-€&. -8 3"- * ^ 8 .^t!%. I UNITED STATES ENVIRONMENTAL PROTECTION AGENCY 3 c WASHINGTON, D.C. 20450 July 31, 1985 HOH. Lee M« ThomaS THE ADMINISTRATOR Administrator U. S* Environmental Protection Agency 401 M Street, S. ¥. Washington, D.C. 20460 Dear Mr. Thomas: In late November , 1984 , the Science Advisory Board was asked to review the technical basis for the development of a "decision rule" for determining whe- ther or not specific hazardous wastes should be restricted from land disposal. This review was assigned to the Environmental Engineering Committee* In the course of Its review, the Committee examined two proposed approaches to developing the decision rule, one proposed by the Office of Solid Waste, and the other by the Office of Policy Analysis in OPPE. We have already sent you our report on the OSW version, and are pleased to now forward our review of the One proposed by OF A. The Committee agrees that the QPA approach, because of its complexity and data-intensiveness, will not be applicable to all waste-banning decisions. The approach should be useful, however, on a waste- and site-specific basis for comprehensive comparisons of the risks of alternative hazardous waste disposal options* The Committee has been particularly pleased with the cooperation extended by the OPA staff, and we are pleased to note that they have already taken steps to implement some of the Committee's reeomftendations • If you have any questions, or should you wish any further action on our part, please call on -us.- " "••' ' " "•' ' • ;< -~ '-- ••• ~ • -" • -~ - Sincerely, Raymond C. Loehr Chairman, Environmental Engineering Committee Science Advisory Board cc: R. Morgeastern S* Napolitano A. Fisher J, Briskin A. Carson Norton Nelson S. Brorom Chairman, Executive Committee T. Yosie Science Advisory Board ------- REPORT of the ENVIRONMENTAL ENGINEERING COMMITTEE SCIENCE ADVISORY BOARD U. S. ENVIRONMENTAL PROTECTION AGENCY on their review of "COMPARISON OF RISKS AND COSTS OF HAZARDOUS WASTE ALTERNATIVES; METHODS DEVELOPMENT AND PILOT STUDIES" July, 1985 ------- I. EXECUTIVE SUMMARY The Committee finds that the QPPE method of comparative risk analysis has scientific and technical merit and can provide useful information to decision- makers if further developed. However, we do not find the method per se useful for the Nationwide waste banning decision on several hundred chemicals. It is useful on a waste- and site-specific basis in that a. decision can be based on a comprehensive appraisal of comparative risks of alternatives* The method may not have generic applicability* The method is based on the scenario approach, the selection of a specific set of sub-models required, and the output fora, as characterized by comparative risks and costs among the chosen scenarios. Its generality depends on how representative the scenarios can be made* An advantage of the method is that it provides for an explicit statement of uncertainties, if the uncertainties of the component parameters and models are known or estimated* . The choice of model components and the linking mechanisms to arrive at the complete model concerns the Committee in the following ways: (1) While some suggested sub-models are tested and accepted, others are not now verified and may not in practice be verifiable; (2) The data base for some of the models needs careful analysis, for both quality and quantity* Selection of para- meter values based on quality peer-reviewed research is essential to avoid misleading results. The health-effects section of the model» as with other similar models, suffers from the data-base problems already described. In addition, however, the Committee has concerns about the methods used. Among these are the use of a non-threshold model (which introduces problems when considering chemi- cals which may have threshold effects)j the ignoring of pharmacokinetic effects and compound interactions; and inadequate lexicological evaluation and extrapolation techniques, especially simplistic temporal, route~to-route and species-to-species extrapolations. The Committee notes that the modular nature of the model does not restrict it to the use of a non-threshold ap- proach. The overall QPPE method needs upgrading in the area of surface drainage modeling and most importantly in the risk assessments related ^to the handling and transport of wastes with respect to fugitive emissions and probability of" leakage and spills. Finally, the model makes no provision for evaluating non-human environ- mental effects except for a "qualitative" evaluation. Howevert we are in- formed by OPPE that improvements are being made. It is important to note that QPPE has responded at length to many of the comments and concerns expressed by the Committee in written summaries and in discussions with the Committee, and is studying ways of improving the method. The Committee commends OPPS for undertaking this major piece of research and encourages further work. The basic idea, if the concerns ex- ------- pressed can be taken into proper account. Is sound for Identifying the compar- stive risks of hazardous waste disposal options* With Agency policy interest in risk assessment strong, a properly developed method will be of real value. II. INTRODUCTION AMD HISTORY At a meeting of the Environmental Engineering Committee on August 16, 1984, Mr. Alan Gorgon, Office of Solid Wastes, briefed the Committee on the development of a decision rule for restricting certain hazardous wastes from land disposal as governed by the proposed amendments to the Resource Conser- vation and Recovery Act. Two main approaches Were under way: One incorpor- ating a simplified predictive modeling approach (referred to as the OSW model), the other a more complex modeling framework based upon comparative risk assessment (referred to as the QPPE model). In response to an Agency request for review of these approaches, a Sub- committee chaired by Dr. J. William Haun was appointed to conduct the review on aa accelerated time schedule. The Subcommittee was assisted by several consultants (for a full list of the Subcommittee, see Appendix A). The full Environmental Engineering Committee completed a report to the Agency on its review of the OSW approach in April, 1985. By letter dated January 7, 1985, and at a meeting of the Waste Banning Subcommittee (denoted above), on January 31, 1985, Die. Richard D. Morgen- stern, Director, Office .of Policy Analysis, OPPE, presented the draft Final Report on "Comparison of Risks and Costs of Hazardous Waste Alternatives: Methods Development and Pilot Studies'* (EPA Prime Contract No. 68-01-6558, Subcontract No. 130.155, Work Assignment No. 24), which forms the basis for the OPPE model. Dr. Nicholas Nichols, Dr. Ann Fisher, Ms. Jeanne Briskin and several contractor representatives also provided the Subcommittee with details of the method and background. The Subcommittee again, met on February 25 and February 27 with Dr. Fisher and other members of the OPPE project team. As a result of this activity, responding to the urgent need of Agency staff for a preliminary response, a Letter Report on the review to date was issued by pr. Terry Yosie, Sjtaff Director, SA|i} .on behalf of the Environmental Engineering Committee on March 8, 1985. This report constitutes the detailed basis for that Letter Report. The scope of the review as originally suggested by OPPE, which focused primarily on the reasonableness of the sub-models and their integration, is shown in Appendix B. In addition, based on the early discussions, Dr. Loehr developed a more general list of issues for the Subcommittee, This list is shown in Appendix C. Both Lists were used by the Subcommittee in its conduct of the review. ------- -3- III. DESCRIPTION AND EVALUATION OF THE MODEL As a result of the recent (11/84) amendments to the Resource Conserva- tion and Recovery Act (ROM), EPA is required to establish which wastes, of those specified in the Act, are not to be banned from disposal in certain land disposal facilities* It is believed that the required rule must be "generic," i.e., of National scope and not of a site specific nature. The research described in the draft study report was designed toi (a) "test the viability of comparative risk assessment for hazardous waste man- agement alternatives," and (b) "serve as a basis for making land disposal prohibition decisions for hazardous waste streams" (p. 1-2). The study contributed to the development and demonstration of comparative risk assess- ment methods by using a pilot study approach. The model proposed is utilized for a specific waste by selecting a num- ber of possible disposal technologies (scenarios) considered appropriate for the particular waste. For each waste and waste treatment scenario, existing models are used in combination to estimate waste releases, environmental transport of the released components of concernt and to identify the poten- tial population exposed, and estimated doses to exposed individuals in that population. Further, the model then develops dose-response relationships for each waste component based on the best literature data available and, from this estimates human health risks by combining the exposure and dose-response information. Finally, the model is used to qualitatively evaluate ecological Impacts of the selected scenarios. Using estimates of uncertainties in the human health risk estimates, an explicit estimate of the uncertainty of each overall estimate is made to permit decision-makers to take these ranges into account. The Pilot Study considers for Illustration three wastes, with four or five scenarios lot management of each. While its potential utility and po- tential versatility were reasonably well represented in the pilot study, the method as presented represents m still-prelimiiwty approach. IV. EVALUATION OF'THE METHOD'AND'ASSUMPTIONS '•' ''' "' ""' '- ' ' A. General Comments Over the next five years, the Agency must determine which, of as many as 450, wastes or waste streams are to be evaluated to determine if they should not be "banned" from landfills. The OPPE states that the approach identified in the draft study report may be able to consider from 20 to 40 wastes or perhaps 5 to 10 percent of the wastes that may have to be evaluated. The implementation of this approach even to a small number of wastes will require significant effort Involving extensive data-gathering and evaluation as well as significant judgmental evaluation of input and results. The effort to apply the QPPE method to selected wastes could be an excellent investment if it prevented suboptimal decisions that Increase risks to human health or to the environment. ------- Tha approach makes an important contribution by attempting to esti- mate the relative risks in all media, since, if the land disposal option is banned, the wastes still have to be handled and disposed of in seme manner. The approach can provide information concerning: (a) pretreat- tnent alternatives that can be considered before land disposal, (b) the relative impact of other disposal alternatives on protection of huoan health and the environment, (c) relative costs involved, and (d) data and research needs that can reduce the uncertainties involved in esti- mating the relative risks, It will readily be appreciated that this approach is necessarily data- and resource-intensive. Analysis of each scenario for each waste requires detailed knowledge of the available technologies of disposal and detailed knowledge of the existence and use of many submodels: fugitive emissions from landfills? solute transport in groundwater? dispersion models for air transport? dose-response and health affects, and many others. OPPE is to be commended for undertaking such an important evalua- tion, for its early judgment to support such a detailed study, and for its wisdon of continued support for the study. The study clearly has had good intellectual input, the individual cernponents appear sound, and reasonable estimates of potential health risks appear to have been ob- tained. The study also has considerable fallout value. Even if rot applied solely to the banning, decision, the technique developed will be useful in many other situations. The task of risk assessment is to make the most credible possible statements about definable relationships, reducing uncertainty, and making explicit whatever uncertainty remains. Ihis study accomplished these goals. However, in terms of EPA needs relative to the waste banning deci- sions, it does not appear that the study approach can serve as the sole basis for the final decisions. The Study approach can be used with one or more other approaches or methods to provide a broader perspective on those major wastes that may be banned from land disposal in order to protect human health and the envirorment. Such major wastes could be . those .that-are o.f large volume,' are of unique .characteristics, may have • an apparent adverse economic impact on an industry if banned fran land disposal, and/or appear likely to cause a potential adverse human health and environmental impact if disposed of in another way. Other approaches may be able to more quickly evaluate a larger number of wastes and iden- tify those for which more detailed evaluation is needed. The study approach would provide the Mninistrator with a richer array of information on relative risks, intemedia transfer or costs when making decisions about which wastes should not be banned from land disposal. ------- -5- B. Components o£ the Model and Model Linking I, Randan Walk Solute Transport Model The Random Walk model was selected to predict two-dimensional contaminant transport in groundwater aquifers, The horizontal flow field in the aquifer is computed using a finite difference tech- nique. Solute transport is calculated using a population balance technique in which many particles are released and their fate simu- lated. The adveetive transport for each particle follows the flow field ard dispersion is incorporated by randan displacements. As the number of particles released becomes large, the spatial distri- bution of particles corresponds to the concentration profile of the constituent. The technique produces a solute concentration profile which should in theory (as the manber of particles becomes large) be identical to that obtained with more traditional, two- dimensional advective-dispersive models .using finite difference or finite element solution methods. Linear equilibrium adsorption is incorporated through the use of a retardation coefficient, and reaction is represented with a generalized half-life (i.e., multi- ply results by Ihe Random Walk model has been teed in many applications and is a well recognized tool, for example, through its use as part of the program of the Bolccmb Qroundwater Research Institute* This pro- vides confidence and credibility for its use. There are, however, limitations to the model that should be recognized, particularly as regards its treatment of chemical transformations of contaminants in the soil . A detailed review of Random Walk and other groundwater transport models was performed as part of a study sponsored by the Electric Rjwer Research institute (Kincaid and Morrey, 1984). A sum-nary of this Rardcm Walk review is included as Appendix D, Based on the EPRI analysis and a review of the Handera Walk Model by the Comdttee's consultants, there is a major area of concern about the model's applicability. The model allows for a detailed char- acterization of two-dimensional flow profiles; spatial heterogeneity in hydraulic conductivity,, storatiyity, etc. ,, As mentioned. in ,. ,^ . the QPPE presentation, the model allows incorporation of pumping- remedial action, which is useful for its intended applications. Similarly, a careful representation of dispersion is incorporated. As such, transport mechanisms are well represented, and the model is very appropriate for predicting the fate of "solutes11..* consti- tuents which undergo no chemical transformation. Ihe reaction and adsorption caoponents of the model, however, -are much more limited* In particular: a. In the Random Walk model, both the reaction rate and the adsorption (retardation) coefficient are constant over the agaifer study area. Heterogeneity in soil conditions which might affect these factors is not considered. ------- -6- b. Unlike other numerical models where the chemical inter- actions are formulated as part of the finite difference or finite element equations, the population balance tech- nique used in Random Walk is expressly designed for the case of linear, equilibrium adsorption and first order decay. The Random Walk model would require significant modifications to make it applicable to more complex chemi- cal conditions. While nonlinear, non-equilibrium adsorp- tion and higher order kinetics are not commonly incorpor- ated in applied groundwatei: models today, they may be .. used in the future as our scientific understanding advan- ces. It will be very'difficult to incorporate these advances in the Random Walk model. The limited representation of chemical processes relative to the detail given to transport processes should be recognised. It may reduce the applicability of the Random Walk model for certain kinds of problems in certain locations, particularly when chemical processes.are non-ideal. The level of chemical representation is no better than that provided by the OSW model (and possibly worse, depending on the resolution of item b above). This limitation should be recognized, To summarize, the use of the Random Walk model is acceptable, with the limitations noted. In addition, it should be noted that there are a number of other numerical codes which can simulate two- dimensional adveetive—dispersive transport. Those models utilize more traditional solutions of the material balance equation at a grid point or cell, and like the Random Walk model, can allow for non^homogeneouS flow conditions, pumping wells, etc* Some users may be more familiar with the conceptual basis for these models, arri they may be easier to adapt to situations where the use of more complex chemistry is appropriate. As such, alternative numerical models should be considered in future applications. A final consideration applicable to the use of any numerical groundwater mode 1,_ regards the limited lev^l of validation, j?arti- cuiarly for complex field conditions where constituents undergo chemical transformation. Successful attempts to verify models in the field have been made in recent years, though validation remains difficult and expensive. Some degree of field calibration and verification is recommended. 2. Modeling of Unsaturated 2one Transport In the analysis presented in the QPPE report, the Mcwhorter- Nelson model is used as a basis for modeling transport in the unsaturated Zone* The McWhorter-Nelson model, however, computes only a water recharge rate - no contaminant transport mechanisms are included. Contaminant transport is calculated in the OPPE ------- examples by considering only the hydraulic residence time in the unsaturated zone associated with the computed recharge rate. Shis results in a step break-through profile, with no consideration for the effects of dispersion, adsorption (retardation), or reaction. Significant adsorption or reaction may considerably alter the pollutant washout profile from the unsaturated zone. An analysis which ignores these processes is not consistent with either the current state-of-the-art of unsaturated zone modeling, or the level of sophistication used in other components of the risk assessment* &s noted in the OPPE report and the supporting MRl documents, there are models available for the unsatnrated zone, such as the analy- tical PESTAN model, which incorporate dispersion, retardation, and decay. These should be utilized to generate more realistic esti- mates of the temporal breakthrough profile from the unsaturated zone. 3. Atmospheric Transport/Dispersion Models Air pollution impacts are simulated 'using Gaussian plume mo- dels incorporating wind speed and direction, transverse and verti- cal diffusion (as a function of atmospheric stability class), terrain adjustment in certain eases, and plume-depletion and par- ticle deposition processes. The selection of the Industrial Source Complex (ISC) Long-Term Model for area sources and the ATM model for point sources appears to be based on a careful and credible review of the current state-of-the-art of air modeling, and a full consideration of the capabilities of existing models. The ability to link the ATM model to population exposure estimates through the Graphic Exposure Modeling System (GEMS) is particularly beneficial. it i$ worthwhile to note that long-term average concentration pro- files are sought (rather than short-term "event" concentrations), therefore, long* term versions of the models are utilized. The long-term versions use integrated forms of the Gaussian plume model based on the joint frequency distribution of wind speed, stability class, and wind direction. Although the models. .selected represent in Gaussian plume modeling, there has been concern expressed among SAB members that the level of validation for this class of models has been limited. 4. Uncertainty Analysis The propagatiOn«of -error technique for evaluating uncertainty is formulated on the basis that links between model components occur in a multiplicative fashion. The assessment may effectively be represented by an equivalent, simplified model of the form: Risk = Pollutant x Transport x Exposure x Response x Health Effects Release Factor Dose Factor ------- Uncertainties in each component are assumed to be independent and are routed logarithmically. The method is correctly developed and implemented, assuming the multiplicative assumption is adequate. The technique has a number of desirable features, including its simplicity and direct use with "order^of-taagnitude" judgments of uncertainty bounds. There may, however, be a need for a more care- ful consideration of the implications and limitations of the multi- plicative assumption. This discussion is relevant as well to the integration issue (Appendix B, 2; Have the models been integrated (combined) without violating scientific principles? Is the integra- tion consistent with the state-of-the-art?}* Two issues taay be raised to illustrate possible difficulties with the multiplicative assumption for model linking. The first arises If and when thresholds are incorporated in the dose—response functions for health effects* A more sophisticated uncertainty routing procedure would then be required to account for the proba- bility of zero impact (e.g.» below threshold). The second issue relates to the temporal aspects of the analysis resulting from the stochastic nature of pollutant release in the Pope^Reid landfill liner failure model. (We were asked not to review the Pope-Reid model itself, but the incorporation of the model in the overall framework is important.) In the OPFS report, the results of repli- cations of the Pope-Reid model are averaged to- obtain a nominal temporal profile of pollutant release. It is then assumed that the use of a multiplicative uncertainty factor can capture the full range of uncertainty in both the amount of pollutant released and its temporal distribution. The validity of this assumption is not intuitively obvious, and needs to be demonstrated with a more detailed set of example simulations. In particular, it would be useful to evaluate transport simulations for each of the Pope-Reid replication outputs. The resulting "exact" distribution of concen- tration-exposure can then be compared to the lognormal distribution derived from the multiplicative assumption. This analysis is com- putationally intensive and should not be performed for all cases. Rather, the ^comparison should be demonstrated once to evaluate the adequacy of the simplified integration assumption; to build confi- dence in its use (or provide guidance for a better alternative). The use of "one standard deviation" in the uncertainty analy- sis results in an 84 percentile concentration. This is fine so long as the Scenarios are considered only in a comparative sense. If, however, the absolute level of impact is also evaluated (as is apparently the case from the QPPE report), then 84% seems too low for an "upper limit." The OPPE has indicated that it concurs with this suggestion and intends to use a wider confidence interval. ------- -9- C. Toxicologies! Risk Assessment Conceptually, the methods of health risk assessment as outlined In the OPPE report are appropriate tools. Quantitative risk assessment and quantitative uncertainty analyses are both desirable approaches. The dose/response asses$«nt proposed by OPPE is innovative in that it estimates dose response functions for a spectra of adverse effects. Most other approaches used by the Agency either estimate (Jose/response functions for only a single effect, usually cancer» or are restricted to the estimation of an acceptable exposure limit (e.g.» ADI, ambient water quality criteria, drinking water criteria)* For the purposes of comparing risks with costs of hazardous waste alternatives, the esti- mation of dose/response functions for all significant effects should be encouraged if it leads to more fully and clearly using the available toxicity data* With several significant modifications, the OPPE ap- proach could serve as a useful decision-making tool. As currently written, however, it has some serious flaws and could mislead rather than assist the decision-maker* Concerns related to the OPP1 methodology include: The use of a. non-threshold model for all effects, the use of maximum likelihood esti- mates, and simplistic temporal, route~to-rotite, and species-to-species extrapolations* In addition, several areas in the methodology and application of the methodology require clarification. These include: the rationale for combining effects (i.e., independent vs. graded series); how quantitative estimates of uncertainty are made (matheraatic or judgmental) as well as the validity of such estimates; details of how effects on which incidence data are not available will be handled in the risk assessment (in the case studies, such effects are ignored), and how data on pharnacoklnetlcs and compound interactions will be used (in the case studies, such data are not considered). These concerns have been discussed with OPPE personnel and their contractors. OPPE has indicated a willingness to alter their approach to constructively address these issues. D. Fugitive Emissions, Leaks and Spills The OPPE study attempts to address the risks from production of fugitive emissions, transportation over interstate highways and atmos- pheric emissions from capped landfills. Each of the primary refer- ences Is based on a minimum of information. Indeed, OPPE shares In these SAB concerns (stated in a follow-up letter to the EEC). In addition, as off-site landfill and deep-well injection alternatives force more chemical wastes to be stored, transported and re-stored before ultimate disposal, the probability of risks will be magnified over previous experience. The inclusion of small-generator wastes, all of which will have to be transported, will further exacerbate the problem. The OPPE methodology needs a major effort to gather the information to adequately address these issues and their ramifications. ------- -10- E. Application of the Model The Committee is concerned that many of the specialized models (ot submodels) required to apply the method may not be adequately verified or even verifiable. An example is the estimation of exposures resulting from handling of wastes in transportation, which both the Pilot Study and one's intuition would indicate as a major route of population expo- sures. One advantage of the OPPE framework is that its modular nature permits the substitution of improved models and data as they become available. ------- APPENDIX A U.S. ENVIRONMENTS PROTECTION AGENCY • SCIENCE ADVISORY BOARD ENVIRONMENTAL ENGINEERING SUBCOMMITTEE WASTE BANNING SUBCOMMITTEE CHAIRMAN Dr. J. William Haun Engineering Policy (4SW) General Mills, Incorporated P. O. Box 1113 Minneapolis, MN 55440 MEMBERS Mr. Richard A. Conway Corporate Development Fellow Union Carbide Corporation P. 0, Box 8361 (770/3423 South Charleston, WV 25303 Dr. Benjamin C. Dysart, III Environmental Systems Engineering Department 401 Rhode? Engineering Research Center Clemson University Clemson, SC 29631 Mr. George Green Public Service Company of Colorado P.O. Box 840, Room 820 Denver, CO 80202 Dr. Joseph T. Ling 3M Community Service Executive Office Minnesota Mutual Life Center 6th s Robert llth Floor St. Paul, MN ' 55144 Dr. Raymond C. Loehr Civil Engineering Department 8.614 ECJ Hall University of Tfexas Austin, TX 78712 Dr, Donald J. O'Connor Professor of Environmental Engineering Environmental Engineering Science Program Manhattan College Manhattan College Parkway Bronx, NY 10471 ------- CONSULTANTS D£, Marc Anderson Water Chemistry 660 North Park Street University of Wisconsin Madison, WI 53706 Dr. Paul £. Brubaker, Jrt Paul E. Brubaker Associates, Inc 3 Halstecwl Road Mendham, NJ 07945 Mr. Allen Cywin Consultant 1126 Areturus Lane Alexandria* VA 22308 Dr. Patrick R, Syracuse Research Corporation Merrill Lane Syracuse, NY 13210 Dr. Charles F. Reinharctt Haskell Laboratory for Toxicology & Industrial Medicine I. I* du Pont <3e Nemours and Company Elkton Road Newark, DE 19711 Dr. Mitchell Small Department of Civil Engineering Carnegie-Mellon University Schenley Park Pittsburgh, PA 15213 Dr. Leonard Greenfield 6721 Southwest 69 Terrace South Miami, Florida 33143 , . ... Executive Secretary Mr. Harry C. Torno Executive Secretary, EEC O.S. Environmental Protection Agency Science Advisory Board (A-101 F) Washington, D.C, 20460 ------- APPENDIX B jjuestions for _SAB 1. Because of the many steps in the health risk assessment process, several models were used. Is the tige of each of the following reasonable? a. The Farmer et al. equation for fugitive emissions from land- fills? b. The Random Walk Solute Transport model for ground-water movement; c. The Industrial Source Coaplex Bong-Term model for dispersion of air emissions from area d. The ATM component of GEMS for dispersion of air emissions from point sources; s. The multistage model, with the one-hit model as a backup when data are limited, for dose-response functions; f. The Carcinogen Assessment Group potency factors, Wiebull model, and modified acceptable "daily-intake approach as sensitivity checks for the dose-response function selected; and g. The propagation-of -errors approach for evaluating uncertainty in the health-risk estimates. 2. Have the models been integrated without violating scientific princi- ples? Is the integration consistent with the state of the art? 3. How can this approach be improved to better estimate risks for each management strategy for a given hazardous waste stream? ------- APPENDIX C QUESTIONS RAISED BY DR. RAYMOND C, LOEHR on Proposed OPP! Method Following are basic questions to which the Subcommittee can respond or comments 1 . Specific questions related to the specific model/approach with respect to its scientific, fundamental credibility (basically, is the approach scientifically sound}. 2. Is this model/approach likely to address the important questions facing , i.e. will the correct need be Generally, the SAB is asked to respond to type (1) questions. Shis generally skirts the real basic issue and we should attempt to address the type (2) ques tions < Therefore, in addition to the questions that have been placed before the Subcommittee* the Subcommittee should also consider addressing the following questions : 1 » TO what extent does the Subcommittee feel that this approach can be used for the banning decision — i.e., from the scientific or engi- neering basis and not from the policy aspects? 2. Are the'nodels that are proposed to be used the appropriate ones for the intended use? Have they been adequately peer-reviewed and verified by independent data? 3. Is the data base to be used adecuate from the standpoint of accu- racy, OA/QC, etc*? Is there sufficient data that, can be used with this approach? 4. Are there adeouate other models that can be used for other land disposal approaches, such as land treatment, waste piles, surface impoundments and all land disposal approaches listed in the RCRA amendments, i.e., really address whether the model/aproach can be used for other land disposal methods, besides landfilj.? It seems that these types of questions also should be addressed by the Sub- committee. If the Subcommittee decides to address some of the above issues/ then at a future meeting it would be helpful if OPPE could address the following questions (or provide detailed discussion on these items) - it would be helpful to hear their explicit thoughts on these subjects for our consi- derations 1 . How would this approach be useful for the "banning" decision? 2, To what extent is it possible for the OPPE approach to be used for a generic situation, rather than on a waste~specific/site-$pecific situation? ------- -2- 3. To what extent have the models been verified, cheeked or peer- reviewed? Details should be provided on models such as the Pope-Reid model, or the Random-walk model, as used in their approach, 4. If the Subcommittee is to review/comment on applicability of certain models (see questions asked), then detailed information sbout the models needs to be provided to the Subcommittee. ------- APPENDIX D TRANS - A Random Walk Solute Transport Model for Selected Grouiidwater Quality Evaluations described in GEOHYDRQCHEM1CM, MODELS FOR SOLOTE MIGRATION Volume 2: Preliminary Evaluation of selected Computer Codes- EA-3417, VOluiae 2 Research Projects 2485-2r 1619-1 Final Report, November 1984 prepared by Battelle, Pacific Northwest Laboratories Battalia Boulevard Richland, Washington 99352 prepared for Electric Power Research Institute 3412 Hillview Aventte Palo, Alto, California 94304 ------- CODE: SPONSOR: AUTHORS: PROCESS-AND INTERACTIONS: OPERATIONAL ASPECTS; TRANS - A Random Walk*Solute Transport Model for Selected Groundwater Quality Evaluations Illinois Department of Energy and Natural Resources State Water Survey Division ' • Champaign, IL S1801 Thomas A. Prickett Thomas A. Prickett and Associates 8 Monte!air Road " . Urbana, IL 61801 " • Thomas S. Naymik, and Carl G. Lonnquist , • Illinois Water Survey .' ' ' Champaign, IL 61801 The processes .and interactions addressed by this code are: * Saturated groundwater flow in a singled confined or unconfined aquifer where water'flow is typically horizontal. ' [The code addresses temporal variations in two-dimensional (x-y) flow for a variety of" boundary conditions and arbitrary x-y geometry.] * Adygctio_n. of a chemical contaminant in a saturated groundwater system released from a variety of typical sources. * Hydrodynamic Dispersion-(both lateral and transverse) and diffusion of a chemical contaminant ;n a saturated groundwater system. - - * Retardation of a chemical contaminant when it can be characterized by i constant IC^ and the assumptions of instantaneous and reversible adsorption ire adequate. * Radioactive decay of a "-chemical contaminant. There, are two :m in parts to _the_.TRANS code: 7 flow calculations and..- : transport calculations. Provisions for aqyifer flow {potential or head) calculations are per- formed in four ways. Two methods {subroutines HSOLY2 and HSOLV4) compute heid distributions for simple, analytical problems. The third method is through the K50LYE subroutine, which 1s a subroutine fona' of the Prickett and Lonnquist* flow model. This model 1s a.well- documented, finite difference, groundwater flow model for simulating transient or steady-state groundwater flows in a water table or leaky confined aquifer. The fourth method supplies the aquifer's head dis- tribution through a user-supplied program. Any other acceptable method or model can be used as long as head values are supplied for the same finite difference §rid used in TRANS and for the same * Prickett, T. A. and C. G. Lonnquist, "Selected Digital Computer Techniques for Groundwater Resource Evaluation." Illinois State Water Survey Bulletin 55, 1971. A-67 ------- hydraulic conductivity and effective porosity distributions supplied to TRANS. Velocity at every finite difference grid Is calculated by: V = KI/{7.48 n) • ,(A-4)- where V - interstitial velocity K * hydraulic conductivity I » hydraulic gradient n - effective porosi.ty. Velocity at my other position in the system is Interpolate using Chapeau-basis functions and the values at the finite difference grid points. Stability requirements for the flow portion of the model depend on the method chosen to develop the head (and subsequently the velocity) distribution. HSOLV2 and HSQLV4 are analytical solutions* but one must still ensure that adequate spatial, sampling of these analytical solutions has been selected. HSOLVE uses a finite dif- ference numerical scheme and a modified interative alternating direc- tion method, MlADl, to solve for head distribution to a specified - level of convergence at each timestep. Adequacy of a spatial and temporal time spacing can be-checked in the same manner as for any finite difference or finite element scheme, by reducing grid spacing or timesteps and comparing results* The transport model portion of TRAHS uses a direct simulation tech- nique. The concentration of a chemical constituent in a groundwater system Is assumed to-be represented by a finite number of discrete particles. Each of these particles Is moved according to-the advec- tive velocity and dispersed according to random walk- theory. The miss assigned to each particle represents a fraction of the total mass of chemical constituents involved* In the limit,"as the number of particle approaches the molecular level* an exact solution to the actual situation is obtained. This kind of transport model Is inherently mass conservative* Convergence can be checked by increas- ing the number of particles. There are restrictions, as with any numerical method, which limit the size of "timestep that can be taken for both a time-dependent and spatially dependent problem* Tiroesteps for particles are limited such that advective plus dispersive move- ment 1s no greater than the spacing between velocity (head) nodes. APPLICABILITY , • ASPECTS: The TRANS model allows 'the user to investigate groundwater pollution problems from a vertically averaged viewpoint for contamiants injected into wells, leaching fom landfills or arising from surface- water sources such as ponds, lakes, and rivers. The documentation for the TRANS program illustrates comparisons with theory for six problems; 1. Divergent flow from an injection well in in infinite aquifer without dispersion or dilution 2, Pumping from a well near a line source of contami- nated water, with dilution but without dispersion 3, Longitudinal dispersion in a uniform one-dimensional flew with continuous injection at X = 0 A-68 ------- SECONDARY CONSIDERA- TIONS; - 4. Longitudinal dispersion in uniform one-dimensional • flow with a slug tracer injected at X = 0 5, Longitudinal dispersion in a radial 'flow system produced by an injection well 6. Longitudinal and transverse dispersion in a uniform one-dimensional flow with a slug of tracer injected at X = 0 In addition, the documentation illustrates the use of the model for a real field-scale contaminant problem at Merodosia, Illinois. Purpose and Scope The purpose of this code fTRANS), as stated by- the authors, is to provide a generalized computer code that can simulate a large class . of problems Involving convection and dispersion of chemical contami- nants associated with fertilizer applications, hazardous waste leaeh- ate from 1andf111ed and'Other sources, and injection of chemical waste into the subsurface using disposal wells. TRANS does not address density-induced convection. Concentration distribution in the aquifer represents a- vertically averaged value over the saturated thickness of the aquifer. TRANS addresses only a single aquifer. Spatial and temporal distri- bution of head in the aquifer can be calculated by four methods; 1. Analytic (HSOLV2) solutisi for a uniform 1-ft/d flow in the x direction ' ' 2. Analytic (HSOLY4) solution to the Tneis formula cen- tered at node (IS, 15) 3. Numerical finite difference solutionCHSOLVE) to the two-dimensional (x-y) vertically averaged groundwater • ' flow equation* {this solution is for transient or steady-state flow) 4* User-supplied subroutine for reading or calculating head"on the finite-difference grid used in the TRANS transport model. The TRANS code was designed to solve real pollution problems and to address only single contaminants- TRAMS can also handle (with slight changes in subroutine calls! radioactive decay and chemical retarda- tion. The modifications required for decay are clear but those required for retardation are unclear. The code can handle contami- nant source or sinks at every node as well as a variety of special sources, which include points, rectangles, circles, and lines. The flow model can handle impermeable boundaries {no flow), leaky * Priekett, T. A., T. G- Namik, and C. 5- Lonhnquist, "A Random-Walk Solute Transport Model for Selectee! Groundwater Quality Evaluations." Illinois State water Survey Bulletin 65, 1981. A-S9 ------- artesian source, induced Infiltration {I.e.,, streams, lakes, and rivers), held-head boundary conditions, flow from springs, and evapo- transpiration from the water table. Operational Characteristics^ . The TRANS code is written in FORTRAN and run by the authors on a CDC CYBER-175*" The code, for purposes of this study, was run on a Dig1- . tal Equipment Corporation VAX 11/780, Other, than.changes to the pro- gram header, logical' unit checker, and formatted character strings (which were §11 CDC-specific practices); conversion to the VAX required access to a system specific random number generator; A similar but not equivalent-random number scheme was implemented. The authors indicate that, the test problems Included in the documentation examples took no'more than a few seconds of CPU time* including com- piling and loading on their CDC CYBER-175. TRAKS used 140,800 bytes of virtual memory on the VAX; 10,363" central processing seconds were needed to perform the test problem simulation.. The code, as docu- mented, is dimensioned for.a 29 x 30 finite difference grid and 5000 particles. These dimensions can be changed, however, to accom- modate larger problems.. Some difficulties may be encountered because Instructions for increasing dimensions-are not specifically discussed in the documentation. Input Requirements Input reqyresents for,the:code are explained with both appropriate text and in pictorial form (Figures 9, 10t 11, and 12 of Prickett et al,*). Input requirements for the code are those typically avail- able from standard field or laboratory measurements. For the flow portion of the model they include: / • A variable finite difference grid description » Timestep and number of timesteps to be run » Areal distributions of —Permeability —Source aquifer potential for leaky artesian simulations —Aquifer bottom elevations . —Aquifer top elevations —Head ,{Initial conditions) —Aquitard thickness .and .permeability for leaky- .... _. - . artesian aquifers " —Simulations —Artesian and water table storige coefficients » Pumping and recharge well locations and temporal rites * Stream {river.or lake) node locations, surface-water elevations, stream or lake bed thickness and pensea-. bility* fraction of node area available for transfer * Constant head node locations and elevation for held head * Prickett,, T, A.» T, G. Naiwik, and C. S. Lonnquist. "A Random-Walk Solute Transport Model for Selected Groundwater Quality Evaluations." Illinois State water Survey Bulletin 65, 1981. A-70 ------- * Locations of springs, elevation at which spring flow- begins» and slope of the spring flow versus ground- water head for the spring production line » Locations of nodes where evapotranspiration from the water table is to be considered and the slope of the fate versus head Hne and the water-table elevation at which evapotranspiratIon effects are to be ignored. FOP the transport model, additional .Input requirements include: Longitudinal dlspersivity Lateral dispersivity Effective.porosity Actual porosity Retardation factor or K^ Bulk mass density of.porous medium Location and concentration of sources, description of source geometry, and1 selection- of method for release of particles • Sink locations arid grouping of sink locations for sujwnarizing: outflow versus time results. The model contains no checking of input for consistency and automatic termination for faulty or inconsistent inputs. Output Results ' • ' - x. Results ire printed in1 a~'132-eharacter format w.ith a concise and readable output layout. The code echos input parameters and produces line printer plots of head, numbers "of particles, and concentrations. The code also reports the concentration of water entering • sink nodes and groups of sink nodes versus time. The code produces no contour maps or output fields that can be passed on to other computer system programs for plotting and produces no mass balance summaries for water flow or transport, Numerical Approximations The general flow problem solution available with this code (HSQL-VE) solves for head distribution in a leaky artesian or water-tabT_e aquifer for a heterogeneous, anisotropie -porous medium with irregular boundaries. The ac*-«al system" is approximated by finite difference methods based on a block-centered, variable finite difference grid, Hedium properties in each grid block are assumed to be uniform. Approximation of the partial differential equation at each grid block by finite difference methods results In N equations 1n N unknown, where N is the number of grid blocks representing the aquifer.. Time derivatives are estimated by an implicit finite difference method* ' and the water-table problem that results in nonlinear equations is solved by the modified iterative alternating direction implicit (MIAO!) equation-solving method. Once head distribution at the N grid blocks is obtained, the X and Y direction velocity midway'between each grid block is calculated by . (A-5) A-71 ------- where V = interstitial velocity K. = hydraulic conductivity 1 = hydraulic gradient . " t n - effective porosity, " • The midgrid-block X and Y direction velocities are then used in a bilinear interpolation scheme to estimate the. velocity at any arbitrary location (x,.y}~ Transport is simulated by a direct simulation technique which" involves movement of contaminant mass particles according to the con- vective velocity as interpolated from the convective velocity fie-ld- described above, "Dispersion is simulated by random walk methods. In the .limit, as the number of particles gets extremely large (i.e., when the number of particles approaches the level at .which each particle represents i molecule), an exact' solution to the actual situation is ^obtained. Assumptions and Simplification We address the flow and transport separately* The principal assump- tions regarding flow are: * Darcian flow ts assumed. » Flow in the aquifer ,1s horizontal'and controlled only by hydraulic head- gradients. « Leakage between the simulated aquifer, rivers, lakes, Other aquifers, and springs is a 14near function of head difference with the slope of this relationship determined from the leakance-parameter, K/m, where K is the permeability of the aquitard (or stream bed) and m is the thickness. » Storage in the stream, lake, or river beds and aquitards is ignored, The principal assumptions regarding contaminant transport are; » The advection-diffusion equation for solute transport is assumed valid. » Dispersion in porous media. 1s a random process, , , * " Retention of a contaminant (or retardation of a con- centration front) may be represented by an instantan- eous and reversible sorption process. Probabilistic or Statistical Aspects The code solves a deterministic problem. Aval 1able Documentation McDonald, M. G.» and W. B. Fleck. '"Model Analysis of the Impact on Sroundwater conditions of the Huskegon County Waste-Water Disposal System, Michigan." U.S. Geological Survey Open-File Report 78-79, 1978. A-72 ------- Priekett, T. A., and C. 6* Lonnquist. "Selected Digital "Computer Techniques for Groundwater Resource Evaluation." Illinois State Water Survey Bulletin 55, 1971. Prickett, T. A,, T« G. Namik, and C. G. Lonnquist, "A Randoo-Walk . Solute Transport Model for Selected Groundwater Quality Evaluations." Illinois State Water Survey Bulletin 65, 1981. Software Quality The modular code consists of a-main program, 20 subroutines, and three functions. The code listing- Is well annotated and the documen- tation report contains a complete description of each modulet along with flow diagrams-. Transfer of this program from one machine to another should be fairly easy. The code lacks any graphical output capability other than line printer plots. In addition, no routines are supplied to dump model output to disk files for use with generally available computer system plotting routines* GENERAL . CRITIQUE: The documented verification test cases we're easy to set up and repeat; however, direct checking of the results is not possible because a different random number generator is used on our computer system. The code'does not produce any mass balance summaries. *' • In order to use the generated flow option of the TRANS code, one roust obtain Bulletin 55 from the^Illinois State Water Survey, which explains the vertically averaged solution for transient or steady flow. From an application point of view, .the, TRAMS documents are difficult to follow. Examples are weak and the narrative descrip- ' tions are not straightforward. However, excellent code annotation compensates for limitations of the user's manuals. - . Most of the data required by TRANS is typical groundwater survey information. The exception is the source term for the transport simulation, which needs a parcel release rate. This rate may be dif- ficult to quantify for someone unfamiliar with 'random walk' models. TRAMS is very flexible with respect to problem configuration; thus, no modifications to-the specified geometry were necessary, There were no problems encountered while running the code. A-73 ------- |