Environmental Sampling for Hazardous Wastes 9780841208841, 9780841210950, 0-8412-0884-0

Citation preview

Publication Date: October 31, 1984 | doi: 10.1021/bk-1984-0267.fw001

Environmental Sampling for Hazardous Wastes

Publication Date: October 31, 1984 | doi: 10.1021/bk-1984-0267.fw001

ACS SYMPOSIUM SERIES 267

Environmental Sampling for Hazardous Wastes Publication Date: October 31, 1984 | doi: 10.1021/bk-1984-0267.fw001

Glenn E. Schweitzer,

EDITOR

U.S. Environmental Protection Agency

John A. Santolucito,

EDITOR

U.S. Environmental Protection Agency

Based on a workshop sponsored by the Committee on Environmental Improvement of the American Chemical Society, the U.S. Environmental Protection Agency, and the University of Nevada—Las Vegas, Las Vegas, Nevada, February 1-3, 1984

American Chemical Society, Washington, D.C. 1984

Library of Congress Cataloging-in-Publication Data Environmental sampling for hazardous wastes. (ACS symposium series; ISSN 0097-6156; 267)

Publication Date: October 31, 1984 | doi: 10.1021/bk-1984-0267.fw001

Includes bibliographies and indexes. I. Environmental monitoring—Congresses. 2. Hazardous wastes—Congresses. I. Schweitzer, Glenn E., 1930. II. Santolucito, John A. III. American Chemical Society. Committee on Environmental Improvement. IV. United States. Environmental Protection Agency. V. University of Nevada, Las Vegas. VI. Series. TDI93.E58 1984 ISBN 0-8412-0884-0

363.7'28

84-20480

Copyright © 1984 American Chemical Society All Rights Reserved. The appearance of the code at the bottom of the first page of each chapter in this volume indicates the copyright owner's consent that reprographic copies of the chapter may be made for personal or internal use or for the personal or internal use of specific clients. This consent is given on the condition, however, that the copier pay the stated per copy fee through the Copyright Clearance Center, Inc., 27 Congress Street, Salem, M A 01970, for copying beyond that permitted by Sections 107 or 108 of the U.S. Copyright Law. This consent does not extend to copying or transmission by any means—graphic or electronic— for any other purpose, such as for general distribution, for advertising or promotional purposes, for creating a new collective work, for resale, or for information storage and retrieval systems. The copying fee for each chapter is indicated in the code at the bottom of the first page of the chapter. The citation of trade names and/or names of manufacturers in this publication is not to be construed as an endorsement or as approval by ACS of the commercial products or services referenced herein; nor should the mere reference herein to any drawing, specification, chemical process, or other data be regarded as a license or as a conveyance of any right or permission, to the holder, reader, or any other person or corporation, to manufacture, reproduce, use, or sell any patented invention or copyrighted work that may in any way be related thereto. Registered names, trademarks, etc., used in this publication, even without specific indication thereof, are not to be considered unprotected by law. PRINTED IN THE UNITED STATES OF AMERICA Second printing 1986

ACS Symposium Series Publication Date: October 31, 1984 | doi: 10.1021/bk-1984-0267.fw001

M. Joan Comstock, Series Editor Advisory Board Robert Baker U.S. Geological Survey

Geoffrey D. Parfitt Carnegie-Mellon University

Martin L . Gorbaty Exxon Research and Engineering Co.

Theodore Provder Glidden Coatings and Resins

Herbert D. Kaesz University of California- Los Angeles

James C . Randall Phillips Petroleum Company

Rudolph J. Marcus Office of Naval Research

Charles N . Satterfield Massachusetts Institute of Technology

Marvin Margoshes Technicon Instruments Corporation

Dennis Schuetzle Ford Motor Company Research Laboratory

Donald E . Moreland USDA, Agricultural Research Service W. H. Norton J. T. Baker Chemical Company Robert Ory USDA, Southern Regional Research Center

Davis L . Temple, Jr. Mead Johnson Charles S. Tuesday General Motors Research Laboratory C. Grant Willson IBM Research Department

Publication Date: October 31, 1984 | doi: 10.1021/bk-1984-0267.fw001

FOREWORD The ACS SYMPOSIUM SERIES was founded in 1974 to provide

a medium for publishing symposia quickly in book form. The format of the Series parallels that of the continuing ADVANCES IN CHEMISTRY SERIES except that in order to save time the papers are not typeset but are reproduced as they are submitted by the authors in camera-ready form. Papers are reviewed under the supervision of the Editors with the assistance of the Series Advisory Board and are selected to maintain the integrity of the symposia; however, verbatim reproductions of previously published papers are not accepted. Both reviews and reports of research are acceptable since symposia may embrace both types of presentation.

PREFACE

Publication Date: October 31, 1984 | doi: 10.1021/bk-1984-0267.pr001

THE IMPLEMENTATION

of the Resource Conservation and Recovery Act (RCRA) and the Comprehensive Environmental Response, Compensation, and Liability Act (Superfund) has underscored a number of the weaknesses in our capabilities to measure the chemical characteristics of wastes. We are now being called upon to identify and quantify with unprecedented sensitivity hundreds of chemicals found in many types of materials within waste sites, near discharges of hazardous contaminants, and in the surrounding environments. Extrapolations from a limited number of measurements must indicate the general environmental conditions near waste sites. The measurements have to be made faster and cheaper than ever before, with the precision and bias of each measurement fully documented. Thus, the technical challenges facing the monitoring community are substantial. The progress to date in responding to these challenges has been impressive. Many governmental, industrial, and academic laboratories have become equipped with a new generation of computer-based instruments, and they bear little resemblance to the laboratories of a decade ago. Field monitoring activities are becoming far more sophisticated in design and in implementation. Remote sensing tools guide our sampling efforts, and computer models help interpret our data. But we have only begun to exploit the promise of technology to penetrate the earth, to discriminate among molecular structures, and to allow us to choose those few samples that will adequately represent the whole. A great deal of operational monitoring activity is underway throughout the nation while the regulatory basis for this activity continues to expand. Many of the monitoring requirements and the associated research needs to support these efforts are clear although new technical issues are constantly emerging as more practical problems unfold. The monitoring responsibility rests with the U.S. Environmental Protection Agency, the states, and the operators of disposal sites. However, the necessary research can only be accomplished through the concerted efforts of a far larger number of organizations. Perhaps the most neglected aspect in our haste to advance rapidly in assessing and cleaning up waste sites has been the technical basis for the design of environmental sampling programs. The costs of analyzing the large numbers of samples being collected is now causing more careful consideraix

Publication Date: October 31, 1984 | doi: 10.1021/bk-1984-0267.pr001

tion of the importance of each sample. A second concern relates to the availability of appropriate sampling equipment and techniques. Sample integrity, with particular attention to cross contamination among samples, preservation, and general handling procedures, has at last been recognized as a key to obtaining reliable monitoring data. Since both sampling and analysis comprise the complete chemical measurement system, the quality assurance aspects of sampling programs, and particularly external evaluations and audits, are no less important than the elaborate quality assurance measures that have received so much attention within the analytical laboratories. In December 1983, the Committee on Environmental Improvement of the American Chemical Society published in Analytical Chemistry "Principles of Environmental Analysis." That paper emphasized the importance of sound quality assurance procedures for chemical analyses in the laboratory. Little attention was devoted to sampling concerns. This book addresses some of the important considerations in designing and implementing sampling programs, with particular attention to surface and subsurface sampling for hazardous wastes. Furthermore, it reflects the experiences of federal and state agencies and of academic and industrial organizations and provides a good introduction to the subject. However, considerable additional research and synthesis of experiences are clearly in order given the costs, and more importantly the environmental stakes, involved in obtaining reliable monitoring information. We hope this book will stimulate greater attention to ensuring that the samples taken to the laboratory or analyzed in the field are indeed the appropriate samples for characterizing contamination problems.

GLENN E. SCHWEITZER JOHN A. SANTOLUCITO U.S. Environmental Protection Agency Las Vegas, Nevada August 1984

1 H a z a r d o u s Waste Questions and Issues from the Field

Publication Date: October 31, 1984 | doi: 10.1021/bk-1984-0267.ch001

H. PATRICIA HYNES U.S. Environmental Protection Agency, Region1,Boston, MA 02203

Notwithstanding federal hazardous waste statutes which address the dangers to human health and the environment posed by hazardous wastes, the compliance engineer is presented with unique and demanding challenges. Neither §7003 of the Resource Conservation and Recovery Act (RCRA) nor §106 of the Comprehensive Environmental Response, Cost, and Liability Act (CERCLA) sets forth the levels or quantities of hazardous substances or wastes which constitute imminent and substantial endangerment. Secondly, although there are a few "action levels" developed under the Toxic Substances and Safe Drinking Water Programs which may be adapted for cleanup, neither of the hazardous waste statutes sets forth cleanup levels to be attained for particular contaminants or for generic environmental situations. A number of site-specific factors must first be evaluated, including (1) the chemical characteristics and amount of hazardous waste, (2) the potential for release to the environment, (3) the sensitivity of the particular environment to the hazardous waste, (4) the proximity of the hazardous waste to humans, and (5) its potential effect on human health. Then the environmental engineer must decide if a field investigation of the site is necessary, whether a feasibility study for remedial action is required, what remedial action is required to mitigate, if not eliminate, the contamination, and finally, what monitoring plan will enable the efficacy of the remedial action to be evaluated. These decisions have substantial economic impact. They involve the rigorous Integration of toxicology and other environmental sciences. Futhermore, they must be defensible in court if challenged. Environmental scientists and engineers have been working almost four years in the federal hazardous waste program conducting field investigation studies, comparative remedial action studies, emergency cleanups, and to a lesser extent, implementation of long-term remedial action measures. The steep learning curve of these few years has generated a suite of questions and issues common to many hazardous waste sites. This chapter not subject to U.S. copyright. Published 1984, American Chemical Society

2

ENVIRONMENTAL SAMPLING FOR HAZARDOUS WASTES

Publication Date: October 31, 1984 | doi: 10.1021/bk-1984-0267.ch001

Types of Technical Concerns With regard to l a n d f i l l s , the following questions a r i s e . Should the contents of a problem l a n d f i l l be relocated to a more environmentally secure l a n d f i l l ; should chemical and thermal destruction be used when f e a s i b l e ; or should wastes be l e f t i n place and preventive measures used to minimize leachate discharge? If the contents of a heterogeneous l a n d f i l l are unknown, what conditions would be appropriate for the direct sampling of the l a n d f i l l ? What remote sensing technology would be most e f f e c t i v e i n the detection of buried containers and contaminated ground water plumes and i n estimating the depth of each? Also, can remote sensing determine whether the l a n d f i l l i s i n direct contact with ground water at any time of year, and the depth to bedrock under the disposal site? Another d i f f i c u l t area i s ground water monitoring. How many wells should be i n s t a l l e d ; where should they be placed with respect to the s i t e ; should they routinely extend to and into bedrock; does an "impermeable" t i l l layer always serve as a v e r t i c a l boundary f o r ground water investigation; what length of well screen i s optimum? Regarding the f i n a l question, as an example of more detailed concerns, a number of factors must be considered. The longer the screen, the better the chance of detecting ground water contamination i n an area where contamination exists i n a saturated zone i f the zone i s f a i r l y deep. However, the longer screen may result i n greater d i l u t i o n of contaminants, i f they are present i n a shallow plume. Low l e v e l contamination would, therefore, be more d i f f i c u l t to detect. A shorter screen could miss the zone of contamination depending on the placement of the screen above or below the zone, or i f there are dramatic seasonal changes i n the water table. Thus, greater accuracy i s required i n the placement of a short well screen. Whether to renovate contaminated ground water, eliminate the point source of contamination and leave the ground water to renovate i t s e l f , or do both i s also a complex decision. For example, e a r l i e r manufacturing plants were usually located on or near major rivers or t r i b u t a r i e s and often discharged l i q u i d and viscous wastes to unlined lagoons. The subsurface hydraulics often resulted i n lagoon d i s charges to ground water with regional flow patterns toward streams or r i v e r s . If the lagoon i s closed f o r further use and i t s contents are removed, can the residual contaminated ground water be l e f t to natural cleansing and limited cleanup funds be more u s e f u l l y spent elsewhere? At a minimum, the decision process must consider (a) present and future uses of the downstream surface water which i s the surface expression of the contaminated ground water; (b) present and future uses of any aquifer which under pumping conditions may be i n f i l t r a t e d by the r i v e r or the contaminated saturated zone; (c) the mass of contaminants remaining i n ground water; (d) the rates of ground water movement and contaminant migration; (e) the combined effect of d i l u t i o n of the contaminants by ground water and surface water; and ( f ) any ambient water and drinking water quality c r i t e r i a which may apply. A l l of the above must be considered under worst-case conditions. In a r u r a l area, where leachate from a hazardous waste disposal s i t e has contaminated an aquifer currently used by only a few resident i a l wells, i t may appear more cost-effective to replace a water

1.

HYNES

Hazardous Waste

3

supply with a pipeline from a nearby municipal system rather than treat the contaminated ground water. However with this option, should planning and finance capacity be designed f o r present or future use, especially i f i t i s an area projected f o r high growth? Further, the quality of the replacement water, i f i t i s treated surface water, may be s i g n i f i c a n t l y lower than that of the o r i g i n a l ground water.

Publication Date: October 31, 1984 | doi: 10.1021/bk-1984-0267.ch001

Generic Issues This paper focuses on issues which are relevant to a l l hazardous waste s i t e investigations, remedial actions, and ongoing surveillance of cleaned s i t e s . Some questions and concerns of f i e l d engineers and s c i e n t i s t s are: o number of samples o length of sampling period o frequency of sampling © indicator parameters and f i e l d screening techniques o methods of quantitative analysis Two examples, taken from actual case studies, can be given i n which the significance and meaning of the f i e l d investigation results hinged e n t i r e l y on the design of the environmental sampling programs. The purpose of the examples i s to shed l i g h t on those aspects of environmental study which need guidance and additional r i g o r , espec i a l l y from the sciences of environmental chemistry and s t a t i s t i c s . The concept of selecting indicator contaminants In environment a l sampling appears l o g i c a l , e f f i c i e n t , and c o s t - e f f e c t i v e . I t i s an approach adopted from the RCRA ground water regulations. These regulations require the owner or operator of a surface impoundment, l a n d f i l l , or land treatment f a c i l i t y f o r hazardous waste to implement a ground water monitoring program capable of detecting an impact on the quality of the uppermost aquifer underlying the f a c i l i t y . The owner or operator must e s t a b l i s h i n i t i a l background concentrations of parameters regulated under the Safe Drinking Water Act, ground water quality parameters (e.g., chloride, i r o n , and phenols), and i n d i c a tors of ground water contamination ( c a l l e d indicator parameters), and s p e c i f i c a l l y , pH, Specific Conductance, Total Organic Halogen (TOH), and Total Organic Carbon (TOC). During the f i r s t year of sampling, for each of the indicator parameters, at least four replicates must be obtained quarterly and the i n i t i a l background arithmetic mean and a variance determined by pooling the r e p l i c a t e measurements. After the f i r s t year, a l l monitoring wells must be sampled semi-annually. For each indicator parameter the arithmetic mean and variance, based on at least four replicate measurements on each sample f o r each well monitored, must be calculated and compared with the i n i t i a l background arithmetic mean. The comparison must consider each of the wells i n the monitoring system and must use the Student's t-test at the 0.01 l e v e l of significance to determine s t a t i s t i c a l l y s i g n i f i c a n t increases (and decreases i n the case of pH) from background. I f the i n i t i a l study reveals elevated levels of any indicator parameter i n a downgradient well, a plan f o r further d e f i n i t i o n and investigation of the contamination must be submitted to the Agency. The concept of indicator parameters as recommended under RCRA for Industries with a known waste stream with a limited number of

Publication Date: October 31, 1984 | doi: 10.1021/bk-1984-0267.ch001

4

ENVIRONMENTAL SAMPLING FOR HAZARDOUS WASTES

compounds i s being proposed for at least two uses by consulting firms: 1. I n i t i a l f i e l d investigations of o l d , heterogeneous disposal s i t e s where waste types are i n s u f f i c i e n t l y documented, thus deferring expensive p r i o r i t y pollutant analysis u n t i l the fact and location of ground water contamination i s established* 2. Design of long-term monitoring f o r ground water contamination at a waste s i t e following either remedial work or a no-action decision. When ground water contaminants are numerous, parameters are considered l e s s - c o s t l y , but "satisfactory," f o r monitoring trends i n ground water quality* An example of the above i s a r u r a l s i t e where f o r ten years waste o i l and solvents from a transformer manufacturing plant were discharged, burned, and buried* This occurred t h i r t y years ago* A recently completed f i e l d investigation revealed that highly concentrated PCBs have remained adsorbed to s o i l i n the v i c i n i t y of the o r i g i n a l disposal and burning area and that eleven v o l a t i l e organic compounds and one base/neutral compound have migrated as f a r as 1/2 mile i n ground water i n the upper aquifer* The wells Installed closest to the disposal s i t e contain a l l of the twelve organics detected i n an I n i t i a l p r i o r i t y pollutant scan and contain the highest concentration of t o t a l v o l a t i l e organics* As the plume attenuates away from the s i t e , the sampling wells contain some but not a l l of the twelve organic compounds* The study concludes that three compounds, 1,1-dichloroethylene, trans 1,2-dichloroethylene, and t r i chloroethylene, "are r e l i a b l e chemical indicators" of seepage from the disposal p i t * There i s no dlcusslon of why these three compounds were selected rather than v i n y l chloride or other compounds which occur throughout the two major ground water plumes* The selection of the three compounds as "indicators" appears arbitrary and p o t e n t i a l l y misleading since the chemical and physical properties of a l l the compounds were not considered* The responsible party has agreed to remove the o r i g i n a l waste and to replace the water supply from private wells with a pipeline from a municipal source, provided they not be obligated to treat the contaminated ground water. Assuming that the ground water i s l e f t untreated to renovate i t s e l f , there i s s t i l l reason to have a caref u l l y designed monitoring plan, s u f f i c i e n t l y comprehensive and r e l i a b l e to v e r i f y the changes i n ground water quality resulting from natural processes. The purpose of a monitoring plan i s to provide r e l i a b l e data to demonstrate with a good degree of certainty whether or not contaminant attenuation results after remedial action* At a minimum, the plan should consider the physical and chemical propert i e s of the compounds i n question as well as the s o i l s i n the upper aquifer, the ground water flow regime, and s t a t i s t i c a l techniques to enable meaningful comparisons. For a ground water monitoring plan to be approved by a regulatory agency, i t should contain sampling and s t a t i s t i c a l design features which permit the determination of, f o r example, whether or not 40 ^g/1 TCE i n Well A i n July i s a s i g n i f i cant increase from 24 |ig/l measured i n March of the same year. A l t e r n a t i v e l y , one should know i f 35 \xg/l benzene i n Well A i s s i g n i f i c a n t l y d i f f e r e n t from 50 \ig/l benzene i n Well B i f both wells were sampled at the time. S t a t i s t i c a l tools are needed to account for the v a r i a b i l i t y of ground water q u a l i t y , the e f f e c t of sampling

Publication Date: October 31, 1984 | doi: 10.1021/bk-1984-0267.ch001

1.

HYNES

Hazardous Waste

5

frequency on the interpretation of data» and the trends i n ground water q u a l i t y . Another example which poses many questions i s a study of PCBs i n r i v e r sediment and f i s h and PCB transport under varying flow condi­ t i o n s . The objectives of the f i e l d study were: ο To describe the extent of PCB contamination of f i s h as a r e s u l t of former discharges of PCBs into the r i v e r by a p a r t i c u l a r induetry. ο To compare the PCB l e v e l s i n f i s h with the Food and Drug Admin­ i s t r a t i o n (FDA) regulatory l e v e l of 5 ppm. ο To establish a set of i n i t i a l conditions for measuring impacts of future remedial action on the r i v e r . For comparison between PCB levels i n f i s h and sediment, a r i v e r was divided into eight f i s h and sediment stations. A t o t a l of 721 f i s h ( p r i n c i p a l species were perch, sunfish, bass, and trout) were collected between 1980 and 1982. F i f t y - t h r e e percent were used for analysis of PCBs, and those remaining were archived for future use. F i l e t s from each of the four f i s h species were composited and ana­ lyzed. The composites comprised from two to twelve f i l e t s , depending on the weight of the f i s h . The a n a l y t i c a l results were reported as a mean concentration i n micrograms of PCB per gram of dry weight f i s h t i s s u e , and consisted of one result per species per s t a t i o n . Certain l i m i t a t i o n s of this study are worth reviewing. The FDA standard or "action l e v e l " f o r PCBs i n f i s h sold i n interstate com­ merce was used even though the r i v e r i s not commercially fished. . In any event, a study of contamination i n f i s h which w i l l result i n com­ parisons with a regulatory "action l e v e l " f o r j u s t i f y i n g a remedial action must provide results which r e t a i n some degree of confidence. The βample population f o r the study (53% of 731 f i s h ) appears at f i r s t glance s u f f i c i e n t l y large for inferences about the average PCB concentration i n the t o t a l population. The investigator should decide ahead of time what degree of certainty and Interval size are desired. With some estimate of the standard deviation of the popu­ l a t i o n , the sample size required to s a t i s f y the investigation s p e c i ­ f i c a t i o n s may be determined. However, the i t e r a t i v e process involved i n selecting an appropriate sample size was not used i n designing this study, nor were the results analyzed and reported with any degree of certainty and i n t e r v a l s i z e . In f a c t , because of the method of compositing samples the sample population i s small—one result per species per station or 8 results per species. In this case, small sample s t a t i s t i c a l techniques are appropriate. The design of this f i s h study centered on sample c o l l e c t i o n , preservation, preparation, analysis, and QA/QC. There was no d i s ­ cussion of the effect of compositing on the sample population. No description was given of s t a t i s t i c a l techniques to be applied to the data f o r reporting results and f o r comparison with "action l e v e l s " and future data. Unfortunately, the omission of a s t a t i s t i c a l frame­ work during planning of the f i e l d study i s the rule rather than the exception i n hazardous waste investigations. The hydrogeological and QA/QC aspects of hazardous waste f i e l d investigations are f a i r l y well advanced. Yet needed, however, i s a systematic approach to the design of f i e l d sampling, to the selection of compounds f o r analysis, and to the methods f o r interpretation of a n a l y t i c a l data. R E C E I V E D August 14,1984

2

Uses o f E n v i r o n m e n t a l Testing i n H u m a n H e a l t h R i s k Assessment 1

CLARK W. HEATH, JR.

Publication Date: October 31, 1984 | doi: 10.1021/bk-1984-0267.ch002

Centers for Disease Control, Atlanta, GA 30333 The process for assessing environmental health risks is complex. Information regarding 1) the environmental agent, 2) the pathways by which exposure occurs, and 3) the biologic effects observed after exposure must be assembled and simultaneously evaluated. Incomplete knowledge or inadequate methodology in any of these three areas can severely inhibit accurate or useful estimates of health risk. Environmental testing is a critical element in this process since it enables the qualitative and quantitative determination of toxic chemicals in the environment and the definition of environmental pathways which may lead to human exposure. This paper briefly reviews the overall process of health risk assessments and the particular role which environmental testing plays. Recent efforts to assess environmental health risks in relation to Love Canal illustrate both the usefulness and the limitations of environmental testing in risk assessment. The Risk Assessment Process The process of performing risk assessment is outlined in Table I. Factors contributing to the nature and degree of exposure are examined first. They include characteristics of the toxic material, environmental pathways, and mechanisms operating in absorption and metabolism in the host. Next, the biologic effects resulting from such exposure are defined in relation to the amount of exposure, to humans and in animal models. Finally, information about exposure and biologic effect characteristics is interpreted in relation to predetermined definitions of biologic safety to establish benchmarks for acceptable exposure risk levels O ) . The toxic chemical(s) of concern must be identified and their physical and chemical characteristics evaluated. The concentrations of each of the chemicals must be measured, ideally in both the environment and in the tissues of exposed humans. Depending on the nature and distribution of toxic material, environmental measurements may be required in air, water, soil, or food, or in combinations of these media. The critical limiting factor at this stage of assessment relates to the degree to which particular chemicals can be identified 'Mailing address: Emory University School of Medicine, Atlanta, GA 30322 This chapter not subject to U.S. copyright. Published 1984, American Chemical Society

8

ENVIRONMENTAL SAMPLING FOR HAZARDOUS WASTES Table I . The Process of Environmental Risk Assessment

Publication Date: October 31, 1984 | doi: 10.1021/bk-1984-0267.ch002

I.

ASSESSMENT OF EXPOSURE A. I d e n t i f i c a t i o n of toxins and measurement of t h e i r levels or concentrations i n the exposure s e t t i n g . 1. In the environment: a i r , water, s o i l , food. a) Laboratory technology: quality assurance, precision, s e n s i t i v i t y , accuracy. b) Interaction of chemicals producing p o t e n t i a l l y toxic byproducts. 2. In the host: l e v e l s of toxin i n serum or t i s s u e . a) Persistence i n tissue: excretion, tissue/organ specificity. b) Metabolic a l t e r a t i o n s : p o t e n t i a l l y toxic byproducts. B. Environmental pathways for exposure. 1. Mechanisms for transmitting toxin to host: ingestion, inhalation, dermal contact, physical and b i o l o g i c vectors. 2. Degree and mode of exposure: contact through human activities. 3. Degree and mode of absorption into host: E f f e c t i v e dose at t i s s u e / c e l l l e v e l depends on nature of toxin, route of exposure, and interaction of target tissue with absorbed metabolized toxin. I I . ASSESSMENT OF BIOLOGIC EFFECT A. Human e f f e c t s : epidemiologic and c l i n i c a l observations. 1. Acute c l i n i c a l e f f e c t s : organ system s p e c i f i c i t y , short latency, high dose exposure. 2. Delayed (chronic) c l i n i c a l e f f e c t s : long and variable latency, lower dose exposure. 3. S u b c l i n i c a l e f f e c t s : c l i n i c a l laboratory test a l t e r ation ( l i v e r function, nerve conduction v e l o c i t y ) , mutagenicity testing, cytogenetic t e s t i n g . Requires estimation of eventual l i k e l i h o o d of c l i n i c a l disease predicted by s u b c l i n i c a l abnormalities. B. Non-human e f f e c t s : experimental observations i n toxicologic testing i n animals or i n b a c t e r i a l or c e l l culture test eyeterns. C. Low dose e f f e c t s : usually not measurable d i r e c t l y i n human or animal observations. Need to extrapolate observed high dose effects to low or zero dose range by theoretical doseresponse models. I I I . SELECTION OF SAFETY STANDARDS Choice of c r i t e r i a f o r defining a "safe" l e v e l of toxin i n the environment based on animal and human observations. a) Potential carcinogenic e f f e c t s : 1/1,000,000 l i f e t i m e cumulative r i s k . b) Non-carcinogenic e f f e c t s : Highest l e v e l of toxin at which no effect i s observed (NOEL), lowered by safety margin of 100 to 1000 f o l d to allow f o r interspecies biologic variation.

Publication Date: October 31, 1984 | doi: 10.1021/bk-1984-0267.ch002

2.

HEATH

Human Health Risk Assessment

9

and measured In each of the media. Testing methodology must be capable of y i e l d i n g reproducible results of known and acceptable p r e c i s i o n and s e n s i t i v i t y . This requirement i s especially important when testing i s undertaken i n more than one laboratory. Although analytic standards and reference materials exist for a wide range of Individual chemicals i n d i f f e r e n t environmental media, interactions among multiple chemicals coexisting i n the environment pose d i f f i c u l t i e s for many testing procedures. Measurement of exposure can be made by determining levels of toxic chemicals i n human serum or tissue i f the chemicals of concern p e r s i s t i n tissue or i f the exposure i s recent. For most s i t u a t i o n s , neither of these conditions i s met. As a r e s u l t , most assessments of exposure depend primarily on chemical measurements i n environmental media coupled with semi-quantitative assessments of environmental pathway8. However, when measurements i n human tissue are possible, valuable exposure information can be obtained, subject to the same l i m i t a t i o n s c i t e d above f o r environmental measurement methodology. Interpretation of tissue concentration data i s dependent on knowledge of the absorption, excretion, metabolism, and tissue specificity c h a r a c t e r i s t i c s for the chemical under study. The toxic hazard posed by a p a r t i c u l a r chemical w i l l depend c r i t i c a l l y upon the concentration achieved at p a r t i c u l a r target organ s i t e s . This, i n turn, depends upon rates of absorption, transport, and metabolic a l t e r a t i o n . Metabolic alterations can involve either p a r t i a l i n a c t i v a t i o n of toxic material or conversion to chemicals with increased or d i f f e r i n g toxic properties. Toxic chemicals can be transported with d i f f e r i n g levels of e f f i c i e n c y to the target host depending upon the transport pathways. Exposure may occur d i r e c t l y by ingestion, inhalation, or dermal contact or through some form of Intermediate vector such as insects or clothing contamination. The r e l a t i v e contribution of d i f f e r e n t pathways must be assessed by examining the nature of human a c t i v i t i e s which may be expected i n p a r t i c u l a r exposure settings. This evaluation w i l l i d e n t i f y both the situations f o r which the greatest exposure may be anticipated (young children Ingesting s o i l while at play, f o r instance) and the safety standards that w i l l eventually be needed. Again, the actual concentration of toxic chemical i n the host c e l l depends partly on assessments of host-environment contact and p a r t l y on knowledge of absorption and metabolism of the p a r t i c u l a r chemicals. B i o l o g i c E f f e c t . Ideally, r i s k assessment i s based on quantitative knowledge of b i o l o g i c effects i n humans. Unfortunately, such d i r e c t human information does not exist for most toxic chemicals. Therefore, prediction of human effects usually depends upon extrapolating the r e s u l t s of experimentally exposed laboratory animals (usually r o dents). Such extrapolations must be performed not only between species, but between observed high dose effects and predicted low dose e f f e c t s . Most animal toxicologic testing and v i r t u a l l y a l l observed human health effects involve r e l a t i v e l y high dosages. Since safety standards are commonly aimed at preventing the potential e f f e c t s of low dose exposure (especially cancer), low dose extrapolations from e x i s t i n g high dose data are a c r i t i c a l phase i n r i s k assessment. S t a t i s t i c a l models f o r predicting low dose effects

Publication Date: October 31, 1984 | doi: 10.1021/bk-1984-0267.ch002

10

ENVIRONMENTAL SAMPLING FOR HAZARDOUS WASTES

e x i s t . Some are based on the assumption that a no-threshold l i n e a r relationship exists between dose and b i o l o g i c response. Others encompass various concepts of threshold effect and curved dose response relationships which predict d i f f e r i n g degrees of dose r e sponse depending upon the influence of host tissue repair or excret i o n mechanisms. The eventual safety standards developed can vary quite widely according to the theoretical model selected i n a p a r t i c ular r i s k assessment. Biologic e f f e c t s should be assessed for both c l i n i c a l and subc l i n i c a l changes. Either can be acute or delayed, with delayed effects often associated with lower exposures. Clinical illness occurs infrequently following chemical exposures, especially at moderate or low dose l e v e l s . To increase the probability of detecting c l i n i c a l endpoints at lower dose levels i n experimental animals or human epidemiologic studies, i t i s desirable to maximize the size of populations examined. Sample s i z e , therefore, rapidly becomes a l i m i t i n g factor for making c l i n i c a l observations. When population size i s l i m i t e d , i t becomes necessary either to measure s u b c l i n i c a l effects which may occur with greater frequency than c l i n i c a l effects at given dose levels ( l i v e r function abnormalities, cytogenetic changes) or to accept theoretical extrapolations downward from high dose c l i n i c a l e f f e c t s . Increasingly, r i s k assessment e f f o r t s have begun to focus more on s u b c l i n i c a l e f f e c t s , both i n humans and i n test animals. Although this trend holds promise f o r greater testing s e n s i t i v i t y , i t w i l l require Improved understanding of the r e l a t i o n ship between s u b c l i n i c a l endpoints and eventual c l i n i c a l i l l n e s s . Such s u b c l i n i c a l - c l i n i c a l extrapolation i s of c r i t i c a l importance f o r r i s k assessment. It i s not at a l l clear at present, for example, that cytogenetic changes observed i n exposed populations necessarily fore-shadow l a t e r increases i n incidence of cancer or genetic disease. Acceptable Risk. Once information i s assembled concerning the c h a r a c t e r i s t i c s of exposure and biologic e f f e c t s , that information must be interpreted i n terms of human safety standards. That i n t e r pretation requires that one establish a set of c r i t e r i a representing acceptably safe conditions for human existence, bearing i n mind that zero concentrations of environmental chemicals are u n r e a l i s t i c . This process of standard-setting i s by nature q u a l i t a t i v e and somewhat a r b i t r a r y . Nevertheless, certain conventions have evolved for setting safety l e v e l targets. In case of carcinogenic or potent i a l l y carcinogenic substances, with the assumption that no dose threshold exists for cancer r i s k , that target has conventionally been set as a l i f e t i m e increased r i s k of one case of cancer i n a population of one m i l l i o n persons. For chemicals presumed to be noncarcinogenic, acceptable r i s k has been conventionally set at the highest dose l e v e l at which no observed b i o l o g i c effect i s observed i n experimental animals (NOEL or "no observed effects l e v e l " ) . The l a t t e r standard i s then adjusted downward by applying a safety factor of 100 to 1000 f o l d to make allowance for uncertainties of extrapol a t i o n s f o r differences between species and dosages.

2.

HEATH

Human Health Risk Assessment

11

Publication Date: October 31, 1984 | doi: 10.1021/bk-1984-0267.ch002

Environmental Testing Once an acceptable "safe" l e v e l of chemicals has been determined for p a r t i c u l a r environmental media based on existing toxicologic and epidemiologic data and on appropriate safety model c r i t e r i a , accurate, reproducible, and economically feasible programs for measurement of environmental chemicals must be implemented. Since most toxic chemi c a l exposure situations involve multiple chemicals, the task i s f a r from simple. Aside from the economic f e a s i b i l t y of testing programs which often involve large numbers of samples, two other considerations are of central concern. These are 1) sampling design and framework, and 2) technical laboratory procedures. With respect to sampling, s u f f i c i e n t numbers of environmental samples should be obtained to permit r e l i a b l e s t a t i s t i c a l and b i o l o g i c interpretation of r e s u l t s . At the same time, the samples c o l lected should be from environmental locations where human exposure i s most l i k e l y to occur (or did occur, i f questions of past exposures require assessment). They should also be targeted f o r those environmental media which can be expected to have the greatest potential f o r human exposure and absorption. F i n a l l y , the samples must be obtained and preserved so that the chemicals which pose the greatest threat for human health i n terms of t o x i c i t y and tissue persistence can be accurately measured. Laboratory quality assurance procedures must be b u i l t into the sampling plan so that reproducibility and precision of test results can be c l e a r l y demonstrated when testing i s complete. This includes defining the precision of the measurement system and sample c o l l e c t i o n procedures and providing f o r adequate numbers of repeat or s p l i t samples, as well as b l i n d l y inserted positive and negative control specimens. If r i s k assessment i s dependent upon assessing potential environmental exposure, overtime arrangements must be made at the s t a r t f o r laboratory consistency for as long as testing i s expected to continue. Because of the complexity and expense involved, even f o r limited environmental testing programs involving few chemical toxins, specimen c o l l e c t i o n and laboratory testing should not be h a s t i l y undertaken. Careful advance planning i s necessary, complete with outside peer review and approval of proposed testing plans. This i s e s p e c i a l l y true for chemicals f o r which laboratory technology and sampling procedures are not yet f u l l y developed. Environmental Testing at the Love Canal When the Love Canal problem came to active public attention, i t was necessary to reconstruct the nature and extent of past exposure as well as address current and future human exposure. Multiple chemicals of uncertain amounts and d i s t r i b u t i o n patterns within the Canal s i t e required varied laboratory technologies, multi-media sampling, and large numbers of samples drawn from a wide and diverse neighborhood s e t t i n g . The problem was of concern to public health agencies at various levels of Government, and several d i f f e r e n t test programs were undertaken by different organizations, p r i n c i p a l l y the State of New York i n the i n i t i a l phases of remedial work and the U.S. Environmental Protection Agency (EPA) during l a t e r phases.

Publication Date: October 31, 1984 | doi: 10.1021/bk-1984-0267.ch002

12

ENVIRONMENTAL SAMPLING FOR HAZARDOUS WASTES

At the s t a r t , the focus was on environmental chemical levels at the Canal s i t e i t s e l f and i n homes Immediately adjoining i t * An extensive multi-media testing program was conducted over several months i n 1978 and 1979 by the New York State Department of Health (2). This work concentrated on the f i r s t two rings of homes adjacent to the Canal i n i t s early phases but l a t e r was extended to include larger neighborhoods* Community concern was focused especially to the east, where Canal chemicals might p o t e n t i a l l y spread through the remaining traces of pre-existing natural drainage channels i n surface s o i l . Although a long l i s t of Canal chemicals was targeted f o r analysis, a select number of chemicals received p a r t i c u l a r attention on the basis of t h e i r r e l a t i v e l y unique presence i n the Canal. For example, chlorobenzene and chlorotoluene had value as marker contaminants. Water, s o i l , and a i r were sampled, with special emphasis given to water and a i r within homes where people might be expected to have had the highest and most sustained exposure. In l a t e r phases of t e s t i n g , chemical levels i n storm sewers and streams draining the Love Canal neighborhood also received attention. When remedial drainage construction work began, environmental sampling was also required to guide the location of underground drainage pipes and to monitor worker exposure conditions. Since simultaneous health effect surveys were conducted i n the Love Canal area, environmental test results i n the homes adjoining the Canal were examined i n an e f f o r t to demonstrate the presence or absence of correlations between environmental chemical levels and frequencies of p a r t i c u l a r health abnormalities. This e f f o r t was l a r g e l y unsuccessful, since the t o t a l exposed population proved to be too small f o r meaningful interpretation of most health endpoints of interest and since d i f f i c u l t i e s i n c o n t r o l l i n g f o r subjective reporting of health symptoms made i t d i f f i c u l t to interpret health survey results OA). The results of environmental testing i n Love Canal homes conducted p r i o r to remedial drainage construction were used as a basis for testing the hypothesis that persistent cytogenetic abnormalities might have resulted from Canal exposure i n persons who had been l i v i n g adjacent to the Canal i n 1978 (5)· For this study, 12 households with the highest concentrations of marker organic chemicals i n basement a i r i n 1978 were selected. Persons who had l i v e d i n several of these homes were then studied f o r frequencies of d i f f e r e n t forms of chromosomal aberration i n comparison with frequencies found i n simultaneous testing of matched households elsewhere i n the Niagara F a l l s urban/suburban area. No s i g n i f i c a n t differences i n frequencies were seen i n this comparison. Whether cytogenetic changes were never Induced by chemical exposure i n homes near the Canal, or i f they had been Induced but did not p e r s i s t , could not be resolved by this study. Extensive t e s t i n g was also carried out to assess future hazards for human habitation and r e s i d e n t i a l use of the area. This testing was carried out i n an extensive program funded by EPA i n 1980-81 a f t e r remedial drainage construction work at the Canal was complete (6). A l l of the survey design and technical laboratory problems described above were encountered i n this program. Despite extensive e f f o r t s to meet requirements f o r sample size and d i s t r i b u t i o n , to provide f o r adequate control sampling away from the Canal area, and

Publication Date: October 31, 1984 | doi: 10.1021/bk-1984-0267.ch002

2.

HEATH

Human Health Risk Assessment

13

to allow for s a t i s f a c t o r y sample c o l l e c t i o n procedures and laboratory testing protocols, time and resource constraints limited the scope of the program. Questions were raised regarding the interpretation of results i n the face of sustained public concern and l i t i g a t i o n over the entire s o c i o - s c i e n t i f i c s i t u a t i o n . From the viewpoint of scien­ t i f i c assessment, however, review of EPA test results by the U.S. Department of Health and Human Services concluded that, despite l i m i t a t i o n i n sample size and questions regarding certain technical laboratory procedures, the data were s u f f i c i e n t to judge the Love Canal r e s i d e n t i a l area safe f o r human r e s i d e n t i a l use. This r i s k assessment, using the environmental test data i n hand, was based on the absence of chemical levels above the low parts per b i l l i o n range. The conclusion was reached with the s t i p u l a t i o n that storm sewer drainage tracks known to contain excess levels of dloxln and other persistent organic chemicals be cleaned, that the Canal s i t e i t s e l f not be used f o r home s i t e s , and that an adequate program f o r monitor­ ing chemical contaminant within the Canal s i t e be established and maintained into the Indefinite future. These recommendations regard­ ing future h a b i t a b i l i t y have not been adopted pending review of the data on which they were based and consideration of the possible need for further evaluative environmental t e s t i n g .

Literature Cited 1. 2. 3. 4. 5. 6.

"Health Risk Estimates for 2,3,7,8-Tetrachlorodibenzodioxin in Soil," Centers for Disease Control, Morbidity and Mortality, Weekly Report, 1984. Kim, C. S.; Narang, R.; Richards, A. et al. Proc. Environmental Protection Agency National Conference on Management of Uncon­ trolled Hazardous Waste Sites, 1980, p. 212. Vianna, N. J . Proc. 10th Ann. NY State Dept. of Health Birth Defects Symp., 1980, p. 165. Heath, C. W., Jr. Envlr. Health Perspect. 1983, 48, 3-7. Heath, C. W., Jr.; Nadel, M. R.; Zack, Μ. Μ., J r . , et al. J . Amer. Med. Assoc. 1984, 251, 1437-40. "Environmental Monitoring at Love Canal," Environmental Protec­ tion Agency, EPA 600/4-82-030a, 1982.

RECEIVED August 6, 1984

3

Assessing C y a n i d e C o n t a m i n a t i o n from a n A l u m i n u m Smelter RICHARD A. BURKHALTER, THEODORE J. MIX, MERLEY F. McCALL, and DONALD O. PROVOST

Publication Date: October 31, 1984 | doi: 10.1021/bk-1984-0267.ch003

Department of Ecology, State of Washington, Olympia, WA 98504 Cyanide contamination of the Spokane aquifer was discovered by Kaiser Aluminum Company personnel in 1978. The Kaiser aluminum reduction facility is located on 240 acres in Mead, Washington, near the northeastern city limits of Spokane (Figure 1). Potliner, which contains cyanide, has been stored at the plant site since 1942. The facility is in a semi-arid region of the state at the foothills of the Rocky Mountain range. The average annual precipitation is 17.5 inches per year with 70 percent occurring from October through April. The annual average evaporation rate is 14 inches per year with a potential of 25 inches per year (1). The facility is located over the Spokane aquifer which has been designated a major sole source aquifer (Figure 2). The depth to the aquifer at the plant site is about 160 feet. The soil over the aquifer is sand and gravel with interspersing clay lenses (2,3). The Spokane aquifer is a highly permeable aquifer with ground water movement estimated to be between 41 and 47 feet per day by the U.S. Geological Survey and Corps of Engineers. The northern part of the aquifer discharges by springs into the Little Spokane River. Under base flow conditions the flow of the river more than doubles because of these springs (4). Process and Facility Description Aluminum is produced by passing an electrical current through a high temperature electrolyte solution (sodium aluminum fluoride) containing aluminum oxide (Figure 3). Aluminum migrates to the carbon cathode (potliner), and oxygen migrates to pre-baked carbon anodes. Aluminum is usually tapped from the pot once every fourth shift. The oxygen supports the combustion of the carbon anode which must be replaced on a routine basis. The potliner lasts a considerable period of time, usually three years, before it fails and is replaced. Approximately 4,500 tons of potliner are discarded each year when operations are at full production levels. In the process, high temperature and reducing conditions result in the formation of cyanide which is absorbed into the cathode carbon block at the bottom and sides of the pot. When a pot falls, the 0097-6156/84/0267-0015$06.00/ 0 © 1984 American Chemical Society

Publication Date: October 31, 1984 | doi: 10.1021/bk-1984-0267.ch003

ENVIRONMENTAL SAMPLING FOR HAZARDOUS WASTES

Figure 1. V i c i n i t y map - Kaiser Aluminum and Chemical Corporation, Mead Smelter.

Publication Date: October 31, 1984 | doi: 10.1021/bk-1984-0267.ch003

3.

BURKHALTER ET AL.

Assessing Cyanide Contamination

17

p o t l i n e r i s removed from the s t e e l pot s h e l l * The s h e l l i s reconditioned by r e l i n i n g with carbon and i n s u l a t i o n before being returned to the p o t l i n e . To recondition the pots, the p o t l i n e r i s dug out and discarded* P r i o r to discovery of the Spokane aquifer contamination, the procedure had been to remove the pot to an outdoor concrete slab where the pot was f i l l e d with water and allowed to soak f o r a few days to f r a c ture and soften the cathode* The contaminated water was presumably reused f o r soaking and not discharged to the i n d u s t r i a l waste t r e a t ment system because of the cyanide content* The pots were jackhammered and the p o t l i n e r dumped on the slab. The potliner was tranefered by a front-end loader to an unprotected p i l e next to the slab. The 8-acre contaminated area i s located i n the north-central portion of the plant s i t e (Figure 4). The p o t l i n e r p i l e volume i s 2.5 m i l l i o n cubic feet (128,000 tone) and contains approximately 0.2 percent cyanide. The Company's Industrial wastewater s e t t l i n g basin (Tharp Lake) was located 200 feet east of the potliner p i l e . It removed suspended s o l i d s , o i l , and grease from several m i l l i o n gallons of cooling water and storm water runoff each day. The domestic wastewater treatment plant i s also located near the s i t e of the i n d u s t r i a l wastewater treatment f a c i l i t y . The 300,000 gallons per day of treated domestic wastewater was discharged to a seepage lagoon located i n the same v i c i n i t y . The main storm water and i n d u s t r i a l 8ewer l i n e i s located between the p o t l i n e r p i l e and the i n d u s t r i a l / domestic treatment systems. Description of the Problem Because of forthcoming state and federal regulations and problems revealed at other plants, the Company decided to d r i l l some test wells to determine i f storage of waste materials at the plant s i t e had resulted i n environmental contamination. Exploratory wells were d r i l l e d by the Company's contractor around the p o t l i n e r disposal p i l e (5). High concentrations of cyanide and f l u o r i d e were found i n these wells. As a result of these findings, existing wells outside the plant boundary were sampled and also found to contain high concentrations of cyanide. The results were reported to the Department of Ecology i n August 1978, and subsequently, additional wells were tested to determine the extent of the contamination. This sampling outlined a pathway from the plant to the L i t t l e Spokane River, as shown i n Figure 5. The plume i s approximately 800 feet wide at the plant and 1,500 feet wide at the L i t t l e Spokane River, a distance of two and one-half miles northwesterly from the plant s i t e . The ground water elevation drops 80 feet from the plant s i t e to the L i t t l e Spokane River. Total cyanide concentrations i n ground water samples ranged from over 300 parts per m i l l i o n (ppm) at the plant s i t e to about 1*5 ppm at a spring located along the banks of the L i t t l e Spokane River* Wells used f o r drinking water, i r r i g a t i o n , and livestock purposes contained t o t a l cyanide concentrations as high as 23 ppm*

ENVIRONMENTAL SAMPLING FOR HAZARDOUS WASTES

Publication Date: October 31, 1984 | doi: 10.1021/bk-1984-0267.ch003

U

Figure 3.

Figure 4.

ANODE BARS

M

Aluminum reduction c e l l .

Area of contamination, Kaiser's Mead Smelter.

3. BURKHALTER ET AL.

Assessing Cyanide Contamination

19

Publication Date: October 31, 1984 | doi: 10.1021/bk-1984-0267.ch003

Remedial Action and Results, Phase I The Company requested, and immediately received, permission to d i vert the treated sanitary wastewater from the seepage lagoon to the i n d u s t r i a l treatment system. At about this time, i t was also d i s covered that Kaiser employees were discharging cyanide-laden sump water d i r e c t l y to the seepage lagoon, and this practice was immediately discontinued. The Company was also ordered to discontinue pot soaking and instead to dig the pots out i n a dry condition. The Spokane County Health Department ordered the Company to cover the e x i s t i n g potliner p i l e . The Company also constructed a temporary double-lined pad for storage of fresh p o t l i n e r u n t i l they constructed a storage building. Drainage from the storage slab was discharged to a new, lined pond f o r treatment. The Company and Spokane County Health Department immediately contacted a l l well owners i n the v i c i n i t y and tested the wells f o r cyanide contamination. The Company made bottled water available to the affected people on a temporary basis u n t i l a permanent uncontaminated supply could be obtained. A ground water monitoring program of selected wells was developed to v e r i f y the expected changes r e s u l t i n g from these remedial actions. Additional wells were i n s t a l l e d around the covered p i l e to support this monitoring program. Because of the low r a i n f a l l and high evaporation rate, other water sources which might carry contamination into the ground water were investigated. The Company was required to check the i n d u s t r i a l water s e t t l i n g basin and a l l existing storm and sanitary sewers i n the potliner area for leaks. A water quality study of the L i t t l e Spokane River was conducted to determine i f cyanide discharge by the springs would have any e f f e c t on the aquatic l i f e . Very low concentrations of cyanide were measured i n the L i t t l e Spokane River near the contaminated springs, and no measurable effect on aquatic l i f e was detected. The Washington Department of Ecology requested the Company to conduct a f e a s i b i l i t y study of pumping the aquifer and treating the contaminated ground water. The cost of pumping and treating contaminated ground water was estimated to be over $4 m i l l i o n i n c a p i t a l costs, with an operating cost of approximately $1 m i l l i o n per year. If the wells were pumped, the treated water would s t i l l contain 200 ppb cyanide and would also need to be disposed of s a f e l y . Inspection and testing of the storm/industrial and sanitary sewers i n the area indicated the sewers were i n good condition and only minor amounts of leakage were occurring. Nevertheless, minor repairs were performed. As a result of removing the sanitary discharge from the seepage lagoon, the shallow well (TH-1) located to the west of the seepage lagoon, showed a decrease i n cyanide concentration. The well located immediately downstream of the potliner p i l e and p o t l i n e r work area did not improve as expected. Remedial Action and Results, Phase I I The remedial actions summarized here are detailed elsewhere (6-9,10) By mid-1980 the cyanide levels i n the ground water had not changed s i g n i f i c a n t l y , and i t was apparent that the corrective action taken

ENVIRONMENTAL SAMPLING FOR HAZARDOUS WASTES

20

had not improved the contamination problem. The Phase I assumptions and actione were re-examined because more information on the subsurface geology was required to determine the pathways of contaminat i o n . One p o s s i b i l i t y considered was that aquatards (clay lenses) were ponding highly contaminated waters below the p i l e and slowly releasing them to the main aquifer. Other contamination pathways theorized were: 1· 2.

Publication Date: October 31, 1984 | doi: 10.1021/bk-1984-0267.ch003

3. 4.

Leaching of cyanide from the uncovered p o t l i n e r p i l e into the ground by r a i n f a l l and snow melt. Discharge of 1 to 2 percent cyanide-laden water from the pot soaking operation onto the surrounding ground. Pumping of 1 to 2 percent cyanide-laden water to the domestic seepage pond. Seepage of uncontaminated water into the contaminated ground, thus leaching the cyanide into the ground water.

The Company and i t s consultant decided more monitoring wells were needed to better define the s i t u a t i o n , and a new d r i l l i n g program was i n i t i a t e d . With the addition of more wells, now t o t a l l i n g over 35, the geol o g i c a l formation under the plant s i t e was better defined. At the p o t l i n e r s i t e the main aquifer s t a r t s about 160 feet below the surface and consists of three zones ("A", "B", and "C"), with "A" zone containing s i g n i f i c a n t l y higher concentrations of cyanide than the other two zones. Water flow through the "A" zone was determined to move much slower than the other two zones. Aquatards were found intermittently throughout the formation to the ground water table, and some of these formed saturated zones of highly contaminated water above them. Almost d i r e c t l y under the p i l e , a ground water mound i n "A" zone was discovered. This mound had the e f f e c t of s h i f t i n g the d i r e c t i o n of the aquifer flow. Downgradient of the potliner p i l e , the clay lenses terminate, and the three zones merge into one (Figure 6). Aquatards could t h e o r e t i c a l l y be r e d i r e c t i n g water flows beneath the covered storage p i l e and leaching out the cyanide from the highly contaminated s o i l . If so, ground water cyanide concentrations should have decreased after covering the storage p i l e . Since this did not occur, some other source of cyanide was suspected. The i n d u s t r i a l s e t t l i n g basin (Tharp Lake) which o r i g i n a l l y was believed not to leak s i g n i f i c a n t l y was re-checked by diverting the Industrial discharge away from the s e t t l i n g basin for 24 hours. The basin was Indeed found to be leaking between 50-60 gallons per minute (Figure 7). The Company Immediately began construction of a new s e t t l i n g basin 2,000 feet to the north of the area of concern. Within 60 days the o r i g i n a l treatment basin was closed, and within one month following closing, the shallow monitoring wells between i t and the p i l e t o t a l l y dried up, some within a matter of days. The aquifer flow has s h i f t e d back to what i s believed to be i t s normal course, and the ground water mound has dissipated. A l l other potential water sources i n the area have been checked to insure no other c a r r i e r i s available to leach cyanide from the s o i l column beneath the p i l e . A l l storm drains, sanitary sewers, and pressurized water l i n e s have been r e checked and sealed i f necessary. The cyanide l e v e l s have been slowly dropping since closing the s e t t l i n g basin (Figure 8).

Assessing Cyanide Contamination

Publication Date: October 31, 1984 | doi: 10.1021/bk-1984-0267.ch003

BURKHALTER ET AL.

2TH-1

Sc«l«lnFMt

(Pope) Well Number and Owner "* Water Level • Screened Section and Cyanide Analysts in pob.

Note: The level* of cyanide shown give the approximate range of concentration at the particular sampling point. The concentrations for well 8Q, (Pope) were determined from samples taken during replacement well drilling operations.

1

Spring and Cyanide ' Analysis in poo. Water Table

Figure 6.

Cyanide flow-path downgradient of plant s i t e sources

Publication Date: October 31, 1984 | doi: 10.1021/bk-1984-0267.ch003

ENVIRONMENTAL SAMPLING FOR HAZARDOUS WASTES

Figure 7.

Schematic cyanide flow paths beneath Mead plant.

3.

BURKHALTER ET AL.

Assessing Cyanide Contamination

23

A leak of 20 gpm i n a pressurized water l i n e located i n the area to the east of the old p o t l i n e r cleaning building was observed i n l a t e June 1983 and corrected. The results of the leak are v i v i d l y shown by an increased cyanide concentration from March 1983 to midOctober 1983 i n well HC-2A (Figure 9). This showed the need to c a r e f u l l y control water usage i n the contaminated area.

Publication Date: October 31, 1984 | doi: 10.1021/bk-1984-0267.ch003

Sampling Problems Defining the plume of the contamination was straightforward by using e x i s t i n g domestic wells and a few monitoring wells. However, determining the nature of the problem beneath the p o t l i n e r p i l e and i t s v i c i n i t y was much more d i f f i c u l t . Wells i n s t a l l e d through the p i l e were blocked by debris and were generally i n e f f e c t i v e . Many wells were d r i l l e d around the p i l e i n an attempt to determine the water p r o f i l e . Preliminary results were deceiving because aquatarde and ground water mounding had altered contaminant pathways. The investigators had d i f f i c u l t y determining whether or not leaks were carrying contamination down to the main aquifer. A n a l y t i c a l Problems There i s a discrepancy between the cyanide c r i t e r i a f o r both aquatic and drinking water standards and the current a n a l y t i c a l technology. The c r i t e r i a are stated f o r free cyanide (which includes hydrocyanic acid and the cyanide i o n ) , but the EPA approved a n a l y t i c a l method­ ology f o r t o t a l cyanide measures the free and combined forms (11). This test probably overestimates the potential t o x i c i t y . An alterna­ t i v e method (cyanides amenable to chlorination) measures those cya­ nide complexes which are readily dissociated, but does not measure the i r o n cyanide complexes which dissociate i n sunlight. This method probably tends to underestimate the potential t o x i c i t y . Other meth­ ods have been proposed, but s i m i l a r problems exist (12). The Depart­ ment of Ecology used the EPA-approved APHA procedure which Includes a d i s t i l l a t i o n step for the q u a n t i f i c a t i o n of t o t a l cyanide (13,14). A modification of the procedure which omits the d i s t i l l a t i o n step was used f o r estimation of free cyanide. Later i n the study, the Company used a microdiffusion method f o r free cyanide (15). Another potential problem with cyanide analysis i s the recom­ mended preservation method. The APHA standard method recommends preservation by adjusting to a pH of 12 using sodium hydroxide. The Department'β laboratory has been using t h i s method which i s e f f e c t i v e for t o t a l cyanide but unsatisfactory f o r free cyanide since the pH adjustment can change the cyanide species present, and thus the f i n a l r e s u l t . There i s no adequate preservation method f o r free cyanide. In addition to the need f o r an adequate method f o r free cyanide and an adequate sample preservation method, a methodology should be developed f o r the d i f f e r e n t i a t i o n of species, e s p e c i a l l y between free (HCN and CN~), metallic complexes, and organic complexes.

ENVIRONMENTAL SAMPLING FOR HAZARDOUS WASTES

24

Publication Date: October 31, 1984 | doi: 10.1021/bk-1984-0267.ch003

10.000.00-

Figure 9.

Cyanide levels i n well HC-2A, 1982-83.

3.

BURK.HALTER ET AL.

Assessing Cyanide Contamination

25

Publication Date: October 31, 1984 | doi: 10.1021/bk-1984-0267.ch003

Regulatory Considerations Federal drinking water standards for cyanide have been withdrawn and are not included i n the l a t e s t publication. The Public Health Service l i m i t f o r drinking water had been 200 ppb. Whether the l i m i t was expressed as free or t o t a l cyanide was i n question at the time. The fresh water aquatic cyanide c r i t e r i o n i s 3.5 ppb as a 24-hour average, not to exceed 52 ppb at any time. Potliner waste i s exempt from federal RCRA regulations although through State of Washington testing procedures, the potliner at the Kaiser f a c i l i t y was c l a s s i f i e d as an extremely hazardous waste under state regulations. Because of the ground water contamination of the sole source aquifer, the Environmental Protection Agency has included the Kaiser s i t e i n Mead on i t s Superfund l i s t i n g . The Department of Ecology strongly recommended against Superfund status on the grounds that the EPA s i t e evaluation included a populat i o n Impact based on the number of people who could have been affected i n a three-mile radius instead of the population actually affected taking into consideration the directions of ground water movement. Providing the affected residences with a potable water supply by the Company and the impacts of t o t a l vs. free cyanide were discussed by EPA but were not used i n the impact analysis. An EPA contractor has prepared a draft remedial action master plan (RAMP) for the Mead s i t e . The contractor recommends further exploration to determine i f any undiscovered p o t l i n e r p i l e s exist and also further geological studies. This recommendation i s contingent on cost versus benefit of the action. Conclusions This cyanide contamination case study has been an interesting experience because ground water problems are often slow to develop, and cleanup can be even slower. The major technical problem was the i n a b i l i t y to define subsurface geohydrologic conditions with the i n i t i a l data. Expertise i n the area of geohydrology was c l e a r l y needed. A lack of s p e c i f i c anal y t i c a l techniques precluded meaningful environmental and r i s k assessments. Cleanup efforts were complicated because poltiners are not regulated under RCRA but are regulated under state law. In the middle of the cleanup e f f o r t , the s i t e became involved i n Superfund a c t i v i t i e s , and to date this involvement has not been c l a r i f i e d . Project management has become very d i f f i c u l t because of the many players and laws involved. As a r e s u l t , public confidence has been affected. Cyanide contamination creates special public information problems, e.g. i t i s d i f f i c u l t to explain why cyanide i s not included i n the current drinking water standards but that aquatic organisms are affected at r e l a t i v e l y low cyanide concentration. There i s confusion on whether fresh water standards are based on free or complexed cyanides. Fortunately, the provision of a permanent drinking water supply to each affected household removed r i s k assessment as a major issue.

26

ENVIRONMENTAL SAMPLING FOR HAZARDOUS WASTES

Literature Cited

Publication Date: October 31, 1984 | doi: 10.1021/bk-1984-0267.ch003

1.

"Normals of Precipitation and Evaporation," State of Washington, U.S. Weather Bureau, February 1961. 2. Cline, D. R. Washington Water Supply Bulletin 1969, No.27. 3. Drost, B. W.; H. R. Seitz. U.S.G.S Open Files Report 1978, 77-829. 4. Burkhalter, Richard A; Cunningham, Richard K; Tracy, Harry B. Technical Report; 1970, No. 70-1. 5. "Effect of Waste Disposal Practices on Ground Water Quality at Kaiser Aluminum and Chemical Corporation: Mead, Washington," Robison and Noble, Inc., April 1978. 6. Dalton, Matthew G. Progress Report, Hart-Crowser and Assoc. March 29, 1983. 7. Dalton, Matthew G. Hart-Crowser and Assoc. Report Sept. 9, 1982. 8. Dalton, Matthew G. Summary Report, Hart-Crowser and Assoc., Dec. 6, 1982. 9. Dalton, Matthew G. Letter, Report; Hart-Crowser and Assoc. April 1, 1981. 10. Dalton, Matthew G. Summary Report, Hart-Crowser and Assoc. Sept. 7, 1983. 11. Ingersoll, D. EPA 600/54-83-054, 1983. 12. "Cyanide: An Overview and Analysis of the Literature on Chemistry, Fate, Toxicity, and Detection in Surface Waters," Ecological Analysts, Inc., 1979. 13. "Standard Methods for the Examination of Water and Wastewater," 15th ed, American Public Health Association: Washington D.C., 1975. 14. "Methods for Chemical Analysis of Water and Wastes," EPA/ 4-79-020, Method 335.2, 1979. 15. Palmer, Thomas Α.; Skarset, James Q. Kaiser Aluminum and Chem­ ical Corp"; 1981, No. 81-39. RECEIVED August 16, 1984

4

23,7,8-Tetrachlorodibenzo-p-dioxin S a m p l i n g M e t h o d s

Publication Date: October 31, 1984 | doi: 10.1021/bk-1984-0267.ch004

DANIEL J. HARRIS U.S. Environmental Protection Agency, Region 7, Kansas City, KS 66115

From October 1982 to October 1983, the Emergency Planning and Response Branch of Region 7 of the United States Environmental Protection Agency and its contractors collected approximately 8,000 environmental samples for analysis of 2, 3, 7, 8-tetrachlorodibenzo-para-dioxin (TCDD). The majority of these samples have been collected and ana­ lyzed at an average cost of $700 per sample. This includes per diem, labor, equipment, expendable supplies, transportation, and $400 per analysis by contract laboratories. An evaluation of this data has suggested that field sampling and sample handling methods have a significant impact upon the precision and accuracy of the resulting data which, in turn, impact the cost and feasibility of various remedial options. Some of the results from sampling at depths to determine the extent of vertical migration of TCDD have been puzzling. Depth samples have been collected in 6- to 12-inch increments down to a maximum depth of 4 feet. For the most part, these samples have been collected along road centerlines where TCDD-contaminated waste liquid oil was sprayed for dust control. Although the surface samples (0­ -tο 6-inch depth) have the highest levels of TCDD, those taken at lower depths have also shown contamination. In some instances, TCDD levels have been higher in the deeper layers than in the overlying ones. With no historical data to offer an explanation for such phenomenon, the data and sampling techniques warrant further examina­ tion. This paper discusses the data resulting from a number of compara­ tive sampling techniques which took place at one well-documented TCDD site. Study Area in Missouri The study site, consisting of about 11 contaminated acres belonging to eight property owners, is in a small community midway between Verona and St. Louis, Missouri. Records indicate that on May 20, 1971, a truck driven by an employee of a salvage waste oil company was ticketed for being 950 pounds overweight. The truck was en route from Verona to St. Louis with a load of TCDD-contaminated s t i l l This chapter not subject to U.S. copyright. Published 1984, American Chemical Society

Publication Date: October 31, 1984 | doi: 10.1021/bk-1984-0267.ch004

28

ENVIRONMENTAL SAMPLING FOR HAZARDOUS WASTES

bottome. Later, i t was revealed that to avoid another ticket at a subsequent weigh station, the owner of the salvage o i l company, who accompanied the driver on the t r i p , directed the driver to stop at the owner's farm which i s adjacent to the community. It i s understood that on the farm the exit valve was opened, and the truck was driven along a gravel road to dispose of excess waste. The t o t a l quantity of s t i l l bottoms off-loaded i n t h i s manner over time i s not known. From records of previous loads hauled and truck capacity, i t i s reasonable to estimate that the quantity offloaded could not have exceeded 3,500 gallons. EPA sampling i n 1982 confirmed that a county road i n the area was sprayed as w e l l . This finding s i g n i f i c a n t l y expanded the area of suspected surface contamination. Although there i s no way of knowing with certainty the concent r a t i o n of TCDD i n the off-loaded s t i l l bottoms, the waste holding tank i n Verona from which the trucks were f i l l e d was sampled i n August 1974 and was found to contain an average of 328 ppm TCDD. In t o t a l , 550 analyses were conducted from samples taken at t h i s s i t e . These data indicate that only 5.8 percent of the 10.9 acres contaminated represented the road surfaces o r i g i n a l l y sprayed. The remaining surface contamination probably resulted from dispersion by wind, vehicular t r a f f i c , runoff, etc. The t o t a l TCDD sprayed was probably about 340 grams, with 74 percent s t i l l on the areas sprayed. Mean, volume weighted, TCDD concentrations i n the sprayed and d i s persed areas were 469 and 31 ppb, respectively. Concentrations i n i n d i v i d u a l composite samples collected from sprayed areas ranged up to 1,800 ppb. About 90 percent of the TCDD was contained i n 13 percent of the s o i l volume. Experimental Sampling and Presentation of Data The sampling took place between August 16 and 24, 1983. Most of the 274 samples collected for shipment to contract laboratories were to be analyzed f o r the purpose of more f u l l y delineating contamination boundaries. Other samples were collected f o r comparing sample c o l l e c t i o n and handling techniques. Areal Variation. One objective of the sampling comparison studies was to determine the v a r i a t i o n i n TCDD concentration over a small area to estimate the error associated with a grab sample concentrat i o n . Accordingly, a one-square-yard area was selected adjacent to a previously sprayed road. The center of the test area was about 6 feet from a sprayed road shoulder. This one-square-yard area was divided into nine one-square-foot areas. Using clean spoons and knives, a single scoop was collected from the center of each onesquare-foot area down to a depth of 2 inches. The data are presented i n Figure 1. A l l the a n a l y t i c a l data are from the same laboratory; consequently, interlaboratory a n a l y t i c a l v a r i a t i o n i s not a f a c t o r . The intralaboratory v a r i a t i o n for that laboratory was 9.1 percent ( i . e . , the r e l a t i v e standard deviation based on r e p e t i t i v e analyses of performance evaluation samples). V e r t i c a l Migration. H i s t o r i c a l l y , surface samples at other TCDD s i t e s have been taken to depths of 0 to 6 inches using picks and

Publication Date: October 31, 1984 | doi: 10.1021/bk-1984-0267.ch004

HARRIS

TCDD Sampling Methods

55.6

8.1

18.2

52.9

24.5

52.6

I • 1 CM ι—I

CO

50.2

12"

-

Figure 1.

57

21.6

-

"JO

Lateral v a r i a t i o n i n TCDD concentration, ppb.

ENVIRONMENTAL SAMPLING FOR HAZARDOUS WASTES

30

shovels. The estimated contaminated s o i l volumes used i n deciding remedial actions are dependent upon the actual depth of penetration* To obtain some idea as to the actual v e r t i c a l penetration of TCDD i n those areas contaminated by dispersion, three sample points were selected near the farm road which was sprayed i n 1971· At each point a single sample was collected from each of three depths, v i z , at the 0- to 2-, 2- to 4-, and 4- to 6-inch depths, using clean knives and spoons* The selection of a 2-inch Increment was based upon the estimated removal on a single pass by a q u a l i f i e d operator of earth-moving machinery. The resulting data are as follows:

Publication Date: October 31, 1984 | doi: 10.1021/bk-1984-0267.ch004

9-B Distance from Road Shoulder (feet) DEPTHS (inches) 0-2 2-4 4-6

SAMPLE POINT 4-C 10-C

20

10

15

CONCENTRATION (ppb) 0.3 0.51 30 pg/dl Potential Exposure YES NO

4 3

0 0

7. CARRA

65

Lead Levels in Blood of Children Table XXIV. Percent Lead Toxicity by Distance and Other Potential Sources of Exposure f o r Dixie Site

Publication Date: October 31, 1984 | doi: 10.1021/bk-1984-0267.ch007

Approximate Distance (miles)

Sample Size Potential Exposure YES NO

Percent with Lead Toxicity Potential Exposure YES NO

0.0 - 0.5

74

32

4.3

0.5 - 1.0

134

90

.6

A multivariate analysis (Table XXV) shows the increased bloodlead l e v e l caused by the RSR smelter contribution and the t r a f f i c con­ t r i b u t i o n to be 5.5 and 1.0, respectively.

Table XXV. Results of S t a t i s t i c a l Test for Smelter and T r a f f i c Contribution f o r Blacks Only at RSR

Factor

Parameters Percent with Percent with Blood Lead Toxicity Lead XH^g/dl

Smelter Contribution Significance Level Traffic Contribution Significance Level

5.7% .001

.025% .494

Mean BloodLead Level ^ g / d l )

11.8%

and control charts (6) to v e r i f y the quality of i t s data output. Data produced by different individuals within a laboratory need to be coordinated. This i s a further r e s p o n s i b i l i t y of laboratory management· Monitoring programs must have t h e i r own quality assurance programs. These may be called project q u a l i t y assurance plans or protocols f o r s p e c i f i c purposes (3). If r e l i a b l e vendors of services are used, the bulk of the quality assurance e f f o r t can be placed on those a c t i v i t i e s unique to the program. Without r e l i a b l e vendors, QA e f f o r t s w i l l be i n e f f e c t i v e since i t i s not cost e f f e c t i v e to police q u a l i t y assurance practices at a l l lower levels nor to screen a l l data f o r i t s v a l i d i t y . Monitoring Program Responsibilities A prime r e s p o n s i b i l i t y of the o v e r a l l program management i s determinat i o n of program objectives. Data q u a l i t y objectives must be set and r e l i a b l e vendor laboratories selected, evaluated for their i n i t i a l q u a l i f i c a t i o n s , and audited for t h e i r ongoing proficiency. The monitoring program must also provide f o r an i n t e r c a l i b r a t i o n program which involves the vendor laboratories. Reliable vendors w i l l maint a i n t h e i r own quality assurance programs and provide evidence of the q u a l i t y of their data. The program QA plan w i l l specify the control charts and other records that should be maintained by the vendors and used i n judging the quality of the data output. The program must require the vendors to measure a number of reference samples and/or duplicates submitted i n a planned sequence. It should require prompt measurement and reporting of these data and should maintain the results i n a control chart format. Prompt feedback and follow-up of any apparent data discrepancies and r e c o n c i l i a t i o n of the results with control charts maintained by the vendors are required to minimize the length of uncertain performance. The q u a l i t y assurance plan should include random sampling of the vendors' data f o r their v a l i d i t y and conformance with quality assurance r e quirements. If quality assurance i s properly practiced at a l l l e v e l s , an inspection of 5 percent of the t o t a l data output should be adequate. System audits of vendors quality assurance programs should be made i n i t i a l l y and at random intervals throughout the duration of a monitoring program. The general procedure described by Gaft and Richards i s recommended for this purpose (_7). The monitoring program plan should specify that vendors conduct internal systems audits s i m i l a r to the above, but at more frequent i n t e r v a l s . Records of v

Publication Date: October 31, 1984 | doi: 10.1021/bk-1984-0267.ch011

11.

TAYLOR

Quality Assurance for a Measurement Program

107

these should be available f o r inspection on request. Such a procedure would minimize the number of system audits that need to be conducted. Since the use of competent vendors i s the key to a successful monitoring program, a system f o r t h e i r p r e - q u a l i f i c a t i o n i s necessary. Only laboratories experienced and s k i l l e d i n the methodology to be used should be considered. Candidate laboratories must present e v i ­ dence of proficiency equal to or exeeding the minimum requirements of the program. Retention of vendor relationships and payment for ser­ vices should be contingent on maintenance of proficiency standards, s p e c i f i e d i n advance. Faulty data are unworthy of payment. Program management w i l l need to be f a i r i n i t s judgment of such matters and may need to e s t a b l i s h a r b i t r a t i o n procedures i n this regard. The program quality assurance plan may need periodic or emergency r e v i s i o n s . Ongoing review of the data should reveal whether any d e f i c i e n c i e s are due to inadequate performance of vendors or to defects i n the quality assurance plan. Defects i n the plan could result from Inadequate quality assessment techniques i f measured levels of contaminants were s i g n i f i c a n t l y different from anticipated l e v e l s , f o r example· Reference Laboratory The monitoring program may find i t advantageous to use a reference laboratory to carry out some of the r e s p o n s i b i l i t i e s outlined above. A reference laboratory may be used for a r b i t r a t i o n and as a t h i r d party participant to provide smooth operations of the program. A reference laboratory should have unquestioned competence and demon­ strated experience related to a l l aspects of the monitoring program. Above a l l , i t should be recognized f o r i t s o b j e c t i v i t y i n s c i e n t i f i c investigations, must have the f a c i l i t i e s and equipment to conduct r e l i a b l e referee measurements, and should have the capability for real-time responses to questions as they a r i s e . To the extent that i t might be delegated overall r e s p o n s i b i l i t i e s , i t should have pro­ gram management experience, as w e l l . Obviously, i t must have i t s own impeccable q u a l i t y assurance program which can serve as a p i l o t for the monitoring program and as a model f o r the vendors. The duties of the reference laboratory might include the follow­ ing t y p i c a l assignments: ο Select methodology ο Pre-teet methodology ο Develop reference materials ο Develop protocols ο Pre-teet protocols ο Set performance standards ο Qualify candidate laboratories ° Conduct performance and systems audits ο Ongoing data appraisal ° Troubleshooting of problems ° Data release decisions ο Training of vendors

ENVIRONMENTAL SAMPLING FOR H A Z A R D O U S WASTES

108

Publication Date: October 31, 1984 | doi: 10.1021/bk-1984-0267.ch011

Conclusion Monitoring operations have considerable analogy to modern manufacturing operations involving a number of subcontractors or vendors of services. The monitoring program should establish i t s own quality standards and specify minimum quality assurance practices that the vendor of services should follow. The monitoring program should also supply reference materials f o r i n t e r c a l i b r a t i o n of the vendors and maintain control charts to monitor i n t e r c a l i b r a t i o n and to detect any deviations from p r o f i c i e n t performance. Only experienced vendor laboratories, pre-qualifled by objective testing procedures, should be used. Their maintenance of proficiency should be monitored throughout the program and t h e i r retention as vendors should be contingent on their a b i l i t y to supply data of specified q u a l i t y . Laboratories should not be paid f o r data of i n s u f f i c i e n t or undocumented quality. On occasion, an adequate number of q u a l i f i e d vendors may not be a v a i l a b l e . In these cases, the monitoring program management must select and t r a i n the most l i k e l y candidates prior to their acceptance. In no case should monitoring programs be i n i t i a t e d with unqualified vendors and pressures to do so should be strongly reelsted.

Literature Cited 1. 2. 3. 4. 5. 6. 7.

Taylor, J . K. Anal. Chem. 1981, 53, 1588A-96A. Taylor, J . K. "In Quality Assurance for Environmental Measurements"; ASTM STP 867: Philadelphia, PA, 1984. Taylor, J . K. In "Statistics in Environmental Science"; ASTM STP 867: Philadelphia, PA. Deming, W. Edwards. "Quality Productivity and Competitive Positions"; MIT Press: Cambridge, MA. Taylor, J . K. J . Testing and Evaluation 1983, 11, 355-7. "Presentation of Data and Control Analysis", ASTM STP 15D: Philadelpia, PA, 1976. Graft, S.; Richards, F. D. In "Evaluation and Accreditation of Inspection and Test Activities"; ASTM STP 814: Philadelphia, PA, 1984.

RECEIVED August 22, 1984

12

N e w W a y s o f Assessing Spatial Distributions o f Pollutants

Publication Date: October 31, 1984 | doi: 10.1021/bk-1984-0267.ch012

ANDRE G. JOURNEL Stanford University, Stanford, CA 94305 In the last few years, detection, characterization, and handling of pollution sites have grown into a new discipline with its specific technology, initially drawn from other disciplines such as statistics, then customized to meet the specificity of pollution control. A body of specific techniques already exists for sample design and data capturing. However, the algorithms used to interpret these data and map them are still very classical. Passkey interpolation algorithms are most often used even though they ignore the pattern of spatial distribution and correlation particular to each pollution plume. Arbitrary risks α and β are used to make critical decisions, as if agricultural standards, for example, are also valid when deal­ ing with health hazards. Geostatistical techniques, such as variography and kriging, have been recently introduced into the environmental sciences (1). A l ­ though kriging allows mapping of the pollution plume with qualifica­ tion of the estimation variance, it falls short of providing a truly risk-qualified estimate of the spatial distribution of pollutants. Ideally, to characterize the spatial distribution of pollution, one would like to know at each location x within the site the probability distribution of the unknown concentration p(x). These distributions need to be conditional to the surrounding available information in terms of density, data configuration, and data values. Most traditional estimation techniques, Including ordinary kriging, do not provide such probability distributions or "likelihood" of the unknown values p(x). Utilization of these likelihood functions towards assessment of the spatial distribution of pollutants is presented first; then a non-parametric method for deriving these likelihood functions is proposed. The Process of Estimation Pollutant concentrations, just like metal grades in a mineral deposit, can be seen as variables which are functions of their spatial coordi­ nates χ " (u,v) or χ " (u,v,w). Consider a particular pollutant ρ and its concentration value p(x) at location χ· 0097^6156/84/0267-0109$06.00/0 © 1984 American Chemical Society

ENVIRONMENTAL SAMPLING FOR HAZARDOUS WASTES

Publication Date: October 31, 1984 | doi: 10.1021/bk-1984-0267.ch012

110

It i s usual to v i s u a l i z e the s p a t i a l d i s t r i b u t i o n of p(x) within an area A by contouring measured or estimated values p*(xjc) at the nodes xfc of a g r i d . The resulting estimated map necessarily d i f f e r s from the "true" map that would have been obtained from the true values p(xjc). Thus, there appears a need to characterize the potential de­ parture p(xjc)-p*(xjc) between each estimate and the corresponding true value. Furthermore, there i s a need to characterize the j o i n t depar­ ture of a l l estimates from a l l true values. Indeed, a map l o c a l l y precise ( f o r each x^) i s not necessarily the most precise f o r a global c r i t e r i o n such as the reproduction of s p a t i a l trends. P r o b a b i l i s t i c techniques of estimation provide some insights i n t o the potential error of estimation. In the case of k r i g i n g , the variable p(x) spread over the s i t e A i s f i r s t elevated to the status of a random function P(x). An estimator P*(x) i s then b u i l t to minimize the estimation variance E{[P(x)-P*(x)] }, defined as the expected squared error (2). The k r i g i n g process not only provides the estimated values ρ(χ^) from which a "kriged" map can be produced, but also the corresponding minimum estimation variances σ£(χ}ς) Ε{[Ρ(χ]ς)-Ρ*(χ^) (xjc)] }. These variances o*£(xje) can also be con­ toured, thus providing a map of iso-values of r e l i a b i l i t y (see Figure 1). 2

2

The Limitation of Data-free Estimation Techniques A great deal has been written about the optimality of the k r i g i n g e s t i ­ mates and about the error characterization provided by iso-variance maps such as that of Figure 1. However, k r i g i n g , and more generally data-free estimation algorithms, have several drawbacks. 1.

2.

3.

F i r s t l y , the k r i g i n g estimator i s optimal only for the least square c r i t e r i o n . Other c r i t e r i a are known which y i e l d no more complicated estimators such as the minimization of the mean ab­ solute deviation (mAD), E{|P(x)-P*(x)I}, y i e l d i n g median-type regression estimates. Secondly, knowledge of the estimation variance E{[P(x)-P*(x)1 } f a l l s short of providing the confidence i n t e r v a l attached to the estimate p*(x). Assuming a normal d i s t r i b u t i o n of error i n the presence of an i n i t i a l l y heavily skewed d i s t r i b u t i o n of data with strong s p a t i a l correlation i s not a viable answer. In the absence of a d i s t r i b u t i o n of error, the estimation or "kriging" variance o*^(x_) provides but a r e l a t i v e assessment of error: the error at l o c a t i o n χ i s l i k e l y to be greater than that at location x_" i f σ£(χ)>σ£(χ/). Iso-variance maps such as that of Figure 1 tend to only mimic data-position maps with bull's-eyes around data locations. T h i r d l y , and most importantly, both the k r i g i n g algorithm and k r i g i n g variance are independent of data values, thus not d i s t i n ­ guishing the estimation of extreme values (usually the most important for pollution control) from the estimation of median or background values. 2

Consider, for example, a s i t e characterized by a highly skewed d i s t r i b u t i o n of pollutant concentrations, as apparent i n the histogram of data values of Figure 2a. These values present a c o e f f i c i e n t of

Publication Date: October 31, 1984 | doi: 10.1021/bk-1984-0267.ch012

JOURNEL

1

Assessing Spatial Distributions of Pollutants

Figure 1. Isopleths of k r i g i n g estimation variances. bull's-eyes r e f l e c t the data locations.)

(The

Coefficient of variation >3

z(ppm Pb)

Mean 300 ppm

2a · Typical example of histogram of pollutant data (note the strong positive skewness)

Figure 2. values.

Ρ(*Ί) =

* P(*i) = 100 ppm

x

"Ό0 pom

ο p(x) unknown

ο p(i') unknown

xp(xj) = 100 ppm

χ p(x.' ) ~ 2.000 ppm

2b - Two medianvalued data

2c - Two widely different data values

2

The weighting algorithm should be dependent on data

112

ENVIRONMENTAL SAMPLING FOR HAZARDOUS WASTES

v a r i a t i o n (σ/m) greater than 3, which i s not an uncommon figure f o r p o l l u t i o n data. Consider then the two i d e n t i c a l data configurations of Figures 2b and 2c, where a nodal point χ i s estimated by two equi­ distant data points at locations χ and x Kriging, just l i k e more t r a d i t i o n a l estimation technique, gives an equal weight (1/2) to each of the two data and the same estimation variance value for both con­ figurations. Next, consider that the f i r s t configuration (Figure 2b) includes two median-type data values, say around 100 ppm Pb, whereas the second configuration (Figure 2c) includes a median-type datum value (100 ppm Pb) with an o u t l i e r datum value (2,000 ppm Pb). Clearly the potential f o r error attached to the second r e a l i z a t i o n i s greater; i n other word8, the estimation variance should be dependent on data values and not only dependent on data configuration. Also, there i s no reason f o r the two weights of the second r e a l i z a t i o n to be equal, that i s the weights should also be dependent on data values. In some cases, e.g. nugget-type gold mineralizations, the very high grades have a smaller range of s p a t i a l correlation than background to median grades; other cases, e.g. high concentrations of pollutants, have a larger range of s p a t i a l correlation linked to the pollution plume whereas the background concentrations tend to be more e r r a t i c i n t h e i r spatial distribution. In the f i r s t cases, the high datum value of Figure 2c should be downwelghted; i n the l a t t e r cases, i t i s the lower datum value that should be downwelghted.

Publication Date: October 31, 1984 | doi: 10.1021/bk-1984-0267.ch012

2

The Conditional D i s t r i b u t i o n Approach One way to introduce the data values into the estimation algorithm i s to consider the conditional probability d i s t r i b u t i o n of the unknown P(x), given the Ν data values used to estimate i t . Denote this conditional cumulative d i s t r i b u t i o n function (cdf) by: F (z|(N)) - Prob{P(x) < ζ|Ρ(χ ) - ρ ,... ,P(xtf) - p } x

χ

χ

N

(1)

This conditional cdf i s a function not only of the data configu­ ration (N locations i-l,...,N) but also of the Ν data values (p*, i«l,...,N). Its derivative with regard to the argument ζ i s the con­ d i t i o n a l probability density function (pdf) and i s denoted by: ÔF (z/(N)) f (z/(N)) - — * δζ x

(2)

x

This conditional pdf can be seen as the l i k e l i h o o d function of the unknown value p(x). Moments of this conditional d i s t r i b u t i o n can be written as stand­ ard Riemann integrals of the pdf f ( z | ( N ) ) or as S t i e l t j e s integrals of the cdf F ( z | ( N ) ) . For example", the conditional expectation i s written: x

x

E{P(x)|(N)} « /~z.f (z|(N))dz x

and i s

a

function

(usually

non-linear)

/7z.dF (z|(N)), x

of

the

Ν

data

(3) values.

12. JOURNEL

113

Assessing Spatial Distributions of Pollutants

Techniques f o r derivation or estimation of the cdf F (z|(N)) are presented i n a l a t e r section. For now, i t i s shown how knowledge of that cdf provides a solution to the three problems previously mentioned. x

1.

F i r s t l y , various c r i t e r i a f o r estimation, d i f f e r e n t from the least square E{[P(x)-P*(x)] }, may now be considered. Consider a general loss function L(e), function of the error of estima­ t i o n e - p(x) -p*(x). The objective i s to build an estimator that would minimize the expected value of that loss function, and more precisely, i t s conditional expectation given the Ν data values and configuration.

Publication Date: October 31, 1984 | doi: 10.1021/bk-1984-0267.ch012

2

E{L(P(x)-P*(x))|(N)}

- /7L(z-p*(x)).f (z|(N))dz 2

(4)

For any predetermined loss function L, this conditional expec­ t a t i o n appears as a function of both the estimated value p*(x) and the Ν data values, , • i«l,...,N. The optimization process consists of determining the p a r t i c u l a r value p*(x) that would minimize expression ( 4 ) . The solution i s straightforward f o r some p a r t i c u l a r functions L: 2

If L(e) • e , i . e . the loss i s proportional to the squared error, the least square c r i t e r i o n i s apparent, and the best estimator P(x) i s the conditional expectation defined i n (3). Note that this estimator i s usually different from that provided by ordinary k r i g i n g f o r the simple fact that expression (3) i s usually non-linear i n the Ν data values. If L(e) - |e|, i . e . the loss i s proportional to the abso­ lute value of the error, the best estimator i s the condi­ t i o n a l median, i . e . the value: q^ (x) such that: 5

F ( q ( x ) | ( N ) ) - .5 x

e5

(5)

Again t h i s conditional median estimator i s usually a non­ l i n e a r function of the Ν data values. If L(e) « 0 f o r e • 0, and i n f i n i t y f o r any error d i f f e r e n t from zero, the best estimator i s the maximum l i k e l i h o o d estimator, i . e . the value z f o r which the conditional pdf f ( z | N ) ) i s a maximum. Q

x

More generally, the loss function need not be symmetric: L(e) Φ L(-e). Indeed, underestimation of a pollutant concentration may lead to not cleaning a hazardous area with the r e s u l t i n g health hazards. These health hazards may be weighted more than the costs of cleaning unduly due to an overestimation of the p o l l u ­ tant concentration. The optimal estimators linked to asymmetric l i n e a r loss functions are given i n Journel (3>). 2. Secondly, knowledge of the conditional cdf F (z|(N)) provides the confidence intervale which are data values-dependent but independent of the p a r t i c u l a r estimate p*(x) retained: x

Prob{P(x)e]q (x), q

(x)]|(N)}

w

1-w

- l-2w,

(6)

ENVIRONMENTAL SAMPLING FOR HAZARDOUS WASTES

114

with q (x) being the conditional w-quantile such that: w

F

x(qw(x)l< » - we[0,l]. N

The estimate p*(x) retained need not be at the center of the confidence i n t e r v a l ] q ( x ) , q _ ( x ) J . Since the confidence i n ­ tervals are obtained d i r e c t l y , ^there i s no need to calculate the estimation variance, nor to hypothesize any model f o r the error distribution. Thirdly, and most importantly, the conditional cdf F (z|(N)) i s dependent on data values accounting f o r the possible d i f ­ ference i n s p a t i a l c o r r e l a t i o n between high-valued concentration data and background concentration data. This fact stems from the process of estimation of F (z|(N)) i t s e l f described i n a l a t e r section. w

Publication Date: October 31, 1984 | doi: 10.1021/bk-1984-0267.ch012

3.

x

x

x

Risks of Incorrect Decisions The a v a i l a b l i l i t y of an estimate of the conditional cdf F (z|(N)) at each nodal j o i n t χ allows an assessment of the r i s k s a(x_) or β(χ) of making wrong decisions. Consider the contour map of a p a r t i c u l a r estimate p*(x), χ ε A (see Figure 3a). Suppose that the threshold value 500 has been selected to declare any sub-area of A hazardous. The contour l i n e 500 on Figure 3 delineates the zones which are candidates f o r cleaning. Within these zones, the probability that the concentration i s actually under 500, i . e . the r i s k 500| (N)|. expressed in percent (thin lines) plotted for i : p*(g)>500 ppm (thick line)

Figure 3. Isopleth maps of the concentration estimate and the associated probability (1 - a(x)) to make a correct decision to clean.

116

E N V I R O N M E N T A L S A M P L I N G FOR

H A Z A R D O U S WASTES

model i s the multi--normal d i s t r i b u t i o n which assumes that the normal score transform random function U(x) - φ(Ρ(χ)) i s multinomial d i s t r i b u t e d . As a consequence, the multivariate d i s t r i b u t i o n of U(x) and hence of P(x) i s f u l l y characterized by the covariance function C(h) • E{U(x).U(x + h)}, experimen­ t a l l y inferred from the normal score data. Knowledge of the multivariate d i s t r i b u t i o n of P(x) allows exact derivation (not an estimation) of any conditional cdf F ( z | ( N ) ) , (4,5). The lognormal version (φ * Ln) of this approacR i s of current use i n the mining industry (6). Of course the whole procedure i s d i s ­ tribution-dependent, c a p i t a l i z i n g on the i n i t i a l multi^-normal d i s t r i b u t i o n hypothesis. The second approach consists of estimating sequentially the values F (zfc|(N)) for a series of cut-off values z^ covering the range" of v a r i a b i l i t y of the concentration values, usually [0, z < 100%]. The key idea i s to interpret the conditional cdf F (z|(N)) as the conditional expectation of an indicator transform I ( x j z ) which can be estimated from the corresponding indicator data. Indeed, consider the indicator transform:

Publication Date: October 31, 1984 | doi: 10.1021/bk-1984-0267.ch012

x

x

m a x

x

1, i f P(x) < ζ

|

Kx;z) 0, i f not

(9)

Then: E{l(x;z)} « 1 χ Prob{P(x) < ζ} + 0 χ Prob{P(x) > z} » Prob{P(x) < z} - F(z) Similarly: E{l(x;z)|(N)} - Prob{P(x) < z|(N)} - F (x|(N)) x

(10)

The projection theory indicates that a projection estimator of the unknown I(x;z), such as a k r i g i n g estimator, i s also a projection estimator (kriging) of i t s conditional expectation F ( z | ( N ) ) , (7). Consider then the indicator k r i g i n g estimator: x

[I(x;z)l*

«

Ν l i»l

\ (ζ).ηχχ;ζ) ±

(11)

The weights \±(z) are cut-off z-dependent, and are determined as solutions of a l i n e a r system (8). In f a c t , the indicator estimators used for the case-study underlying Figures 3 were obtained by a- s l i g h t l y more elaborate technique called "prob­ a b i l i t y k r i g i n g or PK" (9, 10). The k r i g i n g process (11) i s repeated as many times as there are d i f f e r e n t cut-offs (z^) retained to d i s c r e t i z e the i n t e r v a l of v a r i a ­ b i l i t y of the concentration P(x). The d i f f e r e n t kriging estimates i ( x ; z k ) * are then pieced together to provide an estimate of the con­ d i t i o n a l cdf F ( z | ( N ) ) . x

12.

JOURNEL

117

Assessing Spatial Distributions of Pollutants

Note that i f the kriging weights λ^(ζ) are data valuesindependent, the indicator estimates (11) are not; hence, the f i n a l estimate of the conditional cdf F (z|(N)) i s data values-dependent. For each cut-off z^, the k r i g i n g estimator (11) requires a d i f f e r e n t indicator covariance model: x

C^hjz^

2

- E{l(x;z ). K x +h;z )} - F ( z ) k

k

k

(12)

These various covariance models are inferred d i r e c t l y from the corresponding indicator data ±(^χ;ζ^) i - l , . . . , N . The indicator k r i g ­ ing approach i s said to be "non-parametric," i n the sense that i t draws solely from the data, not from any multivariate d i s t r i b u t i o n hypothesis, as was the case for the multi--normal approach. These indicator covariance models allow d i f f e r e n t i a t i o n of the s p a t i a l correlation of high-valued concentrations (cut-off z^ high) and low to background-valued concentrations ( z low). In the p a r t i c u ­ l a r case study underlying the Figure 3, i t was found that high value concentration data were more s p a t i a l l y correlated (due to the plume of p o l l u t i o n ) than lower value data.

Publication Date: October 31, 1984 | doi: 10.1021/bk-1984-0267.ch012

9

k

Conclusions Assessment of s p a t i a l d i s t r i b u t i o n s of pollutant concentrations i s a very s p e c i f i c problem that requires more than blind mapping of these concentrations. Not only must the c r i t e r i o n of estimation be chosen c a r e f u l l y to allow zooming on the most c r i t i c a l values (the high concentrations), but also the evaluation of the potential error of estimation c a l l s f o r a much more meaningful c h a r a c t e r i s t i c than the t r a d i t i o n a l estimation variance. F i n a l l y , the r i s k s α and β of making wrong decisions on whether to clean or not must be assessed. An answer to the previous problems i s provided by the conditional d i s t r i b u t i o n approach, whereby at each node χ of a grid the whole l i k e l i h o o d function of the unknown value p(x) i s produced Instead of a single estimated value p*(x). This l i k e l i h o o d function allows d e r i ­ vation of d i f f e r e n t estimates corresponding to different estimation c r i t e r i a (loss functions), and provides data values-dependent c o n f i ­ dence i n t e r v a l s . Also t h i s l i k e l i h o o d function can be used to assess the risks α and β associated with the decisions to clean or not.

Literature Cited 1. Flatman, G. T. Proc. of Workshop on Environmental Sampling, EPA Las Vegas, February 1984. 2. Brooker, P. I. In "Geostatistics"; McGraw Hill, 1980; pp. 41-60. 3. Journel, A. G. In "Geostatistics for Natural Resources Charac­ terization"; Verly G. et al., Ed., Reidel: Dordrecht, Nether­ lands, 1984; Part 1, pp. 261-270. 4. Anderson, T. W. "An Introduction to Multivariate Statistical Analysis"; Wiley & Sons: New York, 1958; p. 374. 5. Verly, G. W. Math Geology 1983, 15, 2, p. 263-290. 6. Parker, H. M.; Journel, A. G; and Dixon, W. L. Proc. of the 16th APCOM Symp., AIME, New York, 1979, pp. 133-148, ed. 7. Luenberger, D. L. "Optimization by Vector Space Methods"; Wiley & Sons: New York, 1969; p. 326.

118 8. 9. 10.

ENVIRONMENTAL SAMPLING FOR HAZARDOUS WASTES

Journel, A. G. Math. Geology 1983, 15, 3, p. 445-468. Sullivan, J . In "Geostatistics for Natural Resources Charac­ terization"; Verly, G., et a l . , Ed.; Reidel: Dordrecht, Nether­ lands, 1984; Part 1, pp. 365-384. Isaaks, Ε. H. MSc Thesis, Stanford University, 1984.

Publication Date: October 31, 1984 | doi: 10.1021/bk-1984-0267.ch012

RECEIVED August 6, 1984

13

Detecting Elevated C o n t a m i n a t i o n b y C o m p a r i s o n s with B a c k g r o u n d WALTER LIGGETT

Publication Date: October 31, 1984 | doi: 10.1021/bk-1984-0267.ch013

National Bureau of Standards, Washington, DC 20234 The objective of the studies considered in this paper is detection of excess levels of a common soil contaminant. To meet this objective, these studies must include the sampling and measurement of a background region so that detection can be based on comparison of the background measurements with measurements on the suspected region. Thus, the background measurements are important in the studies considered, much more important than in other studies such as those aimed at determining the spatial extent of a highly contaminated area. To compare the measurements from the two regions, the studies discussed use statistical methods based on the following model: The background measurements are observations drawn at random from some population. Since the background measurements consist of actual background concentrations perturbed by subsampling and measurement error, this model requires that both the actual background concentrations and the subsampling and measurement errors be random. The contamination in the suspected region is the sum of the background contamination and any excess contamination. Measurements on the suspected region are also perturbed by subsampling and measurement error. Thus, in the absence of excess contamination, the measurements from the suspected region are just further random observations from the background population. The two regions can be compared by asking whether the measurements from the suspected region could be observations from the same population as the background measurements. Detection follows from the conclusion that the two sets of measurements could not reasonably be from the same population. In statistical terms, detection is rejection of the null hypothesis that the two sets are from the same population. Background measurements that are obtained from soil samples may have an asymmetrical distribution, that is, a skew distribution. Typically, the asymmetry is characterized by a positive skewness and by a higher probability of unusually large values than of unusually small values. This asymmetry may originate from the distribution of the contaminant over the background region or the procedure used to extract a subsample for laboratory analysis. Several authors discuss the relation of sampling and subsampling to the variability of This chapter not subject to U.S. copyright. Published 1984, American Chemical Society

Publication Date: October 31, 1984 | doi: 10.1021/bk-1984-0267.ch013

120

ENVIRONMENTAL SAMPLING FOR HAZARDOUS WASTES

measurements Ο-Α)· These authors also review other work on sampling and subsampling. Background measurements with a skew probability d i s t r i b u t i o n generally require an appropriate s t a t i s t i c a l method. In p a r t i c u l a r , a s t a t i s t i c a l method based on appropriate assumptions about the upper t a i l of the background d i s t r i b u t i o n i s needed i f the objective i s to detect a few hot spots i n the suspected region on the basis of the highest measurements from that region. In this context, the upper t a i l of the background d i s t r i b u t i o n i s important because detection should not occur when the highest measurements from the suspected region might reasonably be observations from the upper t a i l of the background d i s t r i b u t i o n . Appropriate s t a t i s t i c a l methods can be based on transformation to normality (_5,6)« Other methods might be based on a skew d i s t r i b u t i o n such as the gamma d i s t r i b u t i o n (7). In t h i s paper, we discuss studies based on comparison with background measurements that may have a skew d i s t r i b u t i o n . We discuss below the design of such a study. The design i s intended to Insure that the model for the comparison i s v a l i d and that the amount of skewness i s minimized. Subsequently, we present a s t a t i s t i c a l method for the comparison of the background measurements with the largest of the measurements from the suspected region. This method, which i s based on the use of power transformations to achieve normality, i s o r i g i n a l i n that i t takes into account estimation of the transforma­ t i o n from the data. The Model and Its V a l i d i t y To provide a framework for the design of the study, l e t us specify the model f o r the measurements i n s t a t i s t i c a l terms. The background measurements, which we denote by XBI> i • 1>···>ηΒ> random 8 amp l e from some population. The measurements from the suspected region, denoted by x s i , i • 1,···,η§, are, i n the case of no excess contamination, a second, independent random sample from this same population. The study should be designed so that this model can be applied i n the data analysis. This goal has various ramifications. First, the background region should be generally the same as the suspected region except that the background region must have no excess contami­ nation. Second, the same sampling, subsampling, and measurement procedures must be used to obtain a l l the measurements. Third, these procedures should minimize skew i n the background d i s t r i b u t i o n and should reduce s e n s i t i v i t y of the measurements to any unavoidable extraneous differences between the two regions. Consider f i r s t the choice of the background region. For concretenees, consider the case of large areas characterized by shallow pollutant deposition, one of the cases discussed by Mason (8). C l e a r l y , the background region should have the same average contam­ i n a t i o n l e v e l as the suspected region would have were there no excess contamination. The background region should not d i f f e r from the suspected region i n some c h a r a c t e r i s t i c such as s o i l type or vegeta­ t i o n that i s extraneous to the question of excess contamination but might cause the two sets of measurements to d i f f e r . The contamination i n the background region should have no large-scale s p a t i a l variations since two regions are not l i k e l y to be comparable i f one exhibits a

r

e

a

Publication Date: October 31, 1984 | doi: 10.1021/bk-1984-0267.ch013

13. LIGGETT

Detecting Elevated Contamination

121

large-scale variations. If the suspected region exhibits large-scale s p a t i a l variations, then i t should be divided into more homogeneous subregions and a matching background region found f o r each subregion. Ideally, the two regions should be i d e n t i c a l i n the absence of excess contamination. However, at best, the two regions w i l l exhibit i r r e g u l a r s p a t i a l variations that are d i f f e r e n t . Under these circumstances, the points at which the samples are collected could be chosen i n such a way that the comparison would be i n v a l i d . One way to guard against this p o s s i b i l i t y i s random selection of the sampling points. If the regions d i f f e r only i n their patterns of i r r e g u l a r s p a t i a l v a r i a t i o n , then random selection could plausibly produce sets of observations that would appear to be from the same population. Although random selection from the background region seems reasonable, random s e l e c t i o n from the suspected region might be rejected i n favor of a sampling plan based on knowledge of how the excess contamination could have been deposited. The design goal of insuring that the s t a t i s t i c a l model i s appropriate obviously requires that the same physical procedures f o r sampl i n g and subsampling be used i n both regions. This design goal has other Implications for the choice of the sampling and subsampling procedures. In p a r t i c u l a r , a sampling procedure that c a l l s for the c o l l e c t i o n of a large amount of s o i l may result i n measurements that are less sensitive to extraneous differences between the two regions Another reason to choose large samples i s to reduce the skewness of the background measurements. The size of the samples i s limited by the size of the hot spots that are to be detected. In samples that are too large, the presence of the excess contamination might be washed out. Smith and James (J_) discuss the biases inherent i n the use of various sampling instruments. If these biases are the same for both regions, these biases are of less concern In a comparison than i n other cases such as the estimation of an average. Generally, only a subsample of each sample collected i s chemic a l l y analyzed. To obtain these subsamples, we need a carefully designed subsampling procedure. In some cases, the difference i n mass between the sample and the subs amp l e i s very large (3>). For this reason, the subsampling procedure must include vigorous grinding and mixing so that the subsampling does not introduce too much v a r i a b i l i t y and asymmetry. A review of various design c r i t e r i a f o r the minimum eubsample mass has been made (J_). This mass i s given i n terms of the s i z e to which the p a r t i c l e s i n the sample have been reduced and the amount of error that can be tolerated. These design c r i t e r i a assume thorough mixing even though this i s sometimes hard to achieve (9). The lack of thorough mixing may cause the subsampling procedure to be biased. In a comparison, t h i s bias w i l l balance out i f i t i s the same for both regions. However, the bias might depend on some spurious factor that i s not the same i n both regions. The choice of a chemical analysis procedure i s also part of the design. We w i l l not discuss the additional d i f f i c u l t i e s the chemical analysis procedure can introduce. In t h i s paper, we assume that the chemical analysis procedure has adequate s e n s i t i v i t y and r e l a t i v e l y small e r r o r . The l i t e r a t u r e on sampling bulk materials i s more s p e c i f i c and more quantitative than the general discussion presented above. Nevertheless, experiments on the sampling and subsampling procedures

ENVIRONMENTAL SAMPLING FOR H A Z A R D O U S WASTES

122

are often a necessary part of the design process, Liggett, et a l . (10) discuss experiments on subsampling procedures. They consider one-to-twenty gram subsamples of plutonium-contaminated s o i l and show that despite vigorous grinding and mixing, the subsamples exhibit asymmetrical subsampling error. The question of what steps to take to reduce asymmetry i n the d i s t r i b u t i o n of the background measurements involves a cost-benefit tradeoff that can only be decided by experi­ ments that show how much reduction i s given by various modifications of the sampling and subsampling procedures.

Publication Date: October 31, 1984 | doi: 10.1021/bk-1984-0267.ch013

A Method f o r Detecting Hot Spots Various s t a t i s t i c a l methods can be used to compare the measurements from the suspected region with the background measurements. The power of these methods depends on how the excess contamination i s d i s t r i ­ buted over the suspected region. Comparison of the means of the measurements from the two regions i s appropriate i f a uniform d i s t r i ­ bution of excess contamination i s expected. Comparison of the maximum measurements from the suspected region with the background measure­ ments i s appropriate i f a few hot spots with high contamination levels are expected. This l a t t e r case, which i s the one considered i n this section, i s one i n which positive skewness i n the d i s t r i b u t i o n of the background measurements cannot be ignored. The best known approach to measurements with positive skewness i s transformation. In environmental data analysis, the measurements are often transformed to their logarithms. In this paper, we consider power transformations with a s h i f t , a set of transformations that includes the log transformation and no transformation at a l l ( 5 ) . These transformations are given by λ

[(χ - τ ) - 1]/λ log (χ - τ)

for λ > 0 f o r λ - 0.

(1)

This set of transformations depends on two parameters, the power parameter λ and the s h i f t parameter τ. We r e s t r i c t λ to non-negative values to avoid having a f i n i t e upper bound on the y values. (When λ i s negative, the point y « -Ι/λ corresponds to χ » «>.) Also, τ must be less than the smallest value of x. When τ i s positive, we can think of the measurements as the sum of a constant component with l e v e l τ and a variable component. Since the constant component can­ not have a negative l e v e l , r e s t r i c t i n g τ to positive values makes sense i n some contexts. In s o i l sampling, we need a more general set of transformations than just the log transformation because the amount of skewness and therefore the proper transformation depends on the choice of sampling and subsampling procedure. The transformations given by Equation 1 seem to f u l f i l l this need. Although we usually do not know the values of λ and τ that trans­ form the background measurements to normality, l e t us consider the case i n which we do. Using Equation 1, we transform the measurements X B i and to y^i and respectively. To compare the maximum

xsi

ysi>

13.

LIGGETT

123

Detecting Elevated Contamination

measurement from the suspected region with ments, we form the t s t a t i s t i c s (

*J " ? S j " V where m

B

^

( 1

+

n

1/2

( 2 )

!/ B> >'

and sg are given by *B * ll

v

n

Bi/ B > 2

SB «

Publication Date: October 31, 1984 | doi: 10.1021/bk-1984-0267.ch013

B

the background measure­

(li(YBi - m ) / B

(n -l))

1 / 2

,

B

(3)

and perform the test using maxj t j as the test s t a t i s t i c . To perform the t e s t , we need a c r i t i c a l value c with which to compare the test s t a t i s t i c . As i n the usual s t a t i s t i c a l hypothesis testing, c i s chosen so that the probability of the test s t a t i s t i c exceeding c when there i s no excess contamination i s a. This i s the probability of f a l s e detection. Although an exact value of c can be obtained by numerical integration, an approximate value that i s just a l i t t l e too high can be obtained more simply. This value i s given by the 100(1 - α/ns) percentile of the t d i s t r i b u t i o n with n - 1 degrees of freedom (11). Denoting the cumulative t d i s t r i b u t i o n function by F t ^ S ΠΒ-1), e have B

w

F (c; t

n - l ) - 1 - a/n . B

(4)

s

Since we do not know the proper values for λ and τ, we need a way of judging plausible values of λ and τ from the data. We do this by testing the transformed background measurements for normality. Our choice of a test f o r normality i s the probability plot c o r r e l a t i o n c o e f f i c i e n t r (12). The c o e f f i c i e n t r i s the c o r r e l a t i o n between the ordered measurements and predicted values for an ordered set of normal random observations. We denote the ordered background measure­ ments by y ( i ) , where y ( i ) < y ( 2 ) ·'· Β(η )· denote the predicted values with which the ordered measurements are correlated by Μι· The computation of these values i s discussed by F i l l i b e n ( 12). A conceptual d e f i n i t i o n of the M^ follows from consideration of a set of numbers drawn at random from the standard normal d i s t r i b u t i o n , the one with mean zero and standard deviation one. Ordering this set of numbers gives a sequence called order s t a t i s t i c s . The M^ are the medians of the distributions of these order s t a t i s t i c s . Since M^ predicts the value of the i t h order s t a t i s t i c from the standard normal d i s t r i b u t i o n , μ + dti± predicts the value of the i t h order s t a t i s t i c from the normal d i s t r i b u t i o n with mean μ and standard deviation σ. Thus, i f the ordered transformed measurements are from a normal d i s t r i b u t i o n and they are plotted versus the M^, they should f a l l near a straight l i n e . How close they come to a straight line i s measured by the probability plot correlation c o e f f i c i e n t
c(r) where k [(1 - .97r) •1/2

for r >r 005

1]

c(r) c(r

.005

) Π

- 2 (r

.005

-r)/(r

-r

f o r

)1

r

< r

7

» () .005

Λ Λ

F (r Q05) " 0.005 and r ^ i s the minimum value of r ( 12). The function c ( r ) was chosen through Monte Carlo experiments. The value of k i s chosen so that the probability of t j > c ( r ) , when xgj i s drawn from the same population as the background measurements, i s a. This probability i s given by 1 - / i F ( c ( r ) ; n - l ) d F ( r ) . The simultaneous test given by Equations 6 and 7 leads to a test appropriate for (λ,τ) unknown. The (\,x)-unknown test rejects the n u l l hypothesis that xgj belongs to the background population i f t j > c ( r ) f o r a l l (λ,τ). Since this test rejects the n u l l hypothesis only i f Equation 6 i s s a t i s f i e d for the true value of (λ,τ), this test has no greater probability of false detection than the simul­ taneous t e s t . Thus, the (λ,τ)-unknown test i s conservative i n the sense that the probability of a false detection i s less than α i f the probability of false detection for the simultaneous test i s a. To test a l l the measurements from the suspected region, we choose the value of k i n Equation 7 that s a t i s f i e s r

m

#

n

t

jj F ( c ( r ) ; n t

B

B

r

-1) d F ( r ) - 1 - a/n . r

(8)

s

and proceed as above. We reject the n u l l hypothesis that a l l the xgj are from the background population i f maxj t j > c ( r ) f o r a l l (λ,τ). This generalization from ng « 1 to arbitrary ng i s based on the same p r i n c i p l e as the use of Equation 4 i n the (λ,τ) known case. This generalization i s conservative i n the sense that the actual probabil­ i t y of f a l s e detection i s less than or equal to a. This test requires two computations. F i r s t , we must determine k, the parameter i n c ( r ) . To do t h i s , we must choose a numerical method f o r evaluating Equation 8. If α/ng i s not too small, we can use the tabulation of F ( r ) provided by F i l l i b e n (12). We approx­ imate Equation 8 by r

0.0025 + I 0.5

[F (c(ri+i)) + F ( c ( r i ) ) ] t

t

[F (r r

1 + 1

) - F ( r ) ] - l-a/n r

t

s

(9)

13.

LIGGETT

125

Detecting Elevated Contamination

where the r± are the points at which F i l l i b e n (12) evaluates F . To solve Equation 9 f o r k, we need an algorithm for evaluating F (13) and an algorithm f o r finding roots. Since the l e f t side of Equation 9 i s an increasing function of k, a simple root-finding algorithm such as repeated halving of an Interval known to contain k i s adequate. Since F i l l i b e n (12) only tabulates F ( r ) down to 0.005, very small values of a/ng require a Monte Carlo evaluation of Equation 8. The second computation i s the test i t s e l f . This can be arranged as the computation of a c r i t i c a l value x i with which maxj xsj i s compared. The c r i t i c a l value χχ i s max\, x ^ where r

t

r

x

τ + (1 + X y )

1 / X

for λ > 0

Publication Date: October 31, 1984 | doi: 10.1021/bk-1984-0267.ch013

u

*u τ + exp(y )

for λ - 0

u

(10)

\

and y

u

« m

B

+ c(r)(l + l / n ) B

1 / 2

s .

(11)

B

To perform the maximization over (λ,τ), we need an algorithm such as the Nelder-Mead simplex search (14). An alternative that i s adequate i n many cases i s a simple search over a (λ,τ) g r i d . The c r i t i c a l value χχ has an interpretation of i t s own. It i s the upper bound on a simultaneous prediction i n t e r v a l f o r ng as yet unobserved observa­ tions from the background population. Evaluation of the Method The previous section presents a test that i s based on the assumption that f o r some (λ,τ) Equation 1 transforms the background measurements to nomallty but i s otherwise conservative. The test contains no e x p l i c i t r e s t r i c t i o n on n or ng except n > 2. However, the test cannot be expected to be satisfactory f o r a l l values of n and ng. F i r s t , the test i s based on extrapolation of the d i s t r i b u t i o n of the background measurements to higher values than are represented i n the data. If extrapolation i s carried too f a r , the results w i l l not be s a t i s f a c t o r y . Second, the test i s conservative and may be too conser­ vative f o r some values of n and ng. In this section, we present an example based on analogous data that shows values of n and ng f o r which the test i s useful. Before looking at the example, consider a very simple alternative to the test just described. This alternative i s v a l i d regardless of what the d i s t r i b u t i o n of the background measurements i s . The alterna­ t i v e i s to declare a detection i f the largest of the measurements from the suspected region i s larger than the largest of the background measurements. The probability of f a l s e detection with this procedure i s ng/(n + ng) ( 15). This formula shows that njj must be much larger than ng i f the probability of false detection i s to be suitably small· The data considered are blank measurements made as part of a study of trace quantities of heavy metals dissolved i n the water of the Chesapeake Bay (16). While obtaining and processing the Bay B

B

B

B

B

B

Publication Date: October 31, 1984 | doi: 10.1021/bk-1984-0267.ch013

126

ENVIRONMENTAL S A M P L I N G FOR

H A Z A R D O U S WASTES

samples, samples of high-purity sub-boiling d i s t i l l e d water were handled i n the same way. These blank samples were exposed to the same sources of processing contamination as the Bay samples. Thus, the question of whether the measurements on the Bay samples are due to no more than the processing contamination might be raised. This question i s analogous to the detection of excess s o i l contamina­ t i o n . (Fortunately, the measurements on the Bay samples were well above the measurements on the blank samples.) We consider three metals, Cobalt, Iron, and Scandium. The twenty-four blank measure­ ments f o r each of these metals are given i n Table I. In our application of the transformation given i n Equation 1 to these data, we r e s t r i c t τ to positive values. This r e s t r i c t i o n i s based on models of the various sources of processing contamination. Possible sources of contamination include the chemical reagents which might add a constant l e v e l to the blank and a i r borne p a r t i c l e s which might add a variable l e v e l with positive skewness. There does not seem to be any reason to include a constant l e v e l that i s negative. Therefore, we have adopted t h i s r e s t r i c t i o n . Each of these data sets i s skewed, yet each can be transformed to normality. With no transformation applied, the probability plot c o r r e l a t i o n c o e f f i c i e n t s f o r the Co, Fe, and Sc data sets are 0.855, 0.857, and 0.987, respectively. For Co and Fe, the hypothesis of normality i s rejected at the 0.5 percent l e v e l (12). On the other hand, the maximum probability plot c o r r e l a t i o n c o e f f i c i e n t s are 0.993, 0.990, and 0.993 for Co, Fe, and Sc, respectively. The maxima occur at (λ,τ) - (0,0.0048), (0,0.42), and (0.457,0), respectively. These maxima are so high that they provide no evidence that the range of transformations i s inadequate. Note that the (λ,τ) values at which the maxima occur correspond to log transformations with a s h i f t for the Co and Fe and nearly a square-root transformation f o r the Sc. Pretend these data sets are the background measurements, and l e t ng • 5. If we use the d i s t r i b u t i o n - f r e e test and compare the maximum measurement from the suspected region with the maximum measurement from the background region, we w i l l have a probability of f a l s e detection of 5/29 (- 0.17). This probability may be too high. To obtain a probability of false detection of 0.05, we need the test discussed i n the previous section. From Equation 9, we f i n d that k - 0.7753. From Equations 10 and 11, we obtain x j « 0.228, 6.10, and 0.00033 f o r Co, Fe, and Sc, respectively. These values are sub­ s t a n t i a l l y higher than the largest of the background measurements. This i s due i n part to the conservativeness of the t e s t . Monte Carlo experiments on the test discussed i n the previous section suggest that the k value 0.7753 corresponds to a/ng • 0.005 instead of 0.010. Thus, these values of χχ correspond to ng • 10 Instead of 5. The test presented i n the previous section i s useful when a smaller p r o b a b i l i t y of f a l s e detection i s needed than i s provided by the d i s t r i b u t i o n - f r e e t e s t . However, the test i n the previous section i s no panacea. Reduction of the skewness through proper choice of sampling and subsampling procedures i s an alternative that may have much more potential f o r improving the study.

13.

LIGGETT

Detecting Elevated

127

Contamination

Table I. Blank Concentrations f o r Dissolved Cobalt, Iron, Scandium (ng/mL) Given as Stem-and-Leaf Displays (Tukey 1977) (6)

Publication Date: October 31, 1984 | doi: 10.1021/bk-1984-0267.ch013

Cobalt

.00**

53,53,58,58,63,66,72,73,75,78,81,87,95,95

.01

00,10,10,23,47,60

.02

40,80

.03

00

.04

00 Iron

Scandium

0. **

59,68,75,79,80,89,90,95

.0000*

5,7,7,8,8,9,9

1.

00,02,10,10,10,20,24,25,30

.0001

0,0,0,1,1,2,2,3,4,4,4

1.

50,50,53,70,88

.0001

5,5,6,6

2.

03

.0002

0,0

2. 3. 3.

70

Note that the actual values of the measurements can be obtained by combining the entries on the two sides of the v e r t i c a l l i n e . For example, the f i r s t Cobalt measurement i s .0053 ng/mL.

128

ENVIRONMENTAL SAMPLING FOR HAZARDOUS WASTES

Literature Cited 1. 2. 3. 4. 5.

Publication Date: October 31, 1984 | doi: 10.1021/bk-1984-0267.ch013

6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16.

Smith, R.; James, G. V. "The Sampling of Bulk Materials"; The Royal Society of Chemistry: London, 1981. Kratochvil, B.; Taylor, J . K. Analytical Chemistry 1981, 53, 924A-938A. Ingamells, C. O. Talanta 1974, 21, 141-155. Ingamells, C. O.; Switzer, P. Talanta 1973, 20, 547-568. Box, G. E. P. and Cox, D. R. Journal of the Royal Statistical Society, 1964, Series B, 26, 211-252. Tukey, J . W. "Exploratory Data Analysis"; Addison-Wesley Publishing Company: Reading, Mass., 1977. Bain, L. J. "Statistical Analysis of Reliability and Life-Testing Models"; Marcel Dekker: New York, 1978. Mason, B. J . "Preparation of Soil Sampling Protocol: Techniques and Strategies," Report EPA-60014-83-020, U.S. Environmental Protection Agency, 1983. Williams, J . C. Powder Technology 1976, 15, 245-251. Liggett, W. S.; Inn, K. G. W.; Hutchinson, J . M. R. Environmental International 1984, to appear. Hahn, G. J . Journal of the American Statistical Association 1970, 65, 1668-1676. Filliben, J . J . Technometrics 1975, 17, 111-117. Kennedy, W. J.; Gentle, J . E. "Statistical Computing"; Marcel Dekker: New York, 1980. Nash, J . C. "Compact Numerical Methods for Computers: Linear Algebra and Function Minimization"; John Willey and Sons, New York, 1979. Hahn, G. J.; Nelson, W. Journal of Quality Technology 1973, 5, 178-188. Kingston, H. M.; Greenberg, R. R.; Beary, E. S.; Hardas, B. R.; Moody, J . R.; Rains, T. C.; Liggett, W. S. "The Characterization of the Chesapeake Bay: A Systematic Analysis of Toxic Trace Elements," Report NBSIR 83-2698, National Bureau of Standards, 1983.

RECEIVED August 14, 1984

Publication Date: October 31, 1984 | doi: 10.1021/bk-1984-0267.ix001

Author Index Barth, Delbert S., 97 Burkhalter, Richard Α., 15 Carra, Joseph S., 53 Flatman, George T., 43 Harris, Daniel J., 27 Heath, Clark W., Jr., 7 Hynes, H. Patricia, 1 Journel, Andre G·, 109 Liggett, Walter, 119 Mason, Benjamin J., 97 McCall, Merley F., 15 Mix, Theodore J., 15 Provost, Donald 0., 15 Provost, Lloyd P., 79 Spittler, Thomas M., 37 Taylor, John Κ., 105 Thomas, Ralph Ε., 67

Subject Index A Aluminum production, 15,17 reduction c e l l , I8f Aluminum smelter, Kaiser—See Kaiser aluminum smelter Analytic study, description, 79 Assessment biologic effect of hazardous wastes, 8t,9-10 exposure to hazardous wastes, 7,8t,9

Β

Background measurements, s o i l samples, 119-20 Blood-lead levels in preschool children, Dallas lead study, 59

C Chlordane, sensitivity of f i e l d method, 39,4lf Chromatogram, Arochlors, 37,38f,39,41f Comprehensive Environmental Response, Cost, and Liability Act, 1 129

Concentration estimate and associated probability, isopleth maps, 115f Conditional distribution approach, assessment of spatial distribu­ tions of pollutants, 112-14 Conditional distribution of pollutants, estimation, 114,116-17 Conditional expectation of pollutant concentrations, Rlemann and Stieltjes integrals, 112 Confidence intervals averages and standard deviations of sample characteristics, 92 spatial distributions of pollutants, 113-14 Contaminated s o i l , f i e l d measurements, 39,42f Cost-effective sample size model, 88-90 Critical difference, 99,100 Cyanide contamination from Kaiser aluminum smelter analytical and sampling problems, 23 assessment, 15-25 regulatory considerations, 25 theories of pathways, 20 Cyanide flow paths beneath plant, Kaiser aluminum smelter, 22f

Publication Date: October 31, 1984 | doi: 10.1021/bk-1984-0267.ix002

Author Index Barth, Delbert S., 97 Burkhalter, Richard Α., 15 Carra, Joseph S., 53 Flatman, George T., 43 Harris, Daniel J., 27 Heath, Clark W., Jr., 7 Hynes, H. Patricia, 1 Journel, Andre G·, 109 Liggett, Walter, 119 Mason, Benjamin J., 97 McCall, Merley F., 15 Mix, Theodore J., 15 Provost, Donald 0., 15 Provost, Lloyd P., 79 Spittler, Thomas M., 37 Taylor, John Κ., 105 Thomas, Ralph Ε., 67

Subject Index A Aluminum production, 15,17 reduction c e l l , I8f Aluminum smelter, Kaiser—See Kaiser aluminum smelter Analytic study, description, 79 Assessment biologic effect of hazardous wastes, 8t,9-10 exposure to hazardous wastes, 7,8t,9

Β

Background measurements, s o i l samples, 119-20 Blood-lead levels in preschool children, Dallas lead study, 59

C Chlordane, sensitivity of f i e l d method, 39,4lf Chromatogram, Arochlors, 37,38f,39,41f Comprehensive Environmental Response, Cost, and Liability Act, 1 129

Concentration estimate and associated probability, isopleth maps, 115f Conditional distribution approach, assessment of spatial distribu­ tions of pollutants, 112-14 Conditional distribution of pollutants, estimation, 114,116-17 Conditional expectation of pollutant concentrations, Rlemann and Stieltjes integrals, 112 Confidence intervals averages and standard deviations of sample characteristics, 92 spatial distributions of pollutants, 113-14 Contaminated s o i l , f i e l d measurements, 39,42f Cost-effective sample size model, 88-90 Critical difference, 99,100 Cyanide contamination from Kaiser aluminum smelter analytical and sampling problems, 23 assessment, 15-25 regulatory considerations, 25 theories of pathways, 20 Cyanide flow paths beneath plant, Kaiser aluminum smelter, 22f

130

ENVIRONMENTAL SAMPLING FOR HAZARDOUS WASTES

Cyanide flow-path downgradient of plant site sources, Kaiser aluminum smelter, 21f Cyanide levels in wells and springs, Kaiser aluminum smelter, 22f,24f

Publication Date: October 31, 1984 | doi: 10.1021/bk-1984-0267.ix002

D Dallas lead study blood chemistry data, 56,59-65 blood-lead levels i n preschool children, 59 design, 53-54 dust lead, 55-56 erythrocyte protoporphyrin levels in preschool children, 60 isoarea maps of pollution estimates and confidence intervals, 48,50f,51 isopleth maps of pollution estimates and standard deviations, 46,48,49f s o i l lead, 56,57-58 t r a f f i c density, 55 Data-free estimation algorithms, limitations, 110-12 Double-sampling strategy, example, 94-95

ε Elevated contamination detection by comparisons with background, 119-27 detection model, 120-22 Enumeratlve study, description, 79 Environmental pathways for exposure to hazardous wastes, 8t,9 Environmental risk assessment process, hazardous wastes, 7-10 Environmental sampling characteristics of interest and methods of measurement, 80-81 degree of precision required, 81 determination of study population, 80 expected confidence intervals for a parameter mean, 82t Love Canal, 11-13 models, 83-84,87f monitoring program responsibilities, 106-7 number of tests required for specified maximum probable error, 85

Environmental sampling—Continued objectives, 80 quality assurance, 105-8 reference laboratory, 107 sample size selection, 84-88 stages of study, 81,83 statistical methods, 79-95 study design, 11,81,90,92-94 technical laboratory procedures, 11 uses in human health risk assessment, 7-13 Erythrocyte protoporphyrin levels in preschool children, Dallas lead study, 60 Exploratory study, importance in s o i l sampling quality assurance, 100-1 F Factorial tables, multifactor experiments, 72-73,74t Finite population correction, 84 G Gas chromatograph, portable, 37,38f Geostatistical logic and tools, 43 Ground water monitoring, technical concerns, 2-3 Groundwater cyanide boundary and water table elevations, Kaiser aluminum smelter, 17,21f H Hazardous wastes assessment of biologic effect, 8t,9-10 assessment of exposure, 7,8t,9 environmental pathways for exposure, 8t,9 questions and issues from the f i e l d , 1-5 safety standards, 8t,10 Hierarchical trees, multifactor experiments, 73-76 Hot spot detection, 122-27 evaluation of method, 125-27 order statistics, 123 probability plot correlation coefficient, 123 t distribution function, 123 t statistics, 123 transformations, 122 Human health risk assessment, 7-13

131

INDEX

Publication Date: October 31, 1984 | doi: 10.1021/bk-1984-0267.ix002

I Indicator contaminants case study examples» 4-5 uses in environmental sampling, 3-5 Indicator covariance model, spatial distributions of pollutants, 117 Indicator kriging estimator, spatial distributions of pollutants, 116 Indicator transform, spatial distribu­ tions of pollutants, 116 Isoarea maps, pollution estimates and confidence intervals, Dallas lead study, 48,50f,51 Isopleth maps concentration estimate and associated probability, 115f kriging estimation variances, 111f pollution estimates and standard deviations, Dallas lead study, 46,48,49f

J Judgmental approach, sampling site selection, 101-2 Κ Kaiser aluminum smelter contaminated area, I8f cyanide contamination analytical anc sampling problems, 23 cyanide contamination regulatory considerations, 25 cyanide flow paths beneath plant, 22f cyanide flow-path downgradient of plant site sources, 21f cyanide levels in wells and springs, 22f,24f f a c i l i t y description, 17 groundwater cyanide boundary and water table elevations, 17,21f potliner, 15,17 remedial action and results, phases I and II, 19-23 sampling description and findings, 17 vicinity map, I6f Kriging isopleths of estimation variances, 111f limitations, 110-12 use i n geostatistics, 46

L Landfills, technical concerns, 2 Lead contamination near smelters, assessment using geostatistics, 43-51 Lead content of paint, determination, 54-55 Lead levels, blood of children around smelter sites, 53-66 Loss function, spatial distributions of pollutants, 113 Love Canal, environmental testing, 11-13

M Mathematical models, multifactor experiments, 76-77 Maximum probable error, averages of independent samples, 84,86 Models cost-effective sample size, 88-90 elevated contamination detection, 120-22 environmental sampling, 83-84,87f Monitoring program responsibilities, environmental sampling, 106-7 Multifactor experiments factorial tables, 72-73,74t hierarchical trees, 73-76 interdisciplinary design, 67-77 mathematical models, 76-77 predicted results, 68 pruning the tree, 76 team approach, 69-72 Multivariate distribution of pollutants, 114,116

Ν Nomographs determination of optimum number of replications, 90,91f determination of sample size, 86,87f t distribution, 92,93f Nugget effect, 45

0 Order statistics, hot spot detection, 123

132

ENVIRONMENTAL SAMPLING FOR HAZARDOUS WASTES

Publication Date: October 31, 1984 | doi: 10.1021/bk-1984-0267.ix002

Ρ Pollution monitoring regional variable, 43-44 semi-variogram, 44 Polychlorinated biphenyls f i e l d measurement using a portable gas chromatograph, 37-42 river sediment and f i s h , 5 transport under varying flow conditions, 5 Potliner, Kaiser aluminum smelter, 15,17 Probability plot correlation coefficient, hot spot detection, 123 Pruning the tree, multifactor experiments, 76

Q Quality assurance, environmental sampling, 105-8 R Random approach, sampling site selection, 102 Random semi-variogram model, 45-46,471 Recovery, Arochlor 1242, 39,40f Reference laboratory, environmental sampling, 107 Regional variable, pollution monitoring, 43-44 Resource Conservation and Recovery Act, 1,3 Riemann integrals, conditional expec­ tation of pollutant concentrations, 112 Risk assessment process, hazardous wastes, 7-10 S Safety standards, hazardous wastes, 8t,10 Sample mean, total site area, 94 Sample size selection, environmental sampling, 84-88 Sampling design and framework, environmental testing, 11 Sampling resources, allocation, 88-90 Sampling site selection, judgmental, random, and systematic approaches, 101-2

Semi-variogram, pollution monitoring, 44 Soil sampling analysis and interpretation of data, 104 background measurements, 119-20 number and locations of sites, 101-2 quality assurance, 97-104 sample handling, 102-4 Spatial distributions of pollutants, 109-17 assessment using conditional dis­ tribution approach, 112-14 conditional cumulative distribution function, 112 confidence intervals, 113-14 estimation, 109-10 indicator covariance model, 117 indicator kriging estimator, 116 indicator transform, 116 loss function, 113 risks of incorrect assessment, 114 Spherical semi-variogram model with no nugget, 46,47f Spherical semi-variogram model with nugget, 44-45,47f Spokane aquifer description, 15 north arm, I6f Statistical methods, environmental sampling, 79-95 Stieltjes integrals, conditional expectation of pollutant concentrations, 112 Study population, determination in environmental sampling, 80 Systematic approach, sampling site selection, 102

Τ Technical laboratory procedures, environmental testing, 11 2,3,7,8-Tetrachlorodibenzo-D-dioxin contamination in Missouri study area areal variation of concentration, 28,29f,31,33 depth sampling, 30-31 experimental sampling and presentation of data, 28 mean concentrations, 33,34f road centerline concentrations, 32-35 sampling methods, 27-35 vertical migration, 28,30

INDEX

133

Toxic chemical Identification and measurement, 7,8t,9 Transformations, hot spot detection, 122

W Water table elevations and groundwater cyanide boundary, Kaiser aluminum smelter, 17,21f

V

Publication Date: October 31, 1984 | doi: 10.1021/bk-1984-0267.ix002

Variance, concentration estimates, 83,89

Production and indexing by Karen McCeney Jacket design by Pamela Lewis Elements typeset by Hot Type Ltd, Washington, D.C. Printed and bound by Maple Press Co., York, Pa.