 |
||
|
Great Lakes Contaminated Sediments |  |
 |
||
|
|
|
|
|||||||||
R/V Lake Guardian
|
|
ASSESSMENT AND REMEDIATION OF CONTAMINATED SEDIMENTS (ARCS) PROGRAM
RISK ASSESSMENT AND MODELING OVERVIEW DOCUMENTUS Environmental Protection Agency. 1993. Assessment and Remediation of Contaminated Sediments (ARCS) Program Risk Assessment and Modeling Overview Document, EPA 905-R93-007. Chicago, Ill.: Great Lakes National Program Office (GLNPO). ACKNOWLEDGMENTSThe initial draft of this report was prepared through support from U.S. Environmental Protection Agency (USEPA) to AScI Corporation of Athens, Georgia, under USEPA Contract Numbers 68-C1-0012 and 68-C0-0054, and administered by USEPA's Environmental Research Laboratory of Athens, Georgia. The primary authors of the initial draft report were Dr. James Martin of AScI and Dr. Judy Crane, formerly of AScI, now of E.V.S. Consultants of Vancouver, Canada. This report was edited and produced by PTI Environmental Services (PTI) of Bellevue, Washington, for Battelle Ocean Sciences of Duxbury, Massachusetts, under USEPA Contract Number 68-C2-0134. Additions to the text of the initial draft were made by Dr. Robert Pastorok and Ms. Lisa Yost of PTI.
This report was prepared for the Risk Assessment/Modeling Work Group as part of the Assessment and Remediation of Contaminated Sediments (ARCS) Program administered by USEPA's Great Lakes National Program Office (GLNPO) in Chicago, Illinois. Dr. Marc Tuchman of GLNPO served as chairman of the Risk Assessment/Modeling Work Group. Mr. David Cowgill of GLNPO and Dr. Tuchman served as the ARCS Program managers. This document provides an overview of risk assessment and modeling methods as applied to areas with contaminated sediments in the Great Lakes region. The document was prepared under the Assessment and Remediation of Contaminated Sediments (ARCS) Program, administered by the U.S. Environmental Protection Agency's (USEPA) Great Lakes National Program Office (GLNPO), in Chicago, Illinois. The goal of the risk assessment and modeling studies was to develop and demonstrate a comprehensive human health and ecological risk assessment framework for use in the evaluation of alternative remedial actions for contaminated sediments. As part of that effort, risk assessment and modeling studies were performed at selected Areas of Concern in the Great Lakes region. The goal of those studies was to provide estimates of potential changes in exposure and risk that may occur either under a no-action alternative or following implementation of various remedial alternatives for contaminated sediments. The risk estimates may then be used to aid in the selection of an appropriate remedial action. This document does not provide detailed guidance on risk assessment and modeling methods, but refers the reader to pertinent source documents for further information. TABLE OF CONTENTS
DISCLAIMER
1. INTRODUCTIONThis document provides an overview of risk assessment and modeling methods as applied to areas with contaminated sediments in the Great Lakes region. It was prepared under the Assessment and Remediation of Contaminated Sediments (ARCS) Program, administered by the U.S. Environmental Protection Agency's (USEPA) Great Lakes National Program Office (GLNPO) in Chicago, Illinois. BACKGROUNDAlthough toxic discharges in the Great Lakes and elsewhere have been reduced in the last 20 years, persistent contaminants in sediments continue to pose a potential risk to human health and the environment. High concentrations of contaminants in bottom sediments and associated adverse effects have been well documented throughout the Great Lakes and associated connecting channels. The extent of sediment contamination and its associated adverse effects have been the subject of considerable concern and study in the Great Lakes community and elsewhere. Contaminated sediments can have direct toxic effects on aquatic life, such as the development of cancerous tumors in fish exposed to polycyclic aromatic hydrocarbons in sediments. The bioaccumulation of toxic contaminants in the food chain can also pose a risk to humans, wildlife, and aquatic organisms. As a result, advisories against consumption of fish are in place in many areas of the Great Lakes. These advisories have also had a negative economic impact on the affected areas. To address concerns about the deleterious effects of contaminated sediments in the Great Lakes, Annex 14 of the Great Lakes Water Quality Agreement between the United States and Canada stipulates that the cooperating parties will identify the nature and extent of sediment contamination in the Great Lakes, develop methods to assess impacts, and evaluate the technological capability of programs to remedy such contamination. The 1987 amendments to the Clear Water Act, in section 118(c)(3), authorized GLNPO to coordinate and conduct a 5- year study and demonstration projects relating to the appropriate treatment of toxic contaminants in bottom sediments. Five areas were specified in the Act as requiring priority consideration in conducting demonstration projects: Saginaw Bay, Michigan; Sheboygan Harbor, Wisconsin; Grand Calumet River, Indiana; Ashtabula River, Ohio; and Buffalo River, New York. To fulfill the requirements of the Act, GLNPO initiated the ARCS Program. In addition, the Great Lakes Critical Programs Act of 1990 amended the section, now section 118(c)(7), by extending the program by 1 year and specifying completion dates for certain interim activities. ARCS is an integrated program for the development and testing of assessment techniques and remedial action alternatives for contaminated sediments. Information from ARCS Program activities will help address contaminated sediment concerns in the development of Remedial Action Plans (RAPs) for all 43 Great Lakes Areas of Concern (AOCs), as identified by the United States and Canadian governments, as well as lakewide management plans. To accomplish the ARCS Program objectives, the following work groups were established:
This document is intended to provide an overview of the risk assessment and modeling methods developed by the ARCS RAM Work Group and to provide general guidance on their application to other Great Lakes AOCs. Sediment contamination is of concern primarily because of the potential risks it poses to humans, wildlife, and aquatic organisms. Therefore, the management of contaminated sediments includes the overall process of risk management. For this project, risk management is defined as the process of integrating findings from a risk assessment with engineering, policy, and nontechnical concerns to make decisions about sediment remediation at a specific site or to set remediation priorities among sites. Risk management should be distinguished from risk assessment, which is the process of producing qualitative or quantitative estimates of the potential risks associated with exposure to specific concentrations of contaminants under specific current or future exposure conditions at a site. The general objective of the ARCS RAM Work Group was to develop and demonstrate a comprehensive risk management framework for: 1) identifying existing risks to human health and ecological receptors at sites with contaminated sediments, 2) estimating the potential impact of various sediment remedial alternatives on contaminant concentrations in various media and their associated risks, and 3) comparing existing and potential future risks to aid in the selection of sediment remedial alternatives. Steps in the overall risk management process are illustrated in Figure 1-1. A general discussion of each of the steps is provided below, followed by a more detailed description of the use of risk assessment and modeling in the ARCS RAM studies.
The studies of the ARCS RAM Work Group have provided support for Steps 4-8 of the risk management process. Other ARCS studies deal specifically with other aspects of the decision-making process. The final results of the ARCS RAM studies are estimates of contaminant concentrations and potential risks associated with various sediment remedial alternatives that may then be used, along with other information collected at a site, to select the appropriate remedial action from among the various alternatives. The following sections of this document provide an overview of the use of risk assessment and modeling in the ARCS RAM studies. As part of the ARCS RAM studies, baseline risk assessments were performed to estimate the current health risks to humans and wildlife exposed to sediment-derived contaminants in the absence of any remediation. The results of these assessments can be used by ARCS risk managers in prioritizing sites and making decisions concerning the need for sediment remediation. The risk assessment approach developed and used in the ARCS Program is intended to produce conservative estimates of risk in order to ensure adequate protection of human health and the environment. This approach to risk assessment is specifically designed not to underestimate risks and, therefore, is likely to overestimate risks at many sites. The following sections provide an overview of the risk assessment approach used under the ARCS Program. This approach may be used to assess potential human and ecological risks at other sites with contaminated sediments. HUMAN HEALTH RISK ASSESSMENTIndividuals in the Great Lakes region may be exposed to sediment contaminants through various activities that result in intake of contaminants through dermal, ingestion, and/or inhalation pathways. For the ARCS risk assessments, human health risk estimates were determined for both carcinogenic effects (i.e., increased probability of an individual developing cancer over a lifetime) and noncarcinogenic effects (i.e., chronic or subchronic effects other than cancer) over a range of exposure scenarios. The risk estimates were calculated by using conservative exposure assumptions and USEPA-verified toxicity values called cancer slope factors (CSFs) and reference doses (RfDs) (for noncarcinogenic effects). The primary guidance used to conduct these risk assessments was obtained from USEPA's Risk Assessment Guidance for Superfund--Volume I: Human Health Evaluation Manual (Part A) (USEPA 1989c), although the use of additional USEPA guidance for risk assessment (USEPA 1988b, 1989a,b, 1991) is also described herein. The following sections describe the main components used in performing the ARCS human health risk assessments (Figure 2-1), including specific examples and recommendations. Baseline human health risk assessments were conducted under the ARCS Program for the five priority AOCs: Saginaw River, Michigan (Crane 1992b); Sheboygan River, Wisconsin (Crane 1993a); Grand Calumet River, Indiana (Crane 1993b); Ashtabula River, Ohio (Crane 1992a); and Buffalo River, New York (Crane 1993c). The same human health risk assessment framework can be applied to any site with contaminated sediments. Data ReviewPrior to beginning the human health risk assessment, site data must be collected and analyzed to determine contaminant concentrations in the media of interest (e.g., air, water, sediments, biota) and potential routes of exposure to contaminated media. Available historical data, including information on site use and possible contaminant sources, should be reviewed to focus sampling efforts on contaminants known or likely to be present. Such data may include analytical data from previous sampling efforts, descriptions of the past uses of the site or other site records, and interviews with site personnel that may suggest what contaminants may be present. For example, RAPs contain information about monitoring and scientific studies that have been conducted at Great Lakes sites. USEPA's storage and retrieval (STORET) database is also a good source of water quality data that are routinely collected at U.S. Geological Survey (USGS) and statewide gaging stations. In addition, to determine current or future uses of the site, personnel at various local, State, and Federal agencies that deal with public health, natural resource, and fish and wildlife issues should also be contacted for information about the site. In particular, applicable zoning regulations, land use plans, and restrictions on site uses (e.g., fishing or hunting bans) should be described in the risk assessment. The risk assessor should visit the site, preferably during the period of greatest activity or during several seasons, to observe recreational and business uses of the AOC. It may also be helpful to interview game wardens, lifeguards, and local officials regarding site use. A thorough sampling program for sediments, water, fish, and other important media should be conducted using appropriate quality assurance and quality control (QA/QC) procedures, such as those identified in Guidance for Data Usability in Risk Assessment (USEPA 1990). Water, sediment, and fish samples should be collected preferentially in areas where people are known to be using the site, such as public beaches, or are likely to use the site in the future, such as areas near shorelines or along access roads. Samples should also be collected at one or more reference sites (i.e., areas that are unlikely to be influenced by sediment contaminants within the AOC or by other anthropogenic sources). In addition, several species of fish should be sampled, including species that feed on the bottom, such as carp and catfish, and species that feed in the water column, such as walleye. Both whole body and skin-on or skin-off fillets should be analyzed for various organic and inorganic chemicals, especially those chemicals detected in the sediments and known to bioaccumulate in fish tissue. The need for air sampling should be evaluated on a case-by-case basis. Exposures to contaminants in air are likely to contribute much less to overall site risks than exposures via other pathways such as consumption of fish. Thus, air sampling is not generally required for screening-level evaluations and was not conducted as part of the ARCS risk assessments. However, the air pathway may be important at sites with volatile contaminants of concern, where upland soils are exposed, or where areas of sediments are dry for a substantial proportion of the time and thus may be a source of airborne particulates. At some such sites, simple models may be used to estimate exposures to contaminants via the air pathway.
Once data from current and historical sources have been reviewed, the most appropriate data available for the media of interest should be selected for use in the risk assessment. The adequacy of QA/QC procedures followed in generating the analytical data should be a key criterion in selecting data. However, a complete QA/QC review may not be possible, particularly when analyses are conducted using historical data. For example, the ARCS risk assessments relied primarily on historical data, and in many cases, little QA/QC information was supplied with the data. Thus, risk assessment staff should work with the regulatory agency's project manager in determining whether a data set is adequate for a specific risk assessment application. In addition, the implications of any limitations in available data should be discussed in the risk assessment document. For example, unsuitable detection and quantification limits are often a major limitation in the use of historical data sets. A list of all the contaminants detected in the media of interest at the site should be made. Inorganic chemicals present at concentrations near background levels and chemicals that are infrequently detected or that may be present as laboratory contaminants may be excluded. Where the list of contaminants of concern is extensive, a screening step can be conducted to exclude contaminants that only contribute a minimal amount to the overall site risk. For example, risks associated with the maximum detected concentration can be calculated using toxicity data available in USEPA's Integrated Risk Information System (IRIS) database and exposure assumptions that assume a higher degree of exposure than is likely to occur at the site. Such an approach is considered to be conservative because it incorporates assumptions that may overestimate risks in order to ensure that risks are not underestimated. Using this approach, contaminants can then be excluded when they contribute an individual risk of less than 1x10-7 (for carcinogens) or a hazard quotient of less than 0.1 (for noncarcinogens). A carcinogenic risk of 1x10-7 corresponds to a one-in-ten-million chance of an individual developing cancer during their lifetime. Use of these conservative target risk levels and worst-case exposure assumptions (e.g., use of the maximum detected concentration) generally ensures that chemicals with significant risks due to the cumulative effects of multiple contaminants and multiple exposure pathways are not prematurely excluded from the risk assessment. Exposure AssessmentIn the exposure assessment, the magnitude, frequency, duration, and route of direct and indirect exposures of individuals to sediment-derived contaminants from an AOC are determined. Populations that may be exposed (i.e., receptor populations) should first be identified by considering the site's proximity to population centers, the accessibility of the site, and any features such as beaches or fishing piers that would attract visitors. The predominant types of receptor populations to consider are residents, workers, and recreational visitors. Recreational uses of AOCs may include fishing, swimming, boating, or beach activities. Current and potential future exposures should be evaluated. In particular, future exposures should be evaluated when future uses of a site may increase the potential for exposure to site contaminants. For example, if a site may be used for residential purposes in the future, exposure to site soils and sediment could be greatly increased in comparison with a current recreational scenario. Exposures to contaminants can potentially occur via three exposure routes: ingestion, dermal contact, and inhalation, each of which is in turn part of numerous exposure pathways. Ingestion of contaminants can result from inadvertent consumption of contaminated soils or sediment, or through consumption of drinking water, surface water, or wildlife. Dermal contact involves direct contact of the skin with either contaminated sediments, riverplain soils, or overlying water. Inhalation of airborne vapors or dust may introduce contaminants of concern into the respiratory system. The ingestion exposure pathways often result in higher exposure estimates than the dermal or inhalation pathways because of the greater absorption of contaminants through the gastrointestinal tract as compared with absorption through the skin, and the relatively high levels of intake of contaminants in soil, water, and food as compared with inhalation of contaminants. The potential pathways by which people may be exposed to contaminants from an AOC are then examined to determine whether they are complete or incomplete. An exposure pathway is complete if there is: 1) a source and mechanism of chemical release, 2) a retention or transport medium (or media) whereby chemicals are transferred between media, 3) an exposure point where contact occurs, and 4) an exposure route by which contact occurs (USEPA 1989c). An exposure pathway is incomplete if any of these conditions is not met. The exposure pathways that were complete for most of the five priority ARCS sites included: 1) consumption of contaminated fish, 2) dermal contact with contaminated water, 3) limited dermal contact with contaminated sediments, and 4) limited ingestion of surface water while swimming. Incidental ingestion of sediment may also be of concern at some sites. All complete exposure pathways should be evaluated in the exposure assessment unless certain criteria apply. These criteria include: 1) the potential magnitude of exposure from a pathway is low, or 2) the probability of the exposure occurring is very low and the risks associated with the occurrence are not high (USEPA 1989c). For example, at the Saginaw River AOC, there are no beaches along the river and swimming may occur only infrequently when people jump off recreational boats into the water. In addition to contacting the water, these people could ingest some water while swimming. In this case, the risk from ingesting surface water was considered insignificant, and an assumption was made that the health risk from dermal contact would be even lower than the health risk associated with ingestion (Crane 1992b). At some sites, it may be reasonable to assume that no fishing takes place because of the absence of edible fish or shellfish or because sites are physically inaccessible or remote. Once the exposure pathways to be quantitatively evaluated are selected for a site, contaminant concentrations and exposure parameters are used to calculate the chronic or subchronic intake level of each contaminant (in mg of chemical per kg body weight per day) (Table 2-1). For each current and potential future exposure scenario, exposure parameters may be selected to represent typical, reasonable maximum, and, in some instances, worst-case exposure conditions. Typical, or average, exposures and reasonable maximum exposures (i.e., the maximum exposure that is reasonably expected to occur at a site) are usually evaluated for each complete pathway. In general, an average exposure case is calculated using site concentrations and exposure parameters that best represent the central tendency of the data. Under the reasonable maximum exposure case, 95th or 90th percentile values are used for contact rates, intake rates, and exposure frequency and duration variables, and the upper 95 percent confidence limit on the average concentration is used for the exposure point concentration in the contaminated media. (See also USEPA [1992c] for further clarification of calculation of exposure point concentrations.) Site-specific information is often not available for many exposure parameters; thus, assumptions about the types and frequencies of exposure may be made based on recommended USEPA values or on professional judgment. The following documents provide useful information on estimating exposure parameters and conducting the exposure assessment: Superfund Exposure Assessment Manual (USEPA 1988b); Exposure Factors Handbook (USEPA 1989b); Assessing Human Health Risks from Chemically Contaminated Fish and Shellfish: A Guidance Manual (USEPA 1989a); and Standard Default Exposure Factors: Interim Final (USEPA 1991), which is a supplement to the Superfund risk assessment manual (USEPA 1989c); and Dermal Exposure Assessment: Principles and Applications (USEPA 1992a). In some cases, it may be appropriate to determine the fractional intake of exposure that occurs at a site. The fractional intake, which is the proportion of all exposure of a given type (e.g., the fraction of all fish consumed) that comes from the site, is generally estimated based on best professional judgment of factors such as the site size and accessibility and any restrictions on site use (e.g., warning signs, fishing bans, or barriers to the site). In calculating fractional intake for fish consumption, the abundance of edible fish and shellfish at a site should be considered. Some sites may not have any fish or may not have edible aquatic species. However, although it is important to consider these limits on site use, they may not be sufficient to prevent access at a site, and thus the risk assessment should not assume that exposure will not occur. For example, although all of the five priority AOCs examined for the ARCS Program had fish advisories in effect, some people continued to fish from the river. Because recreational fishing is very popular in the Great Lakes region and consumption of contaminated fish is an important exposure pathway, several researchers have gathered data on consumption rates of fish by Great Lakes populations. A survey of the angler population in the AOC should be made to obtain a better estimate of local fish consumption rates and patterns. If these data cannot be obtained, the results of a survey of Michigan anglers and their families by researchers at the University of Michigan (West et al. 1989) may be used to estimate more "localized" consumption patterns. The survey results can be obtained from Patrick West at the University of Michigan. An important result of this survey was that Michigan anglers and their families had an average fish consumption rate of 19.2 g/person-day (West et al. 1989), nearly 3 times the average fish consumption rate of people in the United States as a whole (USEPA 1989c). Anglers in other Great Lakes states may be consuming fish at a comparable rate to Michigan anglers. Additional data on fish consumption rates for sport anglers on Lake Michigan are provided as part of a nationwide survey reported in Rupp et al. (1980).
In selecting appropriate consumption rates, subsistence fishing or hunting should be considered (in addition to the average and reasonable maximum exposure cases) at sites with special subgroups of people who rely on locally caught fish, waterfowl, or other aquatic-related wildlife as their main source of protein. Examples may include members of a particular ethnic community who traditionally rely on fish as an important part of their diet (e.g., the southeast Asian community of Hmong in Sheboygan, Wisconsin) or indigent people who spend time in the area and may rely on locally caught fish for their main source of protein. In a toxicity assessment, available data are reviewed to determine and quantify the relationship between the level of exposure to a contaminant (dose or intake level) and the increased likelihood and/or severity of adverse effects. This relationship is termed the dose-response relationship and provides the basis for deriving quantitative toxicity values used in the risk assessment. For carcinogenic health effects, CSFs are used to estimate the risk of developing cancer that corresponds to estimated exposure concentrations. This risk is in addition to the risk of developing cancer due to other causes and thus is often termed excess cancer risk. The potential for noncarcinogenic health effects from oral exposures is typically evaluated by comparing estimated daily intake levels with RfDs, which represent daily intake levels at which no adverse effects are expected to occur. For assessment of inhalation exposures, USEPA has recently begun issuing reference concentrations (RfCs) that represent exposure concentrations at which no adverse effects are expected to occur. Carcinogens and systemic toxicants are treated differently, because according to current scientific theory it is plausible that for any dose of a carcinogen there could be some finite increase in cancer risk. Systemic toxicants are considered to act via a threshold mechanism, which allows for the identification of a safe dose. Hazard identification and dose-response evaluations for more than 600 chemicals have been conducted and verified by USEPA work groups; additional chemicals are awaiting review. USEPA-verified toxicity values can be obtained by accessing USEPA's IRIS database. The IRIS User Support group can provide technical assistance and information on how to access IRIS and can be reached at (513) 569-7254. Brief toxicity profiles on contaminants of concern should be prepared as part of the toxicity assessment. At a minimum, such profiles should contain information on the derivation of toxicity values for the contaminants of concern and should describe any uncertainties associated with the toxicity values. The following data should be gathered, to the extent available, for all contaminants of concern that have been shown to be carcinogenic in experimental animals or in human populations:
Pertinent data to be identified and discussed in the baseline risk assessment for contaminants associated with noncarcinogenic effects include the following:
Inclusion of these background data in the toxicity assessment assists risk managers in interpreting the findings of the risk assessment. The purpose of the risk characterization step is to combine the exposure and toxicity estimates into an integrated expression of human health risk. Three means of expressing carcinogenic and noncarcinogenic risks are presented in the risk assessment. First, chemical-specific risks are estimated for each exposure pathway. Second, these chemical-specific risks are added to estimate a cumulative pathway-specific risk. Finally, risks are added across all chemicals and relevant pathways to estimate the total human health risks to individuals exposed to contaminants from the AOC. Table 2-2 illustrates how a summary table of risk estimates may be arranged. The approaches used to quantify carcinogenic and noncarcinogenic health risks are described below. Carcinogenic risk is expressed as the upper-bound excess probability of an individual developing cancer over their lifetime following exposure to a given chemical concentration for a specified period of time. Carcinogenic risk estimates are computed by multiplying the chronic daily intake prorated over a lifetime of exposure by the CSF for each carcinogen of interest. Carcinogenic effects are summed for all chemicals in an exposure pathway (e.g., consumption of fish, incidental ingestion of sediments). This summation of carcinogenic risks assumes that there are no synergistic or antagonistic chemical interactions and that all chemicals produce the same effect. USEPA believes it is prudent public health policy to consider actions to mitigate or minimize exposures to contaminants when estimated excess lifetime cancer risks exceed the 10-5 to 10-6 range (USEPA 1988a). Noncarcinogenic effects are evaluated by calculating the ratio, otherwise known as the hazard quotient, of a site-specific exposure level for a specified time period to an RfD derived from a similar exposure period. Unlike cancer risk estimates, hazard quotients are not expressed as a probability. A hazard quotient of less than 1 indicates that exposures are not likely to be associated with adverse noncarcinogenic effects. As the hazard quotient approaches or exceeds 10, the likelihood of adverse effects is increased to the point where action to reduce human exposure should be considered (although the magnitude of the uncertainty factors used to derive the RfD should also be considered). Because of the uncertainties involved with these estimates, values between 1 and 10 may be of concern, particularly when additional significant risk factors are present. However, because RfDs do not have equal accuracy or precision and they are not based on the same severity of toxic effects, evaluation of hazard indices (the sum of two or more hazard quotient values for multiple substances and/or multiple exposure pathways) should take into account the uncertainties associated with specific RfDs.
The consumption of contaminated fish resulted in the greatest human health risk at the five priority AOCs examined in the ARCS Program: Saginaw River, Michigan (Crane 1992b); Sheboygan River, Wisconsin (Crane 1993a); Grand Calumet River, Indiana (Crane 1993b); Ashtabula River, Ohio (Crane 1992a); and Buffalo River, New York (Crane 1993c). Locally caught fish were assumed to accumulate contaminants primarily through the food chain, and in-place contaminated sediments were assumed to be the major source of contaminants to the food chain and water column. In most cases, PCB contamination contributed the greatest degree of carcinogenic risk. Noncarcinogenic risk levels were usually not of concern except for the subsistence exposure case; however, some chemicals (e.g., PCBs) lacked a verified RfD, and thus the noncarcinogenic effects of these contaminants could not be evaluated in the risk assessments. In addition, the consumption of bottom-feeding fish species, like carp, usually resulted in carcinogenic risks greater than 10-6, whereas the consumption of water column-feeding fish species, like walleye, did not always result in significant carcinogenic risks. A number of assumptions and estimated values are used in baseline risk assessments that contribute to the level of uncertainty about possible human health risks. As with most environmental risk assessments, the uncertainty about the risk estimate is generally at least an order of magnitude or greater (USEPA 1989c). Thus, at a minimum, the risk assessment should include a qualitative uncertainty analysis that identifies the key site- related variables and assumptions that contribute most to the uncertainty inherent in the risk estimates. Some of the major uncertainties in the ARCS risk assessments arose from the following factors:
For each of these assumptions, the level of uncertainty associated with the final risk estimates was estimated as low, moderate, or high. Additional site-specific sources of uncertainty are likely to be important for risk assessments conducted at other contaminated sediment sites. Calibration and fine-tuning of model results after field testing can greatly reduce uncertainties associated with risk estimates at specific sites. The results of the baseline risk assessment can be used by risk managers for several purposes. First, the baseline risk assessment provides a quantitative way to identify the exposure pathways and contaminants that contribute to carcinogenic and noncarcinogenic human health risks at a site. However, the calculated human health risks are not actual values, but are instead estimates that must be interpreted in the context of all the uncertainties associated with each step in the risk assessment process. Second, the baseline risk assessment can be used to identify sensitive subgroup populations (e.g., children, subsistence anglers) within the AOC. Third, the results of the baseline risk assessment can be used to compare the estimated risks of different sediment remedial alternatives with the impact of the no-action alternative during the comparative risk assessment. Additional applications of the baseline risk assessment are discussed in subsequent chapters. ECOLOGICAL RISK ASSESSMENTThe wildlife and aquatic organisms of the Great Lakes region may be exposed to sediment contaminants through various mechanisms, resulting in ingestion, inhalation, or dermal uptake of potentially toxic chemicals. For the Great Lakes AOCs, ecological risk assessments may be conducted to evaluate the likelihood of acute and chronic adverse effects of sediment contaminants on wildlife, aquatic plants, benthic invertebrates, and fish that rely on lakes and streams for habitat, food, and drinking water. Under the ARCS Program, wildlife risk assessments were conducted for two of the priority AOCs, Buffalo River and Saginaw River. Ecological assessments may involve empirical measurements of realized effects using the retrospective approach and theoretical modeling to estimate the probability of effects using the predictive approach. The balance of empirical and modeling approaches depends on the objectives of the assessment, the practicality of measurement methods for the receptors of concern, and data availability. For example, empirical approaches are commonly used to evaluate the effects of existing contamination on species populations and communities that are easily sampled, and on endpoints such as population abundances that can be easily quantified and interpreted. Theoretical models are used to estimate exposure of large-bodied wildlife species and rare species for which direct measurement of population or community endpoints is impractical, and to predict the effects of future conditions. USEPA is currently developing guidelines for conducting ecological risk assessments (Norton et al. 1992) and recently issued their Framework for Ecological Risk Assessment (USEPA 1992b). This framework report describes the elements of an ecological risk assessment and provides the basis for conducting ecological risk assessments within USEPA. Development of specific guidelines for ecological risk assessment is in progress, but will require considerable time (Norton et al. 1992). USEPA Region V has issued its own framework document entitled Regional Guidance for Conducting Ecological Assessments. In addition, the State of Wisconsin recently issued their Guidance for Assessing Ecological Impacts and Threats from Contaminated Sediments (WDNR 1992a,b). In addition to the guidance offered herein, these other guidance documents should be consulted in planning an ecological risk assessment for contaminated sediments. Past ecological risk assessments that have been performed for contaminated sediment sites in Region V are available from USEPA Region V. Because of the variety of habitats and species associated with sediments and interactions between biota and physical-chemical conditions, diverse techniques may be used in ecological risk assessments. The physical and chemical structure of an ecosystem influences the bioavailability and toxicity of contaminants to resident species. Biological interactions may determine the transport and fate of contaminants in the environment as well as species exposure patterns. Thus, the risk assessment process cannot be easily standardized to a "cookbook" approach. The remainder of this chapter presents an overview of ecological risk assessment methods as they might be applied to contaminated sediments. The elements of the approach described below are consistent with USEPA's framework report and other USEPA reports (USEPA 1989d,e; Warren-Hicks et al. 1989). Other related activities in the development of ecological risk assessment guidance for aquatic habitats include: 1) formation of an Ecorisk Group within the USEPA Office of Water to develop a paradigm for ecological risk assessment, 2) development of sediment quality criteria by USEPA based on the equilibrium partitioning model (DiToro et al. 1991), 3) issuance of a review of aquatic risk assessment methods by Parkhurst et al. (1990) for the Water Environment Federation (formerly the Water Pollution Control Federation), and 4) USEPA's development of proposed methods for the derivation of ambient water quality criteria that would be protective of human health, wildlife, and aquatic organisms under the Great Lakes Initiative (40 CFR Parts 122, 123, 131, and 132; 58 Fed. Reg. 20802).
Because of the varied nature of contaminated sediment sites and the objectives of individual ecological risk assessments, only general guidance is offered here. Ecological risk assessors must still rely on their own judgment and expertise when evaluating potential risks to wildlife and/or aquatic organisms. Thus, any ecological risk assessment should include a clear statement of assumptions and an uncertainty analysis. The risk assessment process developed for estimating human health risks generally applies to determination of ecological risks. However, the complexity of ecological systems requires consideration of multiple species and other physical-chemical stressors in addition to toxic chemicals. Ecological endpoints may also differ from those used in human health risk assessment. For example, survival, growth, and reproduction may be emphasized as ecological endpoints, instead of cancer or more subtle sublethal effects. In ecological risk assessment, risks to populations, communities, and ecosystems are often considered more relevant than individual risk. Except in the case of rare, threatened, and endangered species, individual plants and animals are not highly valued because compensatory mechanisms in ecological systems may preclude higher-level effects even if individuals are eliminated from a population. In ecological risk assessment, the ability of the ecosystem to recover from the stress may also be considered. USEPA (1992b) defines ecological risk assessment as "a process that evaluates the likelihood that adverse ecological effects may occur or are occurring as a result of exposure to one or more stressors." In general, both wildlife and aquatic risk assessments follow the basic framework shown in Figure 2-2 (USEPA 1992b):
In ecological risk approaches, assessment endpoints are defined as environmental characteristics or values that are to be protected, such as wildlife population abundance, species diversity, or ecosystem productivity (Suter 1989). For example, maintenance or restoration of valuable natural resources is typically a goal of the remediation process. If protection of a valuable commercial fish stock is the goal, the assessment endpoint may be recruitment rate for the species population. Measurement endpoints are quantitative expressions of an observed or measured biological response related to the valued environmental characteristic chosen as the assessment endpoint. In some cases, the measurement endpoint is the same as the assessment endpoint. When these endpoints differ, a model must be used to express their relationship. The process of estimating ecological risk based on chemical, toxicological, and ecological data is called a forward-mode assessment. Risk assessment procedures may also be used to back-calculate exposure guidelines from an allowable risk level or a no-observed-adverse-effect level (NOAEL). A reverse-mode assessment may be used to derive cleanup levels (e.g., maximum allowable concentrations of contaminants in sediment).
The level of detail required for a given risk assessment depends on remedial action objectives, the complexity of the site, and the difficulty in adequately describing exposure, toxicity, and other properties of the contaminants of concern. An ecological risk assessment can be conducted in tiers with the most basic analyses conducted first. For example, an initial screening-level risk assessment is conducted that uses available data and conservative assumptions about exposure and toxicity. From the results of this screening-level assessment, areas, contaminants, and species of concern are identified and decisions are made about additional data collection. In the next tier, more realistic models are used and additional data may be collected that will better define the relationship between chemical concentrations and adverse effects at the site. The conceptual model developed during the planning phase of an ecological risk assessment illustrates how exposure to sediment contaminants may cause ecological effects. The results of the problem formulation stage clarify the scope of an ecological risk assessment and how the results will be used in developing RAPs for the AOCs. Based on the results of a screening-level assessment, chemicals, species, and endpoints are selected for a detailed assessment that may involve collection of additional field data and risk modeling.
Selection of Contaminants of Concern
The last step in the selection process is to compare measured or estimated environmental concentrations with threshold concentrations such as NOAELs. This analysis should be conservative by incorporating plausible worst-case assumptions regarding bioavailability, exposure, and sensitivity of ecological receptors. The list of contaminants of concern for an ecological risk assessment may differ from that used for a human health risk assessment because of differences in exposure pathways, uptake, and sensitivities between humans and ecological receptors. Therefore, the initial list of contaminants considered for evaluation in the selection process described above should be comprehensive rather than simply the list selected for evaluation in the human health risk assessment.
Selection of Species
These criteria for selection of indicator species were adopted for use in the ARCS Program. In addition, species selected for the assessment should be sensitive to effects of the contaminants of concern and, if possible, should be representative of a group of valuable species. The U.S. Fish and Wildlife Service is evaluating the bald eagle, mink, otter, colonial waterbird group, and lake trout (salmonids) as possible indicator species for Great Lakes water quality (Kubiak and Best 1991). These species can also be used to evaluate the effects of contaminated sediments. In addition, other aquatic biota that should be considered in most risk assessments include benthic macroinvertebrates and bottom-feeding fish. Any of the species or groups just discussed could be considered in a predictive assessment. For empirical assessments of existing conditions, the following ecosystem indicators are recommended:
For a specific risk assessment, the selection of species to be evaluated also depends on the contaminants of concern and the scale of resolution needed to define problem areas within AOCs. For example, assessment of bioconcentratable contaminants such as PCBs and dioxins over a wide portion of Lake Michigan might consider wide-ranging fish and wildlife species that feed at high trophic levels. Relatively fine-scale resolution of problem areas within a tributary might consider benthic macroinvertebrates and localized bottom-feeding fish species like bullhead. The selection of species and endpoints for an assessment should consider whether the contaminants biomagnify and whether they cause direct toxicity to receptors at lower trophic levels. Practical methods for field and/or laboratory measurements must be available for retrospective assessments. Predictive and retrospective assessments both require adequate data on contaminant distributions or appropriate transport and fate models to estimate exposure. Limitations in data or models may influence the final selection of species for the assessment.
Selection of Endpoints Various endpoints may be used for predictive assessments, but their final selection is often affected by the availability of toxicity data in the literature and the quality of the data. Because prediction of community-level responses from survival, growth, and reproductive endpoints involves substantial uncertainties, wherever practical, effects on selected communities should be directly observed in the field. For example, population and community analyses of benthic macroinvertebrate communities may be used to evaluate toxic effects of sediment contamination.
Ecological Risk Assessment Objectives Assessment techniques appropriate for the receptors and contaminants of concern and the level of complexity of the risk assessment are determined on a site-by-site basis. The selection of ecological assessment techniques to be applied at a site depends on the objectives of the risk assessment, site-specific receptor species and contaminants of concern, and the extent of available data. The primary techniques are:
Each combination of tools selected for an AOC should provide adequate data for the assessment and facilitate risk predictions. Figure 2-3 summarizes some of the candidate tools for ecological risk assessments according to habitat, media, and receptors. The problem formulation stage should include development of a strategy for integrating the results of individual assessment tools into the overall approach to risk characterization. Moreover, the risk assessment objectives should be clearly related to remedial action objectives and the decision-making process. The overall assessment strategy may involve both empirical and theoretical approaches. Empirical approaches involve direct measurement of biological effects or derivation of relationships between chemical and biological variables from field data, or toxicity testing of field-collected samples. Empirical approaches rely heavily on observed relationships without attempting to describe theoretical cause-effect relationships. Warren-Hicks et al. (1989) describe empirical assessment approaches used to quantify the ecological effects of contaminants at hazardous waste sites, including: 1) toxicity testing, 2) use of biomarkers, including analysis of tissue contaminants, and 3) community analysis based on field surveys.
Theoretical models are mainly derived from theoretical principles and include explicit mechanistic (cause-effect) relationships. Modeling may be used to support empirical risk assessments and to make risk predictions. Mathematical exposure models are used for dynamic systems (such as river water), for long-term predictions in more stable systems (such as sediment), and for transfer of chemicals through the food web to receptors higher up the food chain. Ecological models are used primarily to extrapolate from measurement endpoints to assessment endpoints in receptor groups such as amphibians, birds, and large mammals. Both exposure models and ecological models may vary from relatively simple extrapolation models with few data requirements to complex mechanistic models with substantial data requirements. Whenever practical, models should be based on site-specific data and validated. In the exposure assessment phase, measurements or estimates are made of the concentrations of contaminants of concern in the environment or the rate of chemical intake by organisms. Analysis of the magnitude, duration, and frequency of exposure is based on information or assumptions about:
For empirical assessments, tissue concentrations of contaminants in key species may be measured as indicators of exposure. To develop estimates of exposure using models, exposure scenarios are developed from the conceptual site model to describe the pathways a chemical may take through various environmental media to reach an organism. For each site, the analysis of several exposure scenarios helps to identify data gaps for transport pathways and key exposure processes, such as chemical transformations or biological uptake. Data gaps for specific chemical forms or processes related to important pathways are filled through estimations from predictive chemical models or the collection of additional site-specific data. The distributions and seasonal activity patterns of receptors are described relative to contaminant distributions in various habitats at a site. Habitats, concentrations of contaminants of concern, species distributions, and exposure variables related to species activities may be mapped and spatial patterns investigated using a mapping/database system such as a geographic information system. Information from the transport and fate analysis for each exposure scenario is used to develop quantitative estimates of exposure that serve as inputs to the risk characterization. Summaries of data for the exposure assessment may include:
Wildlife The exposure assessment for indicator species can be approached in two different ways (Kubiak and Best 1991). The first approach requires site-specific information or estimates of the types of forage items commonly eaten, contaminant concentrations in forage items, and grams of food eaten per kilogram of wildlife predator body weight. This information is used to calculate an ingested dose for contaminants of concern in the indicator species, similar to food web model approaches currently being used for sediments and other media at hazardous waste sites (e.g., Fordham and Reagan 1991; Menzie et al. 1992). The calculated ingested dose is then compared to a NOAEL developed from a model feeding study where known adverse toxicological endpoints were measured. There is considered to be a significant risk to the indicator species if the calculated ingested dose exceeds the NOAEL. Unfortunately, the database necessary for applying this approach is not sufficiently developed for Great Lakes species. The second approach can be conducted for contaminants that bioaccumulate through the food chain (e.g., DDE, PCBs, dieldrin). In this approach, the concentration of a contaminant of interest is measured in a specific tissue of the indicator species. Because the exposure estimate is expressed as a contaminant concentration in tissue, the ecological effects assessment includes an estimate of the NOAEL expressed as a contaminant concentration in the tissue rather than as an ingested dose. The measured tissue contaminant concentration is compared with the NOAEL to evaluate existing risk. Based on the NOAEL and on biomagnification factors calculated from actual field data, Kubiak and Best (1991) also applied this approach to backcalculate the contaminant concentrations in forage fish that would be necessary to result in a contaminant concentration equal to the NOAEL in the tissue of the indicator species. Either approach, appropriately used, could be applied to the calculation of "safe concentrations" or remedial action goals for sediments. The goal would be to determine the contaminant concentrations in sediments that would not result in exceedances of the NOAELs, expressed as an ingested dose to the indicator species in the first approach or as a tissue contaminant concentration in the second approach. For the ARCS Program, the second approach was applied for bald eagle, other fish-eating birds, and mink in the wildlife risk assessments conducted for the Buffalo River AOC (Mann-Klager, in prep.) and the Saginaw River AOC (Kubiak, in prep.).
Aquatic Life
In most cases, the exposure estimate will be expressed as the contaminant concentration in the bulk sediment, pore water, or water column, which is obtained from empirical measurements. Few toxicity data are available to interpret exposure estimates expressed as contaminant concentrations in tissue (e.g., Dillon 1984). Ecological effects assessment includes a hazard identification step to identify the potential effects of chemicals and an exposure-response assessment to characterize the relationship between each stressor and the biological or ecological endpoints. Confounding effects of physical stressors such as currents or sediment grain size must be addressed by using models, reference-area measurements, or experimental designs to separate the effects of physical factors from those of chemicals. Techniques for ecological effects assessment may include the following:
Data from direct field measurements or from laboratory analyses of field samples should be used whenever possible to derive exposure-response relationships. Site-specific properties of sediment and water may modify the bioavailability of chemicals, and literature data may not be appropriate for ecological effects assessment.
Wildlife An estimate of the NOAEL for each contaminant of concern can be derived from the literature or from a model feeding study on a surrogate laboratory species. NOAELs can be expressed as chemical intake rates, which typically result from ingestion of chemicals, or chemical concentrations in tissue. The units for the NOAEL must correspond to the units used for the exposure estimate. When developing NOAELs, the following guidelines developed by PTI (1992) are relevant:
An endpoint for a relatively sensitive tissue, organ, or life stage should be determined to derive a conservative estimate of a NOAEL. Adverse effects on reproductive organs and early life stages are typically good endpoints for risk assessment because they are likely relevant to changes at the population level. For birds, the most sensitive stages are the egg and the developing embryo, at least for chlorinated organic compounds and methylmercury. In contrast, the liver is the most sensitive tissue known for mink (Kubiak and Best 1991).
Aquatic Life
Guidance on the use of toxicity tests, benthic macroinvertebrate community surveys, and fish histopathology investigations in support of ecological risk assessments is provided in the ARCS Assessment Guidance Document (USEPA 1993). In risk characterization, the exposure and ecological effects assessments are combined to estimate the probability of adverse ecological effects. Final risk estimates may be expressed in simple narrative terms or as quantitative values. The risk characterization should summarize:
Mitigating factors, such as reduced bioavailability of contaminants incorporated into sediment particles and mechanisms for possible wildlife avoidance of hot spots, should be discussed. The importance of mitigating factors should be confirmed by field measurements or laboratory experiments on samples from the site. Also, compensatory mechanisms that preclude population- or community-level effects should be acknowledged, even though effects on individuals may be predicted. Approaches to develop quantitative risk estimates (or hazard indices) include the quotient method, joint probability analysis, model uncertainty analysis, and integrated analysis of site-specific empirical data (Barnthouse et al. 1986; Suter et al. 1992; Ginn and Pastorok 1992; Chapman et al. 1992). The quotient method uses a ratio of the value of an endpoint at the site to a toxicity reference value such as a NOAEL as an approximate risk index. The quotient method is useful mainly for screening-level analyses because it does not provide a complete characterization of the magnitude of risk and uncertainties (Bascietto et al. 1990). Joint probability analysis (Barnthouse et al. 1986) can be applied to estimate the risk that exposure exceeds toxicity thresholds or criteria where probability distributions are available for the variables being compared. Model uncertainty analysis (Barnthouse et al. 1986; Suter et al. 1992) may be used to develop risk estimates for a species based on statistical analysis of growth, survival, and reproduction of individuals. Approximate risk estimates may be derived for an ecological system associated with sediments by combining site-specific data for chemicals, toxicity tests, community indices, and possibly other risk indicators (Chapman et al. 1992).
Wildlife The joint action of contaminants should also be considered. Hazard quotients for individual contaminants with similar modes of action may be summed to yield a hazard index.
Aquatic Life
Uncertainty Analysis The baseline aquatic risk assessments for the ARCS Program were designed to complement the baseline human health and wildlife risk assessments so that the exposure pathways leading from sediments to fish to humans and wildlife could be quantified. However, these aquatic risk assessments were difficult to perform for different trophic levels of the aquatic food chain because of data gaps. Conservative assumptions and published data from other studies were used to fill missing information. The problem of data gaps affects most aquatic and wildlife risk assessments and constitutes a major source of uncertainty in the risk estimates.
Uncertainty analysis should be conducted during the problem formulation stage to identify data gaps and plan the approach for the ecological risk assessment. When significant data gaps exist, the assessment should typically include several tiers of analysis, with use of available data in the early screening tier to help define critical data needs to be addressed in further field sampling or laboratory studies and subsequent analyses in the next tier. Mass balance modeling studies were conducted at the Buffalo and Saginaw River AOCs to demonstrate the use of this approach in evaluating remedial alternatives for contaminated sediments. Mass balance modeling studies were applied at these AOCs to estimate changes in contaminant concentrations in water, sediments, and biota that may occur following sediment remediation. The estimated concentrations can then be used to compute changes in human health and ecological risks and to aid in the selection of remedial alternatives. The mass balance modeling studies, as described below, are based on established models and methods and are considered applicable to other sites with contaminated sediments. OVERVIEWWhile there are many possible remedial alternatives for contaminated sediments, only a few may be feasible at a particular site. After the range of feasible remedial alternatives is identified for a site, the potential for reduction of contaminant concentrations in water, sediments, and biota must be considered when selecting a remedial alternative. This selection process requires some method to estimate changes in contaminant concentrations that would result from each remedial alternative. The preferred approach for estimating these changes, and the approach used in the ARCS RAM studies, is based on the application of mass balance models. Mass balance models attempt to describe each of the underlying mechanisms causing change in the system, and are termed mechanistic. In the mass balance modeling approach, the law of conservation of mass is applied in the evaluation of the sources, transport, and fate of contaminants. This approach requires that the quantities of contaminants entering the system (i.e., contaminant loading) equal the quantities leaving the system, less the quantities stored, transformed, or degraded. Thus, the mass balance is simply a bookkeeping of all of the processes affecting the mass of contaminants in a system. After the mass balance has been established for each contaminant of concern, quantitative changes in contaminant concentrations can be estimated. For example, the mass balance can be used to estimate the change that may be expected following removal of some portion of the contaminant mass. Mass balance models have been successfully applied to the Great Lakes and elsewhere in the regulation of toxic and conventional pollutants. Properly applied, a mass balance model can serve as a surrogate for the natural system that is easily manipulated to estimate the system's response to change. Management questions that can be addressed with a mass balance model include:
COMPLEXITY OF THE MASS BALANCE MODELING STUDYMass balance modeling studies vary widely in their complexity. The modeling approach can vary from simple screening calculations to applications of complex computer programs. Modeling studies also vary in the complexity of the field studies used to support the mass balance calculations, from studies relying solely on historical data to large and expensive field efforts. The degree of complexity required depends on the physics of the system, factors affecting the transport and transformations specific to the contaminants of concern, and the management questions the mass balance modeling study will address. The degree of complexity used in particular studies is often dictated by the time and funding available. Modeling studies of contaminants in the Great Lakes have typically been cataloged into groups or study levels depending on the level of effort and complexity in the modeling and supporting field studies. Studies have been categorized as either "screening" or "detailed" studies, as well as "Level 0" through "Level III" studies. The study levels can generally be described as:
Generally, the level of uncertainty is reduced as the studies increase in complexity from Level 0 to Level III, while the time and costs associated with each study level increase. Increases in modeling complexity need not always correspond to increases in data collection, so various combinations of the levels mentioned above are possible. For example, simple models usually predict average conditions, so sufficient data must be collected to provide an accurate estimate of the average condition for comparison with predictions. However, more complex models may predict conditions at a specific point (in time and space), thus reducing the need for averaging. In the ARCS RAM studies, both near-field and far-field modeling studies were conducted. The near-field modeling studies concentrated on the lower portions of the Buffalo and Saginaw Rivers where sediment contamination was the greatest and where remedial actions will likely be implemented. The far-field modeling studies were used to estimate the impact of remediation of the Buffalo and Saginaw Rivers on their receiving waters--Lake Ontario and Saginaw Bay, respectively. While the general mass balance modeling approaches used in the near-field and far-field studies were similar, they differed in the resolution of the models applied and in the level of supporting field studies. The ARCS near-field studies were categorized as Level I or a "Mini-Mass Balance" modeling effort. The primary criterion restricting the modeling studies to Level I was the limited level of supporting field studies.
The far-field modeling was more typical of Level 0, in terms of data collection, and Level I, in terms of modeling. These studies were designed to estimate the long-term impact of changes in loadings from the rivers on their respective receiving water bodies. As such, hydrodynamics and sediment transport were described rather than predicted, and the resolution was not as precise as in the near-field studies. The far-field modeling studies of Lake Ontario and Saginaw Bay relied exclusively on historical data. The application of the mass balance modeling approach involves the quantification of the sources, transport, and fate of contaminants. The components of the modeling approach are illustrated in Figure 3-1. The typical steps in a mass balance modeling study are: 1) predict water and sediment transport, 2) use the predicted water and sediment transport, along with estimates of contaminant loadings from point and nonpoint sources, to estimate the changes in chemical concentrations in water and sediments, and 3) use the predicted contaminant concentrations in water and sediments to estimate the transfer of contaminants through the food chain and their accumulation in fish. The models used for each step of this process are described below. Water Transport ModelsThe first step in characterizing the transport of dissolved contaminants is to characterize the transport of the water, or its motion. Often, much of the variability of contaminant concentrations in the water column can be explained by water transport alone. Water transport models may be descriptive (i.e., based on a balance of the water's mass) or hydrodynamic (i.e., based on a balance of the water's momentum). Characterization of water transport may be qualitative or quantitative, depending on the study level. In a qualitative approach, flow patterns are either measured directly or inferred from measurements of related parameters. The qualitative approach is often adequate where the system is very simple (hydraulically) or where only long-term, relatively rough estimates of water transport are required. A qualitative approach is most often used in Level 0 and Level I studies and is occasionally used in Level II studies. Hydrodynamic models are used to quantitatively predict changes in volumes, depths, and velocities in response to changes in flow or water surface elevations. Hydrodynamic models require data on boundary conditions (i.e., flows or water surface elevations, wind speed and direction), which are applied to predict the resulting flows within the modeled system. Flows are often measured at gaging stations on many Great Lakes tributaries, and water surface elevations are routinely measured at many locations within the Great Lakes. Where such direct measurements of flows in the tributaries of the Great Lakes are problematical because of the complex interactions and relationships between upstream inflows and lake effects, hydrodynamic models can be used to quantitatively predict flows. Hydrodynamic models can also be used to estimate changes in flows that may occur under future conditions, such as in evaluating the effects of changes in dredging patterns. The hydrodynamic models used in the ARCS RAM modeling studies included HYDRO- 3D, a 3- dimensional hydrodynamic and transport model, and RIVMOD, a 1- dimensional model. Both models are maintained and distributed by the Center for Exposure Assessment Modeling at Athens, Georgia. Multidimensional models were required where it was necessary to resolve variations in water transport and chemical characteristics along the longitudinal, vertical, and lateral axes of the rivers. For example, a model that was capable of estimating variations in transport with depth was required in the Buffalo River because of water column stratification during low-flow periods.
A qualitative approach was taken for the far-field studies. For Lake Ontario, historical measurements of inflows and outflows were used to compute water transport, while for Saginaw Bay, water transport was computed from measured flows and chemical data. Because contaminant concentrations in surface sediments are a primary concern in the AOCs, adequate characterization of the movement of surface sediments is a critical step in the mass balance modeling process. There are two primary goals of the sediment transport model: 1) to predict the movement of the sediments to estimate changes that may occur in patterns of erosion, deposition, and transport, and 2) to estimate the transport of the particulate contaminant mass. Sediment transport models are based on a sediment mass balance. As with water transport, sediment transport may be described either qualitatively or quantitatively in mass balance modeling studies. Qualitative approaches have commonly been used in Level 0 and Level I studies. In studies of this type, measured or estimated settling, resuspension, and sedimentation rates are used to compute the transport of sediments and their associated contaminants. Although it is generally assumed that there is no net sedimentation or erosion, the different contaminant concentrations on suspended and settled particulate matter make the quantification of these processes important. The qualitative approach has proven useful in providing rough estimates of the effects of sediment transport on contaminant distributions. However, sediment transport is a very dynamic process and the assumption of no net sedimentation or erosion is a gross simplification. In more quantitative sediment transport models, resuspension and transport are computed using the output of a hydrodynamic model and the measured characteristics of the sediments. The sediment types of primary importance are silts and clays, rather than sands. Silts and clays are classified as cohesive sediments, while sand is classified as a noncohesive sediment. The sediment transport model is used to predict changes in sediment resuspension, deposition, and transport and the effect of these processes on particulate concentrations in the water column. The sediment transport model is also used to estimate changes in the structure of the sediment bed, such as the impact of erosion or deposition on the channel shape and sediment composition. As with hydrodynamic models, sediment transport models can be used to interpolate between existing measurements or estimate sediment transport for conditions for which data are not available. The majority of sediment transport occurs during extreme (rare) events, such as storms on lakes and large runoff events in rivers. Because data are rarely available for these events, sediment transport models must be used to estimate transport under these conditions. For example, these models may be used to estimate whether contaminated sediments may be buried or exposed by a 10- year or 100- year flood. This information can be used in the evaluation of remedial alternatives as well as the no-action alternative. Sediment transport models may also be used to evaluate the impact of removing or immobilizing sediments on subsequent erosion and deposition patterns. For example, if sediments are removed from a particular area, sediment transport models may be used to estimate how long it may take for the area to fill in, as well as changes that may occur in deposition and erosion areas.
In the ARCS RAM modeling studies, a model of the transport of cohesive sediments was applied to both of the near-field study sites (i.e., Buffalo and Saginaw Rivers) to predict the interactions between the transport, deposition, and resuspension of sediments under various meteorological and hydrological conditions. The model applied was a 2- dimensional (longitudinal and lateral), time-variable, hydrodynamic and sediment transport model developed by Wilbert Lick at the University of California, Santa Barbara. This model has been applied at various locations around the Great Lakes, including the Detroit River, Fox River, Green Bay, Lake Erie, and elsewhere. The sediment transport model was coupled with a 3- dimensional, time-variable model of the sediment bed and its properties, the Water Quality Analysis Program, WASP4 (Ambrose et al. 1990). WASP4 integrates predictions from the hydrodynamic and sediment transport model to estimate contaminant concentrations in the water and sediment. The WASP4 model, maintained and distributed by the Center for Exposure Assessment Modeling at Athens, Georgia, has been widely distributed. WASP4 is currently the framework used for modeling studies in Green Bay, Lake Michigan, Lake Ontario, and elsewhere around the Great Lakes. It is generally assumed that the water column is completely mixed vertically because sediment resuspension typically occurs at flows where this would be the case. This model was used to predict variations in suspended solids concentrations, resuspension and deposition rates, and variations in the sediment bed as a function of flows and loadings. The model provided predictions for use in determining the transport of sorbed contaminants and resuspension of toxic sediments. Application of the sediment transport model was of particular importance in these studies because of the lack of historical sediment data. Therefore, these studies relied heavily on the sediment transport model to supply estimates of sediment transport, resuspension, and deposition. The contaminant exposure model is the mass balance model for contaminants. The contaminant exposure model used as a framework for the ARCS RAM mass balance modeling studies was WASP4. In the application of the contaminant exposure model, the rate of change in contaminant mass (accumulation) is a function of the transport of a contaminant into, out of, and within the system via water transport or sediment transport for those materials that sorb to sediments; the mass added to the system via point and nonpoint loadings minus the outputs; and the quantities transformed and degraded within the system via processes such as volatilization, biodegradation, and photodegradation. The output expected from the contaminant exposure model includes estimated contaminant concentrations of both particulate and dissolved forms in water and sediments, as well as estimated changes in mass fluxes due to inflows and loadings, outflows, and degradation and transformation processes. Depending on the level of complexity, the transport via water and sediments may be described or predicted using hydrodynamic and sediment transport models, which are then coupled with the contaminant exposure model. In the ARCS RAM near-field studies, WASP4 was applied to predict the effects of water transport, sediment transport, sorption, and transformation processes on the concentrations of selected contaminants of concern. The WASP4 model was linked to the output of the hydrodynamic and sediment transport models, which together provided the necessary transport information, using data collected during ARCS field studies. The output of the contaminant exposure model included temporally and spatially varying estimates of contaminant concentrations in water (both particulate and dissolved) and sediments for comparison to field data and for projections of the effectiveness of various remedial alternatives. In addition, the output included estimates of the magnitude of all processes that result in gains or losses of contaminants, so that their relative importance could be evaluated. The contaminant exposure model also provided information that could be incorporated into the food chain model to estimate the contaminant body burdens in fish due to varying contaminant concentrations in water and sediment.
The same contaminant exposure model used in the near-field studies (WASP4) was applied in the far-field studies. However, in the far-field studies the resolution was not as precise as in the near-field studies. Therefore, there was no need to apply the hydrodynamic and sediment transport models in the far-field studies. WASP4 was applied to predict steady-state (long-term average) conditions over large spatial scales. The output of the model included load-response relationships. For example, given a particular contaminant loading, the model predicted the average contaminant concentrations in water [both particulate and dissolved] and sediments. This relationship was then used to evaluate the impact of changes in loadings from the two AOCs on Lake Ontario and Saginaw Bay. The food chain model is a mass balance model for contaminants where the rate of change in contaminant mass in each component of the food chain is a function of the transport of a contaminant into and out of that component (e.g., via ingestion, gill exchange, excretion), as well as internal changes that may occur due to growth. The food chain model supports evaluation of the impact of various remedial alternatives including the no-action alternative on contaminant concentrations within the food chain, given variations in contaminant concentrations in water and sediments derived from the contaminant exposure model. Outputs from the food chain model include time-variable estimates of contaminant concentrations in each component of the food chain.
The food chain model applied to both the Buffalo and Saginaw Rivers as part of the ARCS RAM studies was the Water Quality Analysis Simulation for Toxics, WASTOX (USEPA 1985), a predecessor of WASP4 that includes a food chain component. In the Buffalo River, the only fish species simulated was carp, because it is a bottom-feeder with a high fat content and would be expected to have an appreciable contaminant body burden. In the Saginaw River, the food chain model included forage fish and walleye, the latter due to its high importance to recreational anglers. Data collected as part of the ARCS RAM studies were used to construct a simple food chain model. Field data are required to apply any mass balance model. For model application, field data are required to define the characteristics of the site and the mass fluxes of water, sediment, and the associated contaminants into and out of the site. In addition, data are often required to estimate the site-specific values of model parameters and to assess the uncertainty associated with model projections. The confidence that can be placed on those projections is dependent upon both the integrity of the model and how well the model is calibrated to that particular water body. While models can be run with minimal data, the resulting predictions are subject to large uncertainty. Models are best used to interpolate between existing conditions but may be used to extrapolate from existing to future conditions, such as in the evaluation of the effects of remedial or mitigative alternatives.
The types of data required and the necessary frequency of measurement vary with the level of model complexity, the characteristics of the system, and the contaminants of concern. Generally, three kinds of data are required: 1) boundary conditions, 2) initial conditions, and 3) data for calibration/evaluation. Boundary conditions are external to the model (i.e., the model does not predict them). Instead, they are used to "drive" the model. Initial conditions are used to aid in the design of the model application (e.g., to determine segmentation) and to provide a starting point for model predictions. For example, the initial, or existing, sediment contaminant concentrations are required to provide a starting point for model predictions. Temperature data may be used to determine the need for predicting water-column stratification and its effects on contaminant transport. Model calibration/evaluation data include the parameters that the model is designed to predict (e.g., contaminant concentrations and fluxes). These data are compared to the model's predictions to aid in determining the values of site-specific parameters and in estimating the uncertainty associated with those predictions. The types of data required for each component of a mass balance modeling study are described below. Boundary condition data for the water transport models include the system's bathymetry, point source and nonpoint source flows and loadings, upstream and downstream flows and/or water surface elevations, and weather data (e.g., wind speed and direction, temperature). These data can often be obtained from the National Oceanic |
