Ground Water Models: Scientific and Regulatory Applications (1990)

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

s Experience With Contaminant Flow Models in the Regulatory System INTRODUCTION This chapter is divided into two parts: (~) a review of federal regulations and guidance concerning the use of contaminant trans- port models, and (2) five case studies illustrating the site-specific application of such models. These sections are based on the com- mittee's review and interpretation of these regulations and guidance, existing reports on the use of such models, discussions with agency personnel, and the personal experience of the committee members. This chapter focuses on the regulations, guidance, and prac- tices of the U.S. Nuclear Regulatory Commmsion (USNRC) and the U.S. Environmental Protection Agency (EPA). These two regulatory agencies deal with contaminant transport from historic or proposed disposal facilities and recognize the need to evaluate present condi- tions and predict potential migrations. Both agencies have programs in place that require mocleling. However, each agency suffers from unique problems that reflect its particular regulatory concerns. The USNRC has had a number of years to prepare for an am plication for a high-level radioactive waste disposal. As a result, the agency has had the opportunity to develop detailed procedures on reviewing mode} applications. Unfortunately, because of changes in federal programs, the developed procedures are largely untested. In 160

CONTAMINANT FLOW MODELS IN THE REGULATORY SYSTEM 161 contract, EPA has had to evaluate a large number of modeling stud- ies as part of the Superfund program. Because of the rapid increase in sites being evaluated, EPA has not had an opportunity to develop a systematic plan for mode] review or application. In the following sections the two agencies' approaches to the use and review of models are summarized. U.S. NUCLEAR REGULATORY COMMISSION REGULATIONS AND GUIDANCE One of the USNRC's responsibilities is the licensing of facili- ties for the disposal of low-level and high-level radioactive wastes (see 10 CFR Part 61, relicensing Requirements for Land Disposal of Radioactive Waste," and 10 CFR Part 60, "Disposal of High-Leve! Radioactive Wastes in Geologic Repositories; Licensing Procedures," respectively). To be licensed, a facility must meet certain require- ments. For example, one requirement is that the site be capable of being modeled (10 CFR Part 61~. Thus the USNRC has embed- ded into its regulations and guidance general principles concerning contaminant transport modeling. However, this guidance is largely untested because the USNRC has performed only limited licensing for waste disposal facilities. Low-leve} radioactive waste (~LW) is generated by a number of institutions including industries, laboratories, hospitals, and facilities involved in the nuclear fuel cycle. Wastes are packaged and placed in shallow excavations or engineered structures that are then back- fi~led and capped to limit infiltration. The USNRC LLW disposal regulations specify performance objectives and specific technical re- quirements for site suitability that are designed to adequately protect public health (Siefken et al., 1982~. One of the requirements is that "the disposal site shall be capable of being characterized, modeled, analyzed, and monitored" (U.S. Nuclear Regulatory Commission, 1987~. The purpose of this requirement is to ensure that the hydro- geological conditions of the site are adequately understood through field studies. The USNRC has also developed standard review plans (SRPs) (U.S. Nuclear Regulatory Commission, 1987) that direct the USNRC staff in evaluating the potential for migration for a disposal facility. Review plans have been issued to evaluate a number of potential migration pathways including radionuclide movement through the

162 GROUND WATER MODELS ground water and movement of radionuclides resulting from infiItra- tion through the ground surface. The SRPs for ground water and infiltration contain information on the amount of modeling planned by the regulatory agency as well as the type of issues that will be reviewed by the agency. The SRP indicates that the license application will be reviewed to determine whether the use of the input parameters has been justified and whether the data are sufficient to provide a reasonably accurate or conservative analysis regarding ground water pathways. The transport models will also be evaluated for their defensibility, suitability, and basic conservatism. The codes must be based on sound physical, chemical, and mathematical principles and must be correctly applied and sufficiently documented. The applicant must supply the following: a complete description of the contaminant transport path- ways between the engineered disposal unit and the site boundary and existing or known future ground water user locations; ~ estimates and justification for the physical and chemical input parameters used in the transport models to calculate radionuclide concentrations; a description of the contaminant transport models used in the analysis, including modeling procedures and complete documen- tation of the codes as required in NUREG-0856 (U.S. Nuclear Regu- latory Commission, 1987, p. 6.1.5.1-4~; ~ the justification, documentation, verification, and calibration of any equations or program codes used in the analyses; and ~ the description of data and justification for the manipulation of any data used in the analyses (p. 6.1.5.2-3~. The SRP does not attempt to quantify the level of information re- quired to adequately characterize the potential ground water trans- port at the sites nor does it outline the acceptance criteria for ad- equate site modeling. To evaluate the applicant's submittal, the USNRC will use "simple analytical modeling techniques with demon- strably conservative assumptions and coefficients" (p. 6.~.5.1-3~. The SRP does not outline which codes will be used, and no other support- ing documentation was provided that outlined the codes planned for use by the USNRC. The SRP guidance states that If the applicant's results are more realistic than conservative, then the applicant must clearly justify the application and results of the models (p. 6.~.5.2-3~. More

CONTAMINANT FLOW MODELS IN THE REGULATORY SYSTEM 163 sophisticated numerical modeling will be performed by the USNRC when the issues relating to the applicant's modeling studies cannot be resolved. The SRP does not discuss the apparent disparity between requesting field information to characterize the site and the use of conservative data in the modeling process. The LLW program appears to have developed a systematic plan for incorporating modeling into the site evaluation process. The plan has attempted to consider, in a general way, the reliability of the data input, as well as the documentation and reliability of the computer codes. The program, as outlined in the USNRC guidance and the SRPs, appears to emphasize conservatism, although the regulations place equal emphasis on collecting adequate field information. Also, the program has not attempted to direct applicants toward partic- ular computer codes because the codes the USNRC will use are not defined. The USNRC also has published extensive documentation on the codes that are planned for potential use in evaluation of license applications. The publication of this documentation allows license applicants to consider using USNRC codes or to review their code choices against the USNRC-distributed tools. The USNRC guidance is designed to evaluate whether the models accurately simulate the phenomena that are considered and to de- termine whether the numerical approximations accurately solve the mathematical equations. The test problems include analytical and semianalytical solutions, as well as problems based on laboratory or field studies. By providing a standardized process of mode} evaluation, the USNRC is attempting to limit the amount of code comparison that will be required at the time of license application. The USNRC (1982) outlines the level of documentation deemed adequate, i.e., The documentation of mathematical models and numerical methods will provide the basis for USNRC's review of the theory and means of solution used in the code. It should contain derivations and justification for the model. The documentation will help the USNRC in understanding modeling results that are submitted by the applicant during the licensing process and permits the USNRC to install and use the code on its own computer. The USNRC guidance also outlines a computer software man- agement system that will provide a software storage system to ensure

164 GROUND WATER MODELS future retrievability of computer codes and will provide a standard- ized testing process for applied codes. The storage system will include a catalog of modifications and the most updated version of the codes in use. The USNRC is (1) assembling mathematical models for assessing Department of Energy (DOE) demonstrations; (2) cleveloping com- puter software for use in assessing the long-term risk from disposal of radioactive wastes in deep geologic formations, in estimating dose commitments and potential adverse health effects from released ra- dionuclides, and in performing sensitivity and uncertainty analyses; and (3) developing a quality assurance program to ensure adequate quality in computer codes developed and in data generated by these codes, as well as for maintenance of the programs. This program re- quires peer review and management approval to ensure a systematic record of calculations and analyses that are performed. In summary, the USNRC has attempted to define a process that considers not only the problems in evaluating mode! results but also the issues surrounding code selection and application. The guidance documents have attempted to direct applicants to the appropriate level of code review without limiting the choice of code selection. U.S. ENVI:ELONMENTAL PROTECTION AGENCY REGULATIONS AND GUIDANCE The U.S. Environmental Protection Agency uses a wicle variety of contaminant transport moclels en c} has a large number of specific sites where such models are used and will be used. The key EPA regulations and guidance affecting the use of contaminant transport moclels e.g., those in the SuperfuncI, hazardous waste management, ant} underground injection programs-are incluclec! in the following · ~ c .lscusslon. Sup erfund Law and Regulations Superfund is the environmental law that authorizes EPA to ~ identify sites where hazardous substances have been releaser} into the environment; ~ clean up such contamination and recover the costs from the responsible private parties; or

CONTAMINANT FLOW MODELS IN THE REGULATORY SYSTEM 165 ~ in the alternative, (~) order a private party to perform the cleanup or (2) obtain a voluntary agreement from such private party (called potentially responsible party [PRP]) to perform the cleanup.) Superfund Is primarily directed at cleanup of inactive hazardous waste sites. Courts generally have resolved legal uncertainties and issues of statutory interpretation in favor of the government in order to hold private parties liable, because Enliven the remedial nature of . . . iSuperfund] its provisions should be afforded a broad and liberal construction so as to aneroid frustration of prompt response efforts or so as to limit the liability of those responsible for clean-up costs beyond the limits expressly provided.2 If EPA performs the remedy with money from Superfund, the remedy is selected after a review of remedial alternatives. This pro- cess is subject to public comment. At other sites, EPA negotiates the remedy necessary for the site with the PRPs, such as in the S- Area I~ndfi~! case (see Case Study 5, this chapter). These negotiated remedies are then incorporated into a consent decree (a legal docu- ment that resolves a lawsuit without a determination of liability, but requires the defendant to perform an action, e.g., installation of tile drains and a cap, and/or monetary payment). Guidance Modeling may be used in the Superfund program to (1) guide the placement of monitoring wells (Environmental Protection Agency, 1988b); (2) predict concentrations in ground water for an assessment of the present and future risks at the site (Environmental Protection Agency, 19X6a, 198Sb); (3) assess the feasibility and efficacy of re- medial alternatives (Environmental Protection Agency, 1988b); (4) predict the concentration for an assessment of the residual risk after implementation of the preferred remedial action (Environmental Pro- tection Agency, 198Bb); or (5) apportion liability among responsible parties. Contaminant transport modeling is important in the process of estimating exposure and therefore risk. Regardless of the toxicity of the chemical, no injury can occur unless there is exposure. The chemicals must migrate from the source of contamination to a point where they come into contact with humans and interact biologically with the human body. Modeling can be "used as a too} . . . to estimate plume movement . . ." (Environmental Protection Agency,

166 GROUND WATER MODELS 1988b). Models are most helpful when rough estimates are required.3 A worst-case estimate (an estimate where all assumptions are chosen so as not to underestimate the possible exposure) may indicate that little risk exists if significant exposures are not predicted. However, [ads more resources are devoted to an exposure assessment and more studies conducted, a refined assessment is generated. Often there will be several stages of refinement of an assessment, and the degree of refinement and accuracy finally required will be related to the certainty needed to enable risk management decisions [e.g., selecting a ground water cleanup level versus evaluating the most cost-effective method of achieving that leveli.4 The EPA Superfund Public Health Evaluation Manual guidance (Superfund guidance) specifies that realistic exposure assumptions based on the best data available should be used.5 Superfund guidance requires EPA to consider systematically the extent of chemical fate and transport in each environmental medium in order to account for the behavior of all released chemicals (Environmental Protection Agency, 1986a, p. 39; see also 40 CFR §§300.68te]~], 300.68th]~2~tiv], 300.68ti]~1~. A ground water concentration, based on such mode! estimation, is then compared to levels of public health concern, e.g., a drinking water standard or a risk-based cleanup level (Zamuda, 1986~. EPA advises that "caution should be used when applying models at Superfund sites because there is uncertainty whenever subsurface movement is modeled, particularly when the results of the mode! are based on estimated parameters" (Environmental Pro- tection Agency, 1988b, p. 3-22~. Superfund guidance provides a general framework for selecting and applying models (Environmental Protection Agency, 1988b, p. 3-33; 1988c). Superfund modeling guidance recognizes the potential problem posed by the large range of models available and attempts to support users by providing guidelines for mode] choice (Environ- mental Protection Agency, 198Sc). These criteria allow users to more easily justify code choice during discussions with regulators and may provide some common ground for discussing the use of alternative codes. Three types of criteria are recommended for use in mode] se- lection: objective, technical, and implementation. The objective criteria used relate to the level of modeling detail needed to meet the objectives of the study, i.e., (1) performing a screening study or (2) performing a detailed study (Environmental Protection Agency, J

CONTAMINANT FLOW MODELS IN THE REGULATORY SYSTEM 167 1988c). Because the purpose of a screening study would be to ob- tain a general understanding of site conditions or to make general comparisons between sites, a simple mode} may be suitable at that stage. The technical criterion used for mode! selection relates to the model's ability to simulate site-specific transport and fate phenom- ena of interest at the site. There are three areas where technical cri- teria should be developed: transport and transformation processes, domain configuration, and fluid media properties. The third type of criterion used for mode} selection relates to the ability to implement the model. Issues that must be considered include the difficulty of obtaining the model, the level of documen- tation and testing associated with the model, and the ease of mode! use. The budget and the schedule for any project will affect the type of criterion used and ultimately the mode} selected (Environmental Protection Agency, 1988c). The 1988 guidance represents a significant advance in the EPA modeling program because it provides structure to the mode! selec- tion process and will avoid mixing discussions of mode} applicability and mode} results. Dividing these two processes could help simplify interactions between the regulators and the regulated community. Even this guidance represents only a small step toward simplifying the regulatory process. A number of codes used in EPA programs are described in the latter portion of the report. However, information on the level of complexity of these codes and the criteria for their application are not included. Additional clarification of EPA mode! use will be needed to help direct code selection in mode! applications that will be submitted to the agency. If problems arise, EPA personnel are directed to EPA's Center of Exposure Assessment Modeling and the International Groundwa- ter Modeling Center for specific advice (Environmental Protection Agency, 1988b, p. 3-33; 1988c). Ultimately, however, EPA personnel must rely upon their own skills. Resource Conservation and Recovery Act Law and Regulations There are tens of thousands of facilities that handle hazardous waste and therefore must obtain a permit. The Resource Compensa

168 GROUND WATER MODELS tion and Recovery Act (RCRA) establishes comprehensive, "cradIe- to-grave~ hazardous waste management programs. RCRA forbids waste treatment or disposal and limits waste storage for facilities not holding appropriate permits from EPA or a state agency (Section 3005 [a] of RCRA, 42 USC §6925 [a] ~ . The very foundation of any regulatory program is the definition of what is regulated versus what is not. The EPA definition of a hazardous waste determines "whether a waste, if mismanaged, has the potential to pose a significant hazard to human health or the environment due to its propensity to leach toxic compounds" (51 Fed. Reg. 21,653 [1986~6~. EPA has listed industrial waste streams as hazardous based on a lirn~ted sampling of a representative number of plants in the industry. Also, a waste is considered hazardousif it is ignitable, corrosive, or explosive or if the leachable concentra- tions of certain chemicals exceed regulatory health-based lignite (i.e., the extraction procedure [EP] test). A waste is hazardous based on this EP test if chemicals will leach out of the waste in quantities that may cause the ground water concentrations 500 It downgradi- ent to exceed drinking water standards after the waste is placed in a municipal landfill. EPA's original definition of hazardous waste assumed arbitrarily that the leachable concentration of a chemical would decrease by a factor of 100 in the 500 It (45 Fed. Reg. 33,084 t1980171. In 1986, EPA proposed to modify the EP test used to define haz- ardous waste by, among other things, (1) adding 38 organic chemical constituents, (2) substituting a more rigorous leaching test, (3) ap- plying compound-specific attenuation and dilution factors for each organic constituent to evaluate the worst-case potential impact on ground water 500 It downgradient of the location of deposal, and (4) using a risk-based concentration when no drinking water standard is available (51 Fed. Reg. 41,082 t1986~. EPA's proposed new defini- tion uses a subsurface fate and transport model, called EPASMOD (or the Composite Landfill Model), to derive compound-specific at- tenuation and dilution factors. EPASMOD considers the dilution, hydrolysis, and soil adsorption that occur as a chemical migrates from the bottom of a landfill to a drinking water source 500 It away (see 51 Fed. Reg. 1,602 [1986~9 for a more detailed discussion of the EPASMOD). The Environmental Protection Agency has revised EPASMOD and its input data and is considering additional revisions to EPAS- MOD and its input data so that the predicted concentrations would

CONTAMINANT FLOW MODELS IN THE REGULATORY SYSTEM 169 be less overpredictive (53 Fed. Reg. 28,892 t1988~°~. The proposal therefore would incorporate a contaminant transport into the defini- tion of hazardous waste. Contaminant transport models also have been used in other aspects of the RCRA program. For example, EPA uses the vertical- horizontal spread (VHS) mode! to determine when a listed hazardous waste from a particular facility would no longer be subject to RCRA hazardous waste requirements because the particular characteristics of the waste from that facility make the waste nonhazardous wherever it may be disposed (50 Fed. Reg. 48,886 t1985~; see Case Study 1, below). The RCRA regulations also require the permittee to perform ground water monitoring (40 CFR §264.97, 264.98, 264.99) and to clean up contaminated conditions at active facilities in any area where there was historic disposal of either hazardous or solid wastes (40 CFR §264.100~. The corrective action requirements, in essence, convert RCRA into a Superfund-type cleanup statute and expand RCRA's jurisdiction to cover all inactive waste disposal areas on operating facilities. The RCRA regulations require a permitter to clean up the ground water to (1) background levels, (2) the concentrations spec- ified by EPA for drinking water, or (3) a site-specific risk-based action level (the alternate concentration limit, or ACL) (40 CFR §264.94~. To evaluate the potential adverse effects on ground water quality, the permitter must provide information on, among other things, the wastes' potential for migration; the hydrogeological char- acteristics of the facility and surrounding land; the existing quality of ground water, including other sources of contamination and their cumulative impact on the ground water quality; and the potential for health risks caused by human exposure to waste constituents (40 CFR §264.94tb]~. The permittee must also submit an exposure and risk assess- ment. The two key concepts in the ACL process are that (1) the cleanup level must protect the public at the point of exposure (i.e., where ground water is withdrawn to use as drinking water), and (2) the point of compliance (i.e., the point where ground water is moni- tored) must be at the boundary of the regulated unit (Environmental Protection Agency, 1987a). It is necessary to set the ACL at a level (usually monitored at the boundary of the regulated unit) that, based on predictions, will result in ground water exposures that are below health protective

170 GROUND WATER MODELS levels at some distant point (e.g., the nearest drinking water well) and some future time. Guidance Contaminant transport modeling can be used for the same pur- poses as in the Superfund program. EPA's general exposure guidance concerning the use of models (above) is equally applicable here. The RCRA guidance encourages using conservative assumptions "where time and/or resources are limited" (Environmental Protec- tion Agency, 1986b, p. 150~. Numerical models are preferred over analytical models (p. 158~. EPA's RCRA guidance lists publicly available models (p. 158~. The Environmental Protection Agency's RCRA guidance is con- tradictory, however. EPA's RCRA alternative concentration limit guidance (Environmental Protection Agency, 1987a, p. 4-6) states that [although not required for an ACL demonstration, mathematical sim- ulation models of ground water flow and contaminant transport can be extremely useful tools for the applicant. Models are more appropriate for relatively simple geologic environments where conditions do not vary widely; in complex geologic areas, modeling may be less useful. The permit applicant is responsible for ensuring that the mod- els used simulate as precisely as possible the characteristics of the site and the contaminants and minimize the estimates and assump- tions required.... Whenever possible, input parameters and ae~umpt~ona should be conservatioc in nature; worst-case secr~ar~os may eauc much effort. [Emphasis added.] The RCRA ground water monitoring guidance, on the other hand, states "modeling results should not be unduly relied upon in guiding the placement of assessment monitoring wells or in designing cor- rective actions" (Environmental Protection Agency, 1986b, p. 156; emphasis added). Recently, EPA has considered standardizing the steps in the risk assessment/modeling process by "prescribing the types of models that can be used or the assumptions that are incorporated into models" (Environmental Protection Agency, 1987b). Among the standard models being considered is the VHS model. The standard mode! would guide the decision "based on only minimal site-specific data" (Environmental Protection Agency, 1987b). The use of a nationwide database would be contrary to EPA site-specific use on the selection of models.

CONTAMINANT FLOW MODELS IN THE REGULATORY SYSTEM 171 Underground Injection Control (UIC) Program Law and Regulations The Safe Drinking Water Act prohibits underground injection unless such injection is authorized by a perrrut or by rule (Section 1421, 42 USC §300h). The Underground Injection Control (UIC) regulations govern, among other wells, Class ~ wells, those wells used to dispose of hazardous waste below an underground source of drinking water. Class ~ wells are subject to regulations that specify minimum design construction and operating conditions and require continued monitoring of the nearby ground water to ensure that a present or future drinking water supply is not endangered (40 CFR Part 144). The Environmental Protection Agency's recent amendments to the Class ~ well regulations prohibit the injection of hazardous waste into wells unless (1) the waste is treated to the same extent required for hazardous waste disposed of on land, or (2) EPA grants an exemption from the regulation based on a "no migration" petition (53 Fed. Reg. 28,118 1988. The burden is on the permit applicant to prove that no migration will occur. A petitioner must demonstrate that migration outside the injec- tion zone will not occur for 10,000 yr (53 Fed. Reg. 28,155~. Nothing in the statute or its legislative history forbids the use of models or requires their use (53 Fed. Reg. 28,126~. These regulations require the person seeking a "no migration" exemption to submit "predictive models" that are "appropriate for the specific site, waste streams, and injection conditions of the operation, and shall be calibrated for existing sites where sufficient data are available" (53 Fed. Reg. 28,156~. The petitioner also must use "reasonably conservative val- ues," an approved "quality assurance and quality control plarl," and a "sensitivity analysis to deterrn~ne the effect that significant uncer- tainty may contribute to the demonstration" (53 Fed. Reg. 28,156~. The Environmental Protection Agency rejected the contention that one could not accurately mode! over a lO,OOO~yr period (53 Fed. Reg. 2S,126~. In determining the feasibility of a 10,000-yr goal, EPA concluded that (1) the "modeling need not locate the exact point whPr" t.h" wrist.. wilily he ~ -~_ was_ ~ a_ . . . [in 10,000 yr]; deterrn~ning where it would not be [i.e., outside the injection zone] is sufficient. This level of precision is achievable" (53 Fed. Reg. 2S,126~2~; (2) such fluid flow modeling was considered "a well-developed and mature science" that had been "used for many years in the petroleum industry" (53 Fed.

172 GROUND WATER MODELS Reg. 2B,127~; (3) such models had been developed by the DOE for use in the nuclear waste isolation program; and (4) this mode! and its application were peer reviewed by EPA's Science Advisory Board (a group of independent scientists who advise EPA in scientific issues). The Science Advisory Board (1984, cited in 52 Fed. Reg. 32,446) concluded that Modeling for the time periods involved . . . required extension of such . . . techniques well beyond usual extrapolation, however, the extension for 10,000 years can be made with reasonable confidence. From EPA's policy point of view, the precision of these predic- tions is not the only issue. The intent of the statute is to allow deep well injection of chemicals as long as they will not migrate outside the injection zone for a very long period of time. The use of models might be considered a failure if chemicals actually migrate outside the injection zone in 1 or even 30 yr. If a petition were granted, but chemicals migrated outside the injection zone in 8,000 yr instead of 10,000 yr, the overall purpose of the regulation would still be served. The mode! might be considered by many to be satisfactory. As a practical matter in this situation, the only choices other than using contaminant transport models would be to rely on the best professional judgment of a qualified hydrologist or provide for no exemptions to the ban on hazardous waste disposal. Because the statute provides for such exemptions, EPA has attempted to balance the scientific uncertainty. Guidance As of the time this report was written, this program had not yet developed guidance for this use of models. Conclusion The Environmental Protection Agency's guidance is contradic- tory. Some guidance provides a rational scientific framework for selecting models, and other guidance appears to favor use of stan- dardized worst-case models. SELECTED CASE STUDIES The five case studies in this section are presented as examples of how contaminant transport models are currently used as tools to (1)

CONTAMINANT FLOW MODELS IN THE REGULATORY SYSTEM 173 understand ground water systems, (2) predict contaminant migra- tion, and (3) illustrate how models are used by regulatory agencies in the decisionmaking process. The case studies were chosen to involve a large number of the hydrogeologic processes discussed in Chapters 3 and 4, and to demonstrate how this knowledge of hydrogeologic processes is used in actual problem-solving applications. A common theme throughout the case studies is that there is a lack of knowI- edge about system parameters. The case studies illustrate several methods that may be used to deal with this uncertainty. The selection of case studies is inherently subjective. The com- mittee decided that the selection process could not totally exclude controversial examples. By their nature, many of the best-docu- mented uses of models involve problems where there is a factual dis- pute that is longstanding or involves significant issues, e.g., millions of dollars in remedial costs or the right to use a scarce resource such as water. Selection of a particular case study should not be misconstrued as a judgment by the committee concerning whether the particular mode! was appropriately selected or applied. It is not possible to make such determinations without an extensive evaluation of the facts. Such a case-specific, detailed evaluation is beyond the scope of this report and not necessary to accomplish the committee's task. Therefore nothing in the report should be construed as a defini- tive scientific evaluation or endorsement of any particular mode! or modeling approach. The case studies cover a wide range of ground water problems (Table 5.1), involving sites scattered across the United States (Figure 5.1~. The first case study, the VHS model, discusses the use by EPA of a generic mode} (i.e., a mode! that does not require site-specific information) to determine which solid wastes should be treated as nazaroous wastes. The Madison aquifer case illustrates the use of a variety of ground water flow modeling approaches to predict water-level de- clines from large well withdrawals in an aquifer system in which very little is known about the hydraulic properties of the aquifer (Konikow, 1976~. Accurately predicting ground water flow condi- tions is an essential first step in simulating contaminant transport in ground water, and this case study illustrates particularly well tech- niques that can be used to assess the reliability of predicted ground water flow conditions.

174 TABLE 5.1 Synopsis of Case Studies Subject GROUND WATER MODELS Short Title VHS model A generic ground water transport model used by EPA to predict contaminant migration. The use of this model to delist wastes from the Gould, Inc., facility in McConnelsville, Ohio, is discussed. Madison aquifer Evaluation of large ground water withdrawals from an aquifer using a ground water flow model when aquifer parameters are poorly known. Snake River plain A ground water transport model used to predict migration of chloride, tritium, and strontium-90 in basalts. Original modeling study conducted in 1973 predicted concentrations in years 1980 and 2000. Subsequent study conducted in 1980 evaluated accuracy of original predictions. Tucson Airport The use of a ground water flow and transport model to assign liabilities for a multisource plume to specific sources. S-Area A one-dimensional, two-phase flow model used to evaluate the migration of nonaqueous-phase liquids at the S-Area landfill, Niagara Falls, New York. The Snake River plain case study discusses a simulation of chIo ride migration in ground water conducted in 1974 using a numerical ground water transport mode] and a subsequent field study con ducted in 1980 to check the mode! predictions (Lewis and Goldstein, 1982~. This case study is one of a few ground water contamination problems in which field data have been collected almost a decade ~` ~ Madison Aquifer Snake River Plain . ~ _ Tucson Airport FIGURE 5.1 Location of case studies. S-Area A\ Niagara Falls  <: P_Y ~ ':~. ~ =: Gould, Inc. McConnelsville, Ohio rat fit

CONTAMINANT FLOWMODELSIN THE REGULATORY SYSTEM 175 after the modeling was completed to compare the predictions to the observed concentrations. This study illustrates the error that can be expected with predictions of contaminant migration and the prob- lems that result when a three-dimensional ground] water system is modeled as a two-dimensional system. The Tucson Airport study discusses how a ground water mode! wan used to assign liabilities to individual parties for specific ground water contamination incidents.~3 Models are frequently being used for this purpose at Superfund sites (sites on the National Priority List) at which there are several responsible parties. The discussion that follows this case study highlights the conceptual problems that the committee foresees as a result of using the current generation of ground water transport models for this purpose. The S-Area case study examines the use of models to investigate the migration of a immiscible, denser-than-water fluid within an aquifer. The models discussed in this study were developed to help design an appropriate remedial action for the site. The models were presented in litigation involving this site and have been explicitly incorporated into a legally enforceable document as the method to be used in designing the remedial action for the site. To ensure consistent emphasis of particular points, a general format was adopted for the presentation of the case studies. The preparer of each case study was asked to treat in sequence, if pos- sible, the objective of the study, the major hydrogeologic processes considered in the study, a brief description of the mode! used, and the results and conclusions of the study. Each case study is followed by a committee discussion on the strengths and weaknesses of the study and the lessons that can be drawn from it for other studies of a similar nature. Vertical-Horizontal Spread (VHS) Mode] Bacirground As described above, EPA regulations allow a particular plant within the industry to demonstrate that its particular wastes are nonhazardous, i.e., to delist the particular wastes (40 CFR §§260.20, 260.22~. The primary quantitative criterion used to evaluate delisting petitions is whether, assuming worst-case conditions, the leachable chemicals from the solid waste would result in unacceptable ground water quality 500 ft downgradient of the disposal location. As a mat- ter of policy, EPA uses the VHS model (Domenico and Palciauskas, 1982) to determine the concentration of the leachable chemicals 500

176 GROUND WATER MODELS ft downgradient from the location of disposal for delisting purposes. The toxicity of the waste is evaluated by comparing the concentra- tions estimated by the VHS mocle! for a location 500 ft downgradient with EPA drinking water standards or other health-based standards (51 Fed. Reg. 21,666; also 53 Fed. Reg. 18,025 19884. If the concentration at the well is lower than the standard, the waste is considered nonhazardous and will be delisted. Mode! The mode! used in the delisting process considers three basic steps (Domenico and Palciauskas, 1982; 50 Fed. Reg. 48,886 [1985~; 50 Fed. Reg. 7882 t19854~5; 50 Fed. Reg. 41,082 t1986~6~: . generation of a {eachate from the waste; migration of the leachate to an underlying ground water aquifer; and migration of the contaminated ground water in the aquifer to a nearby drinking water well. . The concentrations of the chemical compounds of interest are generated by an appropriate leaching test, e.g., the extraction pro- cedure test (40 CFR §5261.24) or a leaching estimation method such as the organic leachate mode! (OLM) (51 Fed. Reg. 21,653 tI986~. The extraction procedure (EP) and the toxicity characteristic leach- ing procedure (TCEP) are laboratory procedures that are designed to simulate codisposal of the waste with municipal wastes in a sanitary landfill. The OLM is an empirical equation that calculates leachate concentration of a compound on the basis of the compound's solubil- ity and the concentration of the compound in the waste. The second step of the mode} estimates the attenuation that may occur during the migration of the leachate from the waste to the underlying aquifer. EPA assumes that no attenuation occurs because this is a reasonable worst-case characteristic of saturated soil systems, and because the water table is near the bottom of many waste sites. The third step of the modeling process calculates the dispersion of the chemical compound in a drinking water aquifer in the vertical and horizontal directions perpendicular to ground water flow as a result of a continuous source of contamination. The VHS mode! is used to simulate the dispersion of the contaminants and calculate the contaminant concentration at a reception well directly downgradient

CONTAMINANT FLOW MODELS IN THE REGULATORY SYSTEM 177 Disposal Area ~40' ~ ~500' Receptor Well Water Table · v .---..-:-:-:: oT= 2m oz = 0.2 m T ~. : Contami~ ~ FIGURE 5.2 Schematic of vertical-horizontal spread model. of the waste disposal area (Figure 5.2~. The following equation describes the VHS model: Cy = COerf(Z/~2~<xxY)05~)erf(X/~4~aTY)O i), where Cy = contaminant concentration at the receptor well (magi) CO = contaminant concentration in the leachate (high) erf = error function, dimensionless Z = penetration depth of leachate into the aquifer (m) Y = distance from disposal site to compliance point (m) X = length of the disposal site measured in the direction per- pendicular to the direction of ground water flow (m) 0rT = lateral transverse (horizontal) dispersion length (m) ax = vertical dispersion length (m) This equation has three basic terms: (1) the initial concentration of the contaminant in the leachate, (2) a term for the spreading of the concentration in the vertical direction, and (3) a term for the spreading of the contaminant in the horizontal dimension. Vertical and horizontal spreading are the only processes that cause the con- taminant concentration to decrease away from the source. Other processes, such as chemical reactions, precipitation, and biodegrada- tion, that might decrease contaminant concentrations as they migrate away from the source are not considered in the VHS model. The VHS mode! is a steady-state model, and the calculated receptor well concentration is a steady-state concentration. The mode! does not calculate the time required to reach the steady-state concentrations. If the contaminant is strongly sorbed to aquifer

178 GROUND WATER MODELS materials, such as is the case for polychIorinated biphenyIs (PCBs) and dioxins, the time required to reach steady state will be thousands of years. Use of the VHS mode! requires the specification of only six parameters: initial leachate concentration (CO), distance to receptor well (Y), length of disposal site perpendicular to direction of ground water flow (X), horizontal transverse Aspersion length (aT), vertical dispersion length (cat), and the mixing zone depth (Z). EPA uses reasonable worst-case values for these parameters, except for leachate concentration and the length of the disposal site, which vary with each delisting petition. The reasonable worst-case parameter values used by EPA are as follows: . Distance to well: EPA uses 500 It for the distance from the waste disposal facility to the drinking water well. This distance is based on an informal survey that suggested that at 75 percent of the landfi~Is, the closest well is further than 500 It (152.4 m) from the facility. . Dispersion lengths: A value of 6.5 It (2 m) is specified for the horizontal transverse dispersion length, and 0.65 It (0.2 m) is specified for the vertical dispersion length. . Mixing zone depth: A mixing zone depth of 10 It (3.28 m) is used by EPA. The mixing zone depth is related to the width of the disposal area and the ratio of leachate generation to the velocity of ground water. EPA assumed that the average disposal area width was 40 It (12.2 m) and that the ratio was 0.25. . Length of disposal area: The length of the disposal area is cal- culated by using the waste volume specified in the delisting petition and assuming that the waste is placed in a trench 40 It (12.2 m) wide and 8 It (2.4 m) deep, where the long axis is oriented perpendicular to the direction of ground water flow. A minimum of length of 40 It is used if waste volume is small. When the reasonable worst-case parameter values are substi- tuted into the VHS equation, a relatively simple equation relating a dilution factor (CO/Cy) to the waste volume results. A graph of the dilution factor versus waste volume is shown in Figure 5.3. This graph shows that a solution factor of 32 is calculated from the VHS mode} for small waste volumes and that the calculated dilution factor decreases to about 7 for large waste volumes.

CONTAMINANT FLOW MODELS IN THE REGULATORY SYSTEM 179 30 25 A: o ~ 20 11 z o - 15 10 5 _ - 1 O 0 1200 2400 3600 4800 6000 WASTE AMOUNT (yd3) FIGURE 5.3 Predicted dilution factor as a function of waste amount. Application at a Particular Site Gould, Inc. operates an electroplating facility in McConnelsville, Ohio, that annually generates 1,100 y63 of sludge from its wastewater treatment plant. This sludge ~ classified as a hazardous waste in 40 CFR §260.22, and Gould, Inc. petitioned to delist the waste based on the destruction and immobilization of hazardous compounds by its wastewater treatment system. The constituents of concern in the sludge are cadmium, chro- mium, and nickel. Gould, Inc. submitted leaching test data for these compounds, and the VHS mode} was used to determine concentra- tions at the receptor well using a dilution factor of 13.5, which was calculated from the waste volume, and the concentrations of these compounds were set to health-based standards (see Table 5.2~. The calculated concentrations of cadmium, chromium, and nickel at the receptor wed are all below the health-based standard. EPA used this evaluation to delist this waste (50 Fed. Reg. 4S,887 tI985~. TABLE 5.2 RCRA Delisting Data for Gould, Inc., Facility, McConnelsville, Ohio  Calculated Leachate Receptor Well Health-Based Chemical Concentration Concentration Standard Compound (mg/l) (mg/l) (mg/l) . Cadmium <0.1 <0.007 0.01 Chromium <0.5 <0.037 0.05 Nickel 2.5 0.185 0.35

180 GROUND WATER MODELS Regulatory Context The Environmental Protection Agency adopted the use of the VHS mode} in the RCRA delisting program an a matter of po~icy'i7 but claimed that it was not bound to use the model (50 Fed. Reg. 7,882 (1985~. EPA Han "treated the mode} as conclusively disposing of certain moues.... The mode} thus created a norm with 'present- day binding effect' on the rights of . . . ithe companies seeking to have a waste delisted]" (838 F.2d at 1321~. In practice, EPA "evidenced almost no readiness to reexamine the basic propositions that make up the VHS mode! . . . " (838 F.2d at 1321~. As a result, a unanimous pane! of the Court of Appeals ruled that EPA violated the Administrative Procedure Act (5 USC 7=1 706~. EPA is now obligated, in reality, to exercise discretion in individual delisting cases or to issue the VHS as a binding regulation after notice and public comment and an opportunity to challenge the regulation (838 F.2d at 1324~. EPA, however, continues to use the VHS mode! in the RCRA delisting program purportedly now as a truly nonbinding policy (53 Fed. Reg. 21,640 [1988~. Discussion The VHS mode} is a simple generic model. The use of the mode} requires no site-specific data, and therefore it may appear unscientific. EPA (53 Fed. Reg. 7,906 [1988~9) openly acknowledges that the VHS model is more likely to overpredict (rather than underpredict) the receptor concentration of contaminants in any given waste due to the conservative nature of the assumptions underlying the model. EPA also recognizes that all models do not always predict factual values accurately. The Environmental Protection Agency's position (53 Fed. Reg. 7,906 [1988~) is that [uinlese the Agency is able to assure protection of human health and the environment without generic, conservative assumptions, the Agency will employ these assumptions. The Environmental Protection Agency is particularly concerned that once a waste is delisted, there are no restrictions on how or where the waste will be disposed. The model, however, could be im- proved by the addition of chemical-specific terms for biodegradation, precipitation, and other reactions that would cause the concentration of the contaminant to decrease as it migrates from the source area to

CONTAMINANT FLOW MODELS IN THE REGULATORY SYSTEM 181 the receptor well. Health protective values could be used instead of being ignored altogether. As described above, EPA has proposed using the EPASMOD to define hazardous wastes. Although EPA rejected the use of EPAS- MOD for delisting petitions in 1986, it also indicated that it might reconsider use of that mode! once the new test for hazardous wastes was completed (51 Fed. Reg. 41,501 [1986~2°~. More recently, EPA has considered the possible use of EPASMOD in the delisting process (53 Fed. Reg. 28,892 tI988~2~. This mode! would take into account at least some of the factors that decrease concentration in the real world. EPA would still not use site-specific data in the model. This case study illustrates how easily a nonbinding guidance can in reality become an inflexible rule. Madison Aquifer WeB Withdrawals from a Deep Regional Aquifer Background The Powder River basin of northeastern Wyoming and south- eastern Montana contains large coal reserves that have not yet been fully developed. The future development of these energy resources will be accompanied by increased demands for water, which is not abundantly available in this semiarid area. One plan had been for- mulated to construct a coal slurry pipeline to transport coal out of the area; it would have required approximately 15,000 to 20,000 acre-ft/yr of water (20 to 28 ft3/s). In the mid-1970s a plan was proposed to supply water for the coal slurry pipeline by withdrawals from up to 40 deep wells that would be drilled about 3,000 It into the Mississippian Madison limestone in Niobrara County, Wyoming. The Madison aquifer is an areally extensive Paleozoic carbonate roe system that underlies an area exceeding 100,000 mi2 in the Northern Great Plains. Wyorn~ng authorized the withdrawals, but the state of South Dakota was concerned about the cross-boundary effects of the drawdown. Initial Modeling Studies Large ground water withdrawals may cause significant water- leve! declines in the Madison aquifer, perhaps extending into adjacent states, as well as decreases in streamflow and spring discharge in or near the outcrop areas. Thus an ability to predict the effects of

182 GROUND WATER MODELS the proposed ground water withdrawals on potentiometric levels, recharge, and discharge is needed. The Madison aquifer lies at great depths (between 1,000 and 15,000 It) in most of the area and is therefore relatively undeveloped. There are insufficient data available to accurately and precisely define the head distribution and the hydraulic properties of the aquifer. In light of this uncertainty, and as a prelude to a planned subsequent 5-yr hydrogeologic investigation of the Madison aquifer, Konikow (1976) developed a preliminary digital mode} of the aquifer using the two-dimensional finite-difference mode! of Trescott et al. (1976~. The objectives of the preliminary mode! study were to (1) improve the conceptual mode} of ground water flow in the aquifer system; (2) determine deficiencies in existing data and help set priorities for future data collection by identifying the most sensitive parameters, assuming the mode] is accepted as being appropriate; and (3) make a preliminary estimate of the regional hydrologic impacts of the proposed well field. Initial Results The results indicated that the aquifer can probably sustain in- creased ground water withdrawals up to several tens of cubic feet per second, but that these withdrawals probably would significantly lower the potentiometric surface in the Madison aquifer in a large part of the basin. The mode! study and predictions were framed in terms of a sensitivity analysis because of the great uncertainty in most of the parameters. For example, Figure 5.4 shows drawdown predictions made for an area near the proposed well field for an as- sumed reasonable range of values for the storage coefficient (S) and leakage coefficient (Kxm), where Kx and m are the vertical hydraulic conductivity and the thickness, respectively, of the confining layer. The curves show that the range in plausible drawdowns, even after yr, is extremely large. This uncertainty in the nature and magnitude of potential ver- tical leakage was also translated into a disparity of interpretations and opinions in other independent forecasts of these impacts. In the report of the 1975 hearings before the U.S. Congress on pending legislation pertaining to the coal slurry pipeline, a report by the con- sultants to the pipeline company concludes that ". . . the 'leakage' or contribution from beds adjacent to the porous zones in the Madison is sufficient to preclude drawdown at distances more than about 2000

CONTAMINANT FLOW MODELS IN THE REGULATORY SYSTEM 183 o 50 - 100 O 150 600 -Sl 200 C) 300 4t)0 500 0.00001 Kz/m \ '10-12 _= B D E _ .00025 C 700 1 1 1 0.001 0.0 10 13 A _ .00005 o _ \\W 10-1 1 F . \ ~C\ A\ 0.01 0.1 1 10 100 200 TIME (yr) FIGURE 5.4 Time-drawdown curves for model node near hypothetical pump- ing wells in the Madison limestone (modified from Konikow, 1976~. feet." On the other hand, the same report of the 1975 congressional hearings includes a disparate forecast by another expert, based on assumption of nonIeaky conditions, which shows drawdowns after 45 yr exceeding 1,000 It near the pumping center and greater than 200 It at distances, more than 50 my away.  Other Mode} Studies This preliminary mode} analysis helped in formulating an im- proved conceptual mode! of the Madison aquifer. For example, the important influences of temperature differences and aquifer discon- tinuities on ground water flow in the Madison were recognized and documented as a result of the mode! analysis. It could be argued that the importance of these influences could have been (or should have been) recognized on the basis of hydrogeologic principles without the use of a simulation model. However, none of the earlier published studies of this aquifer system indicated that these factors were of major significance. The difference from earlier studies arose from the quantitative hypothesis-testing role of the model; the nature of the inconsistencies between observed head distributions and those calculated using the initial estimates of mode} parameters helped direct the investigators toward testing hypotheses that would resolve or minimize the inconsistencies. Also in this case, the demonstrated high sensitivity to the leakage coefficient highlighted the need to

184 GROUND WATER MODELS 2000 500 400 300 200 100 - - - - - - - - - \ - - - 2 5 10 20 30 40 50 60 70 80 90 95 98 PERCENTAGE GREATER THAN FIGURE 5.5 Calculated probability distribution of drawdowns at the Niobrara well field. reevaluate the system in a true three-dimensional framework so as to better consider vertical components of flow. The effects of the pumping for the coal slurry pipeline were re- examined by Downey and Weiss (1980), and Woodward-Clyde Con- sultants (1981), with a three-dimensional mode! that incorporated the processes found to be important in the initial study. The latter study, which was prepared for an environmental impact statement for the project, used a five-layer mode! and a Monte CarIo simula- tion approach to incorporate and assess the effects of uncertainties  in the parameters. The predicted impacts were then presented as probability distribution curves showing the likelihood of different drawdowns occurring at the specified points (Figure 5.5~. The recog- nized uncertainty in the predictions (i.e., the wide range in predicted drawdowns) was a factor contributing to the fact that the coal slurry pipeline was never built.22 The controversy surrounding the effects of the proposed drawdown in South Dakota was a major factor in the pipeline company's decision to buy Missouri River water from the state of South Dakota to supply the pipeline. Because of falling coal prices and railroad opposition, the project was abandoned in 1984.

CONTAMINANT FLO W MODELS IN THE REGULATORY SYSTEM 185 Regulatory Context The original models developed of the Madison aquifer were used in testimony presented at hearings before a congressional subcommit- tee to support the viewpoints of the proponents and the opponents of the coal slurry pipeline. Because field data on aquifer properties were sparse, a wide range of parameter values was probable. The opponents of the project chose parameter values that resulted in the prediction of widespread impacts, and the proponents chose param- eter values that resulted in the prediction of minimal impacts. The hydrogeologic parameters used in both models were not inconsistent with the available data, but parameter values were clearly chosen to bias the predictions. These modeling approaches illustrate the dis- parity of results that can be predicted using models when a rigorous approach is not utilized to analyze parameter uncertainty. The later mode! developed for the environmental impact state- ment, which was prepared for the Bureau of Land Management, was prepared under fairly rigid guidelines that were designed to produce an objective analysis of the potential environmental impacts of the proposed project. This modeling stiffly is clearly a more objective analysis of the probable impacts of the proposed ground water with- drawals. Discussion Mode} analyses and predictions can lead to an improved under- stancling of an aquifer system and serve as an aid to making decisions or formulating policy. However, the predictions must be clearly pre- sented, together with a realistic assessment of the confidence in them. This case study demonstrates that models can be very useful tools for gaining an understanding of aquifer systems in which little is known about the hydraulic properties of the system and that these models can be invaluable for prioritizing field data collection activities so that a maximum amount of information can be obtained for a given expenditure. Accurately predicting ground water flow conditions is an essen- tial first step in simulating contarrunant transport in ground water, and this case study illustrates, particularly well, techniques that can be used to assess the reliability of predicted ground water flow con- ditions. Because most of the ground water contamination problems of interest to regulators are dominated by convective transport, the uncertainty associated with predictions of ground water transport at

186 GROUND WATER MODELS these sites can be assessed using techniques similar to those used in this case study. Snake Ridered Plain Point Source of Contamination Background The Idaho National Engineering Laboratory (INEL) is located on 890 mi2 of semiarid land in the eastern Snake River plain of south- east Idaho. The facility, formerly called the National Reactor Testing Station, is now operated by the Department of Energy for testing various types of nuclear reactors. Robertson (1974) reports that sev- eral facilities at the site generate and discharge low-level radioactive and dilute chemical liquid wastes to the subsurface through seepage ponds and disposal wells. The two most significant waste discharge facilities, the Test Reactor Area (TRA) and the Idaho Chemical Processing Plant (ICPP), have discharged wastes continuously since 1952. The purpose of this study was to predict the future migration of wastes containing chloride, tritium, and strontium-90. Unlike the other case studies discussed in this section, the modeling analyses of the INEL were not conducted to satisfy the requirements of a regula- tory agency. This discussion of the INEL site and associated mode} is largely extracted and paraphrased from the reports of Robert- son (1974) and Lewis and Goldstein (1982), to which the reader is referred for additional details. Hydrogeologic Setting The eastern Snake River plain is a large structural and topo- graphic basin about 200 mi long and 50 to 70 mi wide. It is un- derIain by 2,000 to 10,000 It of thin basaltic lava flows, rhyolite deposits, and interpolated alluvial and lacustrine sediments. These formations contain a vast amount of ground water and make up the major aquifer in Idaho, which is known as the Snake River plain aquifer. Ground water flow is generally to the southwest at relatively high velocities (5 to 20 ft/day). The principal water bearing zones occur in the basalts, the permeability fabric of which is highly het- erogeneous, anisotropic, and complicated by secondary permeability features, such as fractures, cavities, and lava tubes.

CONTAMINANT FLOW MODELS IN THE REGULATORY SYSTEM 187 Mode} Formulation Concern over ground water contamination from the waste dis- charge prompted Robertson (1974) to develop a digital solute trans- port mode} to simulate the underlying aquifer system. The mode! (that is, the numerical method used to solve the solute-transport equation) was based on the method of characteristics. Robertson first calibrated a flow mode} for a 2,600 mi2 area and then calibrated the transport mode} for a smaller part of that area in which contam- ination was of concern. The calibration of the transport mode! was based on a 20-yr history of contamination, documented by samples from approximately 45 wells near and downgradient from the known point sources of contamination. These data showed that chloride and tritium had spread over a lS-mi2 area and migrated as far as 5 mi downgradient from discharge points. The distribution of waste chIo- ride observed in November 1972 is shown in Figure 5.6. Robertson notes that the degree of observed lateral dispersion in the plumes is particularly large. Results and Conclusion Robertson used the calibrated transport mode} to predict fu- ture concentrations of chloride, tritium, and strontium-90 for the years 1980 and 2000 under a variety of possible future stresses. For the chlorides, assumptions included were that (~) disposal continues at 1973 rates and (2) the Big Lost River recharges the aquifer in odd-numbered years. This scenario came closest to what actually occurred. The projections indicated that by 1980 the leading edges of both the chloride (see Figure 5.7) and the tritium plumes would be at or near the INEL boundary. Lewis and Goldstein (1982) report that eight wells were drilled during the summer of 1980 near the southern boundary to help fill data gaps and to monitor contaminants in ground water flowing across the INEL boundary. They also used the data from the eight wells to help evaluate the accuracy of Robertson's predictive model. The distribution of waste chloride observed In October 1980 is shown in Figure 5.8. A comparison of Figure 5.8 with Figure 5.6 indicates that the leading edge of the chloride plume had advanced 2.5 to 3 mi during that ~yr period ant! that the highest concentrations increased from 85 mg/! to around 100 my/. A comparison of Figures 5.7 and 5.8 indicates that although the observed and predicted plumes show general agreement in the

188 ' - ~,': _ 43030, GROUND WATER MODELS - 113° Disposal / Ponds Am</ RWMC /2EBR~ \ \ ~-~ ~^ \J EXPlANATION Improved roads  · Disposal well 20- Line of equal chloride concentration in milligrams per liter: interval varies. o 0 1 2 KILOMETERS FIGURE 5.6 Observed distribution of waste chloride in ground water in the Snake River plain aquifer, Idaho, ICPP-TRA vicinity in 1972. SOURCE: Robertson, 1974. direction, extent, and magnitude of contarn~nation, some apparently significant differences in detail exist. The observed plume is broader and exhibits more lateral spreading than was predicted and has not spread as far south and as close to the INEL boundary as predicted. Also, the predicted secondary plume north of the Big Lost River, emanating from the Test Reactor Area, was essentially not detected in the field at that time. Lewis and Goldstein (1982) presented a number of factors that they believed contributed to the discrepancy between predicted and observed results. These reasons can be summarized as follows: (1) there was less dilution from recharge during 1977 to 1980 because

CONTAMINANT FLOW MODELS IN THE REGULATORY SYSTEM 189 Disposal / Ponds TRA _ ,~ _ 40030' ~ /ICPP :<,~/~11: ~67 40 RWMC _/ /: ,1 - - r CFA - 20 EXPLANATION Improved roads · Disposal well 0 1 2 MILES  -20- Line of equal chloride concentration, t~-T ~ in milligrams per liter: 0 1 2 KILOMETERS interval varies. FIGURE 5.7 Model-projected distribution of waste chloride in the Snake River plain aquifer for 1980 (ICPP-TRA vicinity), assuming disposal continues at 1973 rates and the Big Lost River recharges the aquifer in odd-numbered years. SOURCE: Robertson, 1974. Of below-normal river flow; (2) chloride disposal rates at the ICPP facility were increased during the several years preceding 1980; (3) the mode} grid may have been too coarse; (4) the mode! calibra- tion selected inaccurate hydraulic and transport parameters; (5) vertical components of the flow and transport may be significant in the aquifer but cannot be evaluated with the two-dimensional areal model; (6) there may be too few wells to accurately map the actual plumes, and some existing wells may not be constructed properly to yield representative measurements; and (7) the numerical method introduces some errors (however, Grove's tI977] analysis of this same

190 GROUND WATER MODELS <0-28 \~ ~- _ 40030, EXPLANATION Improved roads · Disposal well Disposal / Ponds Ah/{ TRA 1~ ~ 20~\ / ICPP 6\9 J RWMC A/ -20- Line of equal chloride concentration. in milligrams per liter: interval varies. 7~  0 1 2 MILES I L I 0 1 2 KILOMETERS FIGURE 5.8 Distribution of waste chloride in the Snake River plain aquifer (ICPP-TRA vicinity), October 1980. SOURCE: Lewis and Goldstein, 1982. system used finite-difference and finite-element methods, and com- parisons of numerical results offer no basis for concluding that the numerical solution algorithm used by Robertson was in itself a sig- nificant source of the predictive errors). Although these factors can be expanded upon, and other factors added, it is extremely difficult to assess the contribution of any single factor to the total error. Re- calibration of the earlier mode! using the now extended historical record could be employed to test some of these hypotheses. Other factors can only be tested if new models are developed that incorpo- rate additional or more complex concepts, such as density differences

CONTAMINANT FLOW MODELS IN THE REGULATORY SYSTEM 191 and three-dimensional flow. Such a recalibration and mode] revi- sion should lead to a mode} that has greater predictive power and reliability. Discussion Whether the errors in this case were significant in relation to the overall problem can be answered best (or perhaps, only) by those who sponsored the mode! study in light of (1) what they expected, (2) what actions were taken or not taken because of these predic- tions, and (3) what predictive alternatives were available. The mode! predictions represented only one hypothesis of future contaminant spreading. The 1980 test drilling was designed, to alarge extent, to test that very prediction. The process of collecting data is most efficient when guided by an objective of hypothesis testing. A major value of the mode! so far has been to help optimize the data collection ant} monitoring process; that is, the predictive mode! offers a means to help decide how frequently and where water samples should be collected to track the plume. Thus, modeling and data collection are an iterative process. Tucson Airport Background Ground water within a zone approximately 6 my long and ~ mi wide in the vicinity of the Tucson Airport is contaminated with organic solvents, primarily trichIoroethene (Figure 5.9) (40 CFR §261.24~. The contaminated ground water is in an extensive alluvial aquifer that Tucson uses as its principal aquifer. The Tucson area, with a population of 517,000, is one of the largest metropolitan areas in the country that is totally dependent on ground water for drinking water, and the trichIoroethene contamination was viewed as a threat to the integrity of the water supply system. The area containing the contaminated ground water is listed on the National Priority List and is known as the Tucson Airport Area Superfund Sites (51 Fed. Reg. 21,054 1986. Several potential sources of the ground water contamination were identified in a remedial investigation conducted at the site (Rampe, 1985~. The available data indicated that several industrial facili- ties had used trichIoroethene in their processes, and that industrial wastewaters had been disposed of in ponds and drainage ditches and

192 EXPLANATION t It GROUND WAI'ER MODELS  ~ \ \ Ircinolon ~Ron \ \ in. 10 ~ d | \  \ \ - AL E ~ C IA \ \ ~  ( rl  to. _ t - at, I I ~ ~ r ~ Approximate limits of TCE contamination during 1984 (dashed where unknown or inferred) Contour of TCE contamination, in pub ADA. art, 1 On ~- . ~) :x'\. _ BOUNDARY OF - U\~  _~W - ~/ _ AIR FORCE LAND N. I Scale miles FIGURE 5.9 Location of Tucson Airport. SOURCE: Adapted from CH2M Hill, 1987.

CONTAMINANT FLOW MODELS IN THE REGULATORY SYSTEM 193 on the ground. There were no data, however, to indicate how much trichIoroethene had been lost to the subsurface at any individual facility. The remedial investigation made some general conclusions re- garding the significance of individual source areas, but these conclu- sions were questioned by the potentially responsible parties (Arizona Department of Health Services, 1986~. As a result, EPA, Region ~X, asked CH2M Hill, an environmental consulting firm, to con- duct an assessment of potential sources. Specifically, CH2M Hill was requested to do the following: . Assess the possibility of contribution to the ground water contamination from the various potential sources that had been iden- tified. Assess the ranges of relative contributions for each potential source and the probability distribution associated with the range. This discussion of contamination in the Tucson Airport area is largely extracted and paraphrased from the draft report by CH2M Hill (1987) and the remedial investigation prepared for the Arizona De- partment of Health Services by Schmidt (1985) and Mock et al. (1985~. Hydrogeologic Setting The Tucson Airport is located within the upper Santa Cruz basin, an alluvial basin bordered by north to northwest trending fault block mountains (Fenneman, 1931~. Basin fill deposits, pre- dominantly sands, sandy gravels, and clayey sands, make up the aquifer system in the vicinity of the Tucson Airport. Three distinct aquifer units are identified in the area: an upper coarse-grained unit that extends to a depth of about 200 ft. a middle fine-grained unit that is about 100 It thick, and a lower unit with lenses of coarse- grained materials whose total thickness is unknown. Ground water contamination is generally confined to the upper unit, which consists mainly of clayey sands interbedded with lenticular deposits of sand and sandy gravels. Water levels in the upper unit are currently about 100 It be- {ow land surface, and the water-level gradient is toward the north- northwest. Water levels in the upper aquifer fell about 30 It from 1952 to 1981, but they have been relatively stable since 1981, possi- bly as a result of reductions in pumpage by the city of Tucson in this

194 GROUND WATER MODELS area (Mock et al., 1985~. Ground water levels in the lower unit differ from those in the upper unit by 60 to 100 ft (CH2M Hill, 1987~. Approach The approach used to address the objectives of this study consists of two aspects: (~) establishing a basis for making assessments of the relative contributions to ground water contamination from multiple sources and (2) developing estimates of the contributions that each source had made to areas of ground water contarn~nation. Models were used to make each of these assessments, and these models are described below. Mode} to Assess Relative Contributions The contribution of a source to contamination of an aquifer can be assessed either by evaluating the quantity of contaminants con- tributed to the aquifer from a source or by evaluating the area of the aquifer affected by the release of contaminants. This study started with the assumption that, for purposes of assigning responsibility for cleanup of an aquifer, it is appropriate to assess relative contributions from sources based on areas of contamination. The reasons given for this assumption were the following: the extraction system required to contain or withdraw con- taminated ground water is not dependent on the levels of contam- ination but, rather, is nearly directly proportional to the area of contamination; ~ treatment costs are influenced more by volume of water treated than by actual levels of contamination; and ~ the quantity of contaminants released cannot be reliably es- timated when only low levels of contamination are observed. The relative contribution of a source area was assessed with the following equation: where ( 1 m A ) RCa= percent relative contribution of source a to the area of contaminated aquifer

CONTAMINANT FLOW MODELS IN THE REGULATORY SYSTEM 195 Ai = area of aquifer where source a had contributed contami- nants, a subarea of the contaminated aquifer Li = number of sources that contributed contaminants to area A. m = number of subareas contaminated by source a AT = total area contaminated The use of this mode! to assess relative contributions is illus- trated in Figure 5.10. In the example given, three sources, a, b, and c, contributed contamination. Source a contributed to the entire area, source b contributed to one-fourth of the area, and source c contributed to one-half of the area. The relative contributions to the contamination of the entire area for sources a, b, and c are 7t, 8, and 21 percent, respectively. The relative contributions calculated for each source take into consideration those sources that affect an area that has also been affected by other sources. This procedure provides a method for considering overlapping contributions from multiple sources. Mode} of Source Contributions The area of contamination resulting from a contaminant release from an individual source could not be determined with the informa- tion available on the distribution of contaminants. Rather, the area of contamination from a source was estimated using a two-dimensional numerical contaminant transport model. The flow transport mode} consisted of two separate steps: first, a finite-difference ground water flow mode} was developed and calibrated; then, a transport mode! based on the method of characterizations (Konikow and Bredehoeft, 1978) was used to estimate contaminant spread in the aquifer. In developing the flow model, CH2M Hill first divided the aquifer, on the basis of aquifer test data, into seven zones of equal permeability. Then an automatic parameter estimation technique, based on the method described by Cooley (1977), was used to refine the perme- ability estimates in each zone. The refined estimates were those that minimized the sum of the squared difference between the simulated levels and the observed 1984 water levels. Once a steady-state flow mode! was calibrated to simulate 1984 conditions, stream lines were calculated from each source area. Be- cause there was good agreement between the distribution of con- tamination and the pattern of stream lines, it was concluded that a steady-state flow mode} is a reasonable representation of the ground

96 INPUT: \ Source a contributed to subareas A I, A2, A3 Source b contributed to subarea A! only Source c contributed to subareas A! and A2 AT = 1 = area of affected aquifer under consideration GROUND WATER MODELS - / Al A2 -  A3 / RC (A Ai ) i0O {A1 A2 A3 ~ 100 (3 ~ 2 + 1) AT {0.25 0.25 0.5\ 100 ~ ~ 3 2 1 ) AT = 0.71- AT = 71% RCb = 8% RCC = 21% FIGURE 5.10 Example calculation of relative contribution to aquifer contam- ination. water flow field and therefore that source releases could be simulated by using the steady-state flow field and transient transport. Prior to simulating mass transport, it was necessary to estimate the quantity of trichIoroethene released from each source and the timing of the releases. The timing of the releases was estimated on the basis of historical records of trichIoroethene usage, and the quantity  released was estimated on the bash of the quantity of trichIoroethene in the aquifer. The effects of potential source releases were simulated

CONTAMINANT FLO W MODELS IN THE REGULATORY SYSTEM 197 TABLE 5.3 Calculated Estimates of Trichloroethene Releases from Source Areas Total Source Trichloroethene Time of Area Release (gal.) Release A 400 1952-1977 B 155 1952-1984 C 155 1952-1984 D 55 1964-1984 E 130 1952-1984 for each source acting alone, and in addition, the combined ejects of the potential sources were evaluated for various release and timing scenarios. A trial and error procedure was used to estimate the most probable release scenario. Results and Conclusions The results indicated the potential for a large area of contami- nation to develop from rather small amounts of trichioroethene, and the release scenario described in Table 5.3 was judged to be most representative of conditions observed in the field. The individual plumes, calculated by simulating each source in- dependently, are shown in Figure 5.~la, and the combined plume is shown in Figure 5.~Ib. On the basis of these simulated plumes, the relative contribution of each source to the total contamination north and south of Los Reales Road was calculated using the rela- tive contribution equation discussed above. The calculated relative contributions are shown in Table 5.4. CH2M Hill noted that the mode! simulations provided results for only specific cases: that is, the best estimate of permeability in each zone and one contaminant release scenario. CH2M Hill stated that if multiple simulations were performed to take into account the plausible variations in permeability and trichIoroethene releases, a range of relative contributions could be developed for each source in a quantitative manner. They noted, however, that the estimates of uncertainty about the release would be subjective owing to lack of data. Therefore only a qualitative assessment was made of the range of relative contributions.

198 I.: i -;-: / ./ / l l l I I IA' .b,, i _ __ ,, 7 . , . , 1 A/ _ . ~ :'.1 ,~  Q ° ° o E: o ~ ~ 1 ,[ ,( ,' i ~-- ~, I l __ ~ ~ so 0 .l .Q .g 0 to a, ~ 5 tt ~ ~ ~ ~ ~ ~ 4) o o o o o o ~ ~ /~. o .Q al _ - - E C o X ~ ID o} _ Q Y E, < L) ~ O it < x

CONTAMINANT FLOW MODELS IN THE REGULATORY SYSTEM 199 TABLE 5.4 Calculated Relative Contributions from Individual Source Areas Source Relative Contribution (percent) South of Los Reales Road A B C D E North of Los Reales Road A B C D E Regulatory Context 74 4 3 19 o 33 14 20 11 22 Ground water contamination in the Tucson Airport area was dis- covered in the early 1950s. Intensive investigations of ground water contamination did not begin until 1979? when a sampling program was initiated at the request of EPA. In March lg81 an investiga- tion conducted by EPA under the authority of the Comprehensive Environmental Response, Compensation, and Liability Act (CER- CLA) identified trichIoroethene and chromium contamination in the ground water. As a result, seven municipal wells were removed from service, and the site was listed on the National Priority List. In November 1982, Hughes Aircraft Company and the U.S. Air Force assumed responsibility for investigating contamination south of Los Reales Road, while EPA assumed responsibility for investigating contamination north of TJoS Reales Road (Figures 5.11a and 5.11b). Hughes and the Air Force conclucled that the contamination beneath the Air Force property was caused by past disposal of waste sol-  vents and claimed that no continuing sources existed on the facility. Consequently, the Installation and Restoration Program (IRP) con- ducted by the Air Force in 1985 did not contain any proposed source control measures. The Tucson Airport Area Remedial Investigation, which was managed by the Arizona Department of Health under a cooperative agreement with EPA, was concluded in 1985. To date, over 100 monitoring wells have been drilled to identify, characterize, model, and monitor the contamination in the area (Environmental Protection Agency, 1988a).

200 GROUND WATER MODELS ~ Eighteen remedial alternatives were designed and analyzed for the Air Force property. Ground water extraction and recharge were chosen as the preferred remedy. This remedy for the Air Force property, which began operation in 1988, involves pumping, treating, and recharging approximately 26 billion gal. of water over a 10-yr period (Environmental Protection Agency, 1988a). This remedy is described in the Record of Decision issued for the site on July 25, 1988. No remedy has yet been selected for the area north of Los Reales Road. The modeling study conducted by CH2M Hill for EPA that attempted to assign liabilities was severely criticized by several of the potentially responsible parties, and as a result no agreement has yet been reached on an appropriate remedial action for this area. Discussion Ground water models are frequently used to determine sources of observed contamination. In general, the information on the current distribution of contaminants and hydrogeologic conditions is insuffi- cient to allow a unique solution for the location of sources and the timing of source releases. This is particularly true when all poten- tial sources are located along the same stream line and there are no marker chemicals for a specific source. Ground water models, however, can be used to help set bounds on the range of possible contributions from individual sources. ~Area, Niagara Falls, New York Background The S-Area landfi~! is located on the southeast corner of Oc- cidental Chemical Corporation's Buffalo Avenue Plant in Niagara Falls, New York. Approximately 63,100 tons of chemical waste was deposited at the site. The S-Area landfill is one of four landfi~Is in the Niagara Falls, New York, area that were operated by Occidental Chemical Corporation (OCC), formerly known as Hooker Chemicals Plastics Corporation. The other landfi~Is are Love Canal, Hyde Park, and the 102nd Street landfill. Ground water flow and transport models have been used ex- tensively at all of these sites, and the use of these models is par- ticularly well documented (Mercer et al., 1985; C. Faust, affidavits in Civil Action Nos. 79-988 and 79-989 in the U.S. District Court for the Western District of New York, 1984 and 1985, respectively).

CONTAMINANT FLOW MODELS IN THE REGULATORY SYSTEM 201 These contaminant transport models have also been incorporated into legally enforceable documents and have been evaluated and ap- proved by a court. For simplicity, this case study focuses primarily on the models used at the S-Area site, because they illustrate the complex processes that can be simulated with the current generation of ground water transport models. A major concern at the landfill is discontinuities in an underlying confining bed that allow dense nonaqueous-phase liquids (NAPEs) to contaminate a bedrock aquifer. The chemicals in the lands! will be contained after remediation by an integrated system of barrier walls, plugs, drains, and a cap that is designed to prevent off-site migration. A conceptual hydrogeologic cross section of the landfill! before and after remediation is shown in Figure 5.12. Prior to remediation, hydraulic gradients are downward, and ground water and NAPL flow into the bedrock where the clay and fill are missing. After remediation, the drains, walls, and cap on the site are intended to create a sufficient upward hydraulic gradient to reverse the flow of ground water and NAPL into the bedrock. A one-dimensional, tw~phase flow mode! was developed by Arthur D. Little, Inc. (ADL) to establish what upward hydraulic gradient would prevent downward migration of NAPL at the S-Area lands! (Arthur D. Little, Inc., 1983; Guswa, 1985; C. Faust, Af- fidavit in Civil Action No. 79-988 in the U.S. District Court for the Western District of New York, 1984 (particularly paragraphs 42-44~. The mode! considers, among other things, the effects of Ethology-dependent capillary pressure functions, hydraulic gradients, and permeability variations. Subsequently, a two-dimensional, two- phase flow mode} was developed by EPA's consultant to ensure that the one-dimensional mode! was appropriate for selecting a remedy for the site. After the initial remedies were selected for the site, a three-dimensional mode! was developed and is currently being used to evaluate conditions at the site and the potential effectiveness of additional remedies at the site. The model's use to design the remedy is discussed in this case study. Site Conditions The NAPL found at S-Area has a specific gravity of approxi- mately 1.5 and consists primarily of trichIorobenzene, tetrachioro- benzene, pentachIorobenzene, tetrachIoroethylene, hexachIorocyclo- pentadiene, and octachiorocyclopentene (S. Fogel, Affidavit in Civil Action No. 79-988 in the U.S. District Court for the Western District

202 Before Rain Lagoon Leaks GROUND WATER MODELS Rain .: . :$ : ;: 8~ ~( t~ NApL : 2 . : ~ ~ ~ ~ ~,,~und W~r ~ NAPL i. · . . . . . %~ ~.. ... ..== ~ ~ ~ j ~j ~ ~ . ~ ~ ~j . j. ... ... . , .: it. ~,.~$::~:~; ~ ~ ~- ' ~ ~ ~ ~ ~ After `~ ` ~ rat Wall ~ ~ I_ ~ G .~ .'.' '.~ ~2 ~.'2. . Mimi ~ ~` ~ ,' .: . ' . . . . .. ~; ~ ! r ,; ~ ~r ~ ~ ~ 7~ ~ ~ ~ ~ ~ ~ ~ ~ ~ -~ ~ ~ _~ ~ it! ~ ~ ~ ~ ;* ; ;; ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ _ r ~ ~ ~ ~ ~' ~ ~ ~ ~ ~ Ground Water & NAPE FIGURE 5.12 A conceptual cross section of the hydraulic containment system to be implemented at the S-Area landfill. SOURCE: Cohen et al., 1978.  of New York, 1984~. These liquids have been observed in discrete discontinuous zones in the landfill. Geologic logs indicate a litho- logic contact between unconsolidated glacial deposits and bedrock (Lockport dolomite) at an elevation of about 541 ft. The base of the unconsolidated glacial deposits is a clay ranging in thickness from about 0.25 to 15 ft. The clay is overlain by a relatively thick (up to 16 It) fine sand layer containing scattered zones of silt and fine gravel. This is overlain by about 14 It of artificial fill. Bedrock water- leve} measurements indicate a potentiometric elevation of about 561 ft. Water levels measured in the overlying unconsolidated deposits

CONTAMINANT FLO W MODELS IN THE REGULATORY SYSTEM 203 indicate a positive head difference between the overburden and the underlying bedrock of between 2 and 5.5 ft. Under these conditions therefore, a vertical downward flow component exists. Mode} Formulation The two models developed to design the remedies used the method of finite differences. The ADL mode! employed the im- plicit pressure-explicit saturation (IMPES) method to solve the two coupled equations of flow for an immiscible nonaqueous phase and water. The air phase is neglected. The ADL mode} also used a mesh-centered grid, whereas the other model, referred to as SWAN- FLOW (simultaneous water and NAPL flow), used a block-centered approach (GeoTrans, 1985~. To evaluate the potential for downward NAPL flow, a vertical column 23 It long was divided into 24 blocks (nodes). The mode! was constructed with a 2-ft negative head difference (downward flow) between the water table and bedrock potentiometric level. The do- main contains three different porous materials. The upper 20 It consists of a fine sand with a hydraulic conductivity of 10-5 cm/s (k = 1.02 x 10-~4 mid. The fine sand is underlain by 1 It of clay (K = 10-7 cm/s; kin 1.02 x 10-~6 mid. The clay is underlain by the Lockport dolomite bedrock (K = 10-3 cm/s; k = 1.02 x 10-~2 mat. The residual saturation values for water and NAPL were assumed to be 20 and 10 percent, respectively. Other simulation data are given in Tables 5.5 and 5.6. Results and Conclusions The results show that a barrier to downward migration of NAPL is provided by capillary pressure differences between the sand and clay (Figure 5.13~. This condition has been confirmed in recent field investigations at the S-Area site (Faust and Guswa, 1989~. A comparison between the results of the two numerical models is shown in Figure 5.13. The saturations calculated by SWANFLOW and the ADL code at approximately 250 days are shown. The results from the two models compare favorably; however, there are some differences, especially just above the clay layer. The differences are probably caused by some combination of instability in the IMPES technique, alternative "ridding and time steps used in the two codes, and slight differences in the relative permeability relationships (the ADL [1983] mode! provided for hysteresis in capillary pressure).

204 GROUND WATER MODELS TABLE 5.5 Capillary Pressure and Relative Permeability Data for ADL Simulation 1 Capillary Water Relative Permeabilities Pressure (N/m2)a Saturation Water NAPL Fine sand and bedrock 103,425.0 0.00 0.00000 1.00000 103,425.0 0.10 0.00000 0.82000 103,425.0 0.20 0.00000 0.68000 27,580.0 0.30 0.04000 0.55000 10,343.0 0.40 0.10000 0.43000 7,585.0 0.50 0.18000 0.31000 7,447.0 0.60 0.30000 0.20000 7,309.0 0.70 0.44000 0.12000 7,171.0 0.80 0.60000 0.05000 7,033.0 0.90 0.80000 0.00000 6,895.0 1.00 1.00000 0.00000 Clay 206,850.0 0.00 0.00000 1.00000 206,850.0 0.10 0.00000 0.82000 206,850.0 0.20 0.00000 0.68000 165,480.0 0.30 0.04000 0.55000 134,453.0 0.40 0.10000 0.43000 110,320.0 0.50 0.18000 0.31000 93,082.0 0.60 0.30000 0.20000 82,740.0 0.70 0.44000 0.12000 75,845.0 0.80 0.60000 0.05000 72,398.0 0.90 0.80000 0.00000 68,950.0 1.00 1.00000 0.00000 aN = newton (i.e., kg-n~s2). SOURCE: Arthur D. Little, Inc.,1983. TABLE 5.6 Data Used in ADL Simulation 1 Parameter Value Porosity Permeability Fine sand Clay Bedrock  Density of water Density of NAPL Water viscosity NAPL viscosity Dz (vertical dispersion length) - SOURCE: Arthur D. Little, Inc.,1983. 0.2 1.02 x 10-~4m2 1.02 x 10- ~6 m2 1.02 x 10- ~2 m2 1,000 kg/m3 1,500 kg/m3 0.001 kg/m-e 0.001 kg/m-e 0.3048 m

CONTAMINANT FLOW MODELS IN THE REGULATORY SYSTEM 205 575 llJ r ~ 570 _ > ~ z 560 _ LIJ O 550 _; i O 540 _ 535 _ BY In 44 ' / '/ ,,, or/ ... :e- ·:~:-: ::: ·:~:-: :-:-:  .... _ Original water table elevation Modeled column (e = node locations) hSR-tWT = 9 It ky fine sand = 10 5 cm/s Initial NAPL distribution ADL ~ 250 days SWANFLOW ~ 275 days Multiple points Model boundary nodes ~ jI "\ L- Fine sand ; ~ 0 20 40 60 NAPL DISTRIBUTION, percent Bedrock FIGURE 5.13 NAPL saturation profiles at one time for the two-layer simula- tion. The effects of a water-phase hydraulic gradient on NAPL migra- tion were also examined via these simulations, where the clay layer was assumed to be musing. As shown in Figure 5.13, the results of this series of simulations indicated that a minimum upward head difference of 9 It between the water table elevation and bedrock po- tentiometric level in the vicinity of a clay layer discontinuity could be sufficient to prevent downward migration of NAPL into the bedrock (Guswa, 1985~. This figure indicates NAPL saturations at about 250 days. As shown, there is a noticeable upward movement of NAPL. Data have been collected as part of a remedy designed to lower the hydraulic head in the overburden sand. These data will be used  to confirm the remedy as well as modeling results. Regulatory Context In December 1979 the federal government filed four lawsuits to obtain cleanup of four OCC landfills. EPA, the state of New

206 GROUND WATER MODELS York, and OCC negotiated an extensive set of remedies. The S- Area Consent Decree incorporated these remedies, including the one-dimensional, two-phase containment transport mode} discussed above. The consent decree was lodged with the court on January 10, 1984.23 Consent decrees are subject to a Sunday public comment period. If the consent decree is adequate, proper, and in the public interest, the Department of Justice and the court finally approve it (see 28 CFR §50.7~. In this particular case, the province of Ontario requested and the court granted a formal evidentiary hearing to re- view the consent decree. These models were subject to close, critical scrutiny during the public comment period and court hearing, in- cluding scrutiny by consultants hired by the province of Ontario.24 The court held that the consent decree was "fair, adequate, and con- sistent with public policy . . . [and] wall adequately protect the public L J Interest In neaten ano tne environment."25 Two consultants employed by EPA, as well as EPA and state personnel, oversaw and peer-reviewed the development of the ADL mode! (G. Pinder, Affidavit in United States v. Hooker Chemicals and Plastics (~5~ Area Landfill}, Civil Action No. 79-988 in U.S. District Court for the Western District of New York, 1984, particu- larly paragraphs 23-25~. The two-dimensional, two-phase flow mode} was developed by one of EPA's consultants to ensure that the one- dimensional mode! was appropriate at the site (C. Faust, Affidavit in Civil Action No. 79-989 in the U.S. District Court for the Western District of New York, 1985~. Discussion This case study illustrates the use of relatively complex models of ground water and NAP L flow to help design a remedial action for a hazardous waste site. Field studies have shown that both of these models were able to simulate observed field conditions. The results of these mode} studies have demonstrated that the current generation of ground water models can be used to investigate the migration of an immiscible, denser than water fluid within an aquifer. Interestingly, this study also shows that a one-dimensional mode} can be just as useful as a two-dimensional mode! in the investigation of the appropriateness of a remedial action.

CONTAMINANT FLOW MODELS IN THE REGULATORY SYSTEM 207 NOTES 1. 42 USC §9601 et seq. and 40 CFR Part 300. National Oil and Hazardous Substances Pollution Contingency Plan, 53 Fed. Reg. 51,394 (1988), contains the proposed new Superfund regulations. 2. United Statue v. Mottolo, 605 F. Supp. 898, 902 (DNH 1985~. 3. Chemical Carcinogens; A Review of the Science and Its Associated Principles, 1985, 50 Fed. Reg. 10,372 (1985~. 4. Guidelines for Estimating Exposures, 51 Fed. Reg. 34,042 (1986~. 5. Ibid. 6. Hazardous Waste Management System; Identification and Listing of Hazardous Waste; Final Exclusion and Final Organic Leach ate Model (OLM). 7. Hazardous Waste Management System; Identification and Listing of Hazardous Waste. 8. Hazardous Waste Management System; Identification and Listing of Hazardous Waste; Notification Requirements; Reportable Quantity Adjust- ments. 9. Hazardous Waste Management System; Land Disposal Restrictions. 10. Hazardous Waste Management System; Identification and Listing of Hazardous Waste; New Data and Use of These Data Regarding the Establish- ment of Regulatory Levels for the Toxicity Characteristic; and Use of the Model for the Delisting Program. 11. Underground Injection Control Program: Hazardous Waste Disposal Injection Restrictions; Amendments to Technical Requirements for Class I Hazardous Waste Injection Wells; and Additional Monitoring Requirements Applicable to All Class I Wells. 12. Citing D. Morganwalp and R. Smith, 1987, Modeling of Representative Injection Sites, EPA report in progress. 13. This discussion of contamination in the Tucson Airport area is largely extracted and paraphrased from the report by CH2M Hill (1987) and the Remedial Investigation prepared for the Arizona Department of Health Services by Schmidt (1985) and Mock et al. (1985~. 14. Hazardous Waste Management System; Identification and Listing of Hazardous Waste: Use of a Generic Dilution/Attenuation Factor for Establish- ing Regulatory Levels and Chronic Toxicity Reference Level Revisions. 15. Hazardous Waste Management System; Identification and Listing of Hazardous Waste. 16. Hazardous Waste Management System; Identification and Listing of Hazardous Waste; Notification Requirements; Reportable Quantity Adjust- ments. 17. McLouth Steel Products flora. v. Thomas, 838 F.2d 1317, 1320 (D.C. Cir. 1988). 18. Hazardous Waste Management System; Identification and Listing of Hazardous Waste. 19. Hazardous Waste Management System; Identification and Listing of Hazardous Waste; Final Exclusion Rule. 20. Hazardous Waste Management System; Identification and Listing of  Hazardous Waste; Final Denials. 21. See supra, note 10. 22. The Washington Post, p. 1 (October 9, 1981~. 23. United States v. Hooicr Chemicals ~ Plastics Corp. (S-ArcaJ, Civ. Act. No. 79-988 (filed January 10, 1984~.

208 GROUND WATER MODELS 24. Unfed Stated v. Hoofer Chcaucal~ ~ Plastics Corp., 607 F. Supp. at 1061. 25. Ibid. at 1070. BIBLIOGRAPHY Arizona Department of Health Services. 1986. Responsiveness Summary, Re- sults of the Tucson Airport Area Remedial Investigation, 15 pp. Arthur D. Little, Inc. 1983. S-Area to Phase Flow Model. Prepared for Wald, Harkrader & Ross (now merged with Pepper, Hamilton & Scheetz), Washington, D.C. CH2M Hill. 1987. Assessment of the Relative Contribution to Groundwater Contamination from Potential Sources in the Tucson Airport Area, Tucson, Arizona. Prepared for U.S. EPA Region IX. Cohen, R. M., R. R. Rabold, C. R. Faust, J. O. Rumbaugh III, and J. R. Bridge. 1978. Investigation and hydraulic containment of chemical migration: Four landfills in Niagara Falls. Civil Engineering Practice (Spring), 33-58. Cooley, R. L. 1977. A method for estimating parameters and assessing reliability for models of steady-state ground-water flow, 1. Theory and numerical properties. Water Resources Research 13, 318-324. Domenico, P. A., and V. V. Palciauskas. 1982. Alternative boundaries in solid waste management. Ground Water 20, 301-311. Downey, J. S., and E. J. Weiss. 1980. Preliminary Data Set for Three- Dimensional Digital Model of the Red River and Madison Aquifers. U.S. Geological Survey Open-File Report 80-756, Denver, Colo. Environmental Protection Agency. 1986a. Superfund Public Health Evaluation Manual. OSWER Directive No. 9285.4-1, Washington, D.C. Environmental Protection Agency. 1986b. RCR~A Ground-water Monitor- ing Technical Enforcement Guidance Document. OSWER Directive No. 9950.1, Washington, D.C. Environmental Protection Agency. 1987a. Alternate Concentration Limit Guid- ance Part 1: ACL Policy and Information Requirements. Interim Final. OSWER Directive No. 9481.00-6C, EPA/530-SW-87-017. Washington, D.C. Environmental Protection Agency. 1987b. Evaluation of Risk-Based Decision- making in ACRE. Annotated Briefing. Internal document, p. 5. Environmental Protection Agency. 1988a. Evaluation of Hughes Aircraft, U.S. Air Force Plant No. 44, Tucson, Ariz. EPA/700-8-87-037, Hazardous Waste Ground-Water Task Force, Washington, D.C. Environmental Protection Agency. 1988b. Final Review Draft Guidance on Remedial Actions for Contaminated Ground Water at Superfund Sites. OSWER Directive No. 9283, Washington, D.C., pp. 3-22. Environmental Protection Agency. 1988c. Selection Criteria for Mathematical Models Used in Exposure Assessments: Ground-water Models. EPA/600/8- 88/075. Washington, D.C. Faust, C. R., and J. H. Guswa: 1989. Simulation of three-dimensional flow of  immiscible fluids within and below the unsaturated zone. Submitted to Water Resources Research. In preset Fenneman, N. M. 1931. Physiography of the Western United States. McGraw- Hill, New York.

CONTAMINANT FLO W MODELS IN THE REGULATORY SYSTEM 209 GeoTrans, Inc. 1985. SWANFLOW: Simultaneous Water, Air, and Nonaqueous Phase Flow, Version 1.0. Documentation prepared for Environmental Protection Agency. Grove, D. B. 1977. The Use of Galerkin Finite-Element Methods to Solve Ma~s-Transport Equations. U.S. Geological Survey Water Resources In- vestigation 77-49, 55 pp. Guswa, J. H. 1985. Application of Multi-Phase Flow Theory at a Chemical Waste Landfill, Niagara Falls, New York. Pp. 108-111 in Proceedings of the Second International Conference on Groundwater Quality Research, published by the National Center for Ground Water Research, Stillwater, Okla. Konikow, L. F. 1976. Preliminary Digital Model of Ground-Water Flow in the Madison Group, Powder River Basin and Adjacent Areas, Wyoming, Montana, South Dakota, North Dakota, and Nebraska. U.S. Geological Survey Water Resources Investigation 63-75, 44 pp. Konikow, L. F., and J. D. Bredehoeft. 1978. Computer Model of Two- Dimensional Solute Transport and Dispersion in Ground Water: Tech- niques of Water-Resources Investigations of the United States Geological Survey. Book 7, Chapter C2, U.S. Geological Survey, 90 pp. Lewis, B. D., and F. J. Goldstein. 1982. Evaluation of a Predictive Ground Water Solute-Transport Model at the Idaho National Engineering Labora- tory, Idaho. U.S. Geological Survey Water Resources Investigation 82-55, 71 pp. Mercer, J., C. R. Faust, R. M. Cohen, P. F. Andersen, and P. S. Huyakorn. 1985. Remedial action assessment for hazardous waste sites via numerical simulation. Waste Management and Research 3, 377-387. Mock, P. A., B. C. Travers, and C. K. Williams. 1985. Results of the Tucson Airport Area Remedial Investigation, Volume III, Contaminant Transport Modeling. Arizona Department of Water Resources, 106 pp. Rampe, J. 1985. Results of the Tucson Airport Area Remedial Investigation, Phase I, Volume III, Evaluation of the Potential Sources of Groundwater Contamination near the Tucson International Airport. Arizona Depart- ment of Health Services. Reeves, M., D. S. Ward, N. D. Johns, and R. M. Cranwell. 1986a. Data Input Guide for S WIFT II; the Sandia Waste-Isolation Flow and Transport Model for Fractured Media. Release 4.84. NUREG/CR-3162, U.S. Nuclear Regulatory Commission, Washington, D.C. Reeves, M., D. S. Ward, N. D. Johns, and R. M. Cranwell. 1986b. Theory and Implementation for SWIFT II; the Sandia Waste-Isolation Flow and Transport Model for Fractured Media. Release 4.84. NUREG/CR-3328, U.S. Nuclear Regulatory Commission, Washington, D.C. Reeves, M., D. S. Ward, P. A. Davis, and E. J. Bonena. 1987. SWIFT II Self- Teaching Curriculum; Illustrative Problems for the Sandia Waste-Isolation Flow and Transport Model for Fractured Media (revised). NUREG/CR-  3925, U.S. Nuclear Regulatory Commission, Washington, D.C. Robertson, J. B. 1974. Digital Modeling of Radioactive and Chemical Waste Transport in the Snake River Plain Aquifer at the National Reactor Testing Station, Idaho. U.S. Geological Survey Open-File Report ID0-22054, 41 PP

210 GROUND WATER MODELS Ross, B., J. W. Mercer, S. D. Thomas, and B. H. Lester. 1982. Benchmark Problems for Repository Siting Models. NUREG/CR-3097, U.S. Nuclear Regulatory Commission, Washington, D.C. Schmidt, K. D. 1985. Results of the Tucson Airport Area Remedial Investi- gation, Vol. I, Summary Report. Arizona Department of Health Services, 113 pp. Science Advisory Board. 1984. Report on the Review of Proposed Environmen- tal Standards for the Management and Disposal of Spent Nuclear Fuel, High Level and Transuranic Radioactive Wastes (40 CFR §191~. Report to the EPA by the High-Level Radioactive Waste Subcommittee. Siefken, D., G. Pangburn, R. Pennifill, and R. J. Starmer. 1982. U.S. Nu- clear Regulatory Commission Low Level Waste Licensing Branch Technical Position Site Suitability, Selection, and Characterization. NUREG-0902, U.S. Nuclear Regulatory Commission, Washington, D.C. Silling, S. A. 1983. Final Technical Position or Documentation of Computer Codes for High-Level Waste Management. NUREG-0856, U.S. Nuclear Regulatory Commission, Washington, D.C. Trescott, P. C., G. F. Pinder, and S. P. Larson. 1976. Finite-Difference Model for Aquifer Simulation in Two Dimensions with Results of Numerical Experiments. Techniques of Water-Resources Investigations of the U.S. Geological Survey, Book 7, Gh. C1. U.S. Government Printing Office, Washington, D.C., 116 pp. U.S. Nuclear Regulatory Commission. 1982. Final Technical Position on Documentation of Computer Codes for High Level Waste Management. NUREG-0856, Washington, D.C., p. 2. U.S. Nuclear Regulatory Commission. 1988. Standard Review Plan for the Review of a Licensed Application for a Low Level Radioactive Waste Disposal Facility: Safety Analysis Report 1988, NUREG-1200, Revision 1. Washington, D.C. Wilkinson, G. F., and G. E. Runkle. 1986. Quality Assurance (QA) Plan for the Computer Software Supporting the U.S. Nuclear Regulatory Commission's High-Level Waste Management Program. NUREG/CR-4369, U.S. Nuclear Regulatory Commission, Washington, D.C.  Woodward-Clyde Consultants, Inc. 1981. Well-Field Hydrology Technical Re- port for ETSI Coal Slurry Pipeline Project. U.S. Bureau of Land Manage- ment. Zamuda, C. 1986. The Superfund record of decision process: Part 1 The role of risk assessment. Chemical Waste Litigation Reporter 11, 847.