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.
Overview, Conclusions, and Recommendations OVl:RVIl:W Mathematical models, used commonly in ground water studies, are an attempt to represent processes by mathematical equations. The precise language of mathematics provides a powerful mechanism for expressing a tremendous quantity of information in an amazingly simple and compact way. Naturally, the starting point in modeling is a clear understanding of the processes involved. In terms of the flow of ground water or multiphase flow (i.e., when a fluid such as water, gasoline, or a dense nonaqueous~phase liquid is moving in the subsurface), one mainly needs to consider two dominant processes: flow in response to hydraulic potential gradients and the loss or gain of water from sinks or sources (e.g., pumping or injection, or gains and losses in storage). In the case of contaminant transport, a much larger number of diverse and complicated processes are involved. These processes can be divided into two groups: (~) those responsible for fluxes and (2) those responsible for sources and sinks for the material. Mass fluxes are prompted by processes like advection, diffusion, and mechanical dispersion. Sources ant! sinks are provided by a host of chemical, nuclear, and biological processes, such as sorption, ion exchange, oxidation/reduction, radioactive decay, and biodegradation. In this report, Chapter 2 is devoted to explaining in a simple way 1
2 GROUND WATER MODELS how the differential equations for ground water flow and mass trans- port are created to embody the various processes. To fully describe a ground water system to be modeled, one needs in addition to the governing equations (1) specific numerical values for parameters that characterize the processes and for simulation parameters that are involved with the procedure to solve the equations and (2) informa- tion about the region, shape, and conditions along the boundaries. Solution of the resulting modeling problem is usually carried out analytically or numerically, depending upon the complexity of the hydrogeologic setting and the number of processes that need to be considered. Few flow and transport problems can be modeled with confi- dence. As the following discussion explains, the most satisfactory results to date have come with models involving the flow of water or the transport of a single nonreactive contaminant in a saturated porous medium. As systems become more complicated through par- tial saturation, the presence of several mobile fluids, fracturing, or the existence of reacting contaminants, many more questions arise about the adequacy or validity of the underlying process models. The nat- ural reaction of researchers is to undertake long-term experimental investigations, which in the scientific tradition will gradually improve our understanding of these processes. Although such research is un- deniably important, it may not provide answers in time to influence many important national and local decisions about ground water contamination. Chapters 3 and 4 of this report, along with many other thought- fu] reports, papers, and articles, reveal major areas of uncertainty about subsurface contamination. Decisionmakers need to confront this uncertainty realistically and not be misled by the ability of com- puter models to always provide answers. Admitting the presence of uncertainty, however, is not enough. There is a need to make decisions, clean up water supplies, remove threats to public health, and devise safer methods for disposing of our wastes. Some of the decisions made in the short term may be inappropriate, inefficient, or even counterproductive, but it is unacceptable to simply wait until poorly understood environmental problems can be solved with more confidence. In order to examine this issue further, it is useful to briefly review those areas where the understanding appears to be relatively good and those areas where there is still much to learn. Each of the major modeling categories discussed in Chapters 3 and 4 is
OVERVIEW, OONOLUSIONS, AND RECO~NDATIONS 3 briefly examined in the following paragraphs. Then the question of what decisionmakers can and should do now with problems requiring immediate attention is revisited. The processes that control saturated ground water flow are rea- sonably well understood, and standard models of these processes are generally believed to be able to give reliable predictions if provided with adequate amounts of data. Nevertheless, the impacts of field- scale heterogeneity are still widely debated, and there are few clear guidelines on how model inputs should be estimated from limited databases or on how hydrologic monitoring programs should be de- signed. While saturated flow modeling is becoming more straightfor- ward than it once was, there is much room for individual judgment, and the experience of the modeler still makes a significant difference in the quality of the results obtained. It is questionable whether this experience will ever be replaced by automated techniques such as expert systems, although such innovations may make the job of the informed modeler easier. Unsaturated flow is less well understood. The basic "laws" that govern such flow are still questioned by some investigators. Much of the conventional theory of unsaturated flow is based on small-scale, one-dimensional laboratory experiments, which may not provide an accurate picture of behavior at larger field scales. There have been very few field studies of unsaturated flow that extend over the scales of interest in most contamination applications, and most of these have focused on one-dimensional transport in the vertical direction. Some investigators believe that unsaturated flow can move horizontally over significant distances, although available evidence is insufficient to either confirm or reject this hypothesis. Even if straightforward extrapolation from the laboratory to the field were possible, current techniques for determining unsaturated soil properties are too expensive and time-consuming to provide adequate descriptions of most contaminated sites. The numerical demands of all but the simplest unsaturated zone simulation mod- els are formidable, and accurate three-dimensional unsaturated flow modeling capabilities are not available to most consultants. Yet many important contamination problems, such as leaking under- ground storage tallks, infiltrating pesticides, and leaching mining wastes, afl.ect the unsaturated zone. Contaminant transport in this zone has only recently been perceived by the hydrologic community as an important research priority. Much remains to be done.
4 GROUND WATER MODELS Flow through fractured media may be either saturated or un- saturated. Both types of fracture flow are difficult to predict at a given site unless extensive information is available about the fracture network. In this sense, true predictive modeling is not yet a reality. Nevertheless, recent research has provided significant advances in the understanding of the relative importance of the fracture and matrix systems in fractured flow. These advances have influenced some anal- yses of candidate radioactive waste disposal sites but have not, for the most part, reached the larger modeling community. The prevail- ing approach is to ignore fracture flow and hope that the effects of individual fractures will, in some sense, "average out." This can be a misleading oversimplification in some applications, where fractures can act as conduits for contaminant flow or can significantly modify subsurface flow patterns. Practical modelers need better guidelines for determining when fracture flow may be important and better methods for incorporating such flow into their mode! predictions. The status of contaminant transport modeling depends greatly on the chemical species and phase of interest. In general, the pros cesses that influence the transport of dilute, nonreactive aqueous phase solutes are wed understood, at least in saturated media. There is, however, still widespread disagreement about the effects of spatial and temporal variability and about the related concept of macrodis- Persian. Until very recently, there were very few controlled field studies of ground water contaminant transport. Recent studies tend to indicate that real-worId contaminant plumes have complex three- dimensional structures. which can be difficult to Predict when soil , properties are very heterogeneous. It can be difficult to simply map an existing plume, given the data typically available at a newly dis- covered contaminated site. Prediction of plume movement over many years is an even more difficult task. The problems associated with transport modeling are greatly compounded when the solutes are reactive. In this case, chemical rather than hydrologic processes may govern the behavior of a con- taminant plume. Ground water chemistry and ecology are relatively new fields that have had to contend with the problems inherent in working in an environment where processes are not readily observed and where samples are costly and difficult to obtain. Most models of reactive solutes are based on small-scale laboratory studies, which may not accurately mimic conditions found in the actual subsurface environment. This raises all of the same scale issues mentioned ear- lier in conjunction with unsaturated flow. Despite these difficulties,
OVERVIEW, CONCLUSIONS, AND RECO~NDATIONS 5 simple reactive transport models are in wide use and many modelers are aware of the need to at least consider sorption, biodegradation, and other chemical effects. It remains to be seen whether these simple models are adequate for d~ecisionmaking purposes. Most ground water contaminant modelers would probably agree that multiphase contaminant transport Is the area where the basic physical mechanisms that control contaminant movement and degra- dation are least well understood and most difficult to model. Yet a wide range of important contaminants probably travel as separate liquid or gaseous phases when they move through the subsurface environment. Field-scale experimental investigations of multiphase transport are very limited, and existing laboratory-scale results in- dicate that this type of transport is influenced by a number of inter- acting factors, including viscosity and density contrasts, capilIarity, and phase transitions. Although models of multiphase transport are available, many of the inputs they require are, as in the re- lated case of unsaturated flow, difficult to estimate in a field setting. Because field data are very limited, it is practically impossible to confirm whether or not these models accurately reflect reality. More- over, existing multiphase modeling techniques are computationally demanding and probably impractical to apply in situations where dozens of different interacting species and phases coexist. Such situ- ations occur frequently. Leaking gasoline storage tanks are just one example. Case studies provide a useful way to illustrate the application of models in (1) understanding ground water systems, (2) predict- ing contaminant rn~grations, and (3) decisionmaking by regulatory agencies. An example of the first type of application relates to the use of the generic vertical-horizontal spread (VHS) mode! by the U.S. Environmental Protection Agency (EPA) to determine when solid wastes need to be treated as hazardous wastes. In the case of the Madison aquifer, modeling studies predicted water-level declines due to large withdrawals by pumping. ~ ~. . ~, . . An example of the second type ot application, modeling In connection with contamination of the Snake River plain, provided a prediction of the future extent of plume development. The third type of application is illustrated by the cases of contamination at the S-Area landfi~! in Niagara Falls and at Tucson Airport, where modeling was an integral part of the legal decisionmaking. The above review of the present state of ground water contam- inant modeling is not really as pessimistic as it may appear at first
6 GROUND WATER MODELS glance. In the last several years there has been substantial progress in such fields as fracture flow modeling, modeling of reactive transport and transformation, and multiphase modeling. The present concern with ground water contarn~nation has stimulated a major increase in research efforts that has resulted in advancement. Moreover, it is the committee's opinion that the needs of decisionmakers are best served by honest and realistic assessments of the modeling state of the art. With such assessments we can set priorities, make difficult decisions, and understand how to deal with pressing short-term problems. The fact that many of the models used in practice have not been validated to a significant extent provides an important source of uncertainty in the predictions that come from the models. Unfortu- nately, even more uncertainty enters the modeling process from, for example, (1) the inability to precisely describe the natural variabil- ity of model parameters (e.g., hydraulic conductivity) from a finite and usually small number of measurement points, (2) the inherent randomness of geologic and hydrogeologic processes (e.g., recharge rates and erosion) over the long term, (3) the inability to measure or otherwise quantify certain critical parameters (e.g., features of the geometry of fracture networks), and (4) biases or measurement errors that are part of common field methods. When all these sources of uncertainty are properly considered, a single model prediction re- alistically has to be viewed as one of a relatively large number of possible system responses. Over the past decade, the development of stochastic modeling techniques has been useful in quantitatively es- tablishing the extent to which uncertainty in model input translates to uncertainty in model prediction. To return to the question posed earlier, what should a deci- sionmaker do now, given existing modeling capabilities? There is obviously no easy or comforting answer to this question. It seems apparent, however, that it would be unwise to rely solely on any single source of information when deciding how to formulate regula- tions, carry out a cleanup, or protect public health. Models should be supplemented by carefully conceived field work, which not only provides data for estimating model inputs but also provides an inde- pendent confirmation of conditions in the subsurface environment. Put simply, the decisionmaker should hedge his bets and distribute his resources, funding different types of modeling efforts and mixing modeling with on-site monitoring. When field data are inconclu- sive or insufficient, model results may have a significant influence
OVERVIEW, CONCLUSIONS, AND RECOM:MENDATIONS 7 on the predicted impact of a given decision. In this case, the deci- sionmaker should request a quantitative and defensible assessment of the model's accuracy in order to evaluate the risk of making a bad decision. In this regard, environmental management is no dif- ferent from any other form of management where uncertainty and risk are important. Modem are not going to relieve us of the burden of mating difficult decisions. They simply provide some additional information to consider. It is unrealistic to expect much more. CONCLUSIONS AND RECOMMENDATIONS Modele and Subsurface Processes Conclusions There is a range of capability In modeling fluid flow In geologic media. Modeling saturated flow In porous media is straightforward with few conceptual or numerical problems. At the present time, conceptual issues and/or problems in obtaining data on parameter values limit the reliability and therefore the applicability of flow models Involving unsaturated media, fractured media, or two or more liquids. As a group, flow processes are among the most widely charac- terized hydrogeologic processes. The theories of flow involving either one or more fluids in porous and/or fractured media are well es- tablished and generally accepted. For the simplest cases involving saturated flow in porous media, the basic theoretical models have been validated in countless field and laboratory studies. The great- est source of uncertainty in prediction lies in supplying values of site-specific parameters. Flow in the unsaturated zone is less well understood, particularly in the case of dry soils, where the transport of water vapor can be significant. As was the case with saturated flow, establishing values for the controlling parameters under natural conditions is difficult, particularly for parameters like permeability that can vary in a complex, nonlinear way with moisture content. Flow models involving two or more liquids in porous media are even more complicated in terms of the processes and parameters. Nevertheless, such models have been used and applied successfully, for example, in the petroleum industry. The greatest source of un- certainty in prediction remains the difficulty in accurately describing the spatial variability in controlling parameters. This problem of data is compounded by the variety of organic liquids that can be
8 GROUND WATER MODELS present as contaminants and for which specific experimental data are scarce. In the case of fractured media, it remains to be shown through field and laboratory experiments that existing conceptual models of fractured systems are valid, particularly for cases involving variable saturation and more than one liquid. In addition, there are probably chases of fractured media that cannot be modeled with continuum theories and for which discrete approaches are impractical. The data problems remain. Many (controlling) parameters are difficult to measure or estimate accurately. Thus predictions for these more complex conditions need to be evaluated carefully and assessed in light of possible limitations. Mass transport is controlled by a variety of physical, chemical, and biological processes. Quantitative descriptions of the processes con- cerned with mass transport Erection, diffusion, and dispersion) along with certain mass transfer processes (radioactive decay and sorption) are wed understood. Multidimensional models of these processes have been used succesefi~ly ~ practice. Work is still re- quired to account for other more complicated chemical processes (e.g., oxidation/reduction, precipitation, hydrolysis, and complexa- tion) and biological processes (e.g., bacterial degradation) in mass transport models. Although prototype models exist for these more complicated systems, they are not yet developed for use ~ practice. Contaminant transport is the outcome of mass transport pro- cesses, such as advection, diffusion, and mechanical dispersion, that move the mass and a multitude of mass transfer processes that redistribute mass within or between phases through chemical and biological reactions. Present-day understanding of mass transport developed from early studies on laboratory columns and more re- cent well-documented tracer studies in the field. The basic theory of advective transport modified by diffusion and mechanical disper- sion is embodied in the familiar advection-dispersion equation, which provides a practical framework for modeling contaminant transport. The main source of uncertainty in prediction lies in establishing values of controlling parameters like velocity, effective diffusion coef- ficient, and dispersivity, which can be difficult to measure or estimate and vary spatially. The complete description of mass transport usually requires that various chemical and/or biological processes also be considered. In
OVERVIEW, OONOLUSIONS, AND CONNATIONS 9 the case of reactions such as radioactive decay, sorption, and hydrol- ysis, kinetic or equilibrium models describe the reactions and the necessary rate parameters or equilibrium constants for the reactions. These reactions can be calculated and measured with reasonable accuracy if not tabulated and can be incorporated in contaminant transport models in a straightforward manner. Although models for important reactions like oxidation/reduction, precipitation, and biodegradation exist, they are complicated to formulate and solve, difficult to characterize in terms of kinetic parameters, and largely unvalidated in practical applications. Thus the transport of multiple reacting constituents such as trace metals and organic compounds cannot be modeled with confidence. As was the experience with flow, fracturing adds considerable complexity to mass transport. The issue of whether fractures are open or highly channel~zed, the importance of diffusion into the matrix, and how mixing occurs at fracture intersections make con- ceptualization of even mass transport processes uncertain. Coupled with the difficulty in formulating the mode! in terms of processes is the general lack of field and experimental data to validate models that are available. Thus transport modeling in fractured systems remains a highly speculative exercise. Models and Decisionmaking Conclusions Properly applied models are useful tools to assist in problem evaluation, design remedial strategy, conceptualize and study flow processes, provide additional information for decisionmaking, and recognize limitations in data and guide collection of new data. Ground water models are valuable tools that can be used to help understand the movement of water and chemicals in the subsurface. The purpose of the models is to simulate subsurface conditions and to allow prediction of chemical migration. When properly applied, models can supply useful information about flow and transport pro- cesses and can assist in the design of remedial programs. The results of a mode! application are dependent on the quality of the data used as input for the model. Generally, site-specific data are required to develop a mode} of a site. The mode} cannot be used as
10 GROUND WATER MODELS a substitute for data collection. However, mode} use can help direct a data collection program by identifying areas where additional data are required. Closely linked data collection and mode} application can provide an adequate representation of site conditions. Incorrect mode! use frequently occurs when the limitations of the data used to develop the mode! are not recognized. When properly applied, the results of a ground water mode} application can help in making decisions about site conditions. Mode! results can be used to supplement knowledge of site conditions but cannot be used to replace the decisionmaking process. The results of the models must be evaluated with other information about site conditions to make decisions about ground water development and cleanup. Generic models are useful as a too] for initial screening but can never be used as a replacement for site-epecific models. Geologic materials are characteristically heterogeneous. The het- erogeneity is seen at all scales, ranging from individual laminae a few millimeters thick to entire formations, aquifers, and drainage basins. In contrast, ground water models commonly incorporate various sim- plifying assumptions. Examples of some simplifications commonly used in ground water models include the assumption that an aquifer consists of a perfectly homogeneous, elastic material, or that the aquifer is made up of a small number of alternating homogeneous layers. The differences between the geologic reality of heterogeneity and the simplifications that may be used in ground water models make it scientifically dangerous and potentially misleading to blindly apply generic ground water models to any specific hydrogeological situation. A generic mode} may be useful in offering some initial guidance to an investigator. However, only the most naive would rely on the predictions of a generic mode! in an attempt to understand the details of the movement of ground water or the behavior of a dissolved pollutant in a specific hydrogeological environment. It is essential that an investigator gather site-specific information to use as input to the ground water mode! of choice and, perhaps' that the mode! itself be modified and adapted to fit the hydrogeologic conditions at a particular site. The rents of mathematical computer models may appear more
OVERVIEW, CONCLUSIONS, AND RECOMMENDATIONS 11 certain than they really are; decisionma~cers must be aware of the limitations. Modelers must contend tenth the practical reality that mode! re- sults, more than other expressions of professional judgment, have the capacity to appear more certain, more precise, and more authorita- tive than they really are. Many people who are using or relying upon the results of contaminant transport models are not fully aware of the assumptions and idealizations that are incorporated into them or of the limitations of the state of the art. There is a danger that some may infer from the smoothness of the computer graphics or the number of decimal places that appear on the tabulation of the calcu- lations a level of accuracy that far exceeds that of the model. There are inherent inaccuracies in the theoretical equations, the boundary conditions, and other conditions and in the codes. Special care there- fore must be taken in the presentation of modeling results. Modelers must understand the legal framework within which their work is used. Similarly, decisionmakers, whether they operate in agencies or in courts, must understand the limitations of models. There are situations where government regulations require the use of contaminant transport modeling. As a general rule, however, it is not necessary for regulations to specify that a mode] must be used. A few existing government regulations require that a mode} be used in the submission to the agency. All of the examples the com- mittee found involved situations where the law required a long-term prediction of the migration potential of wastes. In such situations, there is no alternative but the use of contaminant transport models. A regulation that requires contaminant transport modeling re- flects an implicit decision to require a given level of detail and allow a given level of uncertainty. When regulations require the use of a model, however, they do not imply that the solution to the problem is susceptible to a "black-box" mode} application. Quite the contrary, in the cases examined, the regulations seem to require contaminant transport modeling in the most complicated site-specific problems. Several agencies have guidelines that encourage the use of contam- inant transport models. There are many different Apes of mod- ele, mode} applications, modeling objectives, and legal ~ameworke. Agencies cannot specify a list of government-approved modem. A mode! that is appropriate for one problem may not be, and probably is not, applicable to another problem. Such a list also tends to stifle
12 GROUND WATER MODELS notation and use of newer models. On the one hand, government offlciale become reluctant to accept a nonapproved model. On the other hand, the regulated community may use an agency-approved mode} simply because the costs of getting go~rernmental approval will be lese. Suth a list may also appear to be an Implied warranty of the mode} accuracy and therefore lead to misuse of Me models. It is impossible to specify by a generally applicable regulation a contaminant transport mode} that would be scientifically valid in all applications and over the typical life of a regulation. In some circumstances, it may be appropriate to specify the use of a particular contaminant transport model. For example, after reviewing site-specific data from a hazardous waste site, an agency or private company may determine that a particular mode} could be appropriate to apply at the site and such a mode! may be specified in a consent decree or permit for specific purposes. When a mode! is used in such circumstances, the consent decree, permit, or other legally enforceable procedure should require actual monitoring to confirm the modeling results and be flexible enough to allow the mode! to be updated and modified on the basis of new data and recent scientific developments. Recommendations Models used in regulatory or legal proceedings should be available for evaluation. Models used in regulatory or legal proceedings are required to undergo public comment and review by those whose interest may be affected. The documentation associated with the mode! therefore must enable any reviewer to ~ understand what was done; evaluate the quality of the model, considering issues such as the extent to which the equations describe the actual processes (i.e., mode} validation) and the steps taken to verify that the code correctly solves the governing equations and is fully operational (i.e., code verification); evaluate the application of the mode! to a particular site; and ~ distinguish between the scientific and policy input. A list of approved models should not be sanctioned by a regulatory agency. Agencies should not require that specific models be used for site-specific application by regulation, policy, or guidance. ~stead,
OVERVIEW, CONCLUSIONS, AND RECOMMENDATIONS 13 positive attributes such as good quality assurance (QA) and doc- umentation should be mandated, and government agencies should continue to support and provide resources for the development of ground water modeling codes. The regulatory agencies should not develop a list of sanctioned ground water models. Models are used to evaluate a wide range of subsurface conditions for a variety of purposes. Models can be used to gain an understanding of flow and chemical transport, to eval- uate remedial alternatives, and to determine data collection needs. The type of problem being evaluated and the level of understanding required should dictate the mode! selection. A list of government-approved models would limit the choice of numerical codes available for problem solving. Development of a list of government-sanctioned codes could also inhibit mode} develop- ment and innovation. Because the process of mode} approval would probably be lengthy, approved models are likely to lag behind the available state of the art. As previously discussed, the quality of re- sults is dependent on the quality of the data input and the knowledge of the models. Sanctioning of codes would not eliminate the need for proper mode! application and could develop a false sense of adequacy or accuracy for mode! users. Instead of sanctioning particular models, regulatory agencies should provide detailed, consistent procedures for the proper devel- opment and application of models. Detailed specifications of positive aspects need to be developed but should include (1) good documenta- tion of a code's characteristics, capabilities, and use; (2) verification of the program structure and coding, including mass balance results; (3) mode! validation, including a comparison of mode! results with independently derived laboratory or field data and possibly other computer codes; and (4) independent scientific and technical review. The guidance must also be written to avoid being misconstrued as providing a list of "approved" models. The mere approving men- tion of a mode! in agency guidance may appear to inexperienced and untrained agency personnel as indicating that such models are "approved" or "sanctioned." Agency guidance therefore must stress that the descriptions do not sanction the use of any particular model. Instead the guidance should stress best modeling practices or princi- ples, described above, and ensure that only experienced and properly trained personnel are involved in the development and review of such models.
14 GROUND WATER MODELS Modeling should be considered to be only one of several potable methods of assessing liability In cases of ground water cont~;nation. Models should not replace sound scientific and engineering judgment. Contaminant transport models can provide one of several pos- sible methods for identifying contarn~nant sources or apportioning liability. However, it would be rare for modeling alone to provide an unequivocal answer to the question of whether and to what degree a potential source is in fact a source. Ground water models must not be viewed as "black-box" tools that eliminate or lessen the need for common sense and good scientific judgment. Similarly, while models may be useful tools in regulatory deci- sionmaking, they cannot substitute for the decisionmaking process. Such decisions are almost always based on a wide range of factors. Thus mode! results with attendant uncertainties should be consid- ered along with all other information in order to make informed regulatory decisions. Maintaining Scientific integrity Conclusions Ground water models do and should vary In complexity. The com- plexity of the mode} used to analyze a specific site should be deter- mined by the type of problem being analyzed. While more complex models Increase the range of situations that can be described, ~n- creas~ng complexity requires more input data, requires a higher level and range of skill of the modelers, and may introduce greater un- certa~ n the output if input data are not available or of sufficient quality to specify the parameters of the model. Appropriate and successful models of ground water flow and transport can range from simple analytical solutions for one-dimen- sional flow in a homogeneous aquifer to highly complicated numerical codes designed to simulate multiphase transport of reactive species in heterogeneous, three-dimensional porous media. A useful mode! need not simulate all the physical, chemical, and biological processes that are acting in the subsurface. The mode! that is appropriate for analyzing a particular problem should be determined primarily by the objectives of the study. Unfortunately, there are no set rules for determining the appropriate level of complexity. The selection of an appropriate mode! and an appropriate level of detail and complexity is subjective and dependent on the judgment and experience of the analysts and on the level of prior information about the system of
OVERVIEW, CONCLUSIONS, AND RECOMMENDATIONS 15 interest. Managers and other users of mode} results must be made aware that these trade-offs and judgments have been made and that they may affect the reliability of the model. Models must be matched to the objectives of the study. Efforts should be made to avoid using models that are more complicated than necessary. Overly complicated modem require information that cannot be obtained reliably from field measurements, which intro- duces unnecessary uncertainty into the modeling output, and overly complicated models require more time and money to operate, which wastes resources. Because there are no set rules for selecting an ap- propriate model, it is essential that agencies and companies employ qualified and well-trained personnel. One of the key requirements in successfully applying flow or contam- inant transport models is good-quality, site-specific data. Such data provide feasible bounds on the potable range of controlling parame- tere or boundary conditione, thereby minimizing the impact of data uncertainty as a major source of uncertainty associated with mode] predictions. In cases where particular mode] parameters are not or cannot be characterized, mode] prediction becomes much lese certain because predicted variables like hydraulic potential or concentration could take on a much broader range of possible values. A variety of factors can contribute to uncertainty in mode! pre- dictions. One of the most important is the inability to characterize a site in terms of the boundary conditions or the key parameters describing important flow and transport processes. This uncertainty in data results for two basic reasons. First is the issue of the am solute number of data points providing information about a given parameter. Even a relatively large number of data points may not provide a basis for estimating parameter values at locations between them with total accuracy. As the number of data points decreases, this uncertainty attached to a parameter estimate increases to the point where one finally cannot describe the spatial variability in de- tail and has to resort to a simple estimate like a mean value for a given unit. A second issue with data is the inability in some cases to measure or even accurately estimate values for necessary parameters. This problem is most serious in fractured rocks for both single-phase and multiphase flow, and for mass transport processes involving cer- tain kinetic processes (e.g., biodegradation, redox, and precipitation) whose rates can be extremely variable and site specific. These two problems increase the likelihood that in many model
16 GROUND WATER MODELS studies there are some ciata that cannot be specified with accu- racy. Sensitivity analyses provide one important way of establishing the extent to which uncertainty in a given parameter contributes to uncertainty in a prediction. Such analyses in many instances can pro- vide the justification for carrying out additional field and laboratory studies. In general, data collection and mode} application should not be viewed as sequential tasks but as tasks that should be performed interactively, complementing each other. Good documentation of ground water models throughout the mod- el~ng process is necessary because of the complexities ~n~rol~red. A hydrogeologic computer mode! may be very complex, running to thousands of lines of code. It may include hundreds of separate parameters and equations to mode} the movement of the water and the transport and fate of dissolved components. For these reasons it is essential that a mode! be accompanied by clear and thorough documentation, and that the documentation include a set of test problems that can be employed throughout the history of the mode! to verify that it continues to work properly. Adequate plans for testing and documenting a mode! should become part of any quality assurance program. Technical review should also be included in quality assurance plans to ensure that models have been adequately tested. In addition to the inherent complexity, it is common for any given mode! to undergo repeated modifications and revisions, either by the author or by subsequent users. Unless a record is kept of the modifications that are made to the code, and unless the operational accuracy of the code is periodically tested and verified, serious doubts may develop about the validity and applicability of the code. In addition to the original documentation, at least two types of information should accompany the code throughout its lifetime. First, changes in the structure of the mode! or of the database should be documented. The documentation may be in the form of a written record that ~ appended to the original documentation, or it may be included as comment lines within the noncompiled code. Second, an original set of test problems, including sample input and output, should accompany the code so that all users can periodically verify that the code is functioning properly, especially if changes are made. This periodic verification of the operation and output of the code
OVERVIEW, CONCLUSIONS, AND RECOMA~NDATIONS 17 becomes especially critical if the mode} is to be used as a part of a regulatory or legal action. There is no valid reason to use a mode! that is unavailable for evaluation and testing by other qualified investigators. Similarly, new or revised models should be accompanied by sufficient documenta- tion, history, and test problems to allow other qualified investigators to properly evaluate the mode} and to compare its output with that of other models. As ground water mode] mage has increased, a shortage of qualified staff capable of appropriately applying modem has been identifidl. In order to avoid mode! misuse, it is important that the mode} user have the training and background to understand the many processes occurring in the subsurface. Experienced staff having this training and background are insufficient in terms of the number of sites where modem could potentially be used. Recornrnendations All models must be documented so that the derivation of the mode] can be understood and the results can be reproduced by anyone seeking to use the model. The documentation should include, at a minimum, a description of the underlying problem; ~ a description of the fundamental equations that conceptualize the solution to the problem; ~ a list of all assumptions used in the mode! and the rationale for their use; a description of the code used in the model; ~ a verification of mode} codes against other solutions to the problem to verify the accuracy; · an application of the mode} to a problem with a known so- lution, albeit perhaps a simpler problem, and a comparison of the results with the known results; a sensitivity analysis; the results of a quality assurance program; the validation of the model; ~ a list of prior uses of the model, if any; · a clear identification of the site-specific data used in the ap- plication of the model;
18 GROUND WATER MODELS ~ a characterization of the level of precision, accuracy, and degree of uncertainty in the mode} results; ~ a description of the statutory/policy criteria, if any, used to shape and select the assumptions and the acceptable level of precision, accuracy, and uncertainty; and ~ any other information that is essential to understanding or being able to replicate the results. All models mat state quantitatively, to the extent possible, and if not quantitatively, then qualitatively, the Tepee and direction of uncertain n the mode] results and the time Came over which the model's prediction can be considered acceptable. This description of the uncertainties must be given at the be- g~nning of the documentation of the mode} and wherever the con- clusions of the models are used or discussed; e.g., in the conclusion of the modeler's report, in the briefing memorandum to an agency decisionmaker relying on the moclel, in whole or part, to make a regulatory decision, in the preamble to an agency regulation, and in expert testimony concerning the results of the model. The policy assumptions used in the mode] mat be explicitly listed, and the rationale for making each assumption mast be described In the documentation and wherever the conclusions of the mode] are used or discuseed; e.g., in the conclusion of the modeler's report, in the briefing memorandum to an agency decisionmaker relying on the mode], in the preamble to an agency regulation, in press releases and statements to the public, In presentations to Congress, and in expert testimony concerning the results of the model. To adroit the misuse of ground water flow and transport models, agencies and companies should employ qualified and well-tra~ned personnel. Ground water flow and transport models are complex computer codes. To ensure that the input data are appropriate, and that the output results are properly utilized and interpreted, it is important to employ properly trained and qualified individuals. These personnel must be expert in both ground water science and its mathematical representation. A certain fascination exists among technical personnel regarding the use of these powerful tools, and it is tempting to view them as "black boxes" that somehow produce easy and exact answers to previously difficult problems. This tendency may become even more pronounced as the interfaces between the codes and the users become more "user friendly." Indeed, it could be argued that the lack of a
OVERYlEW, CONCLUSIONS, AND RECO~NDA~ONS 19 user-friendly interface may be a useful safety feature to help prevent inappropriate use of the models by nonqualified personnel. If governmental agencies or private companies make the decision to use computer models in their work with ground water, it ~ em sential that the personnel involved be adequately trained and fully aware of the limitations of the code. In order to use ground water models, an organization may have to hire new personnel or train existing personnel. It is not acceptable, however, to assign modeling projects to existing personnel who may simply be available for such tasks, without intensive and appropriate training. The best procedure to ensure competency may be to designate one or more people as specialists in the modeling efforts within an organization. Such specialists would then have the responsibility to continually maintain and update their knowledge of the models being used and to make certain that others within the organization do not use the models inappropriately. The problem of rapid turnover of personnel within government regulatory agencies must also be addressed. Pressures can be very great on regulatory personnel, without corresponding financial re- wards. The record of high turnover rates within regulatory agencies, especially among younger technical employees, shows that the tempo tation to move into the private sector is very great. The Section to leave government service seems to be made about the time the individual achieves a relatively high level of competence and becomes known to various private companies. To overcome this high rate of attrition, some means of providing appropriate financial compensa- tion must be found to properly recognize, reward, and retain highly skilled individuals. If salaries cannot be raised, it is essential that an active program of recruitment and training be maintained within the agency to ensure that an adequate, high level of competency always exists among the personnel involved in ground water modeling. Research should be undertaken to provide the field and laboratory data necessary to validate How and transport models. Given that some types of models cannot be validated with ex- isting, rather limited knowledge about some types of flow and mass transport processes, it is recommended that research be undertaken to fill in information gaps. The committee recognizes a need for well-controlled field and laboratory experunents involving flow and mass transport in fractured media, and multicomponent transport
20 GROUND WATER MODELS of chemically and biologically active contaminants. Such work is es- sential to establish how well existing mathematical concepts describe actual hydrogeological systems. Recommendations for the Future Governments, academic institutions, and private Industry need to provide financial resources and substantially Increase the pool of qualified personnel In the spectrum of fields essential to ground water modeling. A severe shortage of qualified personnel exists in the areas of hydrogeology, ground water hydrology, and organic and aqueous geochemistry. Most of the new positions are with engineering and environmental consulting firms, and severe recruiting pressure exists among the firms, especially for experienced people. If the challenges posed to our ground water environment by an ever-increasing population and continued industrialization are to be met, significant steps to increase the supply of trained ground water professionals must be taken. It is the strong recommendation of the committee that additional educational resources be committed to these fields as quickly as possible. The committee also recommends that government and private industry join in the effort to increase educational resources and opportunities for students entering the spectrum of fields related to ground water modeling. In addition to providing financial support, governmental agencies and private industry should further help in the education of ground water profes- sionals by developing traineeships and industrial-associates programs to give students the opportunity to obtain practical experience in the field. Government agencies and priorate industry should be aware of the need for and benefits of additional research. Research should be pursued ~ the following areas: ~ gravitation and further development of models involved with (~) ground water flow ~ unsaturated and Factored media, (2) mul- tiphase flow in porous and fractured media, and (3) mass transport coupled with chemical reaction; role of bacteria ~ the transport end removal of contaminants; . models in decisionmaimg, including methods for identifying and presenting uncertainty and for eetablish~ng the reliability of mode! results;
OVERVIEW, CONCLUSIONS, AND RECOMMENDATIONS 21 ~ process Characterization through weB-controlled field and laboratory studies; and ~ development of new approaches for parameter estunation and of new measurement techniques. Although many aspects of ground water modeling have major deficiencies in terms of scientific understanding and the availability of field-relevant databases, research in the five areas listed here offers especially great potential for yielding useful results. In the case of the first area- flow and transport in fractured and cavernous media and multiple-phase flow-the potential benefit is very high because these types of flow situations have a relatively widespread occur- rence, have a strong impact on the movement of large masses of contaminants, and have not been adequately documented, resulting in an utter lack of any reliable databases. The second area- the role of bacteria in the transport and removal of contaminants-is critical because of the increasing recognition that bacteria are present in the subsurface, that most organic and some inorganic contaminants are biotransformed, and that bioremecliation offers a potentially econom- ical in situ cleanup technique. The third area the role of modeling in decisionmaking, including legal and social interactions-must be understood if the courts, enforcement agencies, industries, and the affected public are to obtain the benefits of modeling. The last two areas characterization through welI-controlled field and labo- ratory studies and development of new approaches for parameter estimation and new measurement techniques are essential if fate, transport, and remediation are to be measured in the subsurface, which is otherwise not easily accessible to human observation.
