Advances in Weather Radar: Emerging Applications [3] 1839536225, 9781839536229

Table of contents :
Cover
Contents
About the editors
Preface
Acknowledgments
List of editors
List of contributors
List of reviewers
Introduction to volume 3
1 Radar hydrology
1.1 Introduction
1.2 Radar hydrology and the climatology of extreme rainfall
1.2.1 Radar-rainfall data sets
1.2.2 Stochastic storm transposition
1.2.3 Polarimetric measurements and extreme rainfall: an example
1.2.4 Radar-rainfall biases
1.3 Radar hydrology and flood forecasting
1.3.1 Radar-rainfall hydrologic visibility
1.3.2 Random vs. systematic errors
1.4 Summary and outlook
Acknowledgments
References
2 Quantitative precipitation estimation from weather radars, personal weather stations and commercial microwave links
2.1 Introduction
2.2 Datasets and evaluation metrics
2.3 QPE with ground-based weather radars
2.3.1 Signal processing
2.3.2 Clutter removal
2.3.3 Attenuation correction
2.3.4 Vertical profile of reflectivity correction
2.3.5 Conversion to rainfall intensity
2.3.6 Radar-gauge adjustment
2.3.7 Evaluation
2.4 QPE with personal weather stations
2.4.1 Quality control
2.4.2 Evaluation
2.5 QPE with commercial microwave links
2.5.1 Sources of error and processing chain
2.5.2 Evaluation
2.6 Summary and outlook
2.6.1 Merging opportunistic sensing and radar data
2.6.2 CMLs as ‘weather radars’ for the tropics
Acknowledgements
References
3 Quantitative weather radar: a research to operations perspective in Canada
3.1 Introduction
3.2 Weather radars at ECCC
3.2.1 Early years
3.2.2 Radar renewal I
3.2.3 Dual-polarisation (DP) technology
3.2.4 Radar renewal II
3.3 Quality control (QC)/quality assurance (QA) research
3.3.1 Aspects of quality
3.3.2 Monitoring tools
3.3.3 Antenna pointing
3.3.4 Exploiting dual-PRF’s alternating sampling
3.3.5 Assessing radar siting
3.4 Quantitative precipitation estimation
3.4.1 C-band and S-band radars
3.4.2 Satellite-based QPE
3.5 Particle classification
3.5.1 C-band radar era
3.5.2 S-band radar era
3.6 Linking research to production and services
3.6.1 Radar data assimilation
3.6.2 Global perspective
Acknowledgement
References
4 Volcanic plume retrieval using weather radar
4.1 Introduction
4.2 Theoretical framework
4.2.1 Volcanic solid emissions
4.2.2 Tephra size and composition
4.2.3 Tephra particle size distribution and mass concentration
4.3 Tephra mass and size retrieval
4.4 Tephra velocity field retrieval
4.4.1 Methodology
4.4.2 Velocity regimes identification
4.4.3 Case studies and examples
4.5 Mass eruption rate retrieval
4.5.1 Methodology
4.5.2 Sensor synergy
4.5.3 Case studies
4.6 Conclusions
References
5 The effect of weather on the performance of mm-wave and sub-THz automotive radar
5.1 Introduction
5.1.1 Basic principle of automotive radar
5.1.2 Characterization of sub-THz and mm-wave channels
5.1.3 Methodology
5.2 Atmospheric attenuation at mm-wave and sub-THz frequency bands
5.2.1 Attenuation through rain
5.2.2 Attenuation through snowfall
5.2.3 Analysis of radar performance operating in the presence of fire and smoke
5.3 Attenuation through vehicle infrastructure and natural contaminants
5.3.1 Theoretical modeling of reflectivity and transmissivity through uniform multilayered structure
5.3.2 Measurement methodology of transmissivity through vehicle infrastructure and contaminant
5.3.3 Effect of uniform layers of natural contaminants on transmissivity
5.3.4 Modeling of transmissivity through discrete nonuniform contaminants
5.4 Summary of the results
5.5 Conclusion and discussion
References
6 Spectral interference in weather radars from wireless communication systems
6.1 Introduction
6.2 Problem description
6.3 Impacts on operational services
6.4 Why is interference a prevalent problem for weather radars?
6.5 Mitigation approaches
6.5.1 Regulatory
6.5.2 Technical
Acknowledgements
References
7 Advances in weather radar monitoring of bird movement
7.1 Method development and field validation
7.1.1 The relationship between reflectivity and bird density
7.1.2 Vertical profiling
7.1.3 Validation by collocated field campaigns
7.1.4 Airspeed and heading
7.1.5 Stopover mapping and habitat association
7.1.6 Machine learning advances in identification and classification
7.1.7 Integrative models for radar network data
7.1.8 Data quality and radar processor settings
7.1.9 Forecasting and predictive models
7.2 Monitoring and ornithological applications of weather radar data
7.2.1 Flyway mapping
7.2.2 Migration stopovers
7.2.3 Phenology of migration
7.2.4 Biomass flows and abundance
7.2.5 Bird flight behaviour
7.2.6 Artificial lights and migration
7.3 Importance of ornithological radar data beyond traditional ornithology
7.3.1 Existing applications
7.3.2 Developing applications
7.4 Necessary conceptual, technological, and methodological developments
7.4.1 Standardized data streams and appropriate data infrastructure
7.4.2 Improved classification and identification algorithms
7.4.3 Integration of radar data with auxiliary data
7.4.4 Future outlook
References
8 Complementary operation of Doppler radar and lidar at airports
8.1 Introduction
8.1.1 The terminal Doppler weather radar
8.1.2 Combined Doppler radar/lidar operation
8.2 Regulations and further considerations for integrated wind shear detection with radar and lidar
8.2.1 Requirements on ground-based lidar wind shear monitoring
8.2.2 Wind shear detection with radar and lidar
8.3 Fundamentals of Doppler radar operation
8.3.1 The meteorological radar equation
8.3.2 Velocity measurement
8.3.3 Solid-state radars
8.4 Fundamentals of Doppler lidar operation
8.4.1 Estimation of radial velocity (and turbulence)
8.4.2 The semi-analytic pulsed coherent lidar equation
8.5 Meeting wind shear detection requirements in different conditions
8.5.1 Radar and lidar detection skills: a view from aloft
8.5.2 Detection skills and atmospheric phenomena in Np, rp-space
8.5.3 Visibility, a common ground for radar and lidar?
8.6 Comparison of different classes of X-Band weather radars and their suitability for different airport applications
8.7 Classes of lidars for airport operations
8.7.1 Classes of lidars for airport wind field monitoring and wind shear detection
8.8 Radar and lidar complementary wind products and algorithms
8.8.1 Examples of complementary products for operational combined systems of radar and lidar
8.8.2 Product algorithm examples
8.8.3 Runway-oriented wind shear
8.8.4 Echo classification
8.8.5 Lidar range extension ML algorithm
References
Epilogue
Back Cover

Citation preview

IET ENERGY ENGINEERING 95

Advances in Weather Radar

Related titles on radar: Advances in Bistatic Radar Willis and Griffiths Airborne Early Warning System Concepts, 3rd Edition Long Bistatic Radar, 2nd Edition Willis Design of Multi-Frequency CW Radars Jankiraman Digital Techniques for Wideband Receivers, 2nd Edition Tsui Electronic Warfare Pocket Guide Adamy Foliage Penetration Radar: Detection and characterisation of objects under trees Davis Fundamentals of Ground Radar for ATC Engineers and Technicians Bouwman Fundamentals of Systems Engineering and Defense Systems Applications Jeffrey Introduction to Electronic Warfare Modeling and Simulation Adamy Introduction to Electronic Defense Systems Neri Introduction to Sensors for Ranging and Imaging Brooker Microwave Passive Direction Finding Lipsky Microwave Receivers with Electronic Warfare Applications Tsui Phased-Array Radar Design: Application of radar fundamentals Jeffrey Pocket Radar Guide: Key facts, equations, and data Curry Principles of Modern Radar, Volume 1: Basic principles Richards, Scheer and Holm Principles of Modern Radar, Volume 2: Advanced techniques Melvin and Scheer Principles of Modern Radar, Volume 3: Applications Scheer and Melvin Principles of Waveform Diversity and Design Wicks et al. Pulse Doppler Radar Alabaster Radar Cross Section Measurements Knott Radar Cross Section, 2nd Edition Knott et al. Radar Design Principles: Signal processing and the environment, 2nd Edition Nathanson et al. Radar Detection DiFranco and Rubin Radar Essentials: A concise handbook for radar design and performance Curry Radar Foundations for Imaging and Advanced Concepts Sullivan Radar Principles for the Non-Specialist, 3rd Edition Toomay and Hannan Test and Evaluation of Aircraft Avionics and Weapons Systems McShea Understanding Radar Systems Kingsley and Quegan Understanding Synthetic Aperture Radar Images Oliver and Quegan Radar and Electronic Warfare Principles for the Non-Specialist, 4th Edition Hannen Inverse Synthetic Aperture Radar Imaging: Principles, algorithms and applications Chen and Martorella Stimson’s Introduction to Airborne Radar, 3rd Edition Baker, Griffiths and Adamy Test and Evaluation of Avionics and Weapon Sytems, 2nd Edition McShea Angle-of-Arrival Estimation Using Radar Interferometry: Methods and applications Holder Biologically-Inspired Radar and Sonar: Lessons from nature Balleri, Griffiths and Baker The Impact of Cognition on Radar Technology Farina, De Maio and Haykin Novel Radar Techniques and Applications, Volume 1: Real aperture array radar, imaging radar, and passive and multistatic radar Klemm, Nickel, Gierull, Lombardo, Griffiths and Koch Novel Radar Techniques and Applications, Volume 2: Waveform diversity and cognitive radar, and target tracking and data fusion Klemm, Nickel, Gierull, Lombardo, Griffiths and Koch Radar and Communication Spectrum Sharing Blunt and Perrins Systems Engineering for Ethical Autonomous Systems Gillespie Shadowing Function from Randomly Rough Surfaces: Derivation and applications Bourlier and Li Photo for Radar Networks and Electronic Warfare Systems Bogoni, Laghezza and Ghelfi Multidimensional Radar Imaging Martorella Radar Waveform Design Based on Optimization Theory Cui, De Maio, Farina and Li Micro-Doppler Radar and Its Applications Fioranelli, Griffiths, Ritchie and Balleri Maritime Surveillance with Synthetic Aperture Radar Di Martino and Antonio Iodice Electronic Scanned Array Design Williams Advanced Sparsity-Driven Models and Methods for Radar Applications Li Deep Neural Network Design for Radar Applications Gurbuz New Methodologies for Understanding Radar Data Mishra and Bru¨ggenwirth Radar Countermeasures for Unmanned Aerial Vehicles Clemente, Fioranelli, Colone and Li Holographic Staring Radar Oswald and Baker Polarimetric Radar Signal Processing Aubry, De Maio and Farina Fundamentals of Inertial Navigation Systems and Aiding Braasch Theory and Methods for Distributed Data Fusion Applications Govaers Target Detection: A detection theory framework Zaimbashi and Nayebi

Advances in Weather Radar Volume 3: Emerging applications Edited by V.N. Bringi, Kumar Vijay Mishra and Merhala Thurai

The Institution of Engineering and Technology

Published by SciTech Publishing, an imprint of The Institution of Engineering and Technology, London, United Kingdom The Institution of Engineering and Technology is registered as a Charity in England & Wales (no. 211014) and Scotland (no. SC038698). † The Institution of Engineering and Technology 2024 First published 2023 This publication is copyright under the Berne Convention and the Universal Copyright Convention. All rights reserved. Apart from any fair dealing for the purposes of research or private study, or criticism or review, as permitted under the Copyright, Designs and Patents Act 1988, this publication may be reproduced, stored or transmitted, in any form or by any means, only with the prior permission in writing of the publishers, or in the case of reprographic reproduction in accordance with the terms of licences issued by the Copyright Licensing Agency. Enquiries concerning reproduction outside those terms should be sent to the publisher at the undermentioned address: The Institution of Engineering and Technology Futures Place Kings Way, Stevenage Hertfordshire, SG1 2UA, United Kingdom www.theiet.org While the authors and publisher believe that the information and guidance given in this work are correct, all parties must rely upon their own skill and judgement when making use of them. Neither the authors nor publisher assumes any liability to anyone for any loss or damage caused by any error or omission in the work, whether such an error or omission is the result of negligence or any other cause. Any and all such liability is disclaimed. The moral rights of the authors to be identified as authors of this work have been asserted by them in accordance with the Copyright, Designs and Patents Act 1988.

British Library Cataloguing in Publication Data A catalogue record for this product is available from the British Library ISBN Vol 1: 978-1-83953-622-9 (hardback) ISBN Vol 2: 978-1-83953-624-3 (hardback) ISBN Vol 3: 978-1-83953-626-7 (hardback) ISBN 3 Vol Set: 978-1-83953-628-1 (hardback) ISBN Vol 1: 978-1-83953-623-6 (PDF) ISBN Vol 2: 978-1-83953-625-0 (PDF) ISBN Vol 3: 978-1-83953-627-4 (PDF) Typeset in India by MPS Limited Printed in the UK by CPI Group (UK) Ltd, Eastbourne Cover Image: Silhouette of the University of Iowa X-band polarimetric radar deployed at the Iowa City, Iowa, landfill with fire in the background in 2012. The radar was never in close danger of the burning stock of used tires. Photo taken by Aneta Goska.

This work is humbly placed at the Lotus Feet of Goddess Saraswati, Patron of Knowledge and the Arts. – V.N.B. Dedicated, with the highest reverence and humility, to Mahadeva – the Supreme Being. To my Mom Shraddha Mishra. – K.V.M. Dedicated to my beloved parents. – M.T.

This page intentionally left blank

Contents

About the editors Preface Acknowledgments List of editors List of contributors List of reviewers Introduction to volume 3

xiii xv xix xxi xxiii xxv xxvii

1 Radar hydrology 1 Witold F. Krajewski and James A. Smith 1.1 Introduction 1 1.2 Radar hydrology and the climatology of extreme rainfall 3 1.2.1 Radar-rainfall data sets 4 1.2.2 Stochastic storm transposition 4 1.2.3 Polarimetric measurements and extreme rainfall: an example 8 1.2.4 Radar-rainfall biases 9 1.3 Radar hydrology and flood forecasting 11 1.3.1 Radar-rainfall hydrologic visibility 12 1.3.2 Random vs. systematic errors 13 1.4 Summary and outlook 16 Acknowledgments 16 References 17 2 Quantitative precipitation estimation from weather radars, personal weather stations and commercial microwave links Aart Overeem, Remko Uijlenhoet and Hidde Leijnse 2.1 Introduction 2.2 Datasets and evaluation metrics 2.3 QPE with ground-based weather radars 2.3.1 Signal processing 2.3.2 Clutter removal 2.3.3 Attenuation correction 2.3.4 Vertical profile of reflectivity correction 2.3.5 Conversion to rainfall intensity 2.3.6 Radar-gauge adjustment 2.3.7 Evaluation

27 27 36 38 38 39 40 41 41 42 43

viii

Advances in weather radar, volume 3 2.4

QPE with personal weather stations 2.4.1 Quality control 2.4.2 Evaluation 2.5 QPE with commercial microwave links 2.5.1 Sources of error and processing chain 2.5.2 Evaluation 2.6 Summary and outlook 2.6.1 Merging opportunistic sensing and radar data 2.6.2 CMLs as ‘weather radars’ for the tropics Acknowledgements References 3

4

Quantitative weather radar: a research to operations perspective in Canada David Hudak, Daniel Michelson, Sudesh Boodoo, Norman Donaldson, Paul Joe, Andre´s Pe´rez Hortal, Janti Reid, Peter Rodriguez and Brian Sheppard 3.1 Introduction 3.2 Weather radars at ECCC 3.2.1 Early years 3.2.2 Radar renewal I 3.2.3 Dual-polarisation (DP) technology 3.2.4 Radar renewal II 3.3 Quality control (QC)/quality assurance (QA) research 3.3.1 Aspects of quality 3.3.2 Monitoring tools 3.3.3 Antenna pointing 3.3.4 Exploiting dual-PRF’s alternating sampling 3.3.5 Assessing radar siting 3.4 Quantitative precipitation estimation 3.4.1 C-band and S-band radars 3.4.2 Satellite-based QPE 3.5 Particle classification 3.5.1 C-band radar era 3.5.2 S-band radar era 3.6 Linking research to production and services 3.6.1 Radar data assimilation 3.6.2 Global perspective Acknowledgement References Volcanic plume retrieval using weather radar Luigi Mereu, Frank S. Marzano, Simona Scollo, Mario Montopoli and Gianfranco Vulpiani 4.1 Introduction 4.2 Theoretical framework

45 45 46 48 48 51 51 52 56 57 59

69

69 70 70 72 74 76 78 78 81 88 88 90 92 92 98 101 101 104 110 111 113 114 116 129

130 133

Contents 4.2.1 Volcanic solid emissions 4.2.2 Tephra size and composition 4.2.3 Tephra particle size distribution and mass concentration 4.3 Tephra mass and size retrieval 4.4 Tephra velocity field retrieval 4.4.1 Methodology 4.4.2 Velocity regimes identification 4.4.3 Case studies and examples 4.5 Mass eruption rate retrieval 4.5.1 Methodology 4.5.2 Sensor synergy 4.5.3 Case studies 4.6 Conclusions References 5 The effect of weather on the performance of mm-wave and sub-THz automotive radar F. Norouzian, Y. Xiao and M. Gashinova 5.1 Introduction 5.1.1 Basic principle of automotive radar 5.1.2 Characterization of sub-THz and mm-wave channels 5.1.3 Methodology 5.2 Atmospheric attenuation at mm-wave and sub-THz frequency bands 5.2.1 Attenuation through rain 5.2.2 Attenuation through snowfall 5.2.3 Analysis of radar performance operating in the presence of fire and smoke 5.3 Attenuation through vehicle infrastructure and natural contaminants 5.3.1 Theoretical modeling of reflectivity and transmissivity through uniform multilayered structure 5.3.2 Measurement methodology of transmissivity through vehicle infrastructure and contaminant 5.3.3 Effect of uniform layers of natural contaminants on transmissivity 5.3.4 Modeling of transmissivity through discrete nonuniform contaminants 5.4 Summary of the results 5.5 Conclusion and discussion References 6 Spectral interference in weather radars from wireless communication systems Elena Saltikoff, John Y. N. Cho and Philippe Tristant 6.1 Introduction 6.2 Problem description

ix 133 134 134 136 137 138 139 141 144 144 147 148 154 156

161 161 163 165 166 167 168 182 188 192 194 199 201 202 210 211 213

221 221 221

x

7

8

Advances in weather radar, volume 3 6.3 6.4 6.5

Impacts on operational services Why is interference a prevalent problem for weather radars? Mitigation approaches 6.5.1 Regulatory 6.5.2 Technical Acknowledgements References

224 225 228 228 230 235 236

Advances in weather radar monitoring of bird movement Cecilia Nilsson, Adriaan Dokter, Silke Bauer, Kyle Horton and Andrew Farnsworth 7.1 Method development and field validation 7.1.1 The relationship between reflectivity and bird density 7.1.2 Vertical profiling 7.1.3 Validation by collocated field campaigns 7.1.4 Airspeed and heading 7.1.5 Stopover mapping and habitat association 7.1.6 Machine learning advances in identification and classification 7.1.7 Integrative models for radar network data 7.1.8 Data quality and radar processor settings 7.1.9 Forecasting and predictive models 7.2 Monitoring and ornithological applications of weather radar data 7.2.1 Flyway mapping 7.2.2 Migration stopovers 7.2.3 Phenology of migration 7.2.4 Biomass flows and abundance 7.2.5 Bird flight behaviour 7.2.6 Artificial lights and migration 7.3 Importance of ornithological radar data beyond traditional ornithology 7.3.1 Existing applications 7.3.2 Developing applications 7.4 Necessary conceptual, technological, and methodological developments 7.4.1 Standardized data streams and appropriate data infrastructure 7.4.2 Improved classification and identification algorithms 7.4.3 Integration of radar data with auxiliary data 7.4.4 Future outlook References

241

Complementary operation of Doppler radar and lidar at airports Frank Gekat, Ronald Hannesen, Sebastian Kauczok, Marcus Pool, Christian Schiefer and Albert To¨ws 8.1 Introduction

243 243 244 247 248 248 250 251 251 252 253 253 255 257 258 259 260 261 262 265 268 268 268 269 269 270 281

281

Contents

xi

8.1.1 The terminal Doppler weather radar 8.1.2 Combined Doppler radar/lidar operation 8.2 Regulations and further considerations for integrated wind shear detection with radar and lidar 8.2.1 Requirements on ground-based lidar wind shear monitoring 8.2.2 Wind shear detection with radar and lidar 8.3 Fundamentals of Doppler radar operation 8.3.1 The meteorological radar equation 8.3.2 Velocity measurement 8.3.3 Solid-state radars 8.4 Fundamentals of Doppler lidar operation 8.4.1 Estimation of radial velocity (and turbulence) 8.4.2 The semi-analytic pulsed coherent lidar equation 8.5 Meeting wind shear detection requirements in different conditions 8.5.1 Radar and lidar detection skills: a view from aloft 8.5.2 Detection skills and atmospheric phenomena in Np, rp-space 8.5.3 Visibility, a common ground for radar and lidar? 8.6 Comparison of different classes of X-Band weather radars and their suitability for different airport applications 8.7 Classes of lidars for airport operations 8.7.1 Classes of lidars for airport wind field monitoring and wind shear detection 8.8 Radar and lidar complementary wind products and algorithms 8.8.1 Examples of complementary products for operational combined systems of radar and lidar 8.8.2 Product algorithm examples 8.8.3 Runway-oriented wind shear 8.8.4 Echo classification 8.8.5 Lidar range extension ML algorithm References

282 282

Epilogue

284 285 286 288 288 291 292 294 295 297 302 302 306 314 317 318 319 323 323 330 330 333 338 342

347

This page intentionally left blank

About the editors

V. N. Bringi is an emeritus professor at Colorado State University, USA. He received his PhD from The Ohio State University, then joined the faculty at Colorado University where he spent his entire career. He is a fellow of the American Meteorological Society (AMS) and recipient of the AMS Remote Sensing Prize in 2013. Kumar Vijay Mishra is a senior fellow at the United States DEVCOM Army Research Laboratory. He received his PhD in electrical engineering and MS in mathematics from The University of Iowa, Iowa City. He is the recipient of many best paper awards and fellowships including the IET Premium Award for Best Paper (2021), US National Academies ARL Harry Diamond Distinguished Fellowship (2018), and Andrew and Erna Finci Viterbi Postdoctoral Fellowship (2015–2017). Merhala Thurai is a research scientist at Colorado State University, USA. She received her BSc in Physics from Imperial College, London, and her PhD from King’s College, London. She worked at Rutherford Appleton Laboratory, Oxon, UK, for 14 years, and has been with Colorado State University since 2004.

This page intentionally left blank

Preface

This edited three-volume set of books published by the Institution of Engineering and Technology (IET) on “Advances in Weather Radar” falls under the prestigious SciTech/IET book series on radar, sonar, and navigation. We are delighted to edit this new book consisting of 32 original chapters reviewed by one of the editors and one external reviewer. They are written by exceptionally well-qualified experts from academia, research laboratories, and National Weather Services (NWSs) covering a breathtaking range of topics. The focus of this edited book is on dualpolarization radar because of its important application in rainfall measurement, in elucidating details of cloud physical processes, classification of meteorological and non-meteorological echo types, the validation and evaluation of bulk microphysical processes that predict number density and mixing ratio and radar hydrology among other more recent applications. The chapters were selected by the editors who have a combined experience of 80 years in prestigious research laboratories while the lead editor’s entire career in academia gives strong credibility to our reputation. There are only two other edited volumes available: one on “Radar in Meteorology” published by the American Meteorological Society in 1990 edited by the late David Atlas consisting of 44 chapters. This volume gives detailed accounts on the history of radar in meteorology post World War II, research and operational topics using Doppler radar and one chapter on dual-polarization radar. The other edited volume on “Radar and Atmospheric Sciences” was a collection of essays published by the (AMS) in 2003 in honor of the late David Atlas edited by Roger Wakimoto and Ramesh Srivastava but contained only one devoted to dualpolarization. It is, therefore, very appropriate that the next edited volumes cover the scientific progress made since 2003–2023 a period of two decades when Doppler dual-polarization radar became de rigueur in any NWS agency. Several books have been published since 1983 which give an intermediate to advanced level exposition of polarimetric Doppler radar theory with only small subsections in Doviak and Zrnic (1984) which was expanded to about a chapter in the 2nd edition published in 1993. Material for the 2nd edition came from a course taught by the authors at the University of Oklahoma to Senior and first year graduate students in meteorology and electrical engineering. In 2001, the first book exclusively devoted to polarimetric Doppler radar principles and applications was published (Bringi and Chandrasekar) which soon became the standard reference text. The book was an outcome of a graduate level course in Radar and Electromagnetics taught by the authors for many years at Colorado State University. For the next 15 years, it was the textbook of choice for academics and researchers due to its development of the

xvi

Advances in weather radar, volume 3

theory from first principles in a self-contained format. It was also the time when polarimetric radar research was accelerating at a scorching pace. Two more books appeared recently (Zhang 2017) and Ryzhkov and Zrnic (2019), the former based on sound theory rooted in mathematics and based on a class taught by Zhang at the University of Oklahoma. The book by Ryzhkov and Zrnic is a monograph written at an intermediate level for meteorologists and researchers enabling them to gain a quick understanding of polarimetry without going into the mathematical theory which can be found in Bringi and Chandrasekar (2001) or Zhang (2017). It is our belief which is perhaps obvious that books written and edited by authors themselves are a different breed altogether compared to edited volumes where each chapter is written by experts giving a level of detail that cannot be found in textbooks or monographs. With rapid advances in the application of polarimetric radar to topics that cannot be covered by conventional books, the editors of the current 3-volume set of books felt there was a vacuum in documenting recent advances in dual-polarized radar resulting in the current edited volumes. It is well known that the electromagnetic wave is defined by its amplitude, phase, and polarization state. The interaction of the wave and the precipitation particles changes the amplitude, phase, and polarization state. The magnitude of the square of the amplitudes correspond to the power scattered by precipitation, the phase change due to changes in the electrical path length of moving precipitation along the radial direction which is related to the Doppler frequency, and the scattering matrix due to interaction with precipitation such as raindrops which are nonspherical and oriented in space. These changes in the EM waves when quantified with particle microphysical states is one of the main drivers of dual-polarized Doppler weather radar. Recognizing the importance of dual-polarization the US NWS in a major effort installed the Next Generation Weather Radar (NEXRAD) WSR-88 Doppler network of 160 radars beginning around 2008. The main NWS goals were twofold, (a) detection of tornadoes by the mesoscale circulations aloft in supercell storms prior to damaging winds at the ground with lead times of around 20 min, and (b) measuring hourly rain amounts in pixels 44 km out to ranges of 230 km for hydrological applications (flash floods and landslides) and land falling hurricanes which are the most prolific rain producers causing coastal ecological damage and loss of life. Most of the chapters in this book describe the physical properties from the interactions also known as the “inverse” problem which is poorly constrained. Measuring the change of polarization state due to interactions with precipitation is considered arguably the most important advance since Doppler technology was introduced on a NEXRAD prototype. The few dozen pioneers (one of those pioneers is the lead editor of this three-volume set) who were by and large electrical engineers were astonished by the meteorological and non-meteorological applications that they could not have possibly imagined at that time (1982). This edited 3volume set has many chapters that describe these innovative applications. We hope that these edited volumes would be useful to graduate students, to radar systems

Preface

xvii

designers, to high level managers of National Meteorological Services and other research scientists who need to delve deeper into specific topics which cannot be found elsewhere. This third and final volume on “Emerging Applications” includes applications of weather radars in novel as well as non-meteorological applications. The first volume “Precipitation Sensing Platforms” focuses on major advances in designing, operating, and deploying weather radars across the globe. The second volume “Precipitation Science, Scattering, and Processing Algorithms” considers theoretical milestones achieved in microphysics, electromagnetics, and signal processing of radar meteorology. We thank all contributing authors for submitting their high-quality contributions. We sincerely acknowledge the support and help from all the reviewers for their timely and comprehensive evaluations of the manuscripts that improved the quality of this book. Finally, we are grateful to the IET Press Editorial Board and the staff members Nicki Dennis, Olivia Wilkins, and Sarah Lynch for their support, feedback, and guidance. V. N. Bringi Fort Collins, CO, USA Kumar Vijay Mishra Adelphi, MD, USA Merhala Thurai Fort Collins, CO, USA

This page intentionally left blank

Acknowledgments

V.N.B. acknowledges continuous support from the Atmospheric Sciences Program of the National Science Foundation to Colorado State University, from the Army Research Office, and NASA. K.V.M. acknowledges support from the National Academies of Sciences, Engineering, and Medicine via Army Research Laboratory Harry Diamond Distinguished Fellowship. M.T. acknowledges support by the National Science Foundation, grant number AGS-1901585 to Colorado State University.

This page intentionally left blank

List of editors

V.N. Bringi, Colorado State University, Fort Collins, CO, USA Kumar Vijay Mishra, United States DEVCOM Army Research Laboratory, Adelphi, MD, USA Merhala Thurai, Colorado State University, Fort Collins, CO, USA

This page intentionally left blank

List of contributors

Silke Bauer, Swiss Ornithological Institute Sudesh Boodoo, Environment and Climate Change Canada V. N. Bringi, Colorado State University John Y. N. Cho, MIT Lincoln Laboratory Adriaan Dokter, Cornell University Norman Donaldson, Environment and Climate Change Canada Andrew Farnsworth, Cornell University Marina Gashinova, University of Birmingham Frank Gekat, LEONARDO Germany GmbH Ronald Hannesen, LEONARDO Germany GmbH Andre´s Pe´rez Hortal, Environment and Climate Change Canada Kyle G. Horton, Colorado State University David Hudak, Environment and Climate Change Canada Paul Joe, Environment and Climate Change Canada Sebastian Kauczok, LEONARDO Germany GmbH Witold F. Krajewski, The University of Iowa Hidde Leijnse, Royal Netherlands Meteorological Institute Frank Silvio Marzano, University of Rome Luigi Mereu, Istituto Nazionale di Geofisica e Vulcanologia, Italy Daniel Michelson, Environment and Climate Change Canada Kumar Vijay Mishra, United States DEVCOM Army Research Laboratory Mario Montopoli, National Research Council of Italy Cecilia Nilsson, Lund University Fatemeh Norouzian, University of Birmingham Aart Overeem, Royal Netherlands Meteorological Institute Marcus Pool, LEONARDO Germany GmbH Janti Reid, Environment and Climate Change Canada Peter Rodriguez, Environment and Climate Change Canada Elena Saltikoff, Integrated Carbon Observation System, Finland Christian Schiefer, LEONARDO Germany GmbH Simona Scollo, Istituto Nazionale di Geofisica e Vulcanologia, Italy Brian Sheppard, Environment and Climate Change Canada James A. Smith, Princeton University Merhala Thurai, Colorado State University Albert To¨ws, LEONARDO Germany GmbH Philippe Tristant, FREQONSULT, Vertou, France

xxiv

Advances in weather radar, volume 3 Remko Uijlenhoet, Delft University of Technology Gianfranco Vulpiani, Presidency of the Council of Ministers, Italy Yang Xiao, University of Birmingham

List of reviewers

V. N. Bringi, Colorado State University David Hudak, Environment and Climate Change Canada Scott M. Ellis, National Center for Atmospheric Research, Boulder Ahmad Gharanjik, Databourg Systems, Luxembourg Patrick C. Kennedy, Colorado State University Peter T. May, Monash University Kumar Vijay Mishra, United States DEVCOM Army Research Laboratory M. R. Bhavani Shankar, University of Luxembourg Merhala Thurai, Colorado State University Remko Uijlenhoet, Delft University of Technology Zeen Zhu, Brookhaven National Laboratory

This page intentionally left blank

Introduction to volume 3

The third volume of this book sums up a plethora of novel weather radar applications that have proliferated to several other fields. Chapter 1 “Radar hydrology” describes the advances in the field of radar hydrology that has progressed significantly during the past two decades. It addresses estimation of rainfall extremes for infrastructure design and safety studies, and streamflow forecasting for floods. The chapter discusses the importance of hydrological models in radar-based rainfall estimation. Chapter 2 “Quantitative precipitation estimation from weather radars, personal weather stations and commercial microwave links” reviews the state-of-the-art of land surface rainfall estimation using measurements from weather radars, personal weather stations, and commercial microwave links, including comparisons to rain gauge measurements. These studies are related to recently emerging field of opportunistic weather sensing using the existing communications infrastructure. Chapter 3 “Quantitative weather radar: a research to operations perspective in Canada” continues the discussion on quantitative precipitation estimation by focusing on studies carried out with radars deployed by the Environment and Climate Change Canada (ECCC). It also includes satellite-based radar precipitation observations from the 2006 Canadian CloudSat CALIPSO Validation Experiment (C3VP) and 2012 Global Precipitation Mission (GPM) Cold Season Precipitation Experiment (GCPEx). Chapter 4 “Volcanic plume retrieval using weather radar” captures the usage of weather radar in monitoring and parameter retrieval of volcanic clouds. The detection and quantitative retrieval of ash plumes is of significant interest because of the environmental, climatic, and socioeconomic effects of ash fallout. The possibility of monitoring in all weather conditions at a high spatial resolution (less than a few hundred meters) and every few minutes after the eruption is the major advantage of using ground-based microwave weather radar systems. Such weather radar systems can also provide data for estimating the ash volume, total mass, and height of eruption clouds. Chapter 5 “The effect of weather on the performance of mm-wave and subTHz automotive radar” examines meteorological effects and processing in the context of millimeter-wave (mm-wave) (77 GHz) and sub-terahertz (sub-THz) (100–300 GHz) automotive radars. This part of the spectrum offers ultrawide bandwidths leading to a very high range resolution. The chapter presents experimental measurement results of automotive radar signal attenuation during various intensities of rainfall and snowfall at the current automotive radar frequencies.

xxviii

Advances in weather radar, volume 3

These studies are fundamentally important to investigate the effect of adverse weather conditions on low-THz radar performance in comparison with the current automotive radars operating in the traditional mm-wave band. Chapter 6 “Spectral interference in weather radars from wireless communication systems” explores the emerging problem of spectral interference in weather radars from coexisting communications services. Wireless technology, such as local area telecommunication networks and surveillance cameras, causes severe interference for weather radars because they use the same operational radio frequencies. Radio frequency management and joint access of the spectrum is now of the major concerns for the worldwide weather community. Chapter 7 “Advances in weather radar monitoring of bird movement” covers the recent advances in employing weather radars for noninvasive (i.e., without the need for capture, handling, or fitting of tracking devices) monitoring the aerial habitat of birds and tracking the movements of large numbers of flying birds. Further, the networks of weather radars are able to monitor bird movement over spatial scales that match the continental scale of bird movements. Chapter 8 “Complementary operation of Doppler radar and lidar at airports” focuses on the latest technological trend of combining the measurements of weather radars and Doppler lidars at the airports to mitigate the risks caused by windinduced hazards. A weather radar is unable to detect wind hazards during the period of nonprecipitation because it relies on hydrometeor backscatter to deduce wind parameters. However, a Doppler lidar employs aerosol backscatter and, hence, can complement the weather radar for volumetric wind measurements. The volume concludes with an epilogue authored by the editors.

Chapter 1

Radar hydrology Witold F. Krajewski1 and James A. Smith2

The authors address two central applications of radar data in hydrology: estimation of rainfall extremes for infrastructure design and safety studies, and streamflow forecasting for flood warnings and mitigation. They briefly review the fundamentals of radar-rainfall estimation in the context of challenges and basic limitations. They discuss innovative approaches to increase the size of data records and characterize the special challenges of estimating rainfall extremes. Regarding streamflow forecasting, they focus on the interaction of radar-rainfall uncertainty and river basin scale. They identify the rainfall volume as the single most important variable that needs to be estimated well for successful streamflow prediction. They complement the chapter with citing numerous earlier studies on the relevant topics.

1.1 Introduction Radar networks are increasingly used around the world to monitor weather and to warn populations against weather-caused natural disasters. A key aspect of weather is rainfall and its hydrologic consequences. Once on the ground, rainfall is partitioned into surface storage and runoff, infiltration and evapotranspiration. Hydrologic considerations, through rainfall-runoff modeling, follow and emphasize these components to a varied degree, depending on a particular application, e.g., streamflow forecasting. While the applications cover a large range of space and time scales, from climatological and continental to local and nearly instantaneous, in this chapter, we focus on two applications with great interest to public safety and engineering design. The first is the probabilistic description of extreme rainfall motivated by the needs of engineering design and the other is streamflow simulation motivated by flood forecasting. Radar-based rainfall estimation plays increasingly a key role in both. Below, we begin by giving a summary of the central features of radar-rainfall estimation that we deem relevant to our further discussion on the two hydrologic problems. The research community has been conducting studies exploring radarrainfall estimation capabilities for the last 60–70 years, and we have been contributing 1 2

Civil & Environmental Engineering, University of Iowa, Iowa City, IA, USA Civil & Environmental Engineering, Princeton University, Princeton, NJ, USA

2

Advances in weather radar, volume 3

to the results over the second half of this period. Some 20 years ago, we attempted to synthetize, from a hydrologic perspective, developments to date [1]. That early summary focused on the characterization of uncertainty in radar-rainfall estimation. A few years later, the review was supplemented by Villarini and Krajewski [2] and Krajewski et al. [3]. In a subsequent review paper, Berne and Krajewski [4] noted the relatively slow entry of radar-rainfall products to the operational hydrologic forecasting environment. While radar-rainfall maps were routinely produced by the radar network operators and users, these maps were used only in a qualitative way, usually via a visual inspection. Fully automated systems in which radar-rainfall would enter as quantitative input and be transformed into another quantity of interest, e.g., streamflow, were few. This is changing fast as more and more services around the world are deploying operational, fully automated real-time flood forecasting systems. Examples include Argyle et al. [5]; Georgakakos et al. [6]; Bales and Flowers [7]; Krajewski et al. [8], as well as similar systems in Europe and Asia (e.g., [9,10]). Regarding our central problem, i.e., the description of the probability distribution of rainfall quantities, radar-based rainfall products allow trading space for time. Traditionally, an estimation of rainfall accumulation (over a given period, e.g., 5 min or 1 day) quantiles has been based on long records of rain gauge observations. Rain gauge networks are sparse, and long records of high temporal resolution data are rare. Nevertheless, engineering tools such as the rainfall frequency tools used to develop Atlas 14 [11] in the United States (US) have been developed using regional rain gauge network data. As a result, the final products, still spatially incomplete despite considerable efforts, are subject to large uncertainty. Radar-based rainfall maps provide nearly continuous spatial coverage on an easy-to-manipulate grid. For example, the best-known radar-rainfall product in the US is Stage IV [12], hourly rainfall on an approximately 4 km by 4 km grid covering the entire country but with gaps in the mountainous regions due to the beam blockages. While the resolution may be inadequate for some engineering applications, a statistical analysis can benefit from increased sample size. Researchers have analyzed the adequacy of this data set for rainfall characterization (e.g., [13–16]). Other products developed in real time, with a higher spatial resolution of 1 km by 1 km, also contain numerous artifacts that may affect their use in hydrologic applications. As an example, consider the national map of probability of detection of rainfall calculated for the 7 years period from 2016 to 2022 (Figure 1.1). The product is multi-radar multi-sensor (MRMS) or Multi-Radar Multi-Sensor [17] representing here the hourly rainfall accumulation corrected by rain gauges. In the following, we further scrutinize the challenges that lay in front of the research community before radar-rainfall products become sufficiently reliable to support engineering design for the future. Here we summarize the key aspects of radar-rainfall estimation that need to be kept in mind when discussing the two problems we address in the chapter. Our discussion is at a basic level, suitable for all readers, including the novice. Detailed discussions of the issues we cover are documented in the cited literature, some of them are addressed in other chapters of this book (e.g., Chapter 2). Weather radar networks observe rainfall above the ground and indirectly. The consequence of the first point is that as distance (range) from the radar increase, so does

Radar hydrology

3

Figure 1.1 Probability of detection of rainfall calculated based on 7 years (2016–2022) of MRMS hourly, rain gauge corrected data. It shows numerous artifacts of radar data processing that are inherent in ground radar technology. the height above the ground where observations are taken. As a result, while the final radar-rainfall products often “look” as two-dimensional images, they represent a complicated, although systematic, sampling of the four-dimensional (three-dimensional space plus time as radars scan the evolving atmosphere) rainy space. For a single radar, the interplay of range effect and the vertical properties of the atmosphere combine, affecting the relevance of the estimated rainfall to that on the ground (e.g., [18]). It is the true, but unknown, rainfall reaching the ground that causes the hydrologic consequences. The indirectness of radar-rainfall observations implies the need for estimating it from the observed quantities. Therefore, we know the true rainfall only within certain approximation (error). Quantification of this error, more precisely its statistical properties, has been the subject of intense research over the last 30 years. We know that radar-rainfall uncertainty depends on the properties of rainfall itself, i.e., drop size distribution and its variability in space and time, radar system observables and their quality, and the algorithms used to convert the observables aloft to the quantities of interest, i.e., rainfall rate and accumulation at the ground. Given this, there is no wonder that a complete statistical description of radar-rainfall uncertainty still eludes the research community.

1.2 Radar hydrology and the climatology of extreme rainfall After a decade of research and development, the challenges and opportunities identified by Berne and Krajewski [4] for assessing rainfall and flood extremes using radar have increased in importance (e.g., [19,20]). In addition, new

4

Advances in weather radar, volume 3

challenges have emerged, old ones have crystallized and the range of applications has expanded. One of the major developments over the last decade is that many radar-rainfall data sets have increased in length by 10 years, turning the attention of the research community more concretely to the potential for climatological analyses. Despite the additional data, record lengths remain short for many applications concerning rainfall and flood extremes—a central theme of methodological development for radar hydroclimatology remains “trading space for time.”

1.2.1

Radar-rainfall data sets

The development of long-term radar-rainfall data sets is a necessary first step in creating the radar hydrometeorology machinery for rainfall frequency analysis. Rainfall fields developed for operational weather forecasting and analysis (e.g., [21–25]) have provided a natural path for climatological analyses. The “Stage IV” data set maintained by the National Weather Service provides hourly, 4 km4 km rainfall fields over the continental US with a record length that exceeds 20 years (2002–2022; [23]). Reanalysis data sets have been developed from archived radar fields and algorithms that can be tailored to climatological applications (e.g., [17,26–33]). Operational data sets, like the Stage IV or MRMS rainfall fields, reflect changing measurements over time; the polarimetric upgrade of the US radar network in 2012 provides a notable example. In addition to changes in the basic radar measurements, operational algorithms for rainfall estimation have changed over time, as has the implementation through parameter specifications. The most direct approach to rainfall frequency analysis using radar data sets is to treat observations from each grid as thought they were observations from a rain gauge. The approach underlies studies using annual maximum and peaks-overthreshold analyses for “long” radar-rainfall records [13,15,16,34–40]. A compelling rationale for these studies is that sub-daily rain gauge networks with long records are exceptionally sparse in many settings. The hourly time resolution of Stage IV and the 21-year record opens new avenues for assessing sub-daily rainfall extremes. In regions with large spatial gradients in rainfall extremes, radar provides the potential for resolving spatial heterogeneities that are difficult to address solely through gauge-based analyses (e.g., [41]). Over the last decade, evidence for increasing short-duration rainfall extremes in a warming climate has mounted [42]. The availability of radar-rainfall data sets covering the last two decades creates an important resource for assessing climate change impacts on flood hazards [19]. Direct assessments of changing rainfall extremes based on radar-rainfall data sets will provide increasingly important tools for hydroclimatological analyses. Process-based studies of rainfall response to warming can also contribute to debates on changing extremes (e.g., [43,44]).

1.2.2

Stochastic storm transposition

The core strength of radar for rainfall analysis is the ability to monitor rainfall at high temporal and spatial resolution over large domains. Exploiting this strength is central to stochastic storm transposition (SST) methods that resample observed rainfall from a catalog of extreme events. Foufoula-Georgiou [45] introduced SST

Radar hydrology

5

methods based on storm catalogs developed by the US Army Corps of Engineers for probable maximum precipitation (PMP) studies (see [46]) for a historical synopsis of SST. Radar-based storm catalogs have emerged as central components of SST procedures used for precipitation frequency analysis [47–51] and for advances in estimating PMP ([52]; see also [53]). The remnants of Hurricane Ida (2021) produced PMP-magnitude rainfall accumulations at 1–3-h time scale and basin scales smaller than 500 km2 (Figure 1.2; see also Smith et al. [54]) and resulted in more than 50 flood fatalities in the northeastern

Figure 1.2 Storm total rainfall for the remnants of Hurricane Ida (September 1, 2021) based on polarimetric radar estimates and gauge-based mean field bias correction (see [54] for details). Basin boundaries for the Neshanic River and Stony Brook are shown along with locations of rain gauges in Stockton, Hopewell and Skillman in New Jersey.

6

Advances in weather radar, volume 3

US. Rainfall accumulations derived from polarimetric radar observations include accumulations at 1–3-h time scale that markedly exceed 1,000-year return interval values based on National Oceanic and Atmospheric Administration precipitation frequency analyses. Storm structure, motion and evolution contribute to extreme variability of rainfall in space and time for the regions that experienced most extreme rainfall, as depicted by 2-min rainfall rate fields in Figure 1.3. SST-based methods for computing rainfall extremes at small time and space scales are grounded in the capability to monitor storms at small time and space scales (as illustrated in Figures 1.2 and 1.3). A key assumption often made in the application of SST methods is that the occurrence of extreme rainfall events over a homogeneous region constitutes a spatial Poisson process. To assess rainfall frequency over a

Figure 1.3 Polarimetric rainfall estimates at 2240, 2242, 2244 and 2246 UTC on September 1, 2021. Red circles show mesocyclone locations (with times) for the supercell that produced peak rainfall over the Neshanic River basin. The black circle shows the initiation location of the Stockton tornado.

Radar hydrology

7

particular region in the domain, one develops long records of rainfall over the region by resampling storms from a storm catalog and translating the storms randomly over the domain. For radar-based analyses, the region of interest can be specified by a drainage basin, an arbitrary sub-region of the domain or a single grid (providing a reference for point analyses based on rain-gauge analyses). The Poisson assumption implies that any event that occurred over the domain during the observing period could have happened over the drainage basin or point of interest and justifies the resampling approach in which transformed locations of selected storms are uniformly distributed over the domain. SST provides a direct application of the principle of trading space for time. SST procedures exploit the spatial and temporal structure of rainfall sampled by radar and mesh naturally with flood frequency analyses through pairing with hydrologic models [49,55]. For long return intervals, SST results depend strongly on the most extreme event in the storm catalog [49], placing a premium on accurate estimation of the upper tail of rainfall as we discuss later. For a storm catalog of 20 years, how accurately can we estimate the 100-year rainfall over a drainage basin? The answer depends on several features embedded in the problem formulation—size of the homogeneous domain, storm duration and drainage basin area are important elements. The nature of rainfall extremes is also important. If rainfall extremes are driven by mixtures of storm types and the upper tail is dominated by a rare storm type, the estimation problem is especially challenging [49]. Homogeneous regions can be hard to find, and spatial heterogeneities in rainfall extremes can be large due to orographic effects, land-water boundaries and urbanization. The 100-year, 24-h rainfall accumulation in southeast Texas decreases from 470 to 330 mm over a distance of approximately 100 km [11]. The spatial gradient in rainfall over tens of kilometers is comparable to projected changes over many decades in response to a warming climate. Resampling methods used for SST have been adapted to address spatial heterogeneites in rainfall extremes (see [56,57]). Two simple types of spatial heterogeneities have been considered—spatially varying rates of occurrence of major events and spatial variability in rainfall magnitudes for extreme events. Long radarrainfall data sets are critical for resolving spatial heterogeneities in rainfall extremes, as detailed in Zhou et al. [56]. Rainfall variability in time and space (Figures 1.2 and 1.3) is central to SST methods for frequency analysis and more broadly to hydrometeorological assessments of flood hazards. Le Cam [58] introduced a Lagrangian framework for examining the space–time variability of rainfall, providing a theoretical foundation for precipitation research (see also [59]). Structure, motion and evolution of storm cells (e.g., Figure 1.3) are key ingredients of LeCam’s modeling system, which was developed for design flood estimation for hydropower projects. Radar fields and storm tracking algorithms (e.g., [60–62]) have provided the observational and analysis tools to effectively pursue Lagrangian representations of rainfall variability. Storm tracking based on radar fields has also played a major role nowcasting extreme rainfall (e.g., [63,64]). Analyses of space–time variability of rainfall have received considerable attention over the last decade (e.g., [33,65–68]). Multiple, low-elevation

8

Advances in weather radar, volume 3

scanning strategies, which have been applied principally for monitoring supercells and tornadoes, e.g., Kumjian et al. [69] and Van Den Broeke [70], can improve rainfall estimates for extreme rain [54,71], rainfall estimates for Hurricane Ida exploit low-elevation scans with time scales shorter than 2 min (Figure 1.3). There are striking contrasts between radar and rain gauge networks in the ability to detect major rainfall events (e.g., [33,38]). In many settings and for many storms, rain gauge networks simply do not sample extremes, especially for convective rainfall. The ability to accurately estimate extreme rainfall from radar fields, however, continues to present challenges [14,72–74].

1.2.3

Polarimetric measurements and extreme rainfall: an example

Polarimetric measurements have resulted in marked improvements in estimates of rainfall extremes in the US (e.g., [71,75,76]). They hold potential for significant advances in estimating rainfall extremes, especially through specific differential phase shift and specific attenuation measurements (see [54,75,77]). For Hurricane Ida, rainfall estimates grounded in specific differential phase measurements (Figures 1.2 and 1.3) and gauge–radar intercomparisons (Figure 1.4) point to the capability of accurately estimating extreme rainfall at small time and space scales (see also [54]). 140 100

Stockton Radar Gauge

60 20

15-min Rain (mm/h)

140 100

Skillman Radar Gauge

60 20 140 100

Hopewell Radar Gauge

60 20

21:00

22:00

23:00

00:00

01:00

Figure 1.4 Gauge–radar time series for September 1, 2021 for the three rain gauges shown in Figure 1.2: Stockton, Hopewell and Skillman

Radar hydrology

9

Microphysical ideas have a long heritage in radar hydrology, especially through the development of extensive lists of Z–R relationships for rainfall estimation [4,78]. Contrasts between convective and stratiform precipitation played a central role in this line of research on radar-rainfall estimation. Distinctions between tropical convection and mid-latitude convection have also provided prominent themes in storm typing for rainfall estimation. Rainfall estimation problems for extreme rainfall continue to arise across the convective intensity spectrum. Extreme convective rain, especially for storms producing hail, remains a challenge for the estimation of short-duration rainfall extremes. Once discounted as major contributors to extreme rainfall, supercells are increasingly recognized as agents of extreme short-duration rainfall; rainfall extremes from the remnants of Ida are directly tied to tornadic supercells [54]. Extreme rainfall and flooding in Germany in July 2021 demonstrated the challenges associated with weak convection, especially in characterizing vertical profiles of hydrometeors [79]. Storm dynamics also remain a challenge for estimating rainfall extremes. Pauline Austin [80] pointed to departures of vertical motion from terminal velocity assumptions as a significant component of radar-rainfall estimation error. Vertical motion, especially the relationship between downdrafts and rainfall extremes, remains a poorly understood component of errors in extreme rainfall estimates [81]. Polarimetric radar measurements provide the potential for developing a sound microphysical and dynamical grounding for rainfall estimation [82]. Assimilations of polarimetric radar measurements into numerical weather prediction models [83,84] hold potential for addressing a range of problems that arise in estimating extreme rainfall from radar. Range effects, vertical profile properties and microphysical controls of rainfall rate are challenging issues that will ultimately benefit from coupled measurement and modeling programs. Considerable progresses in modeling, assimilation and measurement are needed to realize these advances [82].

1.2.4 Radar-rainfall biases Conditional bias ([85]; see also [86]), in which error properties vary with rainfall magnitude, is a fundamental problem for rainfall frequency analysis procedures outlined above, all of which are strongly influenced by errors in the upper tail of the rain rate distribution. Conditional bias arises as a component of radar-rainfall estimation error, even for the most advanced procedures [57,87]. Additional attention is needed to address an underestimation of rainfall extremes. Previous studies have provided useful directions for improving operational algorithms (see, e.g., [88,89]). Developing a strong, theoretical understanding of conditional bias remains an important task. Procedures that combine radar and rain gauge observations are central to both operational flood forecasting applications and to the development of climatological procedures based on radar observations. They fall into two broad categories—mean field bias adjustment ([90–92]; for recent developments, see [63,93,94]) and procedures that provide spatially distributed adjustments of radar-rainfall fields based on rain gauge observations ([95,96]; for recent developments, see [41,97–101]). Merging approaches, which often take the form of “adjustment” of radar-rainfall

10

Advances in weather radar, volume 3

estimates using rain gauges, are most effective if implemented after having accounted for all known sources of error in radar-rainfall estimates (e.g., calibration, beam shielding, attenuation, ground clutter, vertical reflectivity profile and Z– R conversion). An example of such an approach is given by Hazenberg et al. [18]. Mean field bias corrections are grounded in ratios of gauge observations to radar observations for rain gauge locations. For reflectivity-based algorithms, mean field bias effectively provides a gauge-based approach to specifying the pre-factor in the Z–R relationship; similar interpretations hold for mean field bias adjustments for polarimetric rainfall estimates. Polarimetric estimates of extreme rainfall from Ida are quite accurate (Figure 1.4), provided that gauge-based multiplicative bias is applied; for Ida, the bias is 1.4 [54]. For extreme rain events, bias values in polarimetric rainfall estimates can exceed 2; bias correction remains a key ingredient of methods for estimating extreme rainfall. Methods that provide local corrections exploit the correlation structure of rainfall fields and the error structure of radar-rainfall fields. Complex terrain provides an important setting for gauge–radar merging, as illustrated in Barton et al. [41]. Errors in rain gauge measurements [102] remain a significant problem for all classes of gauge–radar merging. An enduring challenge for radar-rainfall estimation is the range dependence of radar-rainfall estimates [103,104]. Range dependence is fundamentally linked to the vertical sampling by radar—higher elevations and larger volumes with increasing range from the radar—and vertical structure of precipitation [105–107]; for recent analyses of polarimetric rainfall estimates, see Cunha et al. [108] and Chaney et al. (2022). Range effects create significant challenges for the accurate representation of spatial gradients in rainfall extremes [31]. A related set of problems centers on beam blockage and partial beam blockage [109]. In addition to the distinctions between operational and reanalysis products, radar-rainfall data sets can be further divided into those based on a single radar and data sets for which compositing rainfall fields from a network of radars is employed. The Stage IV data set, for example, is a composite of rainfall fields from the network of Weather Surveillance Radar 88 Doppler radars in the US. There are advantages and disadvantages for both approaches, with rangedependent sampling of the atmosphere central to both. Compositing can diminish some of the impacts of range-dependent rainfall estimates, but they are still important elements of the error properties of the rainfall fields. For analyses based on individual radars, range effects can be addressed in post-analysis mode. For rainfall frequency analyses based on annual maximum series, for example, range dependence can be assessed for derived products, like the 10-year return interval rainfall accumulation, and corrections applied to eliminate systematic dependence on range. Flood hazards associated with short-duration rainfall extremes are particularly important for urban environments, and radar plays an increasingly important role in addressing these problems [110–116]. The prevalence of impervious surface and alteration of the drainage network amplify the importance of rainfall variability in time and space for urban flood response [110,115,117,118].

Radar hydrology 400

120

Neshanic Discharge 300

100 Neshanic Rain 80

200 60

Discharge (cms)

Basin-averaged Rain Rate (mm/h)

140

11

100

40 20 18:00

20:00

22:00

20:00

02:00

04:00

06:00

08:00

10:00

12:00

Sep 1st 2021 (UTC)

Figure 1.5 Basin-average rainfall rate and discharge for the Neshanic River for September 1, 2021

Flood extremes and catastrophic loss of life from the remnants of Hurricane Ida were closely linked to rainfall features that can only be resolved by radar. The flood peak in the Neshanic River (Figure 1.5; see also Figures 1.2 and 1.3) was more than 40% larger than the second largest flood peak from a gaging record of more than 100 years. Rainfall variability over a period of 15 min (Figures 1.3 and 1.5) was the central driver of extreme flood response over the Neshanic River on September 1, 2021. In the next section, we sharpen our focus on radar and drainage basin response to rainfall extremes.

1.3 Radar hydrology and flood forecasting Radar network ability to cover large regions with high resolution in space and in time makes it an attractive tool for real-time monitoring of rainfall for hydrologic forecasting applications. Of particular interest to our discussion is streamflow and thus flood forecasting everywhere, i.e., not only at selected sites at the rivers gauges but also anywhere there are human activities, mainly at the locations of communities, small and large, located by streams and rivers. To accomplish, this takes a fully distributed hydrologic model, for example, the Hillslope Link Model (HLM) used by the Iowa Flood Center for operational forecasting over the state of Iowa [8,119]. The model requires rainfall input for each hillslope, the main building block of the model’s architecture. Rainfall intensity time series can be projected on the locations of the hillslopes from the standard operational products typically given on a regular grid. In this chapter, we consider two such products and refer to them as Iowa Flood Center radar-rainfall product based on specific attenuation (IFCRA) [76,120,121], and the gauge-corrected MRMS [17] mentioned earlier.

12

Advances in weather radar, volume 3

Just as actual basin response is very sensitive to rainfall amount and its space–time distribution, the model response is sensitive to the estimated rainfall, its amount and space–time distribution. This brings us back to the issue of uncertainty in the radarrainfall estimates.

1.3.1

Radar-rainfall hydrologic visibility

From the hydrologic perspective, the quantification of radar-rainfall error is further complicated by the basin characteristics such as size, shape, topography, location with respect to the closest radar and more. As an illustration, consider the case of the Cedar River basin in Iowa (Figure 1.6). The basin, with drainage of about 13,000 km2, is covered by four Weather Surveillance Radar 88 Doppler Polarimetric (WSR-88DP) radars. If you consider the maximum range of the 230 km from each radar, the coverage is complete. If you consider the “good” range of, say, 150 km, the coverage is only partial. For illustration, Figure 1.6 shows the coverage by four WSR-88DP radars, i.e., Davenport (KDVM), Des Moines

Figure 1.6 Cedar River basin with outlet at Cedar Rapids, Iowa and its 150 km range coverage by four WSR-88DP radars. KDMX is located near Des Moines, Iowa, KDVN is near Davenport, Iowa, KRAX is located near La Cross, Wisconsin and KMPX is by Minneapolis, Minnesota. The white dots show arbitrary points located equal distance from the outlet.

Radar hydrology

13

(KDMX), La Cross (KRAX) and Minneapolis (KMPX). This is relevant if you assume, based on many past research studies, that the uncertainty in rainfall estimates increases as a function of distance from radar, especially at distance longer than 120–150 km. Now consider a hydrologically relevant characteristic of the river network draining the basin, i.e., the width function. The width function can be thought of as a distribution (histogram) of distances along flow lines of streams and rivers from the outlet. For the width function of the Cedar River at Cedar Rapids (Figure 1.7), we see that most water arrives to the outlet from the band of distances of 300–500 km from the outlet. This band contributes nearly 50% of the total (the entire range of 570 km is divided into 150 intervals). Therefore, for accurate streamflow prediction, the accuracy of rainfall estimates should be the highest for that domain. This is a general requirement, which we need to make more specific.

1.3.2 Random vs. systematic errors Let us consider the random errors first. Since there are many locations in the basin that are at equal distance from the outlet (see Figure 1.6), random errors of over- or underestimation in rainfall at those locations will likely cancel out when the runoff arrives at the outlet. This is especially true for symmetric error distributions with low to no spatial dependence. Now let us consider the potential impact of systematic errors (bias). For this, take two different radar-rainfall products, i.e., IFCRA and gauge-corrected MRMS, both at the hourly resolution. We use both products to force exactly the same rainfall-runoff model, i.e., HLM, at a number of basins throughout the state of Iowa. Each basin has a United States Geological Survey (USGS) stream

Contribution per interval (%)

1.5

1

2

3

Travel time (days) 4 5

6

7

8

1

0.5

KMPX KDVN

0

100

KDMX KRAX 200 300 Distance to outlet (km)

400

500

Figure 1.7 The width function (in blue) of the Cedar River basin with outlet at Cedar Rapids, Iowa. The color-shaded regions correspond to the various WSR-88DP radar coverages. The travel time is estimated assuming a constant water velocity of 0.75 m/s.

14

Advances in weather radar, volume 3

gauge at the outlet where we can check the model simulated vs. gauge observed streamflow. As the outcome of the simulation is a time series, we use a commonly used Kling-Gupta efficiency (KGE) index [122] to compare the two. The closer the value of the index is to 1 the better the simulation performance. Examine the results in Figure 1.8 where we compare the KGE obtained using the two radar-rainfall products. Each dot on the scatter plot represents a USGS stream gauge (out of over 120). The KGE is calculated annually over a period of 5 years of continuous streamflow simulation for the warm season of March through October giving us over 600 dots. The considerable scatter is evident in the plot, with the number of dots under and over the diagonal line approximately the same thus not favoring either product. Since the two radar-rainfall products force the very same hydrologic model, the scatter must be the result of the differences between the two input products. This may seem to be surprising as both use the volume scan (Level 2) data from the same radars. To examine which aspect of the differences between the two products is the most influential on the simulated streamflow, let us begin by equalizing the estimated basin-wide volume of rainfall, hour by hour, keeping the different spatial rainfall distributions provided by each of the products. Since we do not know which product is a closer representation of the true rainfall, we do the equalization both ways: IFCRA to MRMS and then MRMS to IFCRA. We use the two equalized products (i.e., equal in the mean over each basin) to force the same hydrologic model. The results are shown in Figure 1.9. It is

1.0

IFCRA

0.5

0.0

–0.5

–0.5

0.0 MRMS

0.5

1.0

Figure 1.8 Scatter plot of the KGE index (calculated annually) of the streamflow simulation based on two radar-rainfall products, MRMS and IFCRA for 120 USGS locations and 5 years of data (April–October). The same hydrologic Hillslope Link Model with the same parameter values was used with both products.

15

Equalized to IFCRA

Equalized to MRMS

Radar hydrology

–0.5

0.0 MRMS

0.5

1.0

–0.5

0.0 IFCRA

0.5

1.0

Figure 1.9 The same as in Figure 1.8 but with the radar-rainfall products equalized to each other at the hourly scale for each basin. The scatter is much reduced implying that the difference in the basin average was the dominant source of the KGE scatter in Figure 1.8. obvious that the scatter has been markedly reduced, suggesting that the difference in the rainfall volume over a basin is the most significant aspect of the radar-rainfall applications for streamflow forecasting. Based on the above argument, one may suspect that it is sufficient to provide uniform rainfall over the basin. Ghimire et al. [123] addressed this question and demonstrated that this is not the case. They show that preserving the key aspects of the spatial distribution of rainfall is also important (see also [124]). This seems clear if one recalls the basin response is governed to a degree by its width function and the rainfall distribution with respect to the key features of that function. Since for larger basins, there is more opportunity for the random error in the input to average out, this puts onus further on the systematic errors. While the equalization points to the total basin-wide rainfall volume as an important requirement for skillful prediction, from the demonstration above, we still do not know if accounting for a more complex structure of uncertainty would lead to significant improvement of streamflow prediction. The discussion thus far leads to the conclusion that the most important aspect of real-time radar-rainfall estimation and its application in hydrologic streamflow forecasting is to avoid systematic errors. This has been recognized a long time ago as the removal of mean-field bias was one of the first algorithms in the next-generation radar network of weather radars in the US system [90]. Here, however, we add an important caveat: The bias that matters is that over a given basin and not necessarily over the entire radar umbrella. Also, based on our earlier discussion on radar and rain gauge data merging, it seems that adjusting the rainfall field locally is a better way to go. However, the MRMS contains such an adjustment but its use in the hydrologic simulation does not seem to be superior. A likely culprit here is the relatively low density of the operationally available rain gauge data and/or their quality.

16

Advances in weather radar, volume 3

1.4 Summary and outlook In closing, we call out two key aspects of radar-rainfall estimation the solving of which would pave the way for markedly improving the societal benefits of radar systems. Analyses of extreme rainfall for rainfall frequency and PMP studies can benefit substantially from the application of radar-rainfall data sets. The challenges of developing climatologically useful data sets are tied to advances in radar technology—especially through polarimetric measurements—and algorithms for estimating rainfall and characterizing estimation error. The largest societal benefits are linked to short-duration rainfall extremes (especially time periods shorter than 6 h), which are poorly characterized by rain gauge data sets in many settings. Despite the difficulties in developing long radar-rainfall data sets, important applications will also be found in assessing changing rainfall extremes in a warming climate. Procedures which apply the principle of “trading space for time” will play a central role in these analyses. Another major unsolved problem in radar hydrology is the partitioning of the predictive uncertainty between that due to the radar-rainfall input and that due to the rainfall-runoff model, i.e., its representation of the physical processes involved. Solving this problem would allow the community to focus on reducing the most significant sources of uncertainty. The problem is difficult in part because the two components are intricately related. We do not have error-free rainfall input to force and test hydrologic models. And while radar-rainfall can be assessed to some degree without resorting to a hydrologic model, i.e., using rain gauge data only, such evaluations are limited in value for hydrologic model assessment. They allow limited statistical appraisals of basic quantities such as overall and conditional (at point) bias and scatter.

Acknowledgments The insights shared here were collected over many years of research on radarrainfall applications to hydrology and include the contributions from former students and collaborators. We thank all of them. Development of this chapter was supported by the Iowa Flood Center and the Rose and Joseph Summers endowment to WFK and the NOAA Cooperative Institute for Modeling the Earth System for JAS. We would like to thank Radoslaw Goska and Nicolas Velasquez, IFC and Mary Lynn Baeck, Princeton University, for their help with the included examples.

List of abbreviations Atlas-14 Stage IV MRMS

Precipitation Frequency Atlas of the United States US national radar-rainfall hourly product multi-radar multi-sensor

Radar hydrology SST PMP UTC US Z–R WSR-88DP HLM IFCRA KDMX KDVN KRAX KMPX KGE USGS

17

stochastic storm transposition probable maximum precipitation coordinated universal time United States reflectivity to rainfall relationship Weather Surveillance Radar 88 Doppler Polarimetric Hillslope Link Model Iowa Flood Center radar-rainfall product based on specific attenuation WSR-88DP near Des Moines, Iowa WSR-88DP near Davenport, Iowa WSR-88DP near La Cross, Wisconsin WSR-88DP near Minneapolis, Minnesota Kling-Gupta efficiency index United States Geological Survey

References [1] [2]

[3]

[4] [5]

[6]

[7]

Krajewski, W. F., and J. A. Smith, 2002: Radar hydrology: Rainfall estimation. Advances in Water Resources, 25, 1387–1394. Villarini, G., and W. F. Krajewski, 2010: Review of the different sources of uncertainty in radar-based estimates of rainfall. Surveys in Geophysics, 31, 107–129. Krajewski, W. F., G. Villarini, and J. A. Smith, 2010: Radar-rainfall uncertainties: Where are we after thirty years of efforts? Bulletin of the American Meteorological Society, 91(1), 87–94. Berne, A., and W. F. Krajewski, 2013: Radar for hydrology: Unfulfilled promise or unrecognized potential? Advances in Water Resources, 51, 357–366. Argyle, E. M., J. J. Gourley, Z. L. Flamig, T. Hansen, and K. Manross, 2017: Towards a user-centered design of a weather forecasting decision support tool. Bulletin of the American Meteorological Society, 98, 373–382, doi:10.1175/BAMS-D-16-0031.1. Georgakakos, K. P., T. M. Modrick, and E. Shamir et al., 2022: The flash flood guidance system implementation worldwide: A successful multidecadal research-to-operations effort. Bulletin of the American Meteorological Society, 103, E665–E679, doi.org/10.1175/BAMS-D-20-0241.1. Bales, J., and T. Flowers, 2022: Featured collection introduction: National water model IV. Journal of the American Water Resources Association, 58(1), 1–4, https://doi.org/10.1111/1752-1688.13001.

18 [8]

[9]

[10]

[11]

[12]

[13]

[14]

[15]

[16]

[17]

[18]

[19]

[20]

Advances in weather radar, volume 3 Krajewski, W. F., D. Ceynar, I. Demir et al., 2017: Real-time flood forecasting and information system for the State of Iowa. Bulletin of the American Meteorological Society, 98(3), 539–554. Berenguer M., D. Sempere-Torres, and G. G. S. Pegram, 2011: SBMcast – An ensemble nowcasting technique to assess the uncertainty in rainfall forecasts by Lagrangian extrapolation. Journal of Hydrology, 404, 226– 240, https://doi.org/10.1016/j.jhydrol.2011.04.033. Imhoff, R. O., C. C. Brauer, K.-J. van Heeringen, R. Uijlenhoet, and A. H. Weerts, 2022: Large-sample evaluation of radar rainfall nowcasting for flood early warning. Water Resources Research, 58, e2021WR031591, doi:10.1029/2021WR031591. Perica, S., S. Pavlovic, M. St. Laurent, C. Trypaluk, D. Unruh, and O. Wilhite, 2018: NOAA Atlas 14: Precipitation Frequency Atlas of the United States, Volume 11 Version 2.0 Texas. Tech. Rep., National Weather Service, Silver Spring, Maryland. Lin, Y., and K. E. Mitchell, 2005: The NCEP stage II/IV hourly precipitation analyses: Development and applications. Preprints, 19th Conf. on Hydrology, American Meteorological Society, San Diego, CA, 1.2. Available online at https://ams.confex.com/ams/pdfpapers/83847.pdf. Eldardiry, H., E. Habib, and Y. Zhang, 2015: On the use of radar-based quantitative precipitation estimates for precipitation frequency analysis. Journal of Hydrology, 531, 441–453. Eldardiry, H., E. Habib, Y. Zhang, and J. Graschel, 2017: Artifacts in stage IV NWS real-time multisensor precipitation estimates and impacts on identification of maximum series. Journal of Hydrologic Engineering, 22(5), E4015003. McGraw, D., E. I. Nikolopoulos, F. Marra, and E. N. Anagnostou, 2019: Precipitation frequency analyses based on radar estimates: An evaluation over the contiguous United States. Journal of Hydrology, 573, 299–310. Post, R., and W. F. Krajewski, 2023: Towards using radar data for probabilistic analysis of rainfall: A case study over Iowa. Stochastic Environmental Research and Risk Assessment, https://doi.org/10.1007/s00477-023-02439-8. Zhang, J., K. Howard, and C. Langston et al., 2016: Multi-radar multi-sensor (MRMS) quantitative precipitation estimation: Initial operating capabilities. Bulletin of the American Meteorological Society, 97(4), 621–638. Hazenberg, P., H. Leijnse, and R. Uijlenhoet, 2011: Radar rainfall estimation of stratiform winter precipitation in the Belgian Ardennes. Water Resources Research, 47, W02507, doi:10.1029/2010WR009068. Saltikoff, E., K. Friedrich, and J. Soderholm, 2019: An overview of using weather radar for climatological studies: Successes, challenges, and potential. Bulletin of the American Meteorological Society, 100(9), 1739–1752. Sokol, Z., J. Szturc, J. Orellana-Alvear, J. Popova´, A. Jurczyk, and R. Ce´lleri, 2021: The role of weather radar in rainfall estimation and its application in meteorological and hydrological modelling—A review. Remote Sensing, 13(3), 351.

Radar hydrology [21]

[22]

[23]

[24]

[25]

[26]

[27]

[28]

[29] [30]

[31]

[32]

[33]

[34]

[35]

19

Goudenhoofdt, E., and L. Delobbe, 2016: Generation and verification of rainfall estimates from 10-yr volumetric weather radar measurements. Journal of Hydrometeorology, 17(4), 1223–1242. Kreklow, J., B. Tetzlaff, B. Burkhard, and G. Kuhnt, 2020: Radar-based precipitation climatology in Germany—Developments, uncertainties and potentials. Atmosphere, 11(2), 217. Nelson, B. R., O. P. Prat, D.-J. Seo, and E. Habib, 2016: Assessment and implications of NCEP Stage IV quantitative precipitation estimates for product intercomparisons. Weather and Forecasting, 31(2), 371–394. Panziera, L., M. Gabella, U. Germann, and O. Martius, 2018: A 12-year radar-based climatology of daily and sub-daily extreme precipitation over the Swiss Alps. International Journal of Climatology, 38(10), 3749–3769. Yu, J., X.-F. Li, E. Lewis, S. Blenkinsop, and H. J. Fowler, 2020: UKGrsHP: A UK high-resolution gauge–radar–satellite merged hourly precipitation analysis dataset. Climate Dynamics, 54(5), 2919–2940. Boudevillain, B., G. Delrieu, A. Wijbrans, and A. Confoland, 2016: A highresolution rainfall re-analysis based on radar–raingauge merging in the Ce´vennes-Vivarais region, France. Journal of Hydrology, 541, 14–23. Kirstetter, P.-E., J. J. Gourley, Y. Hong et al., 2015: Probabilistic precipitation rate estimates with ground-based radar networks. Water Resources Research, 51(3), 1422–1442. Krajewski, W. F., A. Kruger, S. Singh, B.-C. Seo, and J. A. Smith, 2013: Hydro-NEXRAD-2: Real-time access to customized radar-rainfall for hydrologic applications. Journal of Hydroinformatics, 15(2), 580–590. Nelson, B. R., D. Seo, and D. Kim, 2010: Multisensor precipitation reanalysis. Journal of Hydrometeorology, 11(3), 666–682. Overeem, A., I. Holleman, and A. Buishand, 2009b: Derivation of a 10-year radar-based climatology of rainfall. Journal of Applied Meteorology and Climatology, 48, 1448–1463. Smith, J. A., M. L. Baeck, G. Villarini, C. Welty, A. J. Miller, and W. F. Krajewski, 2012: Analyses of a long-term high-resolution radar rainfall data set for the Baltimore metropolitan region. Water Resources Research, 48(4). Wright, D. B., J. A. Smith, G. Villarini, and M. L. Baeck, 2014: Long-term high-resolution radar rainfall fields for urban hydrology. Journal of American Water Resources Association, 50(3), 713–734. Lengfeld, K., P.-E. Kirstetter, H. J. Fowler, J. Yu, A. Becker, Z. Flamig, and J. Gourley, 2020: Use of radar data for characterizing extreme precipitation at fine scales and short durations. Environmental Research Letters, 15(8), 085 003. Allen, R. J., and A. T. DeGaetano, 2005: Considerations for the use of radarderived precipitation estimates in determining return intervals for extreme areal precipitation amounts. Journal of Hydrology, 315(1–4), 203–219. de Valk, C., and A. Overeem, 2022: A simple model for predicting the statistics of spatiotemporal extremes of sub-daily precipitation. Weather and Climate Extremes, 36, 100–424.

20 [36]

[37]

[38]

[39]

[40]

[41]

[42]

[43]

[44]

[45]

[46]

[47]

[48]

Advances in weather radar, volume 3 Ghebreyesus, D. T., and H. O. Sharif, 2021: Development and assessment of high-resolution radar-based precipitation intensity-duration-frequency (IDF) curves for the State of Texas. Remote Sensing, 13 (15), 2890, doi.org/ 10.3390/rs13152890. Marra, F., E. Morin, N. Peleg, Y. Mei, and E. N. Anagnostou, 2017: Intensity–duration–frequency curves from remote sensing rainfall estimates: Comparing satellite and weather radar over the eastern Mediterranean. Hydrology and Earth System Sciences, 21 (5), 2389–2404. Molter, E. M., W. D. Collins, and M. D. Risser, 2021: Quantitative precipitation estimation of extremes in conus with radar data. Geophysical Research Letters, 48 (16), e2021GL094697. Overeem, A., A. Buishand, and I. Holleman, 2009a: Extreme rainfall analysis and estimation of depth-duration-frequency curves using weather radar. Water Resources Research, 45, W10424, doi:10.1029/2009WR007869. Overeem, A., T. A. Buishand, I. Holleman, and R. Uijlenhoet, 2010: Extreme-value modeling of areal rainfall from weather radar. Water Resources Research, 46, W09514, doi:10.1029/2009WR008517. Barton, Y., I. V. Sideris, T. H. Raupach, M. Gabella, U. Germann, and O. Martius, 2020: A multi-year assessment of sub-hourly gridded precipitation for Switzerland based on a blended radar—Rain-gauge dataset. International Journal of Climatology, 40 (12), 5208–5222. Fowler, H. J., H. Ali, R. P. Allan et al., 2021: Towards advancing scientific knowledge of climate change impacts on short-duration rainfall extremes. Philosophical Transactions of the Royal Society A, 379 (2195), 20190542. Peleg, N., F. Marra, S. Fatichi et al., 2018: Intensification of convective rain cells at warmer temperatures observed from high-resolution weather radar data. Journal of Hydrometeorology, 19 (4), 715–726. Purr, C., E. Brisson, K. H. Schlu¨nzen, and B. Ahrens, 2022: Convective rain cell properties and the resulting precipitation scaling in a warm-temperate climate. Quarterly Journal of the Royal Meteorological Society, doi:10. 1002/qj.4277. Foufoula-Georgiou, E., 1989: A probabilistic storm transposition approach for estimating exceedance probabilities of extreme precipitation depths. Water Resources Research, 25 (5), 799–816. Wright, D. B., G. Yu, and J. F. England, 2020: Six decades of rainfall and flood frequency analysis using stochastic storm transposition: Review, progress, and prospects. Journal of Hydrology, 585, 124816. https://doi.org/10.1016/j.jhydrol.2020.124816 Thorndahl, S., J. A. Smith, M. L. Baeck, and W. F. Krajewski, 2014: Analyses of the temporal and spatial structures of heavy rainfall from a catalog of highresolution radar rainfall fields. Atmospheric Research, 144, 111–125. Wright, D. B., J. A. Smith, and G. V. M. L. Baeck, 2013: Estimating the frequency of extreme rainfall using weather radar and stochastic storm transposition. Journal of Hydrology, 488, 150–165.

Radar hydrology [49]

[50]

[51]

[52] [53]

[54]

[55]

[56]

[57]

[58]

[59]

[60]

[61]

[62]

[63]

21

Wright, D. B., J. A. Smith, and M. L. Baeck, 2014: Flood frequency analysis using radar rainfall fields and stochastic storm transposition. Water Resources Research, 50 (2), 1592–1615. Andersen, C. B., D. B. Wright, and S. Thorndahl, 2022: Sub-hourly to daily rainfall Intensity-duration-frequency estimation using stochastic storm transposition and discontinuous radar data. Water, 14 (24), 4013. Lengfeld, K., E. Walawender, T. Winterrath, and A. Becker, 2021: CatRaRE: A catalogue of radar-based heavy rainfall events in Germany derived from 20 years of data. Meteorologische Zeitschrift, 30 (6), 469–487. AWA, 2015: Probable Maximum Precipitation Study for Virginia. Tech. Rep., Applied Weather Associates. Smith, J. A., M. L. Baeck, M. Steiner, and A. J. Miller, 1996: Catastrophic rainfall from an upslope thunderstorm in the Central Appalachians: The Rapidan Storm of June 27, 1995. Water Resources Research, 32(10), 3099–3113. Smith, J. A., M. L. Back, Y. Su, G. A. Vecchi, and M. Liu, 2023: Strange storms: Extreme rainfall from the remnants of Hurricane Ida (2021) in the Northeastern US. Water Resources Research, 59, e2022WR033934. Perez, G., R. Mantilla, W. F. Krajewski, and D. Wright, 2019: Using physically-based streamflow simulations to assess local and regional flood frequency analysis methods. Water Resources Research, 55(11), 8384–8403. Zhou, Z., J. A. Smith, D. B. Wright, M. L. Baeck, L. Yang, and S. Liu, 2019: Storm catalog-based analysis of rainfall heterogeneity and frequency in a complex terrain. Water Resources Research, 55 (3), 1871–1889. Yu, G., D. B. Wright, and K. D. Holman, 2021: Connecting hydrometeorological processes to low-probability floods in the mountainous Colorado Front Range. Water Resources Research, 57(4), e2021WR029768. Le Cam, L. M., 1961: A stochastic description of precipitation. Proceedings of the Fourth Berkeley Symposium on Mathematical Statistics and Probability, J. Neyman, (eds.), University of California, Berkeley, 165–186. Waymire, E., V. K. Gupta, and I. Rodriguez-Iturbe, 1984: Spectral theory of rainfall intensity at the meso-b scale. Water Resources Research, 20, 1453–1465. Dixon, M., and G. Wiener, 1993: TITAN – Thunderstorm identification, tracking, analysis, and nowcasting – A radar-based methodology. Journal of Atmospheric and Oceanic Technology, 10(6), 785–797. Johnson, J., P. L. MacKeen, A. Witt et al., 1998: The storm cell identification and tracking algorithm: An enhanced WSR-88D algorithm. Weather and Forecasting, 13 (2), 263–276. Mecklenburg, S., J. Joss, and W. Schmid, 2000: Improving the nowcasting of precipitation in an Alpine region with an enhanced radar echo tracking algorithm. Journal of Hydrology, 239 (1–4), 46–48. Imhoff, R. O., C. C. Brauer, K.-J. van Heeringen et al., 2021: A climatological benchmark for operational radar rainfall bias reduction. Hydrology and Earth System Sciences, 25, 4061–4080, doi:10.5194/hess-25-4061-2021.

22 [64]

[65]

[66]

[67]

[68]

[69]

[70]

[71]

[72]

[73]

[74]

[75]

[76]

Advances in weather radar, volume 3 Pan, X., Y. Lu, K. Zhao, H. Huang, M. Wang, and H. Chen, 2021: Improving nowcasting of convective development by incorporating polarimetric radar variables into a deep-learning model. Geophysical Research Letters, 48 (21), e2021GL095302. Marra, F., and E. Morin, 2018: Autocorrelation structure of convective rainfall in semiarid-arid climate derived from high-resolution X-Band radar estimates. Atmospheric Research, 200, 126–138. Po¨schmann, J. M., D. Kim, R. Kronenberg, and C. Bernhofer, 2021: An analysis of temporal scaling behaviour of extreme rainfall in Germany based on radar precipitation QPE data. Natural Hazards and Earth System Sciences, 21 (4), 1195–1207. Smith, J. A., M. L. Back, L. Yang, J. Signell, E. Morin, and D. C. Goodrich, 2019: The paroxysmal precipitation of the desert: Flash floods in the Southwestern US. Water Resources Research, 55, 10218–10247, https:// doi.org/10.1029/2019WR025480. Thomassen, E. D., S. L. Thorndahl, C. B. Andersen, I. B. Gregersen, K. Arnbjerg-Nielsen, and H. J. D. Sørup, 2022: Comparing spatial metrics of extreme precipitation between data from rain gauges, weather radar and high-resolution climate model re-analyses. Journal of Hydrology, 610, 127915. Kumjian, M. R., A. V. Ryzhkov, V. M. Melnikov, and T. J. Schuur, 2010: Rapid-scan super resolution observations of a cyclic supercell with a dualpolarization WSR-88D. Monthly Weather Review, 138 (10), 3762–3786. Van Den Broeke, M. S., 2015: Polarimetric tornadic debris signature variability and debris fallout signatures. Journal of Applied Meteorology and Climatology, 54 (12), 2389–2405. Chaney, M., J. A. Smith, and M. L. Baeck, 2023: Range dependence of polarimetric radar estimates for extreme flood-producing rainfall in urban watersheds. Journal of Hydrometeorology, 23 (8), 1205–1226. Ba´rdossy, A., and G. Pegram, 2017: Combination of radar and daily precipitation data to estimate meaningful sub-daily point precipitation extremes. Journal of Hydrology, 544, 397–406. Peleg, N., F. Marra, S. Fatichi, A. Paschalis, P. Molnar, and P. Burlando, 2018b: Spatial variability of extreme rainfall at radar subpixel scale. Journal of Hydrology, 556, 922–933. Schleiss, M., J. Olsson, and P. Berg et al., 2020: The accuracy of weather radar in heavy rain: A comparative study for Denmark, the Netherlands, Finland and Sweden. Hydrology and Earth System Sciences, 24 (6), 3157–3188. Ryzhkov, A., P. Zhang, P. Bukovcic, J. Zhang, and S. Cocks, 2022: Polarimetric radar quantitative precipitation estimation. Remote Sensing, 14 (7), 1695. Seo, B.-C., W. F. Krajewski, and A. Ryzhkov, Evaluation of the specific attenuation method for radar-based quantitative precipitation estimation: Improvements and practical challenges, Journal of Hydrometeorology, 21 (6), 1333–1347, 2020.

Radar hydrology [77] [78]

[79]

[80] [81]

[82]

[83]

[84]

[85] [86]

[87] [88]

[89]

[90] [91] [92]

23

Kumjian, M. R., 2013a: Principles and applications of dual-polarization radar. Journal of Operational Meteorology, 1 (19–21), 226–274. Uijlenhoet, R., 2001: Raindrop size distributions and radar reflectivity–rain rate relationships for radar hydrology. Hydrology and Earth System Sciences, 5, 615–627, doi:10.5194/hess-5-615-2001. Chen, J-Y., R. Reinoso-Rondinel, S. Troemel, C. Simmer, A. Ryzhkov, and Alexander, 2023: A radar-based quantitative precipitation estimation algorithm to overcome the impact of vertical gradients of warm-rain precipitation: The flood in Western Germany on 14 July 2021. Journal of Hydrometeorology, 24(3), 521–536. Austin, P. M., 1987: Relation between measured radar reflectivity and surface rainfall. Monthly Weather Review, 115, 1053–1070. Kumjian, M. R., 2013b: Principles and applications of dual-polarization radar. Part II: Warm- and cold-season applications. Journal of Operational Meteorology, 1 (20), 243–264. Ryzhkov, A. V., J. Snyder, J. T. Carlin, A. Khain, and M. Pinsky, 2020: What polarimetric weather radars offer to cloud modelers: Forward radar operators and microphysical/thermodynamic retrievals. Atmosphere, 11 (4), 362. Thomas, G., J.-F. Mahfouf, and T. Montmerle, 2020: Toward a variational assimilation of polarimetric radar observations in a convective-scale numerical weather prediction (NWP) model. Atmospheric Measurement Techniques, 13 (5), 2279–2298. Zhang, G., J. Gao, and M. Du, 2021a: Parameterized forward operators for simulation and assimilation of polarimetric radar data with numerical weather predictions. Advances in Atmospheric Sciences, 38 (5), 737–754. Ciach, G. J., M. L. Morrissey, and W. F. Krajewski, 2000: Conditional bias in radar rainfall estimation. Journal of Applied Meteorology, 39 (11), 1941–1946. Seo, B.-C., and W. F. Krajewski, 2021: Multi-scale investigation of conditional errors in radar-rainfall estimates. Advances in Water Resources, 156, 104041. Sharif, R. B., E. H. Habib, and M. ElSaadani, 2020: Evaluation of radarrainfall products over Coastal Louisiana. Remote Sensing, 12 (9), 1477. Zhang, Y., and D.-J. Seo, 2017: Recursive estimators of mean-areal and local bias in precipitation products that account for conditional bias. Advances in Water Resources, 101, 49–59. Kim, B., D.-J. Seo, S. J. Noh, O. P. Prat, and B. R. Nelson, 2018: Improving multisensor estimation of heavy-to-extreme precipitation via conditional bias-penalized optimal estimation. Journal of Hydrology, 556, 1096–1109. Seo, D.-J., J. P. Breidenbach, and E. R. Johnson, 1999: Real-time estimation of mean field bias in radar rainfall data. Journal of Hydrology, 223 (3–4), 131–147. Smith, J. A., and W. F. Krajewski, 1991: Estimation of the mean field bias of radar rainfall estimates. Journal of Applied Meteorology, 30, 397–412. Borga, M., F. Tonelli, R. J. Moore, and H. Andrieu, 2002: Long-term assessment of bias adjustment in radar rainfall estimation. Water Resources Research, 38 (11), 8-1–8-10, doi:10.1029/2001WR000555.

24 [93]

[94]

[95] [96]

[97]

[98]

[99]

[100]

[101]

[102]

[103]

[104]

[105]

[106]

Advances in weather radar, volume 3 Armon, M., F. Marra, Y. Enzel, D. Rostkier-Edelstein, and E. Morin, 2020: Radar-based characterisation of heavy precipitation in the eastern Mediterranean and its representation in a convection-permitting model. Hydrology and Earth System Sciences, 24 (3), 1227–1249. Seo, B.-C., L. K. Cunha, and W. F. Krajewski, 2013: Uncertainty in radarrainfall composite and its impact on hydrologic prediction for the eastern Iowa flood of 2008. Water Resources Research, 49 (5), 2747–2764. Krajewski, W. F., 1987: Cokriging radar-rainfall and rain gage data. Journal of Geophysical Research – Atmospheres, 92 (D8), 9571–9580. Creutin, J., G. Delrieu, and T. Lebel, 1988: Rain measurement by raingageradar combination: A geostatistical approach. Journal of Atmospheric and Oceanic Technology, 5 (1), 102–105. Delrieu, G., A. Wijbrans, B. Boudevillain, D. Faure, and P.-E. Kirstetter, 2014: Geostatistical radar–raingauge merging: A novel method for the quantification of rain estimation accuracy. Advances in Water Resources, 71, 110–124. Goudenhoofdt, E., and L. Delobbe, 2009: Evaluation of radar-gauge merging methods for quantitative precipitation estimates. Hydrology and Earth System Sciences, 13 (2), 195–203. Ochoa-Rodriguez, S., L.-P. Wang, P. Willems, and C. Onof, 2019: A review of radar-rain gauge data merging methods and their potential for urban hydrological applications. Water Resources Research, 55 (8), 6356–6391. Sideris, I., M. Gabella, R. Erdin, and U. Germann, 2014: Real-time radar– rain-gauge merging using spatio-temporal co-kriging with external drift in the alpine terrain of Switzerland. Quarterly Journal of the Royal Meteorological Society, 140 (680), 1097–1111. Zhang, G., G. Tian, D. Cai, R. Bai, and J. Tong, 2021b: Merging radar and rain gauge data by using spatial–temporal local weighted linear regression kriging for quantitative precipitation estimation. Journal of Hydrology, 601, 126612. Steiner, M., J. A. Smith, S. J. Burges, C. V. Alonso, and R. W. Darden, 1999: Effect of bias adjustment and rain gauge quality control on radar rainfall estimation. Water Resources Research, 35, 2487–2503. Fabry, F., G. Austin, and D. Tees, 1992: The accuracy of rainfall estimates by radar as a function of range. Quarterly Journal of the Royal Meteorological Society, 118 (505), 435–453. Simpson, M. J., and N. I. Fox, 2018: Dual-polarized quantitative precipitation estimation as a function of range. Hydrology and Earth System Sciences, 22 (6), 3375–3389. Andrieu, H., and J. D. Creutin, 1995: Identification of vertical profiles of radar reflectivity for hydrological applications using an inverse method. Part I: Formulation. Journal of Applied Meteorology, 34, 225–239. Seo, D.-J., J. P. Breidenbach, R. Fulton, D. Miller, and T. O’Bannon, 2000: Real-time adjustment of range-dependent biases in WSR-88D rainfall estimates due to nonuniform vertical profile of reflectivity. Journal of Hydrometeorology, 1, 222–240.

Radar hydrology [107]

[108]

[109]

[110]

[111]

[112]

[113]

[114]

[115]

[116] [117]

[118]

[119]

25

Giangrande, S. E., and A. V. Ryzhkov, 2008: Estimation of rainfall based on the results of polarimetric echo classification. Journal of Applied Meteorology and Climatology, 47, 2445–2462. Cunha, L. K., J. A. Smith, M. L. Baeck, and W. F. Krajewski, 2013: An early performance evaluation of the NEXRAD dual polarization radar rainfall estimates for urban flood applications. Weather and Forecasting, 28, 1478–1497. Ryzhkov, A., M. Diederich, P. Zhang, and C. Simmer, 2014: Potential utilization of specific attenuation for rainfall estimation, mitigation of partial beam blockage, and radar networking, Journal of Atmospheric and Oceanic Technology, 31(3), 599–619. Cristiano, E., M.-C. ten Veldhuis, and N. van de Giesen, 2017: Spatial and temporal variability of rainfall and their effects on hydrological response in urban areas – A review. Hydrology and Earth System Sciences, 21 (7), 3859–3878. Hamidi, A., N. Devineni, J. F. Booth, A. Hosten, R. R. Ferraro, and R. Khanbilvardi, 2017: Classifying urban rainfall extremes using weather radar data: An application to the greater New York area. Journal of Hydrometeorology, 18 (3), 611–623. Kingfield, D. M., K. M. Calhoun, K. M. de Beurs, and G. M. Henebry, 2018: Effects of city size on thunderstorm evolution revealed through a multiradar climatology of the central United States. Journal of Applied Meteorology and Climatology, 57 (2), 295–317. Saharia, M., P.-E. Kirstetter, H. Vergara, J. J. Gourley, Y. Hong, and M. Giroud, 2017: Mapping flash flood severity in the United States. Journal of Hydrometeorology, 18 (2), 397–411. Shepherd, J. M., 2005: A review of the current investigations of urbaninduced rainfall and recommendations for the future. Earth Interactions, 9, 1–27. ten Veldhuis, M.-C., Z. Zhou, L. Yang, S. Liu, and J. Smith, 2018: The role of storm scale, position and movement in controlling urban flood response. Hydrology and Earth System Sciences, 22 (1), 417–436. Thorndahl, S., T. Einfalt, P. Willems et al., 2017: Weather radar rainfall data in urban hydrology. Hydrology and Earth System Sciences, 21 (3), 1359–1380. Yang, L., J. A. Smith, M. L. Baeck, and Y. Zhang, 2016: Flash flooding in small urban watersheds: Storm event hydrologic response. Water Resources Research, 52 (6), 4571–4589. Zhou, Z., J. A. Smith, L. Yang et al., 2017: The complexities of urban flood response: Flood frequency analyses for the Charlotte metropolitan region. Water Resources Research, 53 (8), 7401–7425. Mantilla, R., W. F. Krajewski, N. Velasquez et al., 2022: The hydrological Hillslope-Link Model for space-time prediction of streamflow: Insights and applications at the Iowa Flood Center. Extreme Weather Forecasting, M. Astitha, and E. I. Nikolopoulos, (eds.), Elsevier, Amsterdam, 200–238, ISBN 9780128201244.

26 [120]

[121]

[122]

[123]

[124]

Advances in weather radar, volume 3 Seo, B.-C., and W. F. Krajewski, 2020: Statewide real-time quantitative precipitation estimation using weather radar and NWP model analysis: Algorithm description and product evaluation. Environmental Modeling and Software, 132, 104791. Seo, B.-C., W. F. Krajewski, and A. Ryzhkov, 2020: Evaluation of the specific attenuation method for radar-based quantitative precipitation estimation: Improvements and practical challenges. Journal of Hydrometeorology, 21(6), 1333–1347. Gupta, H. V., H. Kling, K. K. Yilmaz, and G. F. Martinez, 2009: Decomposition of the mean squared error and NSE performance criteria: Implications for improving hydrological modelling. Journal of Hydrology, 377, 80–91, https://doi.org/10.1016/j.jhydrol.2009.08.003. Ghimire, G. R., W. F. Krajewski, T. B. Ayalew, and R. Goska, 2022: Hydrologic investigations of radar-rainfall error propagation to rainfallrunoff model hydrographs. Advances in Water Resources, 161, 104–145. Lobligeois, F., Andre´assian, V., Perrin, C., Tabary, P., and Loumagne, C., 2014: When does higher spatial resolution rainfall information improve streamflow simulation? An evaluation using 3620 flood events. Hydrology and Earth System Sciences, 18, 575–594, https://doi.org/10.5194/hess-18575-2014.

Chapter 2

Quantitative precipitation estimation from weather radars, personal weather stations and commercial microwave links Aart Overeem1, Remko Uijlenhoet2 and Hidde Leijnse1

2.1 Introduction Precipitation information is employed for a myriad of purposes, such as climate monitoring and understanding the drivers behind extreme events, weather monitoring and nowcasting, natural hazard early warning (e.g. floods and landslides), crop production modelling and rainfall index crop insurance, extreme value modelling and (sewerage) design purposes, evaluation of extreme events and their impact, and validation of weather prediction and climate model output (Chapters 1 and 3 of this book). Hence, localised, timely, regularly updated and accurate precipitation information is highly relevant. The focus of this chapter is on quantitative precipitation estimation (QPE) by ground-based sensors of liquid precipitation (rain) at the Earth’s surface, which is the dominant precipitation type in the mid-latitudes and (sub)tropics. Traditionally, precipitation is observed in situ by rain gauges or retrieved via remote sensing by groundbased weather radars or satellites. Government rain gauge networks (often nationally operated by national meteorological and hydrological services (NMHSs)) can provide accurate local rainfall estimates, but their densities are often too low to capture the spatial rainfall variability. Their estimates are commonly combined with those from weather radars, which do provide wide coverage but are prone to many sources of error. Figure 2.1 provides worldwide maps with estimated coverage by radars and rain gauges and two so-called opportunistic sensors. The estimated 1308 weather radars provide coverage over large parts of the Americas, China, Europe, India, Japan, Oceania, South Africa and South East Asia (Figure 2.1(a)). Coverage over Africa and large parts of Asia is poor or even non-existent. The coverage by rain gauges from NMHSs and other institutes, as obtained from the Global Telecommunication System (GTS), reveals 1 R&D Observations and Data Technology, Royal Netherlands Meteorological Institute, De Bilt, The Netherlands 2 Department of Water Management, Faculty of Civil Engineering and Geosciences, Delft University of Technology, Delft, The Netherlands

28

Advances in weather radar, volume 3

Figure 2.1 Worldwide coverage of ground-based weather radars (a), GTS rain gauges (b), GPCC rain gauges from the Full Data Monthly Product (c), Netatmo personal weather stations (PWSs) (d) and cell phone towers as proxy for commercial microwave links (CMLs) (e). Details on the construction of the maps are provided in Box 2.1. Figure is continued on the next page. Map data ’ 2023 Google. poor coverage compared to radar (Figure 2.1(b)). They constitute an important part of the (near) real-time rain gauges from government institutes. The coverage of rain gauges used in a non-real-time Global Precipitation Climatology Centre (GPCC) product is much higher [1,2] (Figure 2.1(c)).

QPE from weather radars, personal weather stations and CMLs

29

Figure 2.2 European coverage of ground-based weather radars (a), GTS rain gauges (b), GPCC rain gauges from the Full Data Monthly Product (c), Netatmo personal weather stations (PWSs) (d), cell phone towers as proxy for commercial microwave links (CMLs) (e) and rain gauges from the European Climate Assessment and Dataset project [8,9] (ECAD; https://www.ecad.eu). These provide daily timeseries. Locations were obtained mid-June 2022 over the period 1 September 2019 through 31 August 2020 (f). Details on the construction of the maps are provided in Box 2.1. Maps made with Natural Earth. Free vector and raster map data ’ https://naturalearthdata.com.

Box 2.1:

Construction of global coverage maps

The derivation of worldwide coverage maps is not straightforward. Here, the underlying databases of locations and the chosen settings are discussed. The physical footprint for radars is large and plotted in Figure 2.3. In contrast, the physical footprints for rain gauges and commercial microwave links (CMLs) are much smaller, but due to spatial rainfall correlation, their effective coverage is larger. For rain gauges and CMLs, the plotted coverage on a world map is much larger than the physical footprint or even effective coverage to enhance visibility.

30

Advances in weather radar, volume 3

Radars Most radar locations were obtained in February 2023 from a World Meteorological Organization (WMO) database (https://wrd.mgm.gov.tr/Home/ Wrd). Radars (RAdio Detection And Ranging) were assumed to have a 200-km range, independent of the used radio frequency band. Note that the large majority of radars operate at C- or S-band and will at least meet this range. Research radars were not considered. Apart from the operational ones, also the partly operational (6), planned (16) and pre-operational (4) ones were selected. In total 1037 radar locations were obtained in this manner. On top of this the following radar locations have been added: ●

●

●

●

●

●

●

Belgium (one radar): location of radar operated by Flanders Environment Agency obtained from a radar file from July 2021. Radar is currently operational (June 2023). China (197 radars): obtained from https://www.rainviewer.com/radars/ china.html in May 2023. A clear resemblance is found with the locations presented in Minda and Nakamura [3], who report 217 radars. India (nine radars): found by employing https://mausam.imd.gov.in/ in June 2023 (sometimes coordinates are stated in maps), and only selecting those that provide recent radar images for that location. For six out of nine locations, the presence of a radar was confirmed by a Google Maps satellite image (and for three out of these six locations confirmed by the exact location as given in radar images). The other three locations were found by consulting the contents of the downloaded web page, which usually seems to provide approximate locations. Note that for two radar locations, the coordinates from the WMO radar database were replaced by coordinates obtained from a Google Maps satellite image, because one radar was located in the Bay of Bengal and one radar had the same coordinates as another radar. Latvia (one radar): live radar image shown at https://videscentrs.lvgmc. lv/karte/meteorologiska-radara-informacija in May 2023. Found in the WMO database (status non-reporting). Exact location cannot be found by using Google Maps. Lithuania (two radars): live radar image shown at http://www.meteo.lt/ en/radar-information. Locations obtained from https://www.rainviewer. com/radars/lithuania.html in May 2023 and confirmed by a visible radar radome on a Google Maps satellite image. Mongolia (one radar): live radar image shown at http://tsag-agaar.gov. mn/ in May 2023. Exact location cannot be found by using Google Maps. Location taken from https://www.rainviewer.com/radars/mongolia.html. Morocco (seven radars): found in the WMO database as stand-by radars, but operational according to https://www.marocmeteo.ma/fr/observation.

QPE from weather radars, personal weather stations and CMLs

●

●

●

●

●

●

●

●

●

Myanmar (two radars): found by employing https://www.moezala.gov. mm/radar-image-1 in February 2023 and confirmed by a visible radar radome on a Google Maps satellite image. Nepal (one radar). Old radar image shown at https://www.dhm.gov.np/ mfd/radar in June 2023. Approximate radar location provided by Talchabhadel et al. [4]. Exact location found by using Google Maps satellite image. Radar is currently out of service and needs to be repaired or replaced. New Caledonia (three radars and one additional radar, but this one is currently not available and has not been taken into account): live radar image shown at https://www.meteo.nc/nouvelle-caledonie/observations/ images-radars in May 2023. Exact location cannot be found by using Google Maps. Locations taken from https://www.rainviewer.com/radars/ new-caledonia.html. Locations could not be confirmed by Google Maps. Nigeria (four radars): obtained from https://www.google.com/maps/d/ viewer?mid=1GIb9XpFV1dXq-ISwDeI9hYIJ0mu4QwNc&ll=9.139781 073721903\%2C7.63924831988442&z=7 in February 2023. Porto Santo Island (one radar; Portuguese archipelago of Madeira): approximate location obtained from https://www.ipma.pt/en/otempo/ obs.radar/index.jsp in May 2023, which shows live radar images. Exact location found by using Google Maps satellite image, which reveals a visible radar radome. Re´union Island (one radar; France overseas department): approximate location obtained from https://meteofrance.re/fr/images-radar in May 2023, which shows live radar images. Exact location found by using Google Maps satellite image, which reveals a visible radar radome. Republic of Fiji (three radars): approximate locations obtained by combining https://www.met.gov.fj/index.php?page=radar (Fiji Meteorological Service), which shows live radar images (May 2023), and Google Maps. Saudi Arabia (11 radars): found by employing https://www.rainviewer. com/radars/saudi-arabia.html in May 2023, and only selecting those that agree with the rough locations shown on official radar reflectivity composites on the website of the National Center for Meteorology (NCM) of Saudi Arabia (https://ncm.gov.sa/Ar/Weather/RegionWeather/Pages/ KSA-IRIS.aspx; not accessible outside Saudi Arabia). For 10 out of 11 locations, the presence of a radar was confirmed by a Google Maps satellite image. Note that for the only radar location which was already obtained from the WMO radar database, the coordinates were replaced by coordinates obtained from a Google Maps satellite image. The Philippines (20 radars): data source (February 2023): Climate and Agrometeorological Data Section (CADS), Philippine Atmospheric, Geophysical and Astronomical Services Administration (PAGASA).

31

32

Advances in weather radar, volume 3

●

Vietnam (six radars): found by employing https://www.rainviewer.com/ radars/vietnam.html in February 2023 and only selecting those confirmed by a visible radar radome on a Google Maps satellite image.

Some radars may not be operational or temporarily out of service due to maintenance; other radar locations may still be missed. Likely, the radar coverage is underestimated. GTS gauges Because of incomplete coverage, gauge locations were not obtained from the official WMO OSCAR database (https://oscar.wmo.int/surface/#/search/ station). The locations of 11,076 rain gauges contributing to the GTS, as used in the near real-time operational GPCC Monitoring Product [2], were obtained from January 2023. These represent an important part of the automatic rain gauges from government agencies providing (near) real-time data. GPCC gauges The locations of 32,036 rain gauges were obtained as used in the GPCC Full Data Monthly Product from January 2012 (V.2022) [2]. The large majority of stations are expected to be still operational in 2023, but due to latency of rain gauge data becoming available, the locations from 2012 were taken. Most stations typically provide daily accumulations once a day. PWSs The Netatmo rain gauge locations were downloaded in February 2023 using the Netatmo API (https://dev.netatmo.com/apidocumentation/weather). World wide, the number of automatic rain gauges at personal weather stations (PWSs) from all brands is expected to be at least a few hundred thousand. CMLs The cell phone tower locations were obtained from the OpenCelliD Project (https://opencellid.org/) after registration, which is licenced under a Creative Commons Attribution-ShareAlike 4.0 International License (https://creative commons.org/licenses/by-sa/4.0/). OpenCelliD estimates cell phone tower locations by averaging locations of smartphones that receive radio signals from a specific cell phone tower. The smartphone locations are retrieved from a small number of apps. The following selections were performed, resulting in
34 million cell phone tower locations: ●

● ●

Locations for which the approximate area within which the cell could be is smaller than 10 km. Locations that have been estimated from at least two samples. Locations that were last recorded after 1 January 2015.

So-called macro and small cell backhaul links between cell phone towers are estimated at 9.7 million in 2021. They consist of copper cables (2%), satellite links (2%), glass fibre-optic cables (31%) and CMLs (65%). The estimated 4.6 million CMLs (47%) operating in the 6–56 GHz range are the most useful

QPE from weather radars, personal weather stations and CMLs

33

for rainfall estimation. Their number is expected to increase to 6 million in 2027, although their relative share will decrease (35%), mainly due to an increase in glass fibre-optic cables (42%) and partly due to higher frequency CMLs [5]. The large CML backhaul share underpins that cell phone tower locations can serve as a proxy for CML locations. Nevertheless, note that in some regions, the actual coverage of the backhaul CML network will be much lower due to the use of glass fibre-optic cables. High rain gauge coverage by government agencies is found in Europe and parts of North America and Oceania, and low(er) coverage in South America, Africa and parts of Asia, consistent with [6,7]. The total number of rain gauges from operational meteorological networks is estimated to be at most 0.25 million (often daily manual rain gauges), while the number of those giving useful near real-time observations is estimated to be 8,000–12,000 [7]. The coverage by radars and rain gauges is insufficient for large parts of the Earth’s land surface. Satellites provide the widest coverage and are the only means of precipitation information over vast areas of the Earth’s land surface and oceans. The added value of satellite QPE will be limited for areas with good radar and rain gauge coverage. Precipitation is retrieved indirectly via visible and infrared channels from geostationary satellites at typically
3 km resolution every 15 min, or more directly via radiometers on a constellation of low orbiting satellites. They provide higher accuracy but, in contrast to geostationary satellites with their continuous coverage, have longer revisit times of approximately a few hours for a constellation of satellites (but 24 h or longer for an individual satellite) and lower spatial resolution. Higher accuracy can be achieved by the space-borne precipitation radar aboard the Global Precipitation Measurement (GPM) mission’s core satellite (Chapter 5 of Volume 1; the CloudSat satellite cloud radar and CubeSat satellites being the only other existing space-borne precipitation radars). It has a
5 km spatial resolution, but its revisit time is only
1 day for a given location. Hence, across large parts of the Earth’s continents, there is a need for supplementary precipitation information. This can be obtained from opportunistic sensors, often not even designed for precipitation retrieval. The two most mature techniques for opportunistic sensing of precipitation are personal weather stations (PWSs) and commercial microwave links (CMLs). The PWS or third-party rain gauges are Internet of Things (IoT) devices, which are typically purchased and operated by weather enthusiasts. Figure 2.1(d) shows that the coverage of one brand of these crowdsourced rain gauges, from private company Netatmo, is mainly limited to Europe, the United States of America, Japan, and parts of Australia and Canada. There, the network densities are generally much higher than those from government rain gauges. This is illustrated for Europe in Figure 2.2(b)–(d) and (f), where also locations from the government rain gauges from the European Climate Assessment and Dataset project (ECAD) [8,9] (https://www.ecad.eu) are visualised. Because differences in network densities in world maps are difficult to

34

Advances in weather radar, volume 3

discern due to the size of the plotted points, these zoomed in maps for Europe are given for all sensors (Figure 2.2). An overview of (sub-daily) rain gauge datasets for Europe is provided by Van der Schrier et al. [10]. PWSs are usually found in areas with radar coverage and often lack a heating device, limiting their applicability to liquid precipitation (rainfall) estimation. CMLs are radio connections used globally in cellular telecommunication networks. The rain-induced signal attenuation between the transmitting antenna at one base station (or cell phone tower) and the receiving antenna at another base station can be computed and converted to averaged rainfall intensity over a path of typically a few hundred metres up to a few tens of kilometres. Figure 2.1(e) shows the locations of cell phone towers, which are used here as a proxy for CML locations. Worldwide
5 million CMLs are used in telecommunication networks [11]. Their largest potential for rainfall estimation lies in areas which are mostly not covered by the other ground-based sensors, typically low- and middle-income countries (LMICs) in (sub)tropical regions, but for which the required cellular communication infrastructure is already in place. This is especially the case for Africa and for parts of Asia and South America (Figure 2.1(e)). For QPE, not only knowledge of the sensors is important, but also knowledge of the observed hydrometeors. This concerns both physical and statistical aspects (coverage and correlation in time and space). Rainfall consists of a collection of drops with various sizes, shapes and fall speeds and can be described by a socalled drop size distribution (Chapter 6 of Volume 2), the shape of which is determined by coalescence and break-up (and potentially evaporation) of drops [12]. Precipitation types, their size distribution and fall velocities can be observed with disdrometers, making them important sensors to derive parameter values for precipitation retrieval relationships (Chapter 6 of Volume 2). Disdrometers are not addressed in this chapter because of their much lower density compared to government rain gauges. The ability to approximate the actual rainfall accumulation at the Earth’s surface is determined by many factors and varies from sensor to sensor, each having its own sampling volume or spatial support. Radars provide 3D volumetric data, whereas CMLs provide path-averaged (line segment) data and PWSs provide point observations. Typically, these sensor data are interpolated to 2D gridded rainfall fields. Figure 2.3 illustrates the sensors’ footprints, the typical conversions and the interpolation. QPE is indirect for remote sensing devices, because it depends on the conversion from remote observables aloft to rain rates at the ground. It is governed by the interplay of electromagnetic radiation and hydrometeors: received signal power (radar reflectivity factor Z) or specific differential phase (Kdp ; Chapter 5 of Volume 2) for radars, and specific attenuation (k) for CMLs. For rain gauges, QPE is direct by receiving the precipitation falling at the Earth’s surface. Still, environmental conditions (notably wind effects), device set-up and sensor accuracy can lead to deviations with respect to the actual rainfall. Table 2.1 provides a summary of the characteristics of QPE with radars, rain gauges, PWSs, CMLs and satellites. Other overviews are provided by Barthe`s and Mallet [13] and Priebe [14]. The quality of radar QPE depends on the applied processing, but adjustment with rain gauge data remains usually desirable.

QPE from weather radars, personal weather stations and CMLs Validation dataset

DSD

35

Derived datasets

DSD

DSD

DSD

DSD Z

R k

Z

R R

R

R Ground truth

R

R Radar

CML

PWS

Figure 2.3 From drop size distributions (DSDs) to derived rainfall fields for radars, CMLs and PWSs. Figure taken from [15].

Table 2.1 Overview of sensors for rainfall estimation at the Earth’s surface and their typical temporal and spatial resolutions, coverage and cost (related to installation and maintenance, not to quality control, rainfall retrieval and product dissemination). Sensor

Temporal resolution

Sampling area

Coverage

Cost

Radars (X-band)a

Seconds to 1 min 5–15 min

A few tens of m1 0.4 km1 /1–2 km A few dm2

2,000–30,000 km2 30,000– 650,000 km2 Local

Medium

A few dm2 0.1–20 km 1 – a few tens of km2 A few tens–hundreds of km2 5 km

Local Low Local–regional Low World Very high

Radars (C- and S-band)a Government rain gauges PWS rain gauges CMLsb Geostationary satellites Constellation of satellite radiometersc Space-borne radarsc

1 min to 24 h 5 min 10 s to 15 min 5–15 min A few hours 24 h to days

High Medium

Snapshot

Very high

Snapshot

Very high

a Their effective spatial resolution may be lower, coverage depends on the used wavelength (X-, C- or S-band), and the coverage for which accurate QPE is obtained may be much lower. Costs per km2 are not that high. b Provide rainfall estimates over a narrow path. Costs are low because the infrastructure is already in place. c Only provide a snapshot during their overpass, typically from 60 S to 60 N.

36

Advances in weather radar, volume 3

Government rain gauges provide accurate, but only local, estimates if well maintained. PWSs and CMLs can give reasonable rainfall estimates after careful quality control (QC) or rainfall retrieval, respectively. Satellite QPE has usually poor accuracy, but higher accuracy for space-borne radars. In this chapter, the quality of QPE with radar data, PWS data and CML data is evaluated for the Netherlands, a mid-latitude country with a temperate climate. This will provide insight into the stand-alone performance of opportunistic sensors, where the Netherlands acts as a test bed with relatively accurate reference data. First, the employed datasets are described. Next, three consecutive sections discuss QPE with radars, PWSs and CMLs. Apart from the evaluation in each section, details are provided on each sensor (Boxes 2.2–2.4). For radars, an overview and illustration of sources of error in QPE and of algorithms to improve QPE are provided. For PWSs, factors affecting the quality of rainfall estimates and QC algorithms to remedy these are discussed. For CMLs, the principle of rainfall retrieval and the encountered sources of error are described. Finally, a summary and outlook are provided, focusing on (1) improving radar QPE by merging with CML or PWS rainfall estimates and (2) showcasing the potential of CMLs for rainfall estimation in LMICs.

2.2 Datasets and evaluation metrics Figure 2.4 provides an overview of the locations of the employed radars and government rain gauges from the Royal Netherlands Meteorological Institute (KNMI), PWS gauges from brand Netatmo and CML paths of mobile network operator (MNO) T-Mobile NL. The number of PWS and CML locations, which can potentially provide (near) real-time data, is much higher than the number of KNMI rain gauges. The contrast is even larger when only considering KNMI rain gauges at automatic weather stations (AWSs), which can provide near real-time information. The daily accumulations from KNMI’s manual rain gauge network (
1 gauge/100 km2) are used for evaluation of (merged) radar accumulations and interpolated PWS and CML rainfall maps. Validated and complete AWS and manual KNMI rain gauge records are used (also for the merged radar datasets), which are not available in real-time. The PWS data span the period June–August 2020; the open CML data [16] span the period June–August 2012. These periods are different because no coincident CML and PWS data could be obtained covering the entire Netherlands over the same 3-month period simultaneously. A gauge-adjusted radar precipitation dataset is used to evaluate daily PWS point and CML path-averaged accumulations, interpolated PWS and CML rainfall maps and merged radar datasets. It is also employed to derive a 25-year mean annual precipitation climatology from 1998 to 2022. This climatological radar dataset, with
0.9 km2 spatial resolution (
6 km2 for the 25-year climatology), combines a daily spatial adjustment using the manual rain gauge accumulations with an hourly mean-field bias adjustment employing the AWS accumulations [17,18]. It is denoted in this chapter as ‘radars + gauges’. Also a radar dataset

QPE from weather radars, personal weather stations and CMLs KNMI radar & gauge locations

4ºE

5ºE

6ºE

7ºE

PWS locations

37

CML locations

53ºN

53ºN

53ºN

52ºN

52ºN

52ºN

51ºN

51ºN

51ºN

4ºE

5ºE

6ºE

7ºE

4ºE

5ºE

6ºE

7ºE

Figure 2.4 Map of the Netherlands with locations of 3 KNMI radars, 32 automatic and 320 manual KNMI rain gauges. The number and location of gauge locations are from 30 June 2020 and can slightly vary within and between the three periods in this study. The number of Netatmo personal weather stations (June–August 2020; after applying quality control and accumulation to 1-h rainfall as described in Section 2.4.1) is more than 100 times larger than the number of KNMI automatic rain gauges. For commercial microwave links, data from over 1,500 paths are available (June–August 2012; part of the former network of one of the three MNOs in the Netherlands). Paths are shown for which at least one 15-min rainfall estimate was obtained. Note that data from two radars were used, and that the radar in the middle was replaced by the southernmost radar in January 2017. Map data ’OpenStreetMap contributors 2023. Distributed under the Open Data Commons Open Database License (ODbL) v1.0.

without this gauge adjustment is evaluated (‘radars’) or merged with either AWS, PWS or CML accumulations. Finally, four different (un)adjusted radar datasets are evaluated for the period August 2017 to July 2018 (see Section 2.3.7). Table 2.2 provides an overview of the employed rainfall datasets that are evaluated or used as reference. Several metrics are used to evaluate the performance of the various daily precipitation datasets in scatter density plots. Residuals are computed as the precipitation accumulation from the evaluated dataset minus the reference precipitation accumulation. The relative bias is computed as the average of the residuals divided by the average of the reference, expressed as a percentage. The standard deviation of the residuals is divided by the average precipitation accumulation from the reference to obtain the value for the coefficient of variation (CV). Finally, the squared Pearson correlation coefficient (r2), or coefficient of determination (i.e. the fraction of the variance in the reference that is explained by a linear regression model), is computed between the evaluated dataset and the reference.

38

Advances in weather radar, volume 3

Table 2.2 Overview of the different rainfall datasets that are evaluated or used as reference datasets. Dataset title in figures

Description

Radars Interpolated manual gauges Manual gauges Radars + gauges PWSs PWSs, interpolated values PWSs, point values CMLs CMLs, interpolated values CMLs, path-averaged values Radars + AWSs Radars + PWSs Radars + CMLs

Daily Daily Daily Daily Daily Daily Daily Daily Daily Daily Daily Daily Daily

a

gridded radar dataset gridded manual rain gauge dataset (reference) point accumulations (reference) gridded climatological dataset (reference) interpolated maps at the radar grid interpolated values at manual rain gauge locations in situ valuesa interpolated maps at the radar grid interpolated values at manual rain gauge locations path-averaged valuesb gridded dataset from merged 1-h radar and AWS data gridded dataset from merged 1-h radar and PWS data gridded dataset from merged 1-h radar and CML data

Are compared to grid cell values from ‘radar + gauges’ dataset. Are compared to path-averaged values from ‘radars + gauges’ dataset.

b

2.3 QPE with ground-based weather radars 2.3.1

Signal processing

The chain for QPE with ground-based weather radars starts with signal processing at the radar site. Generally, Doppler filtering is applied to remove the part of the power caused by stationary non-meteorological echoes (Chapter 2 of Volume 1). Horizontally polarised echo powers are converted to radar reflectivity factors by means of the radar equation [25]. In case of dual-polarisation (dual-pol) or polarimetric radars (Chapter 2 of Volume 1), this is also done for vertically polarised echo powers. In addition, dual-pol radars provide phase measurements. Monitoring and calibration in the signal processor are important to detect and correct for hardwarerelated errors, such as antenna pointing offsets, which can seriously affect the quality of QPE (Chapter 2 of Volume 2, [26]).

Box 2.2: Introduction to weather radars Pulsed ground-based weather radars are employed by NMHSs to obtain QPE with large coverage at a spatial resolution of a few km every 5–15 min (Chapters 1 and 2 of Volume 1). This is achieved by a long processing chain, typically spanning from signal processing, application of correction algorithms to 3D radar data, conversion to 2D precipitation fields and adjustment with rain gauge accumulations. These steps are needed not only to convert received signal amplitudes and phases to QPE (Chapter 5 of Volume 2) but also to correct sources of error in radar QPE, many of them indicated in

QPE from weather radars, personal weather stations and CMLs

39

Figure 2.5. Extensive overviews of sources of error in radar QPE and algorithms to address these are provided by Joss and Waldvogel [19], Fabry [20], Rauber and Nesbitt [21] and Ryzhkov and Zrnic [22]. Other studies provide a systematic evaluation of the effect of different correction or retrieval algorithms on radar QPE, e.g., [23,24].

2.3.2 Clutter removal Next, the radar reflectivity factor data from individual scans, together called the 3D or volumetric radar data, can be corrected for sources of error, which are related to environmental conditions and limitations in the scanning and sampling strategy. Some of the most important ones are discussed here. First, remaining non-meteorological echoes, called clutter, can be removed (Chapter 11 of Volume 1). Clutter can be caused by obstacles near a radar site or interference (Chapter 6 of Volume 3; [27]). Clutter can also be a result of reflections by airborne non-meteorological targets, such as aircraft and birds. A common cause of clutter is anomalous propagation: refraction of the beam towards the Earth’s surface in case of specific temperature and/or moisture gradients in the atmosphere. Especially non-stationary targets, such as waves, ships and wind farms, are difficult to remove by Doppler filtering. Clutter can be a serious problem for QPE, since it can lead to large overestimation of precipitation. Often, additional methods are needed to further reduce non-meteorological echoes. Many studies employ fuzzy logic algorithms involving single-polarisation and/or dual-polarisation decision variables [28–32]. Dual-pol radars are especially effective in detecting clutter. In contrast to hydrometeors, non-meteorological scatterers often have an irregular surface, which can be detected by comparing the amplitude and phase between horizontally and vertically polarised backscattered signals provided by dual-pol radars. This can be followed by statistical

Aircraft

Sun Non-uniform vertical profile

Chaff

Attenuation Birds

Propagation conditions

Shielding

Insects

Hail effects Overshooting

Wet radome Ship

Interference

Ground clutter

Calibration Sea clutter Wind turbines

Figure 2.5 Overview of sources of error affecting QPE with ground-based weather radars. Reproduced with permission from Markus Peura (Finnish Meteorological Institute).

40

Advances in weather radar, volume 3

postprocessing, e.g., by creating clutter maps based on unrealistically large or frequent accumulations [33] or by detecting large spatial gradients in radar reflectivity factors and by removing isolated echoes [34]. Also auxiliary data are used to remove clutter, e.g., satellite cloud type [35], or precipitation probability products [36], or a combination of satellite and numerical weather prediction model data [37,38].

2.3.3

Attenuation correction

Precipitation can be severely underestimated due to rain-induced attenuation along the radar beam and wet radome attenuation for X- or C-band radars [39–44], especially in case of heavy, convective precipitation. This is illustrated in Figure 2.6 for two C-band radars. The squall line at 16:45 UTC (Universal Time Coordinated) moves in north-easterly direction over the Netherlands but almost completely disappears at 17:00 UTC, because the radar beams are aligned with the Radar DB

Radar DB

Radar DB

47 39 31 23 15

Rada reflectivity factor (dBZ)

55

7 Radar DH

Radar DH

Radar DH

47 39 31 23 15

Rada reflectivity factor (dBZ)

55

7

Figure 2.6 Example of severe (radome) attenuation. Maps covering the Netherlands and surroundings with the horizontally polarised radar reflectivity factor from the 800-m pseudo-CAPPI from the inland radar in De Bilt (DB; a–c) and the coastal radar in Den Helder (DH; d–f) and for 17 July 2004 16:45 (a,d), 17:00 (b, e) and 17:15 (c, f) UTC. Radar locations are denoted by black circles. Map data ’OpenStreetMap contributors 2023. Distributed under the Open Data Commons Open Database License (ODbL) v1.0.

QPE from weather radars, personal weather stations and CMLs

41

squall line. For one radar, it reappears at 17:15 UTC (Figure 2.6(a)–(c)), because the radar now looks sideways to the squall line, i.e., the total path of two-way propagation through severe rain is much reduced. Given the fairly high radar reflectivity factor values above the radars, wet radome attenuation likely contributes to reducing the received radar reflectivity factor values. Algorithms can be applied to estimate the two-way path-integrated attenuation along the radar beam and correct for this. Hitschfeld–Bordan type algorithms convert Z or R to specific attenuation k using a power–law. Next, for each azimuth, k is integrated over the range of the beam. Hence, the attenuation is computed for each range bin and added to the radar reflectivity factor to correct for rain-induced attenuation in an iterative procedure [39,43,45]. To avoid numerical instabilities that can occur when applying Hitschfeld–Bordan type algorithms, constraints can be applied for a maximum allowed correction [43]. Alternatively, a similar algorithm can be applied, but in the reverse direction, using the estimated path-integrated attenuation to a fixed target as an additional constraint [45,46]. For dual-pol radars, the specific differential phase (Kdp) can be used as an independent variable to compute the path-integrated attenuation [47]. In that case, k is computed from Kdp using, e.g., a linear relationship. This can improve QPE with respect to a Hitschfeld–Bordan type algorithm [44]. Wet radome attenuation correction has received little attention in the scientific literature, e.g., [48–50].

2.3.4 Vertical profile of reflectivity correction Shallow precipitation and orographically enhanced precipitation at longer ranges from the radar can be (partly) missed due to the increased height of the radar sampling volume (‘overshooting’). Moreover, stratiform precipitation starts as (dry) snowflakes and ice crystals resulting in low reflectivities higher up, via a layer with higher reflectivities due to melting snow (‘the bright band’, on radar reflectivity or precipitation images often visible as circles with increased values around the radar site), to moderate reflectivities due to rain below the melting layer. The vertical profile of reflectivity can be estimated close to the radar site using data from multiple elevation scans and can subsequently be used to correct for the low reflectivity at long ranges from the radar [51].

2.3.5 Conversion to rainfall intensity At the core of radar QPE is the conversion of the horizontally polarised radar reflectivity factor Z to rainfall intensity R (Chapter 6 of Volume 2; [12]). The most widely used relation is given by Z ¼ 200R1:6 , which was obtained from raindrop size distribution observations at the ground [52], and is often used in temperate climates. The values of the coefficients may vary considerably between rainfall types, implying that a fixed ZR relationship introduces (large) uncertainties [53]. Hence, sometimes different coefficients are employed depending on the value of Z. Dual-pol radars provide opportunities to improve QPE (Chapter 5 of Volume 2). For instance, in case of moderate or stronger rainfall, the dual-pol variable specific differential phase (Kdp ) can be employed for rainfall estimation. Advantages are that it is immune

42

Advances in weather radar, volume 3

for hardware calibration errors, (radome) attenuation and partial beam blockage, and that the Kdp R relationship introduces less uncertainty because it is not as non-linear as the ZR relationship. It can only be applied in rain, however, requiring a hydrometeor classification algorithm or temperature information from weather models to determine whether Kdp can be used for QPE or not. Applying the ZR relationship assumes that precipitation is liquid. Although rainfall is the dominant precipitation type at the Earth’s surface for most areas covered by radars, aloft the radar often measures the backscatter by solid precipitation or mixtures of precipitation. Applying a ZR relation for liquid precipitation then results in an underestimation of surface rainfall (except for hail). Here, hydrometeor classification algorithms [54–56] can help to select appropriate relationships to convert either Z, Z and the differential reflectivity Zdr (Chapter 5 of Volume 2), or Kdp , to rainfall intensity or to even estimate the surface snowfall intensity.

2.3.6

Radar-gauge adjustment

Typically, data from different radars are combined in 2D reflectivity or precipitation fields to obtain better estimates. A common practice is to complete the radar processing chain by adjusting 2D radar precipitation fields using near real-time or historical rain gauge accumulations [59–65]. This entails the computation of an adjustment factor for each radar grid cell (pixel) and applying this factor to the unadjusted radar data. This can be a constant factor over the entire radar domain, a so-called mean-field bias adjustment, or a spatial adjustment factor which can vary from grid cell to grid cell. Adjustment is called merging in case an adjustment factor field is computed using rain gauge and radar data from the same, typically 60 min or 24 h, interval and subsequently applied to the radar data from that interval. When no recent rain gauge data are available, an adjustment factor field can be computed based on historical radar and gauge data and applied to radar data from the present time interval [64,66]. Typically, for each rain gauge location, the ratio of gauge and radar accumulations is computed. For a given radar grid cell, the ratios from surrounding gauge locations are interpolated to obtain an adjustment factor value, where the ratios from the nearest gauge locations get the highest weight. Alternatively, first the radar and gauge values at rain gauge locations are interpolated separately through distance-weighting. The ratio of these values for a given radar grid cell gives an adjustment factor value. An overview of several common gauge adjustment methods is provided by Goudenhoofdt and Delobbe [67]. A practical limitation for merging rain gauge data in real-time radar products is the latency of rain gauge measurements [66]. Gauge-adjusted radar precipitation products are, e.g., highly relevant for radar hydrology (Chapter 1 of Volume 3). An example of a merged radar-gauge dataset is the 25-year climatology of annual precipitation for the Netherlands presented in Figure 2.7, which is based on the methodology by Overeem et al. [17]. Such climatological radar precipitation datasets [68] also exist for, e.g., Germany [69], Europe [35] and the United States of America [70]. For the unadjusted radar dataset, algorithms to improve QPE are mainly limited to Doppler clutter filtering and some statistical clutter removal algorithms. The highest accumulations are caused by clutter resulting from anomalous propagation

QPE from weather radars, personal weather stations and CMLs Interpolated manual gauges

Radars

43

Radars + gauges

925 850 775 700 625 550 475

Mean annual precip (mm)

1,000

400

Figure 2.7 Maps of mean annual precipitation for the Netherlands for the period 1998–2022, based on interpolated rain gauge data [57] (a), unadjusted radar data (b) and gauge-adjusted radar data [58] (c). Availability is 91.0% for the radar datasets, which is taken into account when computing the mean annual precipitation (map data ’OpenStreetMap contributors 2023. Distributed under the Open Data Commons Open Database License (ODbL) v1.0). and occur in an area part of the Port of Rotterdam, where many metal cranes and containers are located. The huge underestimation (Figure 2.7(b)) with respect to the interpolated rain gauge dataset (Figure 2.7(a)) is effectively removed by merging with data from KNMI’s rain gauge networks (Figure 2.7(c)). This provides a much more detailed perspective on local precipitation, confirming the added value of gaugeadjusted radar accumulations for climate monitoring. For some azimuths pointing away from radar locations, clearly lower accumulations are found. These can be related to obstacles, in this case mostly buildings or orography, near radar sites that partly block the beam of the lowest elevation scan. Algorithms are available to recognise and correct for beam-blockage affected areas using digital elevation models, by employing specific differential phase or by employing specific attenuation [71–73], and can be applied when 3D radar data are available.

2.3.7 Evaluation Now a more quantitative assessment of the quality that can be expected from (gauge-adjusted) radar precipitation products is provided in Figure 2.8. Daily radar rainfall accumulations are evaluated against KNMI’s manual rain gauge network over a period of almost 1 year. Four different datasets are evaluated as follows: ●

●

A dataset obtained with similar methods as the unadjusted radar dataset from Figure 2.7 (‘no radar corrections’). Note that data from the two radars are averaged in logarithmic space, to limit the effect of clutter [74]. A dataset which has undergone clutter removal with a fuzzy logic algorithm [74], attenuation correction via Kdp [44], vertical profile of reflectivity correction [51], quality-based conversion of 3D to 2D data per radar, clutter removal by a statistical filter [34], quality-based compositing of data from the

44

Advances in weather radar, volume 3 two radars, and advection correction to account for fast moving precipitation (‘radar corrections’). Data from the two radars are averaged in linear space, which can be done because of the effective clutter removal. This reduces underestimation compared to averaging in logarithmic space [74]. For both datasets also a spatial gauge adjustment is performed using 1-h accumulations of KNMI’s automatic rain gauge network. Given limitations in latency and update frequency of automatic rain gauge data, such an adjustment is not always possible in real-time.

●

A number of conclusions can be drawn from Figure 2.8. First, the average underestimation is severe without gauge adjustment, even when radar corrections are applied. This underlines that more radar corrections are needed and that gauge adjustment is still desirable for operational applications. Second, the gauge adjustment using only
1 gauge/1,000 km2 is more effective in improving QPE than radar corrections alone. Third, radar corrections further improve the gauge-adjusted radar accumulations. To conclude, the effect of radar corrections is relatively limited but noticeable.

(a)

(b)

(c)

(d)

Figure 2.8 Evaluation of four radar datasets with
0.9 km2 grid cells against daily manual rain gauge point accumulations over
1 year. For unadjusted (a, b) and adjusted (c, d) radar data. For this comparison, the radar grid cell at each gauge location is employed.

QPE from weather radars, personal weather stations and CMLs

45

Note that Figure 2.8 only provides a general evaluation. The outcome could be different when considering hourly accumulations, other evaluation metrics or specific events. For instance, annual precipitation maps (not shown) reveal that the radar corrections lead to more realistic, substantially higher accumulations at long range from the radars, where no KNMI rain gauges are present. Figure 2.8 gives an impression of the quality of (gauge-adjusted) radar QPE one may encounter. The actual performance may be lower or higher for other radar networks, depending on, e.g., environmental conditions, hardware (calibration), applied (additional) radar corrections and density of employed rain gauge networks. Finally, representativeness errors explain part of the variability between radar and gauge accumulations [75]. Radars measure aloft and sampling volumes are huge compared to the typical 2–4 dm2 sampling area of rain gauges, which measure at the Earth’s surface. Because of representativeness errors, the minimum accumulation period for comparing or merging rainfall estimates from radar and rain gauges is usually assumed to be 60 min. This reduces the impact of rainfall variability.

2.4 QPE with personal weather stations 2.4.1 Quality control It is expected that rain gauge data from PWSs are more prone to errors due to placement at sites with more interference from the surroundings, fewer and lessspecialised maintenance to the sensor and lower quality devices compared to AWSs. QC methods have been developed to recognise typical errors in this data source and are commonly applied to PWS rain gauge data. These algorithms are sometimes available as open-source algorithms [82,84,86–88]. Often, inter- and/or intra-station checks are performed, and sometimes auxiliary data from, e.g., government rain gauges [82] or radar data [89] are used. Inter-station filters make use of the fact that rainfall is correlated in space [90]. Applying QC can substantially decrease the number of observations. Here, previously developed QC [84,91] is applied to 6 months of Netatmo PWS data from the Netherlands (the data from March to May 2020 are used as spin-up time). After obtaining precipitation accumulations at regular 5-min intervals, a list of all nearest PWSs (up to a maximum of 20) within a 10-km radius is compiled for each PWS, which feeds into the inter-station QC.

Box 2.3:

Introduction to PWSs

PWSs are typically purchased by weather enthusiasts or citizen scientists and are often located in densely populated areas. PWSs provide observations in urban areas where AWSs are relatively scarce due to the difficulty to meet WMO requirements. Common PWSs are produced by companies such as Alecto, Davis and Netatmo. The Netatmo PWS consists of an indoor and outdoor module, with the possibility to extend with an additional outdoor module for rain and an outdoor module for wind [76]. The rain module is of the tipping bucket type and the observations are sent to the Netatmo platform every
5 min and are visualised on

46

Advances in weather radar, volume 3

the Netatmo Weathermap (https://weathermap.netatmo.com/). Several other platforms exist to which data from a variety of PWSs are uploaded, such as the Weather Observations Website (free access; https://wow.metoffice.gov.uk/; https://wow.knmi.nl/) [77–79], the Weather Underground website (https://www. wunderground.com/wundermap) [78] and the Community Collaborative Rain, Hail and Snow Network (https://www.cocorahs.org/) [80]. The quality of PWS sensors, the set-up in relation to the often urban environment, metadata, connectivity, data transfer, power supply, calibration and maintenance are of vital importance for rainfall estimation [81–83]. Typical sources of error in (PWS) rain gauge estimation are discussed by De Vos et al. [81,84] and Os´ro´dka et al. [85] and references therein. The algorithm distinguishes [84] ●

●

●

●

faulty zero errors caused by malfunctioning of the gauge (e.g. a failing tipping bucket mechanism); high influx errors, which are suspiciously large accumulations that may be caused by, e.g., garden sprinklers and tilting of the gauge while handling the device or signal processing errors (e.g. values are added upon previous observations); station outlier errors where observed rainfall dynamics do not correspond with those at nearby PWSs; bias due to systematic underestimation or overestimation that may be caused by wind or a deviating tipping bucket volume, either physically or due to faulty calibration.

The faulty zeroes, high influx, and station outlier filters and dynamically updated bias correction factors are computed and applied to the 5-min data. Given the spin-up time needed for the station outlier filter and bias correction, only QPE from the most recent 3 months is evaluated. Note that the bias correction starts with a default bias correction factor (1.09), which has been obtained from a one-of comparison to gaugeadjusted radar data from September 2019 [92]. Apart from this default bias correction factor, the default parameters have been obtained from De Vos et al. [84], who employ an older Netatmo rain gauge dataset from the Amsterdam metropolitan area, the Netherlands. For each PWS, the QC’ed 5-min accumulations are aggregated to hourly and daily point accumulations in case the availability is at least 83.3%. The hourly point accumulations are interpolated at the
0.9 km2 radar grid using ordinary kriging with a spherical variogram model [93] employing the RAINLINK software package and its default parameter values [94]. Climatological variogram parameters are used, which vary as a function of day of year, and which are based on a 30-year rain gauge dataset [90]. The 1-h PWS rainfall maps are aggregated to 24-h and 3-monthly rainfall maps.

2.4.2

Evaluation

Interpolated PWS 3-monthly rainfall maps are compared to gauge-adjusted radar precipitation maps (Figure 2.9(a) and (b)). The spatial patterns and values of

QPE from weather radars, personal weather stations and CMLs

47

3-monthly PWS summer precipitation match well with the gauge-adjusted radar dataset. As expected, more detailed patterns are visible in the radar precipitation map, and local extremes are missed by the PWS network, which may be partly caused by the limited PWS coverage. No clear outliers are visible for the PWSs. Interpolated daily PWS rainfall maps and daily PWS point rainfall estimates are validated against KNMI rain gauge point accumulations and gauge-adjusted radar accumulations, respectively (Figure 2.9(c) and (d)). The scatter density plots show

Figure 2.9 Comparison of 3-monthly interpolated PWS precipitation accumulation map (a) against gauge-adjusted radar accumulation map [95] (b) at
0.9 km2 for the period 31 May 2020 08:00 UTC to 31 Aug 2020 08:00 UTC. Evaluation of interpolated daily PWS rainfall accumulations against manual rain gauge point accumulations over the same period (c), and evaluation of PWS point accumulations against gauge-adjusted radar accumulations with
0.9 km2 grid cells [96] over the same period (d). For the PWS point accumulation comparison, the radar grid cell at each PWS location is employed. Map data ’OpenStreetMap contributors 2023. Distributed under the Open Data Commons Open Database License (ODbL) v1.0.

48

Advances in weather radar, volume 3

that the relative bias is within 10%, the dispersion is relatively low (CV64 mm

100–1000 km, 10–100 km
3–20 km,
1–10 km,
1–3 km
0.1–1 km

Day to month Day Tens of minutes, few minutes, seconds to minutes Tens of seconds

Lapilli Blocks

Source: Table adapted from [14].

134

Advances in weather radar, volume 3

residence periods in the atmosphere and the distance from the vent that volcanic debris can reach [14,25]. An important measure of the explosiveness of volcanic eruptions is the Volcanic Explosivity Index (VEI) [26]. The VEI is related to the volume of pyroclastic deposit, to the height of volcanic clouds and to some qualitative volcanological observations. The VEI ranges from 0 to 8 in a logarithmic scale and often correlates to eruptive types ranging from low intensity and mainly effusive activity (VEI=0 – Hawaiian) to Plinian-Ultraplinian eruptions (VEI=from 5 to 8).

4.2.2

Tephra size and composition

The concentration and diameter of particles in volcanic clouds decrease with distance from the vent since larger particles tend to fall quickly. Tephra is generally composed of very fine-grained fragments (