 3,100–2,500 BC) has been hypothesised, coincident with a secondary phase of monument construction. The Boyne embanked enclosures (Sites A, P, Q and V) are believed to be of this date, in part owing to Sweetman’s excavations at Monknewtown (Sweetman 1971; 1976). Best preserved of these is Dowth Henge (Site Q), a large, ovoid earthen enclosure with banks surviving to height of about 4 m and a maximum diameter of about 150 m. The Boyne Valley embanked enclosures form a coherent monument group and are widely distributed along both the Boyne itself and its tributaries (cf. Stout 1991). Also of Late Neolithic date is the ritual pond, Site W, at Monknewtown (Condit 1997). Previously believed to be Late Bronze or Iron Age, material obtained from the base of the enclosing ditch has been radiocarbon dated to the Late Neolithic (Brady et al. 2010). A number of timber circles and associated structures are known from within Brú na Bóinne, both through excavation and geophysical survey, and are also believed to date from the Late Neolithic. One such timber circle was excavated at Knowth and a further two at Newgrange (Sweetman 1985; 1987; Eogan and Roche 1997). A small number of similar monuments have been recorded in excavations in the wider area including those at Coolfore, Co. Louth and Lagavooren and Bettystown, Co. Meath (Eogan, J. 2000; Moore 2003; Ó Drisceoil 2003, 2007). Geophysical survey has also indicated the presence of an ‘avenue’ of large post pits in the field immediately to the east of Newgrange (McCarthy 2002; Kevin Barton pers. comm.) which also contains an earthen cursus monument. While this has also traditionally been seen as dating from the Late Neolithic, recent research indicates that similar monuments in the UK were constructed in the mid- to late fourth-millennium BC (Smyth et al. 2009, fig. 1.35, 22, 34; Barclay and Bayliss 1999; Thomas et al. 2009). Bronze Age/Iron Age The later prehistoric period has generally appeared underrepresented in the Brú na Bóinne landscape, with few diagnostic flint tools from fieldwalking surveys and only one Early Bronze Age cist known at Oldbridge. Recent work using aerial imagery and geophysical survey has identified a range of ring ditch features at

225

226

Stephen Davis, Conor Brady, William Megarry and Kevin Barton various locations which most likely date to the Bronze Age/Iron Age period (Cooney et al. 2001; Brady 2007; Smyth et al. 2009, 37–9). Additionally, the Great Circle of standing stones around the Newgrange passage tomb have been shown to post-date the construction of the timber circle and were erected during the Bronze Age (Sweetman 1985, 208). Interpretations of anomalies identified during geophysical survey on a low-profile D-shaped enclosure about 1 km west of the Newgrange mound, although without exact parallels, suggest a possible date from as early as the Late Bronze Age (Fenwick et al. 2009, 21–2). Evidence from the Knowth excavations suggests an extensive period of remodelling of the main mound in the late Iron Age, converting it into a large ditched enclosure. Several burials from this time were also excavated on the site. A highly unusual hoard of Roman material is known from Newgrange, dating from the early centuries AD. This was apparently deposited over a long time period as a series of votive offerings (Topp 1956; Carson and O’Kelly 1977; Stout 2002, 72). Early Medieval A number of monuments in the area have been suggested to be of early medieval date based upon their morphology. Sites N and R appear to be ringforts, a common settlement enclosure type from the period. Because of their size and prominent location, both are suggested to have been of high status (Stout 2002, 77). A further example is recorded at Rathmullan within the WHS, and there are four others recorded as cropmarking (ibid. 78). Site M, a series of earthwork features to the northeast of the mound at Knowth appeared to also contain one such enclosure and excavation revealed three phases of activity, during one of which the site was used as an enclosed secular cemetery (Stout and Stout 2008b). Knowth became an important settlement focus during the early medieval period, functioning as a royal site, and it is likely that Dowth functioned in a similar way as suggested by the presence of a souterrain, an underground passage or cellar believed to function as a refuge during attacks, or perhaps as a storage area (Eogan 1977, 1990, 2007; O’Kelly and O’Kelly 1983; Stout 2002; Smyth et al. 2009; Swift 2008). Several other souterrains are known within the WHS and indicate settlement at the time (Stout 2002, 81).

Further evidence of early medieval settlement has recently been uncovered at Stalleen and Rossnaree (Stephens 2009; Brady 2010, 2011). The site at Rossnaree, a multi-vallate enclosure measuring 250 m by 120 m, had no obvious surface expression either in the lidar data or aerial photographs, and was first identified through geophysical survey. The nearby Hill of Slane became a major ecclesiastical centre at this time and historical sources refer to a number of monastic foundations in the area at Dowth and Rossnaree (Stout 2002, 2007; Seaver and Brady 2011), for example. Medieval and post-medieval During the 12th century, Brú na Bóinne fell under the control of the Cistercian monastery at Mellifont, and saw a major reorganisation of the landscape into granges or farms. Several weirs on the Boyne may date to this period, and there are historical references to a number of mills along the river (Stout 2002; Smyth et al. 2009). The site at Stalleen (Stephens 2009), as well as producing evidence of early medieval occupation, also included evidence of 14th century settlement, and may have been one such grange. The Anglo-Norman lord Hugh de Lacy was granted the lordship of Meath by Henry II in 1172. A series of earthwork castles was erected across the lordship at this time, one at Knowth and a possible second at Dowth. Manorial villages were often developed at such centres and one may have developed at Dowth (Stout 2002, 95–6). A church which may have been attached to this village still stands there. Other small manorial villages would have existed in the wider landscape. Two small castles or tower-houses were constructed in the manor of Dowth, one adjacent to the church mentioned above which still stands and another at Proudfootstown which had disappeared by the end of the 19th century (Stout 2007, 348–9). In the 18th century there was extensive development of a demesne or parkland landscape at Dowth Hall. The Dowth estate included a range of landscape features, including a race course, numerous ‘tree rings’ (perhaps reflecting pre-existing archaeological features), a deer park and a ‘summer house’ on top of Dowth mound itself (O’Kelly and O’Kelly 1983, 138). The stone circle at Cloghalea, much admired by antiquarian writers, was destroyed by quarrying as late as the 19th century (Stout 2002, fig. 39, 37).

18  Lidar survey in the Brú na Bóinne World Heritage Site Remote sensing in Brú na Bóinne: aerial survey While good aerial imagery for Brú na Bóinne exists, not least through the online archive of the Ordnance Survey Ireland (three vintages are currently available – 1995, 2000 and 2005, with 2010 imagery imminent) no systematic aerial review of the area has been compiled. Some excellent oblique imagery exists, most notably through the work of Leo Swan (see Condit 2005) and Gillian Barrett (pers. comm.). Several previously unrecorded sites have been identified within the WHS through aerial survey, including the ‘D-shaped enclosure’ described by Swan and Condit (2000), a curvilinear enclosure associated with Site M described by Barrett and a series of cropmarks to the north of Site P, again described by Barrett (G. Barrett, pers. comm.). Remote Sensing in Brú na Bóinne: lidar survey In 2006 an area of about 90 km2 was captured by the UK Environment Agency (EA) at the behest of Meath County Council. This was post-processed by the EA and supplied in the form of ASCII XYZ files, gridded to a 1 m2 per pixel resolution. These lidar data have been incorporated into a comprehensive GIS, which includes multiple diverse datasets such as Ordnance Survey Ireland (OSI) vertical aerial imagery, historical mapping, Record of Monuments and Places (RMP) information, geological information (Geological Survey of Ireland) and archaeological datasets (e.g. lithic scatter distribution, and geophysical survey data (Brady 2007; Lewis et al. 2009; Brady and Barton 2010; Davis et al. 2011). The area flown extends well beyond the current confines of the WHS, stretching almost to the village of Slane in the west and to the outskirts of Drogheda to the east (Figure 18.1). This area included a total of 279 previously recorded monuments. Visualisation initially focused on standard analytical hillshading with low solar incident angle and varying azimuth. Other visualisation methods were also employed, including multiple direction hillshades (cf. Crutchley 2006) and Principal Component Analysis (PCA – Devereux et al. 2008), Local Relief Modelling (LRM; Hesse 2010) and SkyView Factor (SVF; Zakšek et al. 2011; Kokalj et al. 2011). These techniques have recently been evaluated and compared by Challis et al. (2011) and Kokalj et al. (this volume).

Embanked enclosures Of the three largest embanked enclosures within Brú na Bóinne, Sites A and P are similar in form, incorporating a semi-circular ‘annex’ feature to one side of the main enclosure, clearly discernible through lidar. At Site P this faces eastward enclosing the site’s most obvious point of entry. An apparent second entrance, visible on a 1995 vertical aerial photograph as a small gap in the western bank, has no lidar expression and may point to a variation in bank construction rather than a physical entrance. At Site A the annex faces northeast and seems to have been subject to some quarrying (noted by Stout 1991), making the site more difficult to interpret. While the exact relationship between the annex features and the main enclosures is unclear, the lidar imagery implies (at least at Site P) that either the larger enclosure overlies the smaller or that the structure as a whole was built as a composite. While Site P does not enclose any obvious central feature, a possible passage tomb is located centrally within the main enclosure of Site A (see above). However, it is impossible to distinguish the sequence of construction without further ground-based investigation. While Herity (1974, 145) suggests that the embankment was a later feature surrounding an earlier cairn it is equally possible that the mound post-dates the enclosure (similar to Bradley 1998, 113). Site A and environs (Figure 18.2) Analysis of the lidar data has revealed significant further detail to these monuments, in particular to Site A (Figure 18.2A–D). In addition to the known annex feature, a second, smaller and very low-profile embanked enclosure with a diameter of about 64 m is evident to the north-northwest of the main enclosure, almost abutting it (Figure 18.2A). This is only evident with azimuths in the northeast quadrant (i.e. 0–90o) and is not significantly enhanced by other visualisation techniques. Application of SkyView Factor (16 directions, 10 pixel) accentuates a third annex structure to the northwest of the main enclosure (Figure 18.2B), in form much like the known annexes at Sites A and P. The relationship between these annexes is difficult to define as both are very low profile; however, there is clearly an overlap between them, representing phased construction. Finally, the northern section of the main enclosure appears to have been intersected by a long, curvilinear cut, running approximately northeast-southwest (marked on Figure 18.2D),

227

228

Stephen Davis, Conor Brady, William Megarry and Kevin Barton

Figure 18.2: Site A and its environs. From Top Left: A) Standard analytical hillshade; B) SkyView Factor (16 direction, 10 pixel); C) Local Relief Model (indicating additional possible enclosure to west – see text) and D) 16-direction hillshade. In LRM, depressions are indicated in blue, slight elevations in red and greater elevations in yellow. Note strong visibility of subsidiary enclosure in A) compared with D) and additional annex feature indicated in B)

again best visible with specific azimuths at low solar incident angle. Clearly this last feature, cutting across all of the others, is the latest phase of development. In the case of both Site A and Site P, slight bank deformities are evident. This is particularly clear at Site P where the annex meets the main enclosure to the south and at Site A the deformity exists to the northeast, again where the annex meets the main enclosure. These may represent associated structures (such as the north and south barrows at Stonehenge – cf. Bradley 1998, 96 – see also Site LP1 below) and require additional ground-based study. Site Q (Figure 18.3) The third of the great Boyne enclosures is Dowth Henge (Site Q – Figure 18.3A–B). The site currently has two known entrances (as postulated for Site P), one to the northeast, the other directly opposite to the southwest. There has been debate as to whether these represent original features, with Stout (1991, 259) suggesting that the northeast entrance is a later addition based on early mapping, bank morphology and the form of the immediate exterior. The incorporation of Dowth Henge into a heavily modified demesne or designed parkland landscape raises a number of questions regarding the apparently

well-preserved state of this monument and the large number of cropmarks in its immediate landscape. In early mapping these are shown as ‘tree rings’ (i.e. circular wooded areas), and many of them have limited lidar expression. It is clear that further ground-based investigation is required to determine their nature. The interior structure of Dowth Henge differs markedly from Sites A and P: these have saucer-shaped profiles with no trace of inner scarping, while at Dowth an internal ring approximately 20 m wide has been scarped in the construction of the bank (cf. Stout 1991). This leaves the monument with a noticeably domed central portion. A small enclosure (c.50 m diameter) occupying the area immediately inside, but not directly aligned with, the southwest entrance is visible in the lidar data. The southwest portion of this internal enclosure is missing, possibly owing to scarping in the construction (or perhaps renovation) of the main bank, giving the enclosure a ‘D-shaped’ appearance (Figure 18.3B). Two other features in the Dowth estate are noteworthy: a previously recorded large mound located to the southwest of Dowth Henge and the area to the west of this monument (Figure 18.3C–D). Although not assigned a letter by either Coffey or later authors, the mound is of similar stature to those at Knowth, Dowth

18  Lidar survey in the Brú na Bóinne World Heritage Site

229 Figure 18.3: Dowth Henge and its environs. From Top Left: A) Standard analytical hillshade; B) Detail of Dowth Henge, with interior enclosure indicated; C) Sky View Factor (16 direction, 10 pixel) image of large mound and indistinct enclosure to west; D) Local Relief Model highlighting the same enclosure in LRM, depressions are indicated in blue, slight elevations in red and greater elevations in yellow

and Newgrange. It measures about 80 m across and has a ‘stepped’ edge, best preserved on the northern side, such as might be expected where there was a kerb (although no kerbstones are currently visible at ground level). The scale of this monument is such that these features are difficult to discern from the ground, but are clearly visible through lidar. The positioning of the site is of interest in that it is neither in a particularly prominent landscape position (contra the mounds at Newgrange, Knowth and Dowth) nor does it command particularly expansive views (as might be expected for a motte-type feature). Approximately 120 m to the west, a previously unrecorded possible embanked enclosure is evident, occupying a prominence overlooking the Boyne. Visualisation of the site is complicated by a former field boundary which intersects its northern edge. However, both LRM and SVF significantly enhance this feature and appear to confirm its presence (Figure 18.3C–D). Site B and its environs (Figure 18.4) A prominent mound on the lowest terrace of the Boyne, Site B lies to the southeast of Newgrange and has been interpreted as a passage grave (Stout 2002). Immediately to the west of Site B there is a second recorded mound (Site B1 – O’Kelly, C. 1978, 50–1), described as a ‘slightly raised circular

area with one stone on what might be its perimeter (diam. c.15 m)’ (RMP record, held at http://www. archaeology.ie). From the lidar is it evident that the mound is placed slightly off-centre within a large embanked enclosure, in excess of 100 m in diameter. A second similar but even lower profile mound and enclosure are suggested immediately to the north. As one of the lowestlying areas within the WHS, the presence of three palaeochannels have previously been noted here, and suggested as implying a relationship between the siting of monuments along multiple active watercourses (Lewis et al. 2009). However, lidar data reveal that the enclosure surrounding Site B1 overlies the central palaeochannel. Given the presumed Late Neolithic date of such enclosures, this implies that these channels have been inactive for several millennia. Site LP1 (Figure 18.5) Situated on the second terrace of the Boyne, directly above and to the northeast of Site B, a large, low-profile embanked enclosure was noted, encircling an equally low profile central mound (Figure 18.5A–C). This measures about 120 m in diameter with a vertical expression of some 20 cm and was named Site LP (Low Profile) 1. Visibility of Site LP1 was enhanced through PCA of a 16-direction hillshade and

230

Stephen Davis, Conor Brady, William Megarry and Kevin Barton

Figure 18.4: Site B and its environs. A) Standard analytical hillshade; B) PCA of 16-direction hillshade. Note visibility of palaeochannels in B) and enclosure to the west of main mound, surrounding Site B1 and overlying central palaeochannel

Figure 18.5: Site LP1. From Top Left: A) Standard analytical hillshade; B) PCA of 16-direction hillshade; C) Local Relief Model and D) Magnetic gradiometry survey overlain on 16-direction hillshade. In LRM depressions are indicated in blue, slight elevations in red and greater elevations in yellow. Magnetic gradiometry data acquired on a 1m × 0.25m grid

revealed a bank deformity towards the eastern side of the monument. The site was investigated using magnetic gradiometry, earth resistance and electrical resistivity tomography (ERT) (Barton 2011, 91–100). Gradiometry results revealed two parallel curving ditches to the north side

of the site. The remainder of the enclosure has little magnetic expression. Two previously unrecognised features were also recorded: to the south there is a sinuous ditch feature, while to the east is a circular ditched enclosure about 15 m in diameter, possibly surrounded by a ring of pits

18  Lidar survey in the Brú na Bóinne World Heritage Site

231 Figure 18.6: Site LP2. From Top Left: A) Standard analytical hillshade; B) Local Relief Model; C) Magnetic gradiometry survey overlain on Local Relief Model; D) Magnetic gradiometry and earth resistance survey overlain on analytical hillshade. In LRM depressions are indicated in blue, slight elevations in red and greater elevations in yellow. Magnetic gradiometry data acquired on a 1 m × 0.25 m grid. Earth resistance data acquired on a 0.5 m × 0.5 m grid

or post-holes with a diameter of about 30 m. This latter feature does not directly correspond to the previously noted bank deformity but lies slightly to its south. Site LP2 (Figure 18.6) This previously unrecorded site was identified through lidar as one of a pair of low-profile mounds (maximum height c.20 cm, diameter c.30 m) adjacent to one of the so-called ‘ritual ponds’ between Newgrange and the Boyne (Figure 18.6A–B). Slight traces of an outer enclosure exist, about 140 m in diameter and defined by an extremely low bank, in which the mound is centrally located. This slight bank is particularly striking given the presence of substantial embanked enclosures of similar dimensions in Brú na Bóinne, and may represent an unfinished enclosure, abandoned prior to the construction of the main bank. The site was investigated using magnetic gradiometry (Figure 18.6C) and targeted earth resistance (Figure 18.6D). Gradiometry survey revealed a cut feature, 16 m long by 2 m across with splayed ‘terminals’ each 7 m long (Barton 2011, 101–12). Surrounding this is a circular cut feature about 120 m in diameter. The earth resistance survey focused on the central mound and identified the same feature recorded through gradiometry,

as well as outlining a higher resistance zone probably representing the footprint of a destroyed covering mound. The arrangement of mound incorporating a cut feature with splayed terminals suggests that this may represent the remains of a former passage tomb. Dowth routeways (Figure 18.3A; Figure 18.7) A sinuate raised feature was identified running from the banks of the Boyne at Dowth, past the destroyed stone circle at Cloghalea, through Dowth Henge and, passing underneath Dowth Hall, towards the tumulus of Dowth itself, a distance of about 1.6 km. The winding nature of this feature, coupled with its association with the three principal monuments of the Dowth landscape (Cloghalea, Dowth Henge, Dowth) are suggestive of possibly earlier ritual associations, particularly given the feature apparently predates the demesne landscape. As a linear ridge, this feature is, unsurprisingly, strongly enhanced by SVF. Additionally, a sinuate hollow-way, defined by two low parallel banks was identified, running from northwest of Dowth mound towards the wetland area of Ballyboy Lake, a distance of about 700 m (Figure 18.7A–B). At Ballyboy the feature becomes difficult to trace. However, a second hollow-way extends southwards from

232

Stephen Davis, Conor Brady, William Megarry and Kevin Barton

Figure 18.7: Dowth Hollow-way (eastern), located NW of Dowth and running SE-NW: A) 16-direction hillshade; B) PCA of 16-direction hillshade

the western side of the lake towards a second former lake basin at Balfeddock. The antiquity of these features is impossible to determine without excavation; however, once again they do not respect field boundaries and traverse large areas of the ritual landscape, with a focus on wetland areas within Brú na Bóinne. Visualisations of the Brú na Bóinne lidar: an assessment Multi-direction hillshade Results obtained using multi-direction hillshade algorithms were mixed. While the visualisation of certain features (in particular the Dowth hollow-ways – Figure 18.7) was enhanced, in most cases little improvement was evident, and some known monuments were less visible using this approach owing to their shadows being effectively removed (e.g. Site A – Figure 18.2D). Low relief circular features (such as the low profile embanked enclosures) were particularly poorly served. A more useful approach to multidirection hillshading was to focus on a particular subset of azimuths, rather than a regularly distributed 16-direction approach. Application of PCA to these multi-direction hillshade models has produced some promising results, strongly enhancing visibility of some low profile features (e.g. Site LP1 and its subsidiary enclosure; Site B1 – Figure 18.4B; Figure 18.5B).

SkyView Factor A number of recent papers have discussed the application and utility of SVF algorithms as an aid to archaeological interpretation of lidar datasets (Challis et al. 2011; Zakšek et al. 2011; Kokalj et al. 2011, this volume). Owing to the character of many monuments within Brú na Bóinne (i.e. relatively gently sloping mounds, banks and ditches) SVF proved of limited utility for archaeological prospection beyond what could achieved using analytical hillshading. Although visibility of some features within the WHS was enhanced (e.g. the second annex at Site A – Figure 18.2B; the low profile embanked enclosure at Dowth – Figure 18.3C), SVF performed best with stone-built structures and foundations which lay outside of the WHS but within the lidar area. Local Relief Models (LRM) This approach (after Hesse 2010) highlighted a large number of potential archaeological features, including quarry or pond features, mounds, enclosures and field boundaries. Two areas where the technique proved particularly useful were in highlighting the presence of a possible small enclosure between Sites A and P (circled on Figure 18.2C), and apparently confirming the presence of an embanked enclosure on the promontory to the west of Dowth (Figure

18  Lidar survey in the Brú na Bóinne World Heritage Site 18.3D). While this approach has significant potential, it produces outputs which are noisy and difficult to interpret on a landscape scale. Many features which are flagged using LRM alone are impossible to distinguish from natural hillocks and require ground-based investigation to clarify their nature. Within the Brú na Bóinne landscape it has proved most useful in exploring areas where low-profile monuments are hinted at through other methods, offering a further means of testing their validity. This echoes the views of Challis et al. (2011, 238) who note the utility of the method in ‘visualising and isolating welldefined archaeological earthworks in low-relief landscapes’. Beyond prospection: lidar and the Brú na Bóinne landscape The principal difficulty with integrating new archaeological data obtained through lidar survey into a landscape-scale narrative at Brú na Bóinne is the lack of chronological context for the majority of sites within the WHS. For example, discussion of the Late Neolithic landscape of Brú na Bóinne is hampered by our lack of securely dated Late Neolithic monuments (e.g. embanked enclosures). Beyond glimpses of the middle Neolithic landscape of the passage tomb builders and the early medieval landscape of Knowth Site M and Rossnaree there is a sizeable gap in our current understanding. One approach which we have explored is the use of viewshed analysis. There has been much discussion of the validity of visibility studies in landscape archaeology (e.g. Thomas 2004; see also Chapman and Geary 2000), in particular questioning the value of computers in exploring how past peoples interacted with their landscapes, and commenting on their use in the context of past environmental change. It can be argued that Brú na Bóinne is a better candidate than most for the application of these methods. As the river here occupies a rock-cut channel, its mobility throughout the Holocene has been restricted. This is evident at Site B1 and its enclosure, discussed above, indicating that even the lowest terraces of the Boyne pre-date much of the archaeology. While environmental data from Brú na Bóinne are limited, insect analysis of turf samples from the main mound at Knowth suggest that the immediate environment was, at the time of construction, cleared pasture, with low scrub occupying more marginal areas (Davis forthcoming; O’Donnell forthcoming).

It is unlikely that dense woodland has existed here since pre-Neolithic times. As such it is probable that the large mounds would have been as visible in the past as at present (perhaps even more so given later enclosure and hedgerow development). Using lidar-derived topographic data, a number of research questions were posed. These included: how different was the view of the landscape from the three large tombs to the view of the tombs from the landscape, and how did the view of the landscape change as one journeyed along the River Boyne? These questions addressed important issues about the visual power of the monuments, and the role of the river as a communications route. Exploring the views of and views from the tombs Ten points were digitised on each of the three tombs of Newgrange, Knowth and Dowth. A viewing radius of 5 km was set, accounting for the immediate landscape. By setting the landscape offset at 1.5 m, and leaving the viewer offset at 0 m, a viewshed was generated, showing where in the landscape each of the tombs was visible from. Knowth has the largest viewshed of 38 km2 and is visible from 48% of all land within 5 km of the site. Dowth has the second largest viewshed at 37 km2 or 47% while Newgrange is visible from 32 km2 or 41% of the surrounding region.The three viewsheds were then combined to generate a cumulative viewshed, showing where one, two or all three of the tombs could be seen from (Figure 18.8). This demonstrates that the tombs were more visible from the southern banks of the river and north of the River Mattock rather than from within the WHS itself. A second set of viewsheds were then generated over the same distance from the entrances of the tombs. The offsets were reversed to account for views from the tumuli. It is important to note that views were taken from two points at the entrances of the tombs, and not from the top of the mounds. Here, conversely, Knowth has the smallest viewshed with only 10 km2 visible from the eastern entrance way, compared to Newgrange with 11 km2 and Dowth with the largest viewshed of 15 km2. These views encompass substantially smaller areas than views of the tombs from the landscapes. Exploring how the view changes along the river The Boyne has been a key communications route in the region for millennia, and is hypothesised as having facilitated the transport of building

233

234

Figure 18.8: Cumulative Viewshed analysis for Newgrange, showing areas visible from the tomb vs. areas from where the tomb is visible

Stephen Davis, Conor Brady, William Megarry and Kevin Barton

materials used during the construction of the passage tomb cemeteries (Mitchell 1992, 129). As previously discussed, the river has probably altered its course very little throughout the Holocene, allowing us to explore questions of changing perception of landscape through space. Key points were taken along the river and viewsheds generated from these points. A 5 km radius was set as was a 2 m offset to allow for the draught of the boat and height of the observer. In the absence of vegetation the monuments of the WHS would begin to be visible some distance to the east. However, the presence of low scrub of the type suggested by palaeoenvironmental data is likely to have obscured even the largest monuments until the traveller reached Dowth, where Newgrange would have become visible (approximately 6–7 m of vegetation would be required to obscure this view). The only place where all three tombs would be visible was at the apex of the Bend (adjacent to Site P), at the heart of the ritual landscape. The landscape can be interpreted as becoming increasingly revealed through movement along the river; the form of the river channel is particularly striking in this regard, with the steeply incised sections at

Knowth and Dowth acting to screen the central portion of the WHS.

Conclusions Brú na Bóinne can justifiably claim to be Ireland’s most studied archaeological landscape. While in remote sensing terms it might be argued that it is matched by the Hill of Tara (e.g. Newman 1997; Corns et al. 2008), Brú na Bóinne has a history of excavation, publication and synthesis that is unrivalled. The application of lidar within this well-studied landscape has proved a revelation. New potential monuments include three enclosures (LP1, LP2 and the enclosure at Site B1) each in excess of 100 m diameter. A fourth enclosure at Dowth is over 80 m across. This has potentially doubled the number of embanked enclosures within the WHS. Site A has been revealed to be a highly complex, and most likely multi-phase monument, the large enclosure apparently overlying at least two smaller, previously unrecorded enclosures. Route ways of potential ritual significance also are revealed, forming an almost continuous route from the

18  Lidar survey in the Brú na Bóinne World Heritage Site Boyne at Dowth Henge past Cloghalea, through Dowth Henge itself, to Dowth, and from there to Ballyboy, and finally to Balfeddock, a total distance of approximately 3 km. Other low profile features have also been detected and await the opportunity to conduct further groundbased research. In total, from the 279 recorded monuments known from the total lidar survey area (93 within the WHS) at the outset of our survey, a further 132 potential archaeological features have been noted, including 68 within the WHS itself. While in many ways Brú na Bóinne is a perfect area for lidar prospection owing to its minimal woodland cover and gentle topography, comparison of a range of visualisation techniques (similar to that proposed by Challis et al. 2011) has revealed little that was not already strongly hinted at through standard analytical hillshading (see also Crutchley this volume). That is not to say that some features were not better enhanced using particular methodologies, rather that, as lidar itself is no panacea (as has been demonstrated by recent discoveries at Rossnaree – cf. Brady 2010, 2011), there is also no panacea within these visualisation methods. Clearly there is also a pressing need to follow up the completed analysis of lidar survey with additional geophysical survey and targeted excavation, as well as potentially employing other remote sensing techniques. This can be seen as a natural progression of the work begun by Smyth et al. (2009) in defining a research framework for the WHS, but is complicated by the small proportion of the WHS in state ownership. Within Brú na Bóinne we have also sought to explore the value of lidar for more than solely archaeological prospection. Lidar has begun to help answer questions regarding past human interactions with their landscape through the application of viewshed studies. While such approaches are not without their critics, they at least serve as a starting point for discussion, and provide some fascinating insights into whether monuments were constructed to be seen or to see from, as well as regarding the interaction of the river with the ritual landscape.

Acknowledgements Our thanks to Žiga Kokalj and Keith Challis for useful discussion on lidar visualisation and Ralf Hesse for undertaking the LRM data processing. This research was funded by the Heritage Council INSTAR scheme and Meath County Council,

and we gratefully acknowledge the contributions of all researchers who have participated in this project. We would also like to thank Gillian Barrett for helpful discussion on her aerial surveys at Brú na Bóinne.

References Barclay, A. and Bayliss, A., 1999. Cursus monuments and the radiocarbon problem. In Barclay, A. and Harding, J., (eds). Pathways and Ceremonies: The Cursus Monuments of Britain and Ireland, 11–29. Oxford: Oxbow. Barton, K., 2011. Part 2 – Geophysical Investigations. In Davis, S., Megarry, W., Brady, C., Lewis, H., Cummins, T., Guinan, L., Turner, J., Gallagher, C., Brown, T., and Meehan, R., (eds). Boyne Valley Landscapes Project Phase III Final Report 2010, 91–129. Unpublished report for INSTAR The Heritage Council. Brady, C., 2007. A landscape survey of the Newgrange environs: earlier prehistoric settlement at Brú na Bóinne, Co. Meath. Unpublished PhD thesis, University College Dublin. Brady, C., 2010. Excavation of an enclosure site at Rossnaree (Brú na Bóinne), Co. Meath 10E242: Preliminary Report, Excavation 2010. Unpublished report for the Royal Irish Academy. Brady, C., 2011. Excavation of an enclosure site at Rossnaree (Brú na Bóinne), Co. Meath 10E242: Preliminary Report, Excavation 2011. Unpublished report for the Royal Irish Academy. Brady, C. and Barton, K., 2010. Investigation of Geophysical Properties of Lithic Scatter Sites at Brú na Bóinne, Co. Meath. Interim Report Archaeology Grant 17110. Final Report. Unpublished report for the Heritage Council. Brady, C., Davis, S. and Mulrooney, G., 2010. Part 3 – Palaeoenvironmental investigations: Augering investigation at Site W, Monknewtown ‘Ritual Site – Pond’ at Monknewtown Co. Meath (ME019015--). In Davis, S., Megarry, W., Brady, C., Lewis, H., Cummins, T., Guinan, L., Turner, J., Gallagher, C., Brown, T., and Meehan, R., (eds). Boyne Valley Landscapes Project Phase III Final Report 2010, 130–43. Unpublished report for INSTAR The Heritage Council. Bradley, R., 1998. The significance of monuments. Routledge, London. Carson, R. and O’Kelly, C., 1977. A catalogue of the Roman coins from Newgrange, Co. Meath. Proceed­ ings of the Royal Irish Academy 77C, 35–55. Challis, K., Forlin, P. and Kincey, M., 2011. A Generic Toolkit for the Visualization of Archaeological Features on Airborne LiDAR Elevation Data. Archaeological Prospection 18, 279–89. Chapman, H.P. and Gearey, B.R., 2000. Palaeoecology and the perception of Prehistoric landscapes, Antiquity 74, 316–9.

235

236

Stephen Davis, Conor Brady, William Megarry and Kevin Barton Coffey, G., 1912. New Grange (Brugh Na Boinne) and other incised tumuli in Ireland. Hodges, Figgis & Co.: Dublin. Condit, T., 1997. Late Neolithic ritual – earthen ceremonial enclosures. In Condit, T., (ed.). Brú na Bóinne, a supplement to Archaeology Ireland, 41, 22–3. Condit, T., 2005. Some aerial photos of Leo Swan of sites in County Meath. In Condit, T. and Corlett, C., (eds.). Above and Beyond: essays in memory of Leo Swan. Wordwell: Bray. Cooney, G., 2000. Landscapes of Neolithic Ireland. London: Routledge. Cooney, G., Byrnes, E., Brady, C. and O’Sullivan, A., 2001. A Pilot Archaeological Survey of the Site of the Battle of the Boyne, Oldbridge Estate, Co. Meath. Unpublished report for Dúchas, The Heritage Service. Corns, A., Fenwick, J. and Shaw, R., 2008. More than meets the eye. Archaeology Ireland 22, 34–8. Crutchley, S., 2006. Light detection and ranging (lidar) in the Witham Valley, Lincolnshire: an assessment of new remote sensing techniques. Archaeological Prospection 13, 103–27. Davis, S.R., Forthcoming. The Insect Remains from Knowth. In Eogan, G. and Cleary, K., (eds). Excavations at Knowth 6: The Neolithic archaeology of the large passage tomb at Knowth, Co. Meath. Dublin: Royal Irish Academy. Davis, S.R., Megarry, W., Brady, C., Lewis, H., Cummins, T., Guinan, L., Turner, J., Gallagher, C., Brown, A.G. and Meehan, R., (eds) 2011. Boyne Valley Landscapes Project Phase III Final Report 2010. Unpublished report for INSTAR The Heritage Council. Devereux, B.J., Amable, G.S. and Crow, P., 2008. Visualisation of lidar terrain models for archaeo­ logical feature detection. Antiquity 82, 470–9. Dúchas, The Heritage Service, 2002. Brú na Bóinne World Heritage Site Management Plan. Dublin: Department of the Environment and Local Government. Eogan, G., 1977. The Iron Age-Early Christian settle­ ment at Knowth. In Markotic, V. (ed.). Ancient Europe and the Mediterranean, 69–76. Warminster: Aris and Phillips. Eogan, G., 1986. Knowth and the Passage Tombs of Ireland. London: Thames and Hudson. Eogan, G., 1990. The archaeology of Brugh na Bóinne during the early centuries A.D. Seanchas Ard Mhacha: Journal of the Armagh Diocesan Historical Society 14(1), 20–34. Eogan, G., 2007. Knowth in the Early Christian Period: the cultural and chronological context of the animal bone assemblage. In McCormick, F. and Murray, E., (eds). Excavations at Knowth 3: Knowth and the Zooarchaeology of Early Christian Ireland, 1–6. Dublin: Royal Irish Academy. Eogan, G. and Roche, H., 1997. Excavations at Knowth 2. Dublin: Royal Irish Academy. Eogan, G. and Doyle, P., 2010. Guide to the passage tombs at Brú na Bóinne. Wordwell: Bray.

Eogan, J., 2000. Bettystown: Prehistoric/multiperiod site. In Bennett, I. (ed.). Excavations 1998: Summary accounts of excavations in Ireland, 161. Bray: Wordwell. Fenwick, J., Warner, R. and Eogan, G., 2009. Revealing hidden details of the ancient landscape of Newgrange, Brugh na Bóinne World Heritage Site, Co. Meath. Ríocht na Midhe XX, 5–23. Harbison, P., 2007. In Retrospect: The Royal Irish Academy’s only archaeological excavation: Dowth in the Boyne Valley. Proceedings of the Royal Irish Academy, Section C 107, 205–13. Herity, M. 1974. Irish Passage Graves. Irish University Press: Dublin. Hesse, R. 2010. LiDAR-derived Local Relief Models – a new tool for archaeological prospection. Archaeological Prospection 17, 67–72. Kokalj, Ž., Zakšek, K. and Oštir, K., 2011. Appli­ cation of sky-view factor for the visualization of historic landscape features in lidar-derived relief models. Antiquity 85, 263–73. Lewis, H., Davis, S.R., Turner, J., Foster, G., Brady, C., Gallagher, C., Guinan, L., Meehan, R., Brown, A. G. and Cummins, T., 2009. An integrated, comprehensive GIS model of landscape evolution and land use history in the River Boyne valley Phase II. Unpublished report for the Heritage Council. Lynch, A., 1990. ‘Dowth’, Dowth. In Bennett, I. (ed.). Excavations 1989: Summary accounts of excavations in Ireland. Bray: Wordwell. McCarthy, M. 2002. A preliminary geophysical report on the reputed cursus at Newgrange, Co. Meath. Unpublished report for Dúchas the Heritage Service. Mitchell, G.F. 1992. Notes on some non-local cobbles at the entrances to the passage-graves at Newgrange and Knowth, Co. Meath. Journal of the Royal Society of Antiquaries of Ireland 122, 128–45. Moore, D., 2003. Site 19, Duleek Road, Platin. Neolithic House. In Bennett, I. (ed.). Excavations 2001: Summary accounts of excavations in Ireland, 320. Bray: Wordwell. Newman, C., 1997. Tara: an archaeological survey. Discovery Programme Monographs 2. Royal Irish Academy, Dublin. O’Donnell, L., forthcoming. Charcoal and wood analysis from Knowth. In Eogan, G. and Cleary, K., (eds). Excavations at Knowth 6: The Neolithic archaeology of the large passage tomb at Knowth, Co. Meath. Dublin: Royal Irish Academy. Ó Drisceoil, C. 2003. Balgatheran Site 4: Late Neolithic ritual/settlement site. In Bennett, I. (ed.). Excavations 2001: Summary accounts of excavations in Ireland, 255–7. Bray: Wordwell. Ó Drisceoil, C., 2007. A preliminary report con­ cerning the archaeological excavation of Neolithic houses at Coolfore, County Louth. County Louth Archaeological and Historical Journal XXVI, 3, 260–85. O’Kelly, C., 1978. Illustrated Guide to Newgrange and the Other Boyne Monuments. Third edition. Cork.

18  Lidar survey in the Brú na Bóinne World Heritage Site O’Kelly, M.J., 1982. Newgrange. Archaeology, Art and Legend. London: Thames and Hudson. O’Kelly, M.J., Lynch, F. and O’Kelly, C., 1978. Three Passage-Graves at Newgrange, Co. Meath. Proceedings of the Royal Irish Academy. 78C, 249–352. O’Kelly, M.J. and O’Kelly, C., 1983. The tumulus of Dowth, County Meath. Proceedings of the Royal Irish Academy 83C, 135–90. O’Kelly, M.J., Cleary, R. and Lehane, D., 1983. Newgrange, Co. Meath, Ireland. The Late Neolithic/ Beaker Period Settlement, BAR International Series 190. Oxford: British Archaeological Reports. Ó Ríordáin, S. and Daniel, G., 1964. Newgrange and the Bend of the Boyne. London: Thames and Hudson. Seaver, M. and Brady, C., 2011. The Hill of Slane. Archaeology Ireland Heritage Guide No. 55. Bray: Wordwell. Smyth, J., Brady, C., Chadwick, J., Condit, T., Cooney, G., Doyle, I., Guinan, L., Potterton, M., Stout, G. and Tuffy, C., 2009. Brú na Bóinne World Heritage Site research framework. The Heritage Council, Kilkenny. Stephens, M., 2009. Home on the Grange. Archaeo­ logy Ireland 23, 31–3. Stout, G., 1991. Embanked enclosures of the Boyne region. Proceedings of the Royal Irish Academy 91C, 245–84. Stout, G., 2002. Newgrange and the Bend of the Boyne. Cork University Press, Cork. Stout, G., 2007. The church and tower house at Dowth, County Meath. In Manning, C., (ed.). From ringforts to fortified houses: studies on castles and other monuments in honour of David Sweetman, 335–51. Bray: Wordwell. Stout, G. and Stout, M., 2008a. Newgrange. Cork: Cork University Press.

Stout, G. and Stout, M., 2008b. Excavation of a secular cemetery at Knowth Site M, County Meath, and related sites in north-east Leinster. Bray: Wordwell. Swan, L. and Condit, T. 2000. New enclosure at Brú na Bóinne. Archaeology Ireland 14, 34–37. Sweetman, P.D., 1971.An Earthen Enclosure at Monknewtown, Slane Preliminary Report. Journal of the Royal Society of Antiquaries of Ireland 101, 135–40. Sweetman, P.D., 1976. An earthen enclosure at Monknewtown, Slane, Co. Meath. Proceedings of the Royal Irish Academy 76C, 25–73. Sweetman, P.D., 1985. A Late Neolithic/Early Bronze Age pit circle at Newgrange, Co. Meath. Proceed­ ings of the Royal Irish Academy 85C, 195–222. Sweetman, P.D., 1987. Excavation of a Late Neolithic/ Early Bronze Age site at Newgrange, Co. Meath. Proceedings of the Royal Irish Academy 87C, 283–98. Swift, C., 2008. The early History of Knowth. In Byrne,F.J., Jenkins,W., Kenny, G. and Swift, C. (eds). Historical Knowth and its Hinterland. Excavations at Knowth 4, 5–53. Dublin, Royal Irish Academy. Thomas, J., 2004. Archaeology and Modernity. Routledge, London. Thomas, J., Marshall, P., Parker-Pearson, M., Pollard, J., Richards, J., Tilley, C. and Welham, K., 2009. The date of the Greater Stonehenge Cursus. Antiquity 83, 40–53. Topp, C., 1956. The gold ornaments reputedly found near the entrance to New Grange in 1842. 12th Annual Report of the University of London Institute of Archaeology, 53–62. Zakšek, K., Oštir, K. and Kokalj, Ž., 2011. Sky-view Factor as a relief visualisation technique. Remote Sensing 3, 398–415.

237

19 Immersive visualisation of survey and laser scanning: the case for using computer game engines Keith Challis and Mark Kincey Archaeological metric survey is the primary method of recording the extant physical remains of past landscapes and monuments. Methods of survey have developed dramatically in the recent past with the advent of new digital survey techniques, global positioning systems and laser scanning. The volume of data collected to record monuments and landscapes are now vast, and levels of accuracy and precision unprecedented. This growth in data quality and volume has to some extent been accompanied by a reluctant theoretical debate, largely about method and meaning in the practice of survey. Less well explored is the area of visualisation of survey results, which has tended to remain rooted in traditional approaches, albeit facilitated by new digital media. The ability of modern digital survey to engage with other areas of archaeological debate, for example discussions of sense of place, meaning and interpretation in landscape, as embodied by the phenomenological approach to landscape, for example, has largely been ignored as it is poorly addressed using conventional static visualisation techniques. This paper explores the potential of computer game software to produce accurate, immersive and interactive visualisations of digital survey data of archaeological monuments. We explore visualisation of data from three sources, ground based survey, terrestrial laser scanning and airborne lidar, with a view to producing accurate representations of the landscape “as is” and exposing these visualisations to the phenomenological paradigm. Keywords: archaeology, survey, laser scanning, lidar, visualisation, computer games

Introduction This paper explores the potential of computer game software to enhance our capacity as archaeologists to visualise, explore and understand the data generated by archaeological surveys. We are interested in working at the macro scale of landscape, exploring areas measured in hectares, as we believe that game software uniquely enable the viewer to engage with the huge volumes of data generated by modern archaeological surveys at this scale in a meaningful way. There are two aspects to this dialogue, the first relating to the game paradigm, and in particular the ludic elements of engagement with information and meaning engendered by games and the second the technological facets of game software, in

particular in the form of sophisticated rendering of three-dimensional graphics in real-time. In this paper we focus on the technology of rendering complex survey data in a game environment. Since we focus on the application of first-person games to visualisation we will also consider briefly the ludic nature of first-person engagement with landscape in-game and suggest some ludic mechanisms that might be employed to extend game-based visualisation beyond the realm of moving picture and into engaging, explorative play. We begin by exploring approaches to the recording and representation of archaeological landscapes and past uses of computer game software and the game paradigm in cultural

239

19  Immersive visualisation of survey and laser scanning: the case for using computer game engines heritage. We describe our approach to the display and visualisation of archaeological survey data using a game engine, exploring first our choice of first person game engine before describing our adopted methodology for the processing and visual display of survey data. In the final section we examine the possibilities of phenomenological exploration of survey data using games and discuss some ludic devices that may engender exploration of data, landscape and meaning.

Landscape, survey and representation Metric survey, the process of systematic and accurate measurement and interpretation, is a key aspect of producing a record of and so preserving archaeological structures, monuments and landscapes (Bryan et al. 2009). The techniques of archaeological survey have developed rapidly in complexity and sophistication over the last decade or so, principally due to the introduction of rapid semi-automated and automated means of data collection such as the electronic distance measurer (EDM) and survey instruments based on global positioning systems (GPS) (cf. Bettess 1998; Bowden 1999; English Heritage 2002, 2003; Howard 2007). Recent advances have tended to focus on the ability of GPS and so-called robotic total stations to collect data automatically with little intervention on the part of the surveyor other than to guide the instruments across the landscape (Barratt et al. 2000; Kvamme et al. 2006). Such innovations, coupled with developments in computer processing and storage of survey data, have dramatically increased the volume of data collected in a typical survey and the potential resolution (although not necessarily the accuracy) of the survey (Table 19.1). The arrival of such techniques has engendered a lively debate amongst archaeologists on the relative merits of deliberate versus automatic data collection (e.g. Chapman and Van de Noort 2001; Fletcher and Spicer 1988; Lock 2003, chapter 2). The sheer power, relative ease of use and potential for misuse of modern survey equipment has also fostered a range of technical guidance and specifications of archaeological survey (for example those of English Heritage, e.g. English Heritage 2007a). More recently the development and adaptation of laser scanning equipment for terrestrial survey of both structures and landscapes (an innovation adopted from the

Survey method

Average points m2

GPS/EDM

1–5 points m2

Airborne Laser Scanning

5–12 points m2

Terrestrial Laser Scanning

100+ points m2

earth sciences (Heritage and Hetherington 2007) and the increasing availability of airborne laser scanning data (ALS) has added to the repertoire of landscape survey data and techniques available to the archaeologist and to the ever expanding archive of automatically collected landscape data which contain archaeological content (English Heritage 2007b, 2010). This wealth of information presents both an enormous resource and a significant problem to archaeologists, for example in England and Wales alone the UK Government’s Environment Agency have collected over 16 million hectares of airborne lidar data, covering approximately 66% of the land surface in at least one survey flight. The majority of these data, replete with archaeological content, are entirely unexplored (Challis et al. 2008). The representation of archaeological survey data has undergone a transformation that corresponds with that in data collection. Computer based processing of digitally collected survey data, initially focussing on automatic calculation of the trigonometry associated with survey, has naturally led to the use of specialist computer software for the visualisation and display of survey data (Chapman 2003; Howard 2007; Lock 2003). Representations have tended to veer towards two-dimensional, digitally created, analogues of traditional hand-drawn survey plans, or pseudo three-dimensional representations of varying sophistication (Figure 19.1; English Heritage 2007a). The majority of the threedimensional representations of survey remains are in the form of static images, or occasionally short computer generated animations that offer no means of viewer interaction. Interestingly the pseudo-realistic representation of landscapes and monuments (although not structures) has tended not to rely on metric survey data for its source, and as a result it could be argued that such representations have tended to be viewed as fanciful or unreliable. Significant inroads into the pseudo-realistic representation of survey data, and the use of metric survey to

Data volume/ha 50k points/ha 1mb/ha 1.2 million points/ha 5mb/ha 10 million points/ha 1gb/ha

Table 19.1: Illustrative examples of the point density and data volumes generated by the survey methods discussed in the text

240

Figure 19.1: Typical computer derived visualisations of archaeological survey data, in this instance a contour plan and pseudo three dimensional model of the earthworks of Laxton castle

Keith Challis and Mark Kincey

inform visual reconstruction, have been made in some areas, for example the UK Channel 4 television programme Time Team, which combined intensive archaeological field work (including multi-method metric surveys) with the need for enticing television representation of results, made significant inroads into uniting survey and visualisation (Chapman 2006). Even here, the vast majority of visualisations tend to continue to rely on an elevated viewing angle, giving the viewer an entirely false and god-like overview of sites and landscapes (Johnson 2007). In this paper we argue that to escape from this elevated detachment back into the ‘muddy boots’ of landscape archaeology (Hoskins 1955) gives new meaning to survey data and reinvigorates both the practice of survey and the interpretation and understanding of results. Such insight, we argue, is uniquely the preserve of the first-person computer game.

Computer games and archaeological representation The representation of cultural heritage in computer games is endemic, but the use of game software and the game paradigm for the serious representation of heritage is a relatively new innovation. Popular computer games such as Tomb Raider (Eidos 1995) have used the motif and architecture of heritage and archaeology as a key facet of game play. Past historical settings, often based on considerable primary research to ensure authenticity, have become the benchmarks of such open world games as Assassin’s Creed (Ubisoft 2007; the medieval

world, principally the Middle East and Rome), Red Dead Redemption (Rockstar Games 2010; the Old West of the United States) and most recently LA Noir (Rockstar Games 2011; an authentically rendered 1947 Los Angeles). Other recent games such as Barrow Hill (Lighthouse Interactive 2008) have taken a more structured approach by adopting archaeological themes as part of the game play (Charno 2007). The adoption of heritage purely for play does not concern us particularly here, other than perhaps to observe that the co-option of archaeological motifs for game-play has tended to create a suspicion amongst archaeologists of the use of games and the game paradigm for more serious but equally immersive ends. Intentional, research-led uses of games for representing cultural heritage may be broadly divided into those uses that create virtual repositories for information, virtual museums where disparate and un-associated objects and information are corralled in a single virtual space, rather as in a real museum (Lepouras and Vassilakis 2005), a small but significant set of educational games with the specific intent of teaching archaeology through the medium of a game (Champion 2008) and, our subject here, game-based visualisation, where the sophisticated graphical power of computer games is adopted to render virtual representations of some aspect of cultural heritage, most often an imaginative reconstruction of an archaeological monument (Ch’ng 2009). Interestingly the majority of these examples adopt game software to provide a reconstructed, idealised past. The role of systematically collected archaeological data (let alone metric survey) is rarely explicit in such reconstructions and the aim is usually the engagement of the audience with a storied account of the past based on a reconstructed site or monument. Use of game software for visualisation of landscape is rare, but include Eugene Ch’ng’s reconstruction of pre-inundation Doggerland from oil industry seismic data and Bob Stone’s virtual Stonehenge (Ch’ng et al. 2004; Stone 2009). In this paper we set out to explore the utility of game software to visualise not an imagined reconstruction of the past, but those vestiges of antiquity that survive in the present landscape. Our aim is to explore how the first-person game paradigm might assist in gaining understanding of high density survey data and in the next section we first describe the range of survey data

19  Immersive visualisation of survey and laser scanning: the case for using computer game engines we have used, before discussing choices of game software for its visualisation and outlining our chosen visualisation methods.

Method: visualising archaeological survey Survey techniques and data Archaeological survey typically aims to create a durable record of the physical aspects of a site or monument that through taphonomic processes represents the surviving vestiges of a past entity. The act of survey may take place in a variety of contexts (English Heritage 2007a, 24) – it might be conservation driven (i.e. a baseline survey to allow monitoring of visitor impact on a monument), or a record of preservation prior to destruction. Research driven applications might aim to improve understanding of a monument or monument class. More rarely survey might be motivated purely by the desire to gather baseline data to inform visualisation or reconstruction of a monument, either as is or to a conjectured past state. Contemporary archaeological surveys make use of a variety of field methodologies. In the UK, in commercial archaeology, survey using rapid semi-automatic techniques such as GPS and EDM are typical. Terrestrial laser scanning (TLS) equipment is at present rarely used for survey of archaeological landscapes, more usually it is used for recording upstanding structures, and although the rapid, comprehensive and metrically precise record of elevation provided by TLS has been recognised as a valuable addition to the archaeological surveyor’s toolkit, high capital equipment costs at present limit uptake. ALS, with its ability to produce dense, metrically accurate elevation data, is widely recognised and adapted as a serendipitous source of data for identifying and characterising archaeological monuments and landscapes (Crutchley 2006), but with relatively few exceptions is not widely adopted as a means of primary archaeological survey (English Heritage 2010, 28). Whatever the means of collection, modern digital survey data require processing and visualisation to turn location specific elevation records (usually in the form of xyz coordinates) into meaningful representations of landscape and monument. This process usually involves the conversion of suitably processed field data into continuous surfaces, often termed digital terrain

241

models (DTM) via a process of interpolation and the generation of derived products from those DTM (McCullagh 1988). Some DTM derivatives, such as contour plans, may be analogues of traditional survey products, while others such as pseudo 3D views are an innovation of digital data processing. Integration of survey results in Geographical Information Systems (GIS) allows relationships with other landscape scale data (for example aerial photographs), generation of interpretative layers and some degree of imaginative visualisation (cf. Chapman 2006, fig. 24). However, typically the outputs from all of these processes are static visualisations, either as images or fixed viewpoint animations and it is rarely possible to explore data in an interactive way. Game engines Rendering a ground level view of landscape, in which the viewer may move freely in any direction in real time, requires the facilities provided by first-person gaming software. This hugely popular and diverse class of games is driven by a range of underlying software engines of varying graphical sophistication and hardware requirements (Trentholme and Smith 2008). In order to accurately render an elevation model derived from GIS-based analysis of our original survey it was essential to choose a game engine that could accommodate real-world terrain data. It was also a requirement that the game engine has an accessible editing tool and be freely or cheaply available. We also required tools to handle lighting and atmospheric effects, terrain texturing, incorporation of vegetation and object models derived from external software sources and free-form avatar-based interaction. A range of game engines meet some or all of these criteria (Table 19.2). The usual way for game software to accomm­ odate externally derived terrain data is via a greyscale height map, where variations in elevation are represented by shades of grey tone in a grid array forming a complete terrain block. This approach is adopted by Crytek’s CryEngine, as Engine Unity 3D Leadworks Unreal Engine Source CryENGINE

Table 19.2: First person game software

Uses Height maps

Integrated Editor

yes yes yes no yes

yes yes yes no yes

Cost Free – $1500+ $200 free free free

242

Keith Challis and Mark Kincey well as a number of other game engines, including Unity, Leadworks and Unreal Engine. Game engines such as Source make use of displacement maps, which provide a less direct pathway for incorporation of GIS derived elevation into the game environment. After some experimentation we settled on Crytek’s CryENGINE as our development environment of choice. It’s low cost, graphical sophistication (even though now seven years old), ability to work with greyscale height maps and GIS derived terrain texture maps, and the availability of well-documented opensource tools for the incorporation of externally modelled content were unmatched. We also found the Sandbox editor easy and intuitive to use and stable in use. Most of our development work has been undertaken using CryEngine 1 (FarCry) although some work has now transferred to the second generation CryEngine 2 (Crysis) which includes useful additional features such as dynamic day-night lighting and greater graphical sophistication, at the expense of greater demands on computer hardware. Sandbox provides an easy editing interface for worlds created using CryEngine. The editor provides a multi-pane workspace, allowing a 2D map-view, elevation views and a fully threedimensional real-time editing window, which in the event was the most used. Terrain is constructed from 8- or 10-bit greyscale height maps, and textured using masks generated from the terrain model, via direct painting onto the terrain, or with user generated 8-bit greyscale masks. As well as basic terrain editing there are tools to apply lighting effects, atmospheric and particle effects, vegetation, and to incorporate models of objects and structure, either from the in-game library of assets or (via appropriate conversion routines to CryTeks proprietary model format) from external modelling software. Extensive scripting tools allow for the generation of game scenarios and control player/AI interaction. The software is comprehensively described in CryTek’s Sandbox user manual (CryTek 2006). Using CryEngine In this study we have used data from landscape surveys undertaken using three different technologies. At Laxton Castle, in Nottinghamshire, survey of the earthworks of a substantial medieval motte and bailey castle was carried out over four field seasons by student and community groups using EDM total station and differential GPS (Kincey et al. 2006). In total

12,000 data points were collected over a 5 hectare area, and form the basis of the 3D model of the earthworks (Figure 19.2). At West Burton, also in Nottinghamshire, airborne lidar survey by the Environment Agency was used to generate a 2 m spatial resolution DTM of the earthwork remains of the deserted medieval village (Gaunt 2009), covering in total an area of approximately 26 hectares (Figure 19.4). The earthworks of the eastern entrance of British Camp, an Iron Age hillfort situated on Herefordshire Beacon in the Malvern Hills, Herefordshire (Quigley 2007), were surveyed using a Leica HDS 3000 TLS as part of a student project. Approximately 5 million survey points were collected in a single day of survey and were used to generate the DTM of the earthworks, covering an area of 6 hectares (Figure 19.2). In each case DTM data were prepared for use in CryENGINE using ESRI’s ArcGIS 9.3 to generate a greyscale height map. The terrain shading histogram was adjusted manually to provide optimum contrast between the lowest and highest elevation values, making full use of the 8-bit dynamic range. A simple graphic-mask was employed to highlight an area of compatible dimension with CryEngine’s terrain tools, which require terrain blocks to be sized in multiples of 128 pixels (each pixel may represent one or more metres on the ground) and the DTM image was exported from ArcGIS, trimmed to the area within the graphic mask and resized to the appropriate overall terrain dimensions in pixels, such that one pixel was equivalent to one metre on the ground. In Sandbox a new level of appropriate dimensions was prepared and the GIS-derived greyscale heightmap imported. The raw terrain thus generated in CryEgine is not accurately scaled vertically and careful editing, using the limited tools available in Sandbox, is required to produce aesthetically pleasing terrain with as near to precise vertical scaling as possible – 8-bit heightmaps offer a maximum of 256 elevation variations. The elevation change represented by each change in grey values can be calculated by comparison with the elevation range of the original ArcGIS DTM such that: Heightmax-heightmin/256 = elevation change per pixel

Since no precise vertical scaling tool exists in Sandbox, scaling is achieved by using the calculated per-pixel value and an object of known height as a visual guide. Some estimate

19  Immersive visualisation of survey and laser scanning: the case for using computer game engines

243 Figure 19.2: GIS derived visualisations of the survey data for the three sites discussed in the text, in each case, from the left, greyscale height map, single azimuth hillshade and pseudo three-dimensional view of hillshade

of the overall degradation in precision in the CryEngine based terrain model may be gained by comparing the original GIS-based DTM with the final CryEngine DTM by exporting the results of the final edit from Sandbox and using this for analytical comparison of elevation values in ArcGIS, although it should be noted that accurate impression of landscape, rather than metrical accuracy, is the aim of our research (Figure 19.3). The basic scaled terrain model requires the application of one or more raster based textures of differing types to simulate land cover (grass, dirt, leaves, etc.). Sandbox allows the layering of multiple textures with layer mask generated either via simple built-in DTM analysis tools, via direct painting onto the DTM or via greyscale masks, where the intensity of grey value controls the amount of each texture applied. Since our aim was to recreate as far as possible a visual

impression of the landscape as is, we have usually applied textures using masks derived from analysis of the GIS-based DTM. A variety of DTM derived products, including slope severity, analytical hillshade and solar insolation models (Challis et al. 2011), can be combined to assist in producing naturalistically graded texturing of large open areas. In addition, simple binary masks may be used for specific land use (e.g. woodland or artificial surfaces) and aerial photographs rendered into 8-bit greyscale may be utilised as texture mask to rapidly simulate changes in land cover over large areas (Figure 19.4). Vegetation was added in discrete layers, representing differing vegetation classes, principally ground cover, low scrub, taller scrub and trees, using bespoke European temperate vegetation models. The placing and character of vegetation was guided by reference to terrain texture masks developed from aerial photograph

244

Keith Challis and Mark Kincey

Figure 19.3: Raw unscaled and correctly vertically scaled terrain seen in the Sandbox Editor

and ground level photographs, to indicate type, character and density of vegetation. While complete fidelity in copying existing vegetation, for example matching real-world vegetation species and species diversity, was not feasible, in general it has proven possible to create authentic looking representations of the present landscape and land cover.

Results: engaging with landscape and meaning In this final section we explore some of the issues and insights gained through the application of

game technology to visualising archaeological survey data. In particular we explore four themes: 1) the paradox of the omniscient viewer and the muddy booted avatar; 2) the extent to which game-based visualisation allows exploration of landscape narratives and investigation of meanings; 3) the impact of variations in vegetation cover and atmospheric visual occlusion on views and viewing distance; 4) the ludic potential of landscape visualisation in-game. Much of the meaning that we, as viewers, are able to extract from and attach to landscape is dependent on the scale at which we view it and our viewing perspective. In the British Isles one of the paradigm-shifting advances in the

19  Immersive visualisation of survey and laser scanning: the case for using computer game engines

245

Figure 19.4: Examples of texturing of terrain using GIS derived terrain products, showing on top row from left, air photograph, slope severity map and solar insolation map, middle row, the three products combined to map texture onto the terrain model in Sandbox and bottom row, a finished terrain with texture mapping and vegetation seen in CryEngine 1

appreciation of landscape and heritage came with the work of the early aerial archaeologists, whose photographs revealed not just new sites seen in vestiges visible from the air, but whole integrated past landscapes (Crawford and Keiller 1928; Riley 1980). This change in perspective,

from ground to air, and close to distant, both revealed and emphasised the vastness of the altered landscape of past human endeavour. However, it might be argued that this overarching perspective, which has earlier echoes in the work of antiquarians such as Stukeley and the romantic

246

Keith Challis and Mark Kincey

Figure 19.5: An example of the radically different impressions of landscape given by the typical aerial overview afforded by GIS software and the ground level view of the same landscape seen within a game engine. The example is Laxton castle, viewed in CryEngine 2

poets of the early 19th century, detached us as observer from intimate familiarity with landscape (Johnson 2007). In the mid 20th century landscape historians such as WG Hoskins, working in the historical geographical tradition, emphasised the insights and understandings gained from fieldwork (Hoskins 1955). Late 20th century theorists such as Tilley and Bender (Tilley 1994; Bender et al. 2007), influenced by phenomenologists such as Heidegger (1972), added to this the insights of the emotive personal response of being in a landscape of senses. As we have seen, traditional GIS software, although in general allowing creation of accurate and precise visualisations of digital data, encourage the perpetuation of the distant, detached viewer by acting either as analogues for the paper map or the elevated view of the aerial archaeologists. A number of commentators have challenged this approach (Gillings and Goodrick 1996) and

GIS-based work has acted to immerse viewers in data, even if not in a world of sensuous data (Exon et al. 2000). We suggest that use of gamebased visualisation forces viewers to adopt the point of view of one physically rooted to the land, and thus produces a virtual representation of the experiences of the fieldworker, simulating an encounter with landscape rooted firmly to the ground. Such a view, while removing the advantages of appreciation of broad landscape form and the articulation of elements of a complex landscape into a whole that come from an elevated viewpoint succeeds in rooting the viewer in the embodied experience of the avatar (the computer generated persona of the player of games) and allows experience of the form, scale and potentially the meaning and articulation of landscape derived from survey in a way not achievable from above (cf. Figure 19.5, Laxton castle inner bailey defences, comparison of

19  Immersive visualisation of survey and laser scanning: the case for using computer game engines

247 Figure 19.6: An example of an unfolding visual narrative, in this case two views of eastern entrance of British Camp seen from the perspective of a slow uphill towards the summit. Viewed in CryEngine 2

elevated view and ground-based view, note the impression of scale of defences). We argue that the significance of such an experience is that it reintroduces to the interpretation of survey results the perspective of the fieldworker, allowing others to share that experience in a virtual environment and encouraging the engagement of the senses (cognitive understanding of scale, movement and surroundings) in the act of interpreting. One of the unique facets of the use of game software for visualisation of landscape survey is the potential to recreate the experience of movement through landscape from a first-person perspective. The meanings attached to visual journeys through significant landscapes has been a long-running theme of archaeological debate, fuelled by phenomenologists such as Tilley (1994), expounded as a fusion between traditional archaeology and a sort of theatre of landscape by Bender et al. (2007) and explored digitally,

usually in two-dimensions, by numerous GISbased explorations of visibility and view sheds (Exon et al. 2000; Chapman 2003; Llobera 1996). The first-person game allows recreation of the direct experience of moving through landscape, with both the sensory experience and limited viewing perception of the earth bound viewer. While fixed routes through the landscape can be simulated by programming pre-determined camera tracks, simulating for example the human modified views of key landscape features (Figure 19.7 shows two views from a journey towards the eastern entrance of British Camp) we have found a more effective and revealing method of understanding how viewers react to moving through landscape is to record the unconstrained movements of real users of the game software and examine subsequent patterns of behaviour (cf. Pinchbeck et al. 2006). The role of visual appreciation of landscape

248

Keith Challis and Mark Kincey

A

Figure 19.7a: Conventional GIS derived visibility analysis of the Stonehenge landscape showing in yellow the areas visible to a viewer at a location marked as a green triangle. The field of view shown in Figure 19.7b is indicated by the bold lines Figure 19.7b: Fog, rain and visibility (Stonehenge looking S from the Cursus at point A in Figure 19.7a, View distances of 2000 m, 1500 m and 500 m with and without rain. Bottom image is unlimited view distance, with Stonehenge highlighted). Viewed in CryEngine 1

has taken a leading role in the archaeology of landscape, encouraged both by theory (Bender et al. 2007) and the utility of GIS softwares for automatic analysis of the visual characteristics of large landscapes. Various writers have critiqued GIS studies of visibility and proposed enhancements to and refinements of visibility studies (e.g. Wheatley and Gillings 2000; Ogburn 2006). In essence arguments for the refinement of visibility studies focus on the need to appreciate the impact of range, view direction and target size on visibility. Additionally, a number of authors have pointed out the shortcomings of both theoretical and practical approaches to visibility studies, in particular the fact that many studies ignore or misrepresent the potential impact of vegetation and past vegetation patterns on landscape and visibility. Gearey and Chapman (2005) introduced the concept of ‘digital gardening’ as a method of reintroducing computer models of past vegetation to GIS analysis of visibility. In the present study we have used game software to explore the impact of different vegetation patterns and atmospheric characteristics on visibility. Sandbox allows varying vegetation scenarios for

the same landscape to be stored as layers that may be turned on and off to explore the visual impact of changes in vegetation. Similarly software driven environment controls allow alteration of the character, density and occlusion distance of atmospheric fog. In addition the impact of rain on visibility may be simulated through particle effects. For example, the effect of increasingly dense atmospheric fog and the addition of rain on visibility of Stonehenge are simulated in CryENGINE and compared with a traditional GIS-derived two-dimensional viewshed diagram from the same location (Figure 19.7). Games are, of course, ultimately about play, although precisely what constitutes play in our examples is open to debate. We suggest that ludic exploration of landscape is best facilitated by games that facilitate free and unconstrained movement though landscape, avoid linear patterns of attraction and reward and instead contain features and objects that encourage exploration. Games researchers such as Pinchbeck (2006, 2008) have documented the features and cues in computer games that encourage exploration, direct movement and act as narrative devices, and there is much here

19  Immersive visualisation of survey and laser scanning: the case for using computer game engines

to ponder in constructing ludicly attractive reconstructions of archaeologically significant landscapes. Much depends on the ultimate aims and presentation context of ludic landscape visualisations. Our present work, while largely experimental, has sought to engage both the research objective of determining the role and utility of games in landscape visualisation and the non-specialist general audience, through construction of mini-games for public display. In our present research we have explored the use of sound, atmospheric effects and biots (AI controlled wildlife) to constrain movement and direct attention. Our mini-game, Capture the Castle encapsulates the results of archaeological survey at Laxton Castle in a short computer game which facilitates freeform, multi-player exploration of the surveyed landscape (Figure 19.8). The game affords the player completely unconstrained exploration of the castle earthworks. Encouragement to explore is given in the form of a pre-game checklist of features that might be worth searching for (i.e. scale difference between the earthwork in different parts of the castle, the motte and its unusually shaped summit, vestiges of masonry projecting at various points from the earthworks). Within the game context we have used vegetation to limit movement (e.g. dense undergrowth marks the boundaries of the surveyed areas and so the

limits of the game world) and to selectively and realistically mask some features of the monument (just as in the real work, finding masonry remains requires effort in penetrating undergrowth, and some aspects of the morphology of the defences are masked by their overgrown state). Limited visual cues are provided by strategically placed biots, chiefly birds, whose flight serves to draw players attention towards key facets of the monument such as the motte (a similar device was used in the experimental game Dear Esther to guide player movement (Pinchbeck 2008)). Sound, an acknowledged and increasingly investigated aspect of landscape character and the human perception of landscape (Mills 2005) is used ambiently to create a sense of place (wind, rustling leaves, and occasional bird calls) and locally to provide player feedback. Spot sounds such as animal noises draw players to key landscape features, while startle noises such as a sudden flight of birds mark key discoveries both alarming the player and providing a sense of reaching something hitherto secret and unknown.

Conclusions Modern high precision digital survey techniques produce data in such volumes that the paradigm

249

Figure 19.8: A ludic landscape: scenes from the mini-game Capture the Castle, in which the user driven avatar explores the earthworks of Laxton Castle uncovering and collating evidence for the layout and structure of the monument. Developed in CryEngine 1

250

Keith Challis and Mark Kincey of traditional presentation of survey results, either as paper maps or digital analogues of these is challenged. We suggest that the use of first person computer game software allows realisation of new facets of such high density survey. In particular game software allows users to experience survey results in an immersive way as though their view were that of the landscape or monument surveyed. We argue that, through appropriate presentational devices and use of software tools within game engines, such a view provides new insights into the survey and its meaning, particularly by exploring phenomenological aspects of the sense of landscape and through view and visibility related experiments. These new approaches extend the remit of survey beyond that of simply producing a durable, scientific, cognitively interpreted record of landscape, to producing a record of sense of place. In particular we see applications for this approach to utilising survey within ludicalymediated environments (games) where survey and public meet, for example in museums and visitors centres, where digitally accessed playbased explorations of landscape are likely to be both more accessible and more relevant than many traditional representations of heritage. We view game-based visualisation as providing a new digital medium allowing us as archaeologists to rise to Tilley’s challenge: “To understand a landscape truly it must be felt, but to convey some of this feeling to others it has to be talked about, recounted, or written and depicted” (1994, 31).

References Barratt, G., Gaffney, V., Goodchild, H. and Wilkes, S., 2000. Survey at Wroxeter using Carrier Phase, Differential GPS surveying techniques. Archaeological Prospection, 7, 133–43. Bettess, F., 1998. Surveying for Archaeologists. Penshaw Press: University of Durham. Bender, B. Hamilton, S. and Tilley, C., 2007. Stone Worlds. Narrative and Reflexivity in Landscape Archaeology. Left Coast Press: Walnut Creek, California. Bowden, M., 1999. Unravelling the Landscape: An inquisitive approach to archaeology. Tempus: Stroud. Bryan, P., Blake, B. and Bedford, J., 2009. Metric Survey Specifications for Cultural Heritage. English Heritage: Swindon. Challis, K. Forlin, P. and Kincey, M., 2011. A generic toolkit for the archaeological analysis of airborne lidar. Archaeological Prospection 18(4), 279–89. Challis, K., Kokalj, Z., Kincey, M., Moscrop, D. and

Howard, A.J., 2008. Airborne Lidar and Historic Environment Records. Antiquity 82, 1055–64. Champion, E.M., 2008 Otherness of Place: Gamebased Interaction and Learning in Virtual Heritage Projects. International Journal of Heritage Studies 14:3, 210–28. Chapman, H., 2003. The recognition, interpretation and management of archaeological sites and landscapes using GPS survey and three-dimensional computer modelling. In Robson Brown, K. (ed.). Archaeological Sciences 1999. BAR International Series 1111. Archaeopress: Oxford. Chapman, H.P., 2005. Re-thinking the ‘cursus problem’ – investigating the Neolithic landscape of Rudston, East Yorkshire, UK, using GIS. Proceedings of the Prehistoric Society 71, 159–70. Chapman, H.P., 2006. Landscape Archaeology and GIS. Tempus: London. Chapman, H.P. and Gearey B.R., 2000. Palaeoecology and the perception of prehistoric landscapes: some comments on visual approaches to phenomenology. Antiquity 74, 316–9. Chapman, H. and Van de Noort, R. 2001. Highresolution wetland prospection using GPS and GIS: Landscape Studies at Sutton Common (South Yorkshire), and Meare Village East (Somerset). Journal of Archaeological Science 28, 365–75. Charno, M., 2007. Review of Barrow Hill: Curse of the Ancient Circle PC Game. Internet Archaeology 20. http://intarch.ac.uk/journal/issue20/reviews/ charno.html Ch’ng, E., 2009. Experiential Archaeology: Is Virtual Time Travel Possible? Journal of Cultural Heritage 20, 458–70. Ch’ng, E., Stone, R.J. and Arvanitis, T.N., 2004. The Shotton River and Mesolithic dwellings: recreating the past from geo-seismic data sources. In Cain, K., Chrysanthou, Y., Niccolucci, F. and Silberman, N., (eds). The 5th International Symposium on Virtual Reality, Archaeology and Cultural Heritage. VAST: The Eurographics Association. Crawford, O.G.S. and Keiller, A., 1928. Wessex from the Air. Oxford: Clarendon Press. Crutchley, S., 2006. Light detection and ranging (lidar) in the Witham Valley, Lincolnshire: an assessment of new remote sensing techniques. Archaeological Prospection 13(4), 251–7. Crytek, 2006. CryEngine Sandbox, Far Cry Edition User Manual. Coburg: Crytek. English Heritage, 2002. With Alidade and Tape: Graphical and plane table survey of archaeological earthworks. English Heritage: Swindon. English Heritage, 2003. Where on Earth are we? The Global Positioning System (GPS) in archaeological field survey. English Heritage: Swindon. English Heritage, 2007a. Understanding the Archaeology of Landscapes: A guide to good recording practice. English Heritage: Swindon. English Heritage, 2007b. 3D Laser Scanning for Heritage: Advice and guidance to user on laser scanning in archaeology and architecture. English Heritage: Swindon.

19  Immersive visualisation of survey and laser scanning: the case for using computer game engines English Heritage 2010. The Light Fantastic: Using airborne lidar in archaeological survey. English Heritage: Swindon. Exon, S., Gaffney, V., Woodward, A. and Yorston, R., 2000. Stonehenge landscapes. Journeys through realand-imagined worlds. Archaeopress: Oxford. Fletcher, M. and Spicer, D., 1988. Clonehenge: an experiment with gridded and non-gridded survey data. In Rahtz, S.P.Q. (ed.). Computer and Quantitative Methods in Archaeology, 1988. BAR International Series 446 (ii), 309–25. Gaunt, A., 2009. A Topographic Survey of the Deserted Village of West Burton. In Robinson, C. (ed.). Archaeology in Nottinghamshire 2009. Transactions of the Thoroton Society of Nottinghamshire 113, 151–4. Gearey, B.R. and Chapman, H.P., 2005. Digital Gardening: An approach to simulating elements of palaeovegetation and some implications for the interpretation of prehistoric sites and landscapes. In Evans, T.L. and Daly, P., (eds). Digital Archaeology: Bridging Method and Theory. Routledge: Abingdon. Gillings, M. and Goodrick, G.T., 1996. Sensuous and reflexive GIS: exploring visualisation and VRML. Internet Archaeology 1. Heidegger, M., 1972. On Time and Being. New York: Harper & Row. Heritage, G. and Hetherington, D., 2007. Towards a protocol for laser scanning in fluvial geomorphology. Earth Surface Processes and Landforms 32:1, 66–74. Hoskins, W.G. 1955. The Making of the English Landscape. London: Hodder & Stoughton. Howard, P. 2007. Archaeological Surveying and Mapping: Recording and depicting the landscape. Routledge: Abingdon. Johnson, M. 2007. Ideas of Landscape. Oxford: Blackwell. Kincey, M., Challis, K. and Speight, S., 2006. Laxton Castle. In Robinson, C. (ed.). Archaeology in Nottinghamshire 2005–6. Transactions of the Thoroton Society of Nottinghamshire 110, 124–8. Kvamme, K., Ernenwein, E.G. and Markussen, C.J., 2006. Robotic total station for microtopographic mapping: an example from the Northern Great Plains. Archaeological Prospection 13(2), 91–102. Lepouras, G. and Vassilakis, C., 2005. Virtual museums for all: employing game technology for edutainment.Virtual Reality 8, 96–106.

Llobera, M., 1996. Exploring the topography of mind: GIS, social space and archaeology. Antiquity 70, 612–22. Llobera, M., 2001. Building past landscape perception with GIS: understanding topographic prominence. Journal of Archaeological Science 28, 1005–14. Ogburn, D.E., 2006. Assessing the level of visibility of cultural objects in past landscapes. Journal of Archaeological Science 33, 405–13. Lock, G., 2003. Using Computers in Archaeology: Towards Virtual Pasts. Routledge: London. Mills, S., 2005. Sensing the place: sounds and landscape archaeology. In Bailey, D.W., Whittle, A. and Cummings, V. (eds). (un)settling the Neolithic, 79–89. Oxford: Oxbow. McCullagh, M.J., 1988. Terrain and surface modelling systems: theory and practice. Photogrammetric Record 12 (72), 747–99. Pinchbeck, D., 2008. Dear Esther: an interactive ghoststory built using the Source Engine. Presented at the 1st Joint International Conference on Interactive Digital Storytelling, Erfurt University of Applied Sciences (FH Erfurt), Erfurt Pinchbeck, D., Stevens, B., Van Laar, S., Hand, S. and Newman, K., 2006. Narrative, agency and observational behaviour in a first person shooter environment. In Kovacs, T. and Marshall, J.A.R., (eds). Proceedings of Narrative AI and Games, AISB Symposium. 53–61. Society for the Study of Artificial Intelligence and the Simulation of Behaviour: Bristol Quigley. P., 2007. Visualisation Of The Herefordshire Beacon Hillfort (British Camp). Unpublished MSc Project. Institute of Archaeology and Antiquity: University of Birmingham Riley, D.N., 1980. Early Landscape from the Air. Studies of Cropmarks in South Yorkshire and North Nottinghamshire. University of Sheffield. Stone, R. 2009. Serious game: virtual reality’s second coming? Virtual Reality 13, 1–2 Tilley, C. 1994. A Phenomenology of Landscape. Berg. Trentholme, D. and Smith, S., 2008. Computer game engines for developing first-person virtual environments. Virtual Reality 12, 181–7. Wheatley, D.W. and Gillings, M., 2000. Vision, perception and GIS: developing enriched approaches to the study of archaeological visibility. In Lock, G., (ed.). Beyond the Map: Archaeological and Spatial Technologies. Amsterdam, 1–27.

251

20 The Practice of Collaboration Anthony Beck

We live in ‘interesting times’: the world wide web is undermining traditional work practices while opening up entirely new approaches to research and its communication. Ubiquitous communication and improved access to data and knowledge repositories has an obvious impact on scientific progress. However, the ‘socialization’ of the web means that individuals can become members of a number of loosely coupled formal or informal groups. Members of these groups share the tools of knowledge acquisition: data, algorithms, information, synthesis and, potentially more importantly, experience. Inevitably the web has been the catalyst for the majority of these changes. For some the ability to interact dynamically with different individuals and groups in both a formal and informal capacity has led to the development of new approaches for the acquisition and sharing of scientific data, information and knowledge. This chapter will consider these issues in light of recent social, organisational and technological developments that impact upon collaboration and what can be transferred to the analysis of Airborne Laser Scanning for heritage applications. Keywords: ALS, Lidar, communication, collaboration, Open Data, heritage, archaeology

Introduction Confucious said “Real knowledge is to know the extent of one’s ignorance”. This quote highlights the nature of much research knowledge acquisition: it is about one person. The lonely researcher can gain experience by doing battle with the unknown, and sometimes with themselves, as they travel along the road of knowledge that leads to wisdom. It is important that they achieve some form of Rumsfeldian enlightenment – know what they know, know there are things they don’t know, and recognise there are things that they are completely unaware of which may be extremely important. However, this emphasis on a personal journey may not reflect the reality of research in the forthcoming decades.1 The landscape in which we 2 operate is undergoing fundamental change. The Climate Gate scandal at the University of East Anglia in 2009 (Hulme and Ravetz 2009) means

that researchers and the public are demanding access to both publicly funded research data and the algorithms used to transform that data into information and knowledge. Funding councils, such as the UK EPSRC (Engineering and Physical Sciences Research Council), are specifying that publications derived directly from funded research must be made available as openly accessible documents (EPSRC 2011a). EPSRC’s stated research expectations (EPSRC 2011b) make it increasingly likely that EPSRC will mandate access to data, algorithms and other outputs from the research process in the future something also advocated by the Royal Society (2012). Likewise a number of governments are making publicly funded data openly available. Social media companies, like Twitter (2011), are helping to change the way scientists and policy makers engage with their audiences: consumers are participating in a conversation, not listening to a lecture. The

253

20  The Practice of Collaboration process of knowledge acquisition and research is becoming ever more open and collaborative. These more sophisticated forms of engagement can increase impact dramatically. Impact ensures publicly funded research finds it way to the widest possible sets of uses. These uses, while including traditional commercial exploitation, also encompass the much wider impacts of engagement, education, and even inspiration of the general public. This will significantly change the way we all ‘do business’. However, much publicly funded knowledge is hidden behind paywalls or bureaucratic access requirements (Bevan 2011). Archive size is dramatically increasing but the resources to manage and articulate these resources so they can be effectively re-used are reduced. Both these issues mean that it is difficult to bring all the available knowledge together to help solve a problem (see Figure 20.1). The traditional archaeological publication process produces syntheses which mean that the underlying raw data and transforming methodologies are rarely available. To compound this issue, these syntheses are frozen in time and do not change to reflect new data and knowledge. In this context the process of conducting science cannot be peer reviewed. Global networked communication systems provide the opportunity to share data, algorithms, information and knowledge in ways that were impossible only a generation ago. This raises the question as to whether the traditional system of knowledge acquisition and dissemination are still fit for purpose. This may have a profound impact on those wishing to undertake knowledge led policy and management. This chapter will examine these issues, paying particular attention to collaborative approaches, within the context of heritage applications of Airborne Laser Scanning (ALS).

The heritage environment of the future To provide context for discussion it is germane to consider how technological innovations and social/organisational change might impact on the heritage environment. The way we use software is likely to change dramatically in the next few years. Stable and mature cloud infrastructure will enable persistent data storage and the use of Software as a Service (SaaS: Wikipedia 2012a) to process this data. In some instances users may generate a hybrid

Figure 20.1: How the corpus of knowledge can be applied to problems (Beck 2012a)

model utilising cloud based SaaS and local software to represent the full processing chain. Data can be processed using a variety of open source SaaS, pay per use proprietary SaaS and local packages with the process chain designed and orchestrated in a digital workflow engine. The digital workflow environments will allow the development and sharing of complex processing chains and scripts that invoke software processes locally and remotely (on the web) by calling local Application Programming Interfaces (APIs) and web APIs. This will be an interesting development which will disrupt current software provision models and data processing techniques. Many important innovations will continue to come through collaborations, however, the nature of the collaborations will be different (Townsend et al. 2009). This will be part of a continuing expansion of ‘heritage in society’ and the redefining of relationships between different heritage stakeholders. Many of the collaborators will be amateurs. Many field-activities have built and sustained relationships with volunteers, however, amateurs represent a vastly underutilized resource for many curatorial and research activities. Much like the citizen scientists at Galaxy Zoo (2012), Ancient Lives (2012) or Old Weather (2012) a legion of amateurs can help to reduce the size and enhance the utility of the “littleused and inaccessible data mountain” (English Heritage 1995, 42). Some collaborators may have limited or no interest in the larger heritage objective but specific expertise in a troublesome section of the workflow solution (i.e. they could quickly solve a processing or algorithm problem). Harnessing latent microexpertise from experts distributed around the world in a timely and minimally intrusive manner will transform multi-disciplinary approaches and see a rapid reduction in development and innovation time. This is akin to having all the

254

Anthony Beck right people in the room just when you need to call on their talents. Archaeology will increasingly need to embrace its multidisciplinary nature. For example, is a background in archaeology or heritage appropriate for developing Airborne Laser Scanner (ALS) algorithms? Not many postgraduate students in the heritage sector are skilled in pattern matching and spatial reasoning. Improving ALS detection techniques is more about understanding threedimensional spatial structures and light (signal) propagation issues and less about archaeology. However, heritage knowledge is required to develop the shared praxis under which the collaboration will occur: archaeologists are required to define the phenomena and help conceptualise how this is structured in the data. This is exemplified in the Foldit citizen science experiment (Hand 2010; FoldIt 2012). Foldit aims to improve protein folding through an online game engine. Structural biologists, biochemists and computer scientists developed the rules and interface for the gaming engine which is open for public use. Participants fold the structure of selected proteins, these structures are ‘scored’, the highest scoring configurations are then analysed by researchers. Foldit has been available since 2008 and in that time the top players have demonstrated that they can develop more effective protein configurations than brute-force computer algorithms (Cooper et al. 2010). The use of visualization environments, like Foldit, have great potential for improving both interpretative throughput and public engagement. Data and other resources will continue to be opened up, if only in the name of democratic transparency. This is exemplified in the UK Government led Open Data initiatives, the Semantic Web (as championed by Sir Tim Berners-Lee and Professor Nigel Shadbolt) and the EU INSPIRE directive. These projects focus on making data interoperable, available and accessible to a broad community. In turn these communities can develop tools that aid visualization and interpretation of these large and distributed resources. This will continue in the heritage sector. Issues of access, accreditation, copyright and licensing will be raised and resolved as data and other resources are made available over the semantic web. Linked Data will transform static archives into dynamic resources that fully articulate the impact of change as the supporting knowledge base is refined. For example, the revision of date ranges associated

with artefacts has an impact on phasing and interpretation. By explicitly linking feature dates to supporting evidence, as linked data the impact of change can be evaluated and incorporated into future interpretations. These developments will have a profound impact on the organizational structures within heritage and will challenge management systems, particularly those based on centralisation and control. Cheap, accessible communication tools will be developed that enable formal and informal groups to coalesce around any shared interest, identity or activity. Existing organizations will be transformed through the adoption of these collaborative tools and processes. Cloud computing will mature and provide (open?) access to complex data, processing and visualization frameworks. Interoperable access to data, services and syntheses will provide the building blocks for a range of knowledge ecosystems. Shared workflow systems, such as Taverna, will process data via online services and automatically generate and maintain the processing metadata. The workflow and metadata will provide unambiguous detail on how data is transformed into information. For the resulting data-centric applications data quality and provenance will become increasingly important. Over time a deeper understanding of the underlying data quality will occur on a data-set by data-set basis. This information will be added into the resource discovery metadata. Where appropriate, incorrect data can be corrected at source, resulting in a more accurate knowledgebase. The knowledge concerning data quality will facilitate the production of credence (or authoritative) systems and tools to filter out data based upon quality, robustness and fitness-forpurpose. Such knowledge systems will allow the dynamic aggregation and generalisation of data: the re-faceting of well understood data which is appropriate for a specific problem will provide nuanced policy, planning and research. These changes will occur incrementally and there will be many risks and problems to overcome (not the least the issue of ethics). In summary the future heritage landscape may consist of loosely coupled multi-disciplinary communities that collaborate with a diverse audience in a global setting: professional scientists, their peers, the interested lay-person, the policy maker and the government official. Data is more accessible and the ability to turn this data into knowledge for a variety of different communities

20  The Practice of Collaboration will be transformative and lead to greater, and sometimes unanticipated, impacts. These changes may be highly disruptive and have a profound impact not only on the way we engage with, research into and manage our shared heritage, but on the organisations that mediate engagement.

The nature of collaboration As has been described above collaboration will continue to be central. Wikipedia (2012b) defines collaboration as: “... working together to achieve a goal. It is a recursive process where two or more people or organizations work together to realize shared goals, (this is more than the intersection of common goals seen in co-operative ventures, but a deep, collective, determination to reach an identical objective) – for example, an intriguing endeavor that is creative in nature – by sharing knowledge, learning and building consensus” In this definition, collaboration is an active process. This is the traditional view of collaboration where stakeholders from a single, or closely affiliated, domain ‘work together’ to ‘realize shared goals’. Collaboration can also be a passive process as collaborators who ‘share knowledge’ (which includes data, algorithms, methods and expertise) may not be interested in the ‘vision’ or ‘goal’; they are not primary stakeholders driving the collaborating process but facilitating stakeholders who provide access to critical resource. These facilitating stakeholders may want to achieve maximal impact for their resources – the greatest good to the greatest number. This introduces the concept of serendipitous or passive collaboration. A passive collaborator provides access to resources and data that facilitate or enable the ‘realization of a goal’. If the passive collaborator provides ‘open’ resources then they may only realise they have participated in a collaborative process when their resources are referenced in a publication or other output. Contributors to the Archaeology Data Service (ADS) follow this model. In reality there may be a whole range of different collaborator types. Crowdsourcing (Wikipedia 2012c) and Citizen Science (Wikipedia 2012d) activities have changed the way individuals and organizations can interact providing more subtle and flexible ways for them to collaborate and enabling new players to take an important and active role in research. Reinventing Discovery (Nielsen 2011) describes a range of different

255 successful and unsuccessful collaboration scenarios and gives a clear indication of the future potential of global collaboration frameworks. Given the skills, expertise, and experience that different people, from the interested amateur to the domain expert can bring, how do we ensure that the right tasks are distributed to the right person? Some tasks require practice but not experience, others may require a very specific form of expertise or insight. Some forms of expertise are more rare, and therefore potentially more valuable than others. How do we ensure that the right architecture is built so that all the interested parties can bring their different skills most effectively to the table? How do we “restructure expert attention” in the words of Nielsen to ensure that our most valuable resource, the experience and time of experts, is used effectively? These are some of the challenges of bringing science into society. Collaboration is a social activity. Participants include individuals, professionals and organ­ isations. Each has a different reason to provide data, algorithms, expertise or time. Organisations are mainly driven by policy objectives. These policy driven pull factors also require push factors to encourage engagement. The nature of push factors varies between different stakeholders. Community archaeologists, curatorial archaeo­ logists and academic archaeologists all have different incentives. For example, in academia career advancement is still focussed on the production of synthesis through journal publications. There is little personal merit in engaging in collaborative frameworks where incremental advances are shared between the community and an individual’s contribution is not recognised. This can produce a tension between formal (academic) and informal (public) approaches to advancing scientific progress. It could be argued that the current research framework is not fit for purpose in that the processes and metrics employed to evaluate scientific progress and in turn reward scientists do not facilitate the production of the best, or most efficient, science. It definately does not produce the most ‘open’ science. This is at odds with, for example, the development of open source software where improvements can be implemented comparatively rapidly. It is likely that the incentive structure will change in the future. For example, as the UK funding councils are mandating more open access to data it is likely that good quality data that is regularly re-used (and appropriately cited) will count towards career

256

Anthony Beck enhancement (Royal Society 2012). In addition new journals and collaboration frameworks, such as the Open Research Computation journal (2012) provide alternative environments to document and deposit data, algorithms and methods.

The nature of ALS data and its processing Airborne Laser Scanning (ALS or Lidar) is an active remote sensing technique providing range data that can be mapped as 3D point clouds (see Opitz this volume). The scanner emits thousands of infra-red pulses per second and records the one or more echoes returned to the sensor. Some sensors are able to discriminate up to six individual echoes from a single pulse (Thiel and Wehr 2004). However, approximately 90% of the reflected signal power is contained within the two first echoes. Detection of more than two echoes becomes problematic as the signal to noise ratio decreases. ALS can provide reliable elevation data with high altimetric (