nureg/cr-5927, vol. 1, 'evaluation of a performance ... · m. w. kozak, n. e. olague, r. r....

NUREG/CR-5927SAND91-2802Vol. 1

Evaluation of a PerformanceAssessment Methodology forLow-Level Radioactive WasteDisposal Facilities

Evaluation of Modeling Approaches

Prepared byM. W, Kozak, N. E. Olague, R. R. Rao, J. T. McCord

Sandia National LaboratoriesOperated bySandia Corporation

Prepared forU.S. Nuclear Regulatory Commission

AVAILABILITY NOTICE

Availability of Reference Materials Cited in NRC Publications

Most documents cited In NRC publications will be available from one of the following sources:

1. The NRC Public Document Room, 2120 L Street, NW. Lower Level, Washington, DC 20555-0001

2. The Superintendent of Documents, U.S. Government Printing Office, Mail Stop SSOP, Washington.DC 20402-9328

3. The National Technical Information Service, Springfield, VA 22161

Although the listing that follows represents the majority of documents cited In NRC publications, It Is notintended to be exhaustive.

Referenced documents available for inspection and copying for a fee from the NRC Public Document RoomInclude NRC correspondence and internal NRC memoranda; NRC Office of Inspection and Enforcementbulletins, circulars, information notices, inspection and Investigation notices; Licensee Event Reports; ven-dor reports and correspondence; Commission papers; and applicant and licensee documents and corre-spondence.

The following documents in the NUREG series are available for purchase from the GPO Sales Program:formal NRC staff and contractor reports, NRC-sponsored conference proceedings, and NRC booklets andbrochures. Also available are Regulatory Guides, NRC regulations In the Code of Federal Regulations, andNuclear Regulatory Commission Issuances.

Documents available from the National Technical Information Service Include NUREG series reports andtechnical reports prepared by other federal agencies and reports prepared by the Atomic Energy Commis-sion, forerunner agency to the Nuclear Regulatory Commission.

Documents available from public and special technical libraries include all open literature items, such asbooks, journal and periodical articles, and transactions. Federal Register notices, federal and state legisla-tion, and congressional reports can usually be obtained from these libraries.

Documents such as theses, dissertations, foreign reports and translations, and non-NRC conference pro-ceedings are available for purchase from the organization sponsoring the publication cited.

Single copies of NRC draft reports are available free, to the extent of supply, upon written request to theOffice of Information Resources Management, Distribution Section, U.S. Nuclear Regulatory Commission,Washington, DC 20555-0001.

Copies of Industry codes and standards used in a substantive manner In the NRC regulatory process aremaintained at the NRC Library, 7920 Norfolk Avenue, Bethesda, Maryland, and are available there for refer-ence use by the public. Codes and standards are usually copyrighted and may be purchased from theoriginating organization or, If they are American National Standards, from the American National StandardsInstitute. 1430 Broadway, New York, NY 10018.

DISCLAIMER NOTICE

This report was prepared as an account of work sponsored by an agency of the United States Government.Neither the United States Government nor any agency thereof, or any of their employees, makes any warranty,expresed or implied, or assumes any legal liability of responsibility for any third party's use, or the results ofsuch use, of any information, apparatus, product or process disclosed in this report, or represents that its useby such third party would not infringe privately owned rights.

NUREG/CR-5927SAND91-2802Vol. 1CC, CJ, CO, CY, RW

Evaluation of a PerformanceAssessment Methodology forLow-Level Radioactive WasteDisposal Facilities

Evaluation of Modeling Approaches

Manuscript Completed: July 1993Date Published: August 1993

Prepared byM. W. Kozak, N. E. Olague, R. R. Rao, J. T. McCord

Sandia National LaboratoriesAlbuquerque, NM 87185-5800

Prepared forDivision of Regulatory ApplicationsOffice of Nuclear Regulatory ResearchU.S. Nuclear Regulatory CommissionWashington, DC 20555-0001NRC FIN L1153

Abstract

This report represents an update to our earlier reports on low-level waste performance assessment. This updateaddresses needed improvements and recommended approaches to the existing state of the art in modeling, treatmentof uncertainty, and use of data. Greater attention is paid to developing an integrated approach to performanceassessment than was done in earlier developments of the methodology. Furthermore, insights are being developedby participating in validation exercises, and by evaluating which validation data are needed to improve confidencein the methodology. It is emphasized that the performance assessment methodology update is a work in progress;the recommendations given here will form the general directions toward which the methodology is heading, butsome of the specific approaches may continue to evolve as the research progresses.

iii

Contents

1. Introduction . ..... ...... ............................ .....

1.1 Background ........................................1.2 Scope of the Report .......................................1.3 Structure of the Report .....................................

2. General Considerations ...........................................

2.1 Uncertainty Analysis .......................................

2.1.1 Model Uncertainty ..................................2.1.2 Uncertainty About the Future of the Site ................2.1.3 Parameter Uncertainty ................................2.1.4 Treatment and Reduction of Parameter Uncertainty ...........2.1.5 Incorporating Uncertainty Analysis into the Methodology ........

2.2 Reduction of Uncertainty .....................................

2.2.1 The Process of Performance Assessment ...................2.2.2 Defensibility of Analyses ..............................

2.3 Summary of Uncertainty Analysis Recommendations .................2.4 User Friendliness .........................................

3. Pathway Assessment for Alternative Disposal Technologies ..................

3.1 Role of Pathway Analysis in Performance Assessment ................3.2 Previous Work on Pathway Analysis ............................3.3 General Comments on Pathway Assessment .......................3.4 Alternative Disposal Technologies ..............................3.5 Role of Temporal Progression in Pathway Analysis .................3.6 Summary of Pathway Assessment ..............................

4. Ground-Water Flow and Transport Modeling ........................

4.1 Infiltration Evaluation .......................................4.2 Ground-Water Flow and Transport ..............................

5. Source-Term Modeling ............................................

5.1 Engineered Barriers ........................................

5.1.1 Concrete Structures ..................................5.1.2 Metal Container Degradation ...........................5.1.3 Degradation of Other Materials ..........................

Page

..... •. ..... 1

. . . . . . . . . .. 1

. . . . . . . . .. 1

. . . . . ... . . . . 2

. . . . . . . . .... 5

. . . . . . .. .. . . 5

...... . . . ... 5

..... . . . . . . . 8. . . . . . . . . . . . 11. . . . . . . . . . . . 12. . . . . . . . . . . . 15

. . . . . . . . . . . . 21

. . . . . . . . . . . . 21. . . . . . . . . . .. 25

. . . . . . . . . . .. 26. . . . . . . . . . . . 27

. . . . . . . . . . . . 29

. . . . . . . . . . . . 29. . . . . . . . . . . . 29. . . . . . . . . . . . 29. . . . . . . . . . . . 31. . . . . . . . . . . . 33. . . . . . . . . . . . 34

. . . . . . . . . . . . 37

. . . . . . . . . . . . 37

. . . . . . . . . . .. 37

............ 45

. . . . . . . . . . . . 45

. . . . . . . . . . . . 45

. . . . . . . . . . . . 47

. . . . . . . . . . . . 47

V V NUREG/CR-5927

Contents (continued)

Page

5.2 Leaching Processes and Near-Field Transport

5.2.15.2.25.2.3

Leaching Processes . .Near-Field Transport .Decay Chains ......

............................

............................

............................

............................

............................

............................

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

47

474949

505152

5.3 Gas Production ...........5.4 Geochemistry ............5.5 Source-Term Summary ......

6. Surface-Water Transport, Air Transport, and Exposure Pathway Modeling .......... . . . . . . . . . . .. 55

6.16.26.36.4

Surface-Water Transport ....Air-Transport Modeling ....Exposure Pathway Modeling..Status and Evaluation .......

............................

............................

............................

............................

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

55555656

7. Dosimetry M odeling ........................................

8. Sum m ary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . .

. . . . . . . . . . . . . . 59

. ... . . . . . ............... 61

9. References ....................................................................... 65

Figures

1.1 Processes included in the methodology ........................................... 31.2 Modeling approaches in the original methodology .................................... 42.1 Comparison between two hypothetical dose distributions ...................................... 142.2 Overall approach to uncertainty analysis for low-level waste performance assessment ................. 182.3 Optional approaches for combining information to make the regulatory decision ..................... 192.4 General approach to performance assessment ............................................. 222.5 Uncertainty analysis for models and parameters ........................................... 232.6 The decision support system structure ..................................................... 284.1 Flow processes in and around an intact disposal facility ...................................... 384.2 Flow processes in and near a failed vault ................................................ 394.3 Defining an aquifer stream tube from a flow model ......................................... 425.1 The mixing-cell cascade model ......................................................... 488.1 Updated processes in the methodology .................................................... 628.2 Current recommendations for the methodology ............................................ 63

NUREG/CR-5927 vi

Tables

Page

2.1 Summary of approaches to uncertainty treatment ........................................... 172.2 Least biased distributions for varying amounts of available information ........................... 197.1 Differences in tissue weighting factors between ICRP 26 and ICRP 60 ........................... 608.1 Recommended changes to the models in the methodology ..................................... 64

vii NUREG/CR-5927

FOREWORD

This technical contractor report is a product of Sandia National Laboratories under project FIN L1153.The purpose of this program is to update and improve a performance assessment methodology for low-level radioactive waste disposal facilities previously developed under FIN A1764.

NUREG/CR-5927 is not a substitute for NRC regulations and compliance is not required. Theapproaches and/or methods described in this NUREG/CR are provided for information only.Publication of this report does not necessarily constitute NRC approval or agreement with theinformation contained herein.

NUREG/CR-5927 viii

1. Introduction

1.1 Background

A low-level radioactive waste performance assessmentmethodology was developed by Sandia National Labora-tories (SNL) for use by the U.S. Nuclear RegulatoryCommission (NRC) in evaluating license applicationsunder Section 10 of the Code of Federal Regulations,Part 61 (10 CFR Part 61) [Kozak et al., 1990b]. Thepurpose of the methodology is to allow NRC to confirma licensee's evaluation of postclosure impacts. Theseperformance assessment analyses are the basis for provid-ing reasonable assurance that the performance objectivesin 10 CFR Part 61.41 are met.

The methodology must be flexible enough to handle awide range of potential low-level waste disposal facilities.The performance assessment modeling may need to beeither very simple, or more complex, and models areincluded in the methodology for both of these possibili-ties. Since the methodology is modular, the analyst maysubstitute more complicated models for only part of theanalysis when appropriate.

The components needed for performance assessmentmodeling of a low-level waste facility are shown in Fig-ure 1.1. Before choosing the models to be implementedin the methodology, a literature survey was performed toidentify existing models and codes for each requiredprocess. In addition, general site characterization datarequirements were identified, and significant sources ofuncertainty were discussed [Kozak et al., 1989a]. Thatwork formed the basis for current models in the method-ology. These models are shown in Figure 1.2. Theprimary impetus for choosing many of these codes wastheir flexibility in modeling a wide variety of problems.Further justification and discussion on some modelingareas was provided in Kozak et al. [1990a]. Some areasare modeled very conservatively in the methodology; thisapproach was taken when no adequate model was avail-able, or when details of the processes themselves werepoorly understood.

SNL was subsequently contracted to update and improvethe methodology where necessary, and to build confi-dence in the models in the methodology. This is a reportto assess whether the current models in the methodologyare adequate, and based on this assessment, to identifyadditional models and codes that may be useful to includein the methodology. This report represents an update tothe discussions found in Kozak et al. [1989a, 1989b,1990a]. The intent is to update the information in thesereports by including discussions on new models and

codes that have become available since the preparation ofthe original methodology. In addition, since developingthe methodology, we have applied it several times toperformance assessment test cases [Chu et al., 1991;Kozak;' Kozak and Rao; 2 Kozak and Feeney3], and thisadditional experience allows an improved assessment ofthe modeling needs.

Furthermore, insights are being developed by participat-ing in validation exercises, and by evaluating whichvalidation data are needed to improve confidence in themethodology. Volume 2 of this report covers the valida-tion needs for the modeling areas in the methodology[Olague et al., 1993]. Priorities are set for the mostimportant validation problems that need to be addressed.

1.2 Scope of the Report

This report is an update to our earlier work, described inShipers [1989], Shipers and Harlan [1989], and Kozak etal. [1989a, 1989b, 1990a, 1990b], as well as other perti-nent documents and papers on low-level waste perform-ance assessment [e.g., Starmer et al., 1988; Deering andKozak, 1990]. It must be read in that context, since it isnot intended to be a stand-alone guide to performanceassessment of low-level radioactive waste disposal facili-ties. Rather, it is a summary of needed improvementsand recommended approaches to the existing state of theart in modeling, treatment of uncertainty, and data avail-ability. In addition, the performance assessment method-ology update is a work in progress; the recommendationsgiven will form the general directions toward which themethodology is heading, but some of the specific ap-proaches may continue to evolve as the research pro-gresses.

The goal of the methodology is to enable the NRC toevaluate postclosure, off-site doses from a low-levelradioactive waste disposal facility for comparison withthe regulatory performance measures of 10 CFR Part 61.Inadvertent intruders receive adequate protection through

'Kozak, M.W., "Preliminary Analysis of Cases la and lb," FIN A1764letter report, submitted to F.W. Ross, NRC/NMSS, June 1991.2Kozak, M.W., and R.R. Rao, "Analysis of NSARS Case 1," FINA1764 letter report to NRC, submitted to F.W. Ross, NRC/NMSS,August 1991.3Kozak, M.W., and T.A. Feeney, "Analysis of NSARS Case 2a," FINA1764 letter report to NRC, submitted to F.W. Ross, NRC/NMSS,September 1992.

I NUREG/CR-5927

Introduction

the waste classification scheme developed as part of theEnvironmental Impact Statement for 10 CFR Part 61[NRC, 1981]. As a result, intruder analyses will only berequired in a low-level waste license application underspecial circumstances, when an exemption from the wasteclassification system is proposed [Kozak et al., 1990a].Consequently, little attention will be given to intruderanalyses in this report; the focus of the methodology ison evaluation of off-site doses to the maximally exposedindividual of the public.

1.3 Structure of the Report

This volume covers three primary topics. First, a gener-al assessment of the methodology and of performance

assessment as a whole is given. In particular, revisionsare discussed for the areas of uncertainty analysis anduser friendliness. A pathway assessment is presented fora variety of disposal options in Chapter 3. The intent ofthis pathway assessment is to ensure that the methodologycontains adequate coverage of all types of modeling thatit might be faced with. In essence, the evaluation inChapter 3 is a review of the methodology for complete-ness. The second aspect of the review of the methodolo-gy is to reevaluate the models in the methodology, giventhe results of the completeness review, and the passage oftime. This review of the modeling needs is given inChapters 4 through 8 for each modeling area of the meth-odology. Chapter 8 contains a summary of the evalua-tion of the methodology.

NUREG/CR-59272 2

Introduction

source term

Figure 1.1 Processes included in the methodology

3 3 NUREGICR-5927

Introduction

source term

Figure 1.2 Modeling approaches in the original methodology

NUREG/CR-5927 4

2. General Considerations

In our ongoing evaluation of the methodology, we areassessing aspects of performance assessment that werenot stressed in the original methodology development.These areas are (1) uncertainty analysis, and (2) im-proved user friendliness. These areas, which are outsidethe strict scope of modeling updates but are of generalimportance, are discussed in this section.

2.1 Uncertainty Analysis

The methodology is intended to be used to allow theNRC to determine if a candidate low-level waste site willmeet the performance objectives specified in 10 CFR Part61. These performance objectives specify that an off-siteperson may not receive a committed annual dose equiva-lent of more than 25 mrem whole body, 75 mrem thy-roid, or 25 mrem to any other critical organ. These aredeterministic performance measures, implying that as-sessing the performance of each site will produce onedose that can be compared to the standard. However,uncertainty analyses invariably lead to a suite of doses,which must then be compared with the deterministic stan-dards. The purpose of this section is to point out someof the challenges, and possible solutions, associated withcomparing uncertainty analysis results with a determinis-tic performance objective. This section is an expansionand refinement of earlier work on this subject [Kozak etal., 1991].

The uncertainties in performance assessment have beenclassified as model uncertainty (which spans conceptualmodel uncertainty and mathematical model uncertainty),uncertainty about the future of the site, and parameteruncertainty [Davis et al., 1990a]. These inherent uncer-tainties are dealt with using uncertainty analysis, which isa way of formally documenting, treating, and reducingthe inherent uncertainty of a system. The analysis isnothing more than an identification of how much or howlittle confidence the analyst has in his knowledge of themodeled system [Finkel, 1990].

If all uncertainties are addressed, the result would bealternative possible doses depending on our conceptual-ization of the site, the mathematical models we use torepresent the conceptualization, the parameters used inthe mathematical model, and the conceptualization of theevolution of the site in time. To address these uncertain-ties, NRC currently recommends that the licensee for alow-level waste site give dose estimates with a range ofminimum and maximum values; this range should takeinto account all uncertainties in the calculations [Starmeret al., 1988). However, no explicit guidance is given on

how the ranges should be calculated. NRC also suggeststhat the doses be presented as a function of time, butdoes not identify a method for taking into account chang-es in the low-level waste site as it evolves in time, norfor how long the performance assessment should beconducted. Thus, there is currently no official positionon the treatment of uncertainty for low-level waste per-formance assessment. Consequently, the original per-formance assessment methodology does not contain for-mal uncertainty analysis.

In the following sections, we give recommendations onhow to incorporate uncertainty analysis into the method-ology. First, we give background on how to address andreduce the three different types of uncertainty (model,parameter, and future) based on a literature review.From the literature review, we have determined the bestmethods currently available and the ones that we recom-mend for inclusion in the methodology. Second, andmore important, we have developed a strategy for imple-menting our recommendations into the methodology.

2.1.1 Model Uncertainty

The process of developing a site-specific model beginswith the perceived real world, as defined by site-specificdata [Davis and Olague, 1991]. (It is important to notethat we are never able to completely perceive reality,particularly in ground-water modeling where the systemcannot even be directly observed, since this incompleteperception is a primary source of uncertainty in model-ing.) Next, simplifying assumptions are made to developa conceptual model, which is a qualitative description ofthe processes, geometry, and boundary conditions associ-ated with a site. These qualitative ideas are translatedinto a quantitative mathematical model, which is a set ofequations that represent the behavior of the conceptualmodel. The mathematical model can then be solved,with site-specific input parameters, analytically, or with anumerical approximation (in a computer code) for thequantity of interest (e.g., effective dose equivalent).

Model uncertainty encompasses both the uncertainty inthe conceptualization of the system, the uncertainty in itsmathematical representation, and the uncertainty in thesolution of the mathematical representation [Bonano andCranwell, 19881. Conceptual model uncertainty arisesfrom a number of different sources. There may be un-certainty associated with the characterization of the per-ceived "real system," such as misinterpretations of thedata or inadequacy of data reduction techniques. Uncer-tainty will also be introduced with the simplifying as-

5 NUREG/CR-5927

General Considerations

sumptions that are necessary to make the problem tracta-ble. For instance, the transient, three-dimensional realworld is usually modeled as a steady-state one-dimension-al process. In addition, models are most commonlydeveloped by a single analyst or a small group of analystsusing only their professional judgment to resolve avail-able data into a model. The model is therefore limitedby the abilities and imagination of the developer, in addi-tion to limitations in the available data. Also, despite theusefulness of experience with sites other than the onebeing considered, one can develop a poor conceptualmodel for an apparently similar site based on precon-ceived ideas [Bonano and Cranwell, 1988]. An exampleof this is using a saturated porous media flow model witha few modifications for saturated fracture flow.

In many cases, conceptual model uncertainty is the domi-nant type of uncertainty in a performance assessment: ifan inadequate conceptual model is being used, uncertaintyassociated with the mathematical model and the modelinput parameters becomes irrelevant. In some cases eventhe processes at work are not well understood. Forexample, the conceptual model cannot be identified forfracture flow in porous media. There is potentiallysimultaneous flow in the porous matrix and in fracturesin the media, but no one understands the conditions thatcause one or the other to dominate the performance ofthe system. The physics behind the processes are notunderstood well enough to allow us to write down theequations describing the exchange of material betweenfractures and the matrix, except in a heuristic mannerthat has not been substantiated (or refuted) by experi-ments [Updegraff et al., 1991].

Besides the uncertainties associated with the underlyingconceptual model, uncertainty in mathematical modelsarises from the methods required to arrive at a solution tothe equations of interest [Davis and Olague, 1991] andfrom the inability to represent a conceptual model in asuitable mathematical form [Bonano and Cranwell,19881. If an analytical technique is used, uncertainty canbe introduced if the solution to the equations is incorrect,or from the truncation of a mathematical infinite seriessuch as an error function used as part of the solution. Ifa numerical solution method is used, it will almost al-ways be implemented in a computer program. This canintroduce two sources of uncertainty: errors from thenumerical approximation of the equations and codingerrors in the computer program. An additional source ofuncertainty for computer codes is user error.

2.1.1.1 Treatment and Reduction of Model Uncer-tainty

Very little work has been done to treat conceptual modeluncertainty [Kozak et al., 1991]. The only availableapproach to treating conceptual model uncertainty is tospan the range of conceptual models that are consistentwith site-specific data. Some have suggested that theformal elicitation of expert opinions may be a good wayof spanning this range [Kerl et al., 1991, Chhibber et al.,1991a] by creating an exhaustive list of possible alterna-tive conceptual models that are consistent with availabledata. By broadening the base of expertise from whichthe conceptual models are developed, there is increasedlikelihood that a conceptual model will be included thatcaptures some potentially adverse characteristic of thesite, and to the extent possible, conceptual model uncer-tainty is addressed. The disadvantages of this approachinclude increases in cost and time and reduction in theflexibility associated with formalizing expert judgment[Bonano et al., 1990].

One way to implement this approach into performanceassessment has recently been proposed [Chhibber et al.,1991a,b; Heger et al., 1991]. This method associates aprobability with a given conceptual model, which isinterpreted as a measure of the degree of belief that theconceptual model is appropriate for the given purpose.However, Chhibber et al. [1991a] recognizes a numberof difficulties in this technique. Perhaps the most impor-tant constraint is that to apply probability theory, themodels should be defined so that they are mutually exclu-sive, exhaustive, and independent. This difficulty seemsinsurmountable since all the conceptual models are basedon the same site-specific data. Other difficulties arisewhen combining and aggregating expert opinion, andwhen incorporating new information into probabilityestimates. Given these problems, we conclude that thisapproach is an interesting area of research, but manysignificant issues need to be addressed before it can beused in performance assessment.

We suggest a simpler approach. All conceptual modelsconsistent with data should be used for performanceassessment of low-level waste sites. If the models cannotbe distinguished from each other by acquiring additionalsite-specific. data, then each of the models should beconsidered credible. The performance assessment mustthen be conducted using each model, and the results usedto establish the model that is the most conservative. In

NUREG/CR-5927 6


general, it is not possible to establish conservatism of themodel a priori. The reason for this is that most modelsused in licensing are mixtures of conservative and non-conservative assumptions. True "bounding" analyses arerarely used, because it is usually considered that such anapproach is excessively conservative. Conservatismamong models can only be established by posterior com-parisons of the calculated performance objective of thedifferent models.

Model intercomparison can be used to some extent todetermine conservatism. Model intercomparison is de-fined as a comparison of codes that implement differentconceptual models of the same processes (i.e., the mod-els may not implement the physics or chemistry in anidentical manner, or may have somewhat different as-sumptions). For instance, one might compare resultsfrom a one-dimensional, single layer transport analysis ofradionuclide migration to results from a multidimension-al, multilayer model. The intercomparison can be usedto identify crucial assumptions in the two approaches tomodeling radionuclide migration, and these assumptionscan then be the focus of validation studies. It should benoted that intercomparison is different from benchmark-ing, which is a comparison of computer codes that havethe same conceptual model.

Overall model uncertainty can be reduced, but not elimi-nated, by site-specific model validation. Site-specificdata is the most defensible evidence for determining thereliability of a model, since it represents the real systemto be modeled. However, as discussed by Davis et al.[1991], it is not practical to conduct validation experi-ments for the full range of conditions of interest in per-formance assessment because of time and funding con-straints and because extensive testing at a site may inter-fere with the site's geologic integrity. Therefore, valida-tion can be used to build confidence that the uncertaintiesare reduced to the extent practicable. In general, theappropriateness of any performance assessment modelshould always be determined based on site-specific data.

We note that "conservatism" of a model is always rela-tive to something. Ideally, we want the model to beconservative with respect to actual (perceived) site behav-ior. Unfortunately, we will rarely have the luxury ofestablishing conservatism compared to any single aspectof site behavior, much less conservatism of the overallperformance assessment. Consequently, model conserva-tism will usually be defined with respect to other possiblemodels or combinations of parameters. In this sense,model intercomparison must play an important role in

evaluating the conservatism (hence reliability) of theanalysis results.

For performance assessment, mathematical model uncer-tainty is usually not propagated to the results, since it isbelieved to be negligible compared to the other uncertain-ties [Davis and Olague, 1991]. Therefore, our approachto treating mathematical model uncertainty is that itshould be reduced to the extent practicable, but otherwiseignored in the propagation of uncertainty. This can beaccomplished through a variety of methods. If an analyt-ical solution is used, it can be compared to other avail-able solutions for accuracy or checked by an availableexpert. For numerical or semianalytical solutions imple-mented in computer codes, several uncertainty reductionmethods should be used. Careful quality assurance pro-cedures should be followed during the development of thecomputer code to avoid the introduction of coding errors.Quality assurance activities should also include verifica-tion exercises. Verification gives assurance that themodel equations are implemented correctly in the com-puter code. This is accomplished through careful evalua-tion of the program and by comparison of the programoutput with analytical solutions or other computer pro-grams that implement similar physical processes (i.e.,benchmarking). Once the program is in operation, aconfiguration management system should be followed toensure that no haphazard modifications to the programare made. Error reduction techniques should be fol-lowed, such as using a finer discretization of the domainor using more terms to represent an infinite series. Theuser will know that the error has been reduced to accept-able levels once a convergent solution is achieved; i.e.,the discretization is made finer without the solutionchanging.

The uncertainty associated with conceptual models can betreated by trying to better understand the physics behindthe relevant phenomena. Alternatively, for site-specificmodeling an empirical model with well-chosen parame-ters could be adequate for the intended purpose whencompared to a more complex, physics based model. Thisissue can only be resolved with site-specific modelvalidation [Davis et al., 1991]. However, it is veryimportant to realize that empirically based models cannotbe extrapolated outside of the domain in which theircoefficients were determined. This makes empiricalmodels of intrinsically limited usefulness in performanceassessment, since most conditions that are important in aperformance assessment are not ones that can be ob-served at today's conditions.

7 NUREG/CR-5927


Validation is the process in which confidence is gainedthat a model is an accurate representation of the physicalprocesses for which the models are intended [Davis andOlague, 19911. It offers a chance to build confidence inthe conceptual model, the mathematical model, and thesite-specific parameters by seeing how well the hypotheti-cal representation of the site compares with the perform-ance of the actual site. Validation also tests how well themathematical representation of the physical processescompares to site-specific data. One drawback of valida-tion is that since it tests the whole system, it does not tellyou which component of the overall model is incorrect.Thus, if a site-specific model compares unfavorably withdata, it is not readily apparent if the disagreement iscaused by using the wrong conceptual model or incorrectparameter ranges.

We recommend that all available procedures be used toreduce the uncertainty in low-level waste performanceassessment mathematical models. Careful quality assur-ance procedures should be followed in the developmentof a solution to the mathematical model. Verificationshould be used to ensure that the correct equations arebeing solved and validation should be used to ensure thatthe model does represent the behavior of the site.

2.1.2 Uncertainty About the Future of theSite

Uncertainty about the future of the site is the result ofour inherent lack of knowledge about how the site willevolve in time. The climatic, geologic, and populationconditions that will prevail in the future of the site arenot known. Consequently, we must determine somemethod for accommodating alternative future conditions.

At the present time it is not clear what conditions need tobe evaluated to meet the low-level waste regulations.NRC has suggested that different future scenarios shouldbe modeled, but has given no explicit guidance on theselection process [Starmer et al., 19881. Also, there iscurrently no regulatory guidance concerning the perform-ance assessment time period or how to account for futureconditions at a low-level waste site. NRC has stated thatthe low-level waste "source term inventory should beused to justify the necessary duration of the performanceassessments" [Starmer et al., 1988]. This may imply thatsource terms containing large amounts of long-livedradionuclides, such as 4̀C, 9Tc, etc., should be modeledfor longer times than source terms containing only short-lived nuclides such as 'Co and tritium.

If explicit regulatory guidance was given, it would pro-mote consistency between analyses done by regulators,states, and compacts, which in turn would tend to makethem more defensible. Uncertainty about future condi-tions is not primarily a technical issue, but rather must beresolved by the regulators. The purpose of the followingdiscussion is to provide possible approaches based onrelevant technical information from which regulatorydecisions can be made.

2.1.2.1 Treatment and Reduction of Future Uncer-tainty

The generalized approach to addressing future uncertaintyis to span the range of high probability, high risk futureconditions of the site. Identifying all important eventsand processes is a creative task that may depend exclu-sively on expert judgment [Bonano et al., 1990]. Span-ning all important possible future conditions can mostreadily be accomplished through formalized elicitation ofexpert opinion. This can be done in an analogous man-ner to the conceptual model elicitation. However, asignificant amount of elicitation has already been donefor scenarios, and comprehensive lists of possible disrup-tive events and processes exist [Cranwell et al., 1990].

Two issues need to be addressed regarding uncertaintyabout the future. First, we need to determine how longthe performance assessment will need to be carried out.Second, we must decide on a way to incorporate alterna-tive views of how the site will change in time into theperformance assessment methodology. We emphasizethat the intent must be to span the range of likely behav-ior of the site over the time frame of concern; we are notinterested in addressing improbable events, but ratherincluding extreme, but very likely events in the analysis.

One logical strategy to quantifying the time frame forlow-level waste performance assessment would be tomodel the site until the peak dose is obtained. Becauselow-level waste contains long-lived radionuclides (mainly14C, 1291, 99Tc, and the actinides), and because of the

current emphasis on long-lived engineered barrier sys-tems, this time period can become relatively long. Forlong time periods, it may become important to includesuch events as glaciation, which would not be pertinent toshort time periods. Assumptions would have to be madeabout the future of the site, and these assumptions woulddepend greatly on how long a time period needs to bemodeled.

Another possible strategy for quantifying a time framefor low-level waste performance assessment would be for

NUREG/CR-5927 8


NRC to mandate a specific time period that needs to bemodeled, similar to what was done for high-level wastein 40 CFR Part 191. Once a time period has been desig-nated, a method needs to be developed to account for allprobable, high consequence, future states of the sites.For high-level waste performance assessment, futureuncertainty is accounted for by explicitly acknowledgingpossible alternative future scenarios. These possiblescenarios are then incorporated into the overall perform-ance assessment by assigning likelihoods to each scenar-io. This approach was developed in response to therequirement in 40 CFR Part 191 that high-level wasteperformance assessment must include all significantevents and processes over a period of.10,000 years. Thisrequirement is not explicitly included in 10 CFR Part 61.

One way to address future uncertainty for low-levelwaste performance assessment is to use the scenarioapproach that has been developed for U.S. high-levelwaste regulation. Scenarios are future likely conditionsof the site that affect the performance assessment results.The intent of scenarios is to provide a simpler surrogateapproach for modeling the likely transient future historyof the site. Scenarios are treated by modeling the systemas a steady-state process, then weighting the consequenc-es of that scenario by its likelihood. The rationale forweighting the scenario consequence is to identify thecontribution of that scenario to the overall performancemeasure, which is integrated discharge for the U.S. high-level waste regulation. For example, consider a climatechange that produces increased infiltration that lasts for1000 years. A high-level waste scenario analysis wouldmodel the steady-state integrated discharge using thehigher infiltration rate, then weight it by 0.1 to incorpo-rate its contribution to the overall integrated dischargeover 10,000 years. It is important that the likelihoodassociated with this scenario is not related to its probabil-ity of occurrence (the probability of a wetter time periodduring the next 10,000 years is essentially unity), butrather is meant to identify the amount of contribution tothe performance measure. It is clear that this type ofrationale cannot be applied to doses, which are not inte-grated over time, but rather are point estimates in time.If transport from the site were fast enough under theseconditions, that waste could be flushed from the disposalfacility to the receptor in this wetter period, making thereceptor the maximally exposed person. In this case,there does not seem to be a reason to weight a doseproduced during the wet period with a small likelihood,because the probability of that dose occurring is nearunity (if all the other assumptions made in the analysisare met).

On the other hand, the scenario approach has two advan-tages: (1) it traces assumptions about the future in aformal manner, and (2) it addresses the uncertainty asso-ciated with the future. There are two problems associat-ed with using this approach for low-level waste perfor-mance assessment. First, because most low-level wastesites are located near the land surface, surficial eventsand processes that could reasonably occur over long timeperiods may become important (e.g., flooding, erosion,glaciation). Considering such processes would complicatelow-level waste performance assessment significantly,and it is questionable as to whether or not a near-surfacefacility could meet the regulations with these types ofevents and processes occurring. For instance, in manyparts of the country, if it is reasonable to assume thatglaciation may occur at a site within 10,000 years, doesthat mean that glaciation should be included in a low-level waste performance assessment?

To illustrate the problem with using extreme events andprocesses to compare with the performance objectives, letus analyze direct exposure of an individual to wasteexposed at the surface by glaciation, and assume that theconsequence analysis of this scenario (and only thisscenario) results in exceeding a dose performance mea-sure. The regulations specify a maximum 25 mrem/yeardose, so let us assume that the consequence of this analy-sis significantly exceeds the standard, and the personreceives 100 mrem/year. Given a regulatory philosophythat requires all consequences fall below the performancemeasure, this facility would fail the safety assessment.Furthermore, since no near-surface facility can reason-ably be expected to survive the onslaught of a glacier, itcan be expected that no facility at that site could meet theobjectives, and the entire location could be eliminatedfrom consideration. However, let us consider a counter-argument. A person living near the moraine of a glacierduring an ice age would undoubtedly have much largerhealth problems than those posed by receiving 100 mrem/year, which poses immeasurably small health risks[Gershey et al., 1990]. We have, therefore, beentrapped by our inclusion of the extreme tall of the distri-bution into making an arguably poor regulatory decision.This example identifies the potential problem with usinga combination of a full scenario analysis together withrequiring that the entire dose distribution must fall belowthe performance objective.

The full scenario approach can be salvaged for use inlow-level waste performance assessment by choosing anintermediate confidence limit for comparison with thedeterministic performance objectives. For instance, theEPA provided guidance that suggested using the larger of

9 NUREG/CR-5927


the mean or median value of the probability curves forassessing compliance with the Individual Protection Re-quirement and Ground-Water Protection Requirement in40 CFR Part 191 [EPA, 1985a]. Alternatively, theregulator may choose to use some higher confidence limitof the dose probability curve to compare against the 25mrem objective. This approach is self-consistent, sinceall events and processes of possible importance are in-cluded, but it omits the tails of the distribution, andfocuses the decision maker's attention on the centraltendencies of the distribution. However, it discards thetails of the distribution, which will include some condi-tions of the site that are likely to occur.

If NRC decides that only a relatively short-term timeperiod needs to be modeled for performance assessment,a full scenario approach may not be warranted. Theapproach will then be one of including events and pro-cesses that occur for the arbitrarily defined short timeperiod, which would probably be small perturbationsaround the current state of the site. Analyses conductedin this way will be focused on events and processes thatare likely to occur (e.g., transport to a well), rather thanevents distant in the future. However, this relativelyshort time period may not be sufficient for characterizingthe peak dose because of the longevity of some constitu-ents in low-level waste. Also, the use of concrete vaults,which may last hundreds of years, in low-level wastedisposal facilities may result in moving the peak dosepast the mandated time period.

One way of implementing this approach would be toconduct the performance assessment until the peak doseis reached, but only using conditions that may be reason-ably expected to occur during, say, the first 100 years ofthe postclosure performance time period. The modeling,therefore, includes only relatively minor perturbationsabout the current state of the site. The question stillarises as to how large the perturbations should be. As anexample, consider the rainfall at the site, which is impor-tant for assessing recharge, and hence, degradation ratesof engineered barriers and release rates from the facility.There is an intrinsically probabilistic aspect in definingwhat rainfall will be included in the analysis, for rainfallis usually treated as being stochastically distributed intime. If the analyst decides to use the 100-year "maxi-mum probable" precipitation year in the analysis, itshould be understood that there will remain a finite prob-ability that this value will be exceeded, since by defini-tion "maximum probable" means there is a chance thisvalueswill be exceeded. In addition, significant vegeta-tive progression can occur even over relatively short(100-year) time periods, so the span of conditions that

may need to be included in the analysis is still quitebroad.

Another short time period approach is to maintain thecurrent conditions at the site for the entire performanceassessment time period. This is similar to the approachsuggested by EPA for assessing compliance with theIndividual Protection Requirement and Ground-WaterProtection Requirement contained in 40 CFR Part 191.They assume that current conditions will exist at the sitefor 1,000 years [EPA, 1985a]. Use of this approachrequires a definition of what "current conditions" means.Low-level waste sites may only have a few years ofdetailed monitoring data at the site from which to identifycurrent conditions; these data may fortuitously only spana range of unusual conditions, such as particularly dry orwet years. Use of this data may, therefore, lead tomisinterpretations about the long-term behavior of thesite.

An alternative approach would be to model the projectedtime-dependent future history of the site; this approachhas been adopted in the United Kingdom for use in com-paring with a deterministic regulatory performance objec-tive. The rationale for this approach is that, unlike thescenario approach, extreme (but likely) conditions are notmodeled as steady-state conditions for the entire timeperiod of the analysis. Recall that in the scenario ap-proach, the 10,000-year maximum annual rainfall wasmodeled as a steady-state process that lasts for 10,000years, but it is then weighted by a 10' likelihood (actual-ly a relative frequency), since it is presumed to affect10' of the integrated discharge. In a time-dependentanalysis, the 10' climate events would only affect thedose if significant radionuclide migration occurs duringthat single year of the analysis.

An obvious disadvantage to the time-dependent modelingapproach is that transient modeling is much more difficultand time-consuming than steady-state modeling. Inaddition, the future history of the site is not known, eventhough it is highly likely that some extreme events willoccur. This means that in spite of significant additionalmodeling effort, little will be done to reduce the uncer-tainties about the future. The danger, therefore, exists ofintroducing more complicated analyses without improvingthe confidence in the results.

To summarize, there are four approaches that might betaken for quantifying uncertainties about the future of thesite in the context of low-level waste performance assess-ment. (1) Conduct the performance assessment untilpeak dose or for an NRC-mandated relatively long time

NUREG/CR-5927 10


period, and to include possible future events and process-es through a scenario evaluation. However, to use thisapproach in an appropriate way for decision making, theregulators may have to allow some of the tails of thedistribution of dose estimates to exceed the deterministicperformance objective. The choice of the particularvalue of the percentile for comparison with the perfor-mance objectives is entirely a regulatory decision. Al-though the scenario approach provides a means for sys-tematically treating future conditions at a site and allow-ing for possible alternative scenarios, there are seriousissues that need to be addressed. (2) Define well-estab-lished design-basis conditions, in which only events andprocesses that are reasonable for a NRC-mandated shorttime period are included in the analysis. This approachfocuses attention on the events and processes that aremost likely to occur in low-level waste performanceassessment, but does not address the issues associatedwith the long-lived radionuclides. (3) Use current condi-tions at the site to extrapolate to the longer time periodneeded to characterize peak dose or for the NRC-mandat-ed short time period chosen for performance assessmentcalculations. This approach focuses attention on the mostlikely conditions of interest, but does not take into ac-count that extrapolation of the design-basis conditions tolonger times has progressively less physical meaning asthe time period expands. (4) Model the time-dependenceof the future history of the site. A choice of one of thesefour approaches must be made by the regulators based onregulatory philosophy; there is not a clear-cut best candi-date based on purely technical considerations.

2.1.3 Parameter Uncertainty

Parameter uncertainty relates to an incomplete knowledgeof the performance assessment model(s) constitutivecoefficients. In part, this uncertainty is identified withuncertainty in the actual values and the statistical andspatial distributions of data used to infer the model pa-rameters. Sources of parameter uncertainty includemeasurement error, insufficient data, and inconsistencybetween measurement scale and prediction scale. In thissection, we discuss each of these issues as though theywere independent of model uncertainty. However itshould be recognized that the two types of uncertainty arenot completely independent.

2.1.3.1 Measurement Errors

Measurement errors are generally considered to be oftwo types: random and systematic. In the context of thelarger uncertainties in performance assessments, random

errors in parameter measurements are relatively minor,and will not be discussed further here. Systematic errors,however, are unavoidable in many important measure-ments needed for site characterization, and this is a great-er concern.

Many hydrological measurements must be taken on dis-turbed samples, and in situ measurements are impossiblefor most parameters. For example, the only extant ap-proach for fully evaluating unsaturated-zone characteristiccurves consists of laboratory evaluations of core samples.The core sample will never have the same characteristicsas the same soil resting undisturbed in the field; therewill always be a systematic error associated with themeasurement. The issue is whether the error is signifi-cant. Other common hydrological parameters sufferfrom this same potential systematic error; porosity, de-gree of anisotropy, and dispersivity, among others, areall measured primarily in the laboratory. This source oferror is consistently ignored in the literature, largelybecause it cannot easily be evaluated: there are not usual-ly independent methods for deriving the same informationabout the site to compare to the laboratory test. Whenfield-scale nonintrusive geophysical measurements exist,such as ground-scanning radar, they are often very diffi-cult to interpret.

2.1.3.2 Data Insufficiency

Site characterization for performance assessment requirescollecting enough data to define, to the extent necessaryfor performance assessment, (1) the naturalhydrogeologic environment at the facility, (2) the likelyfuture climatic conditions at the site, and (3) the long-term behavior of both the natural and engineered systemsunder both current and future climatological conditions.This task is difficult even for engineered systems, such asconcrete vaults, which are (relatively) spatially homoge-neous, and which evolve in time in a relatively wellunderstood manner. However, there is generally a largeinherent spatial and temporal variability associated withnatural systems which must be accounted for in datacollection, data interpretation, and modeling. In fact,one of the crucial tasks in site characterization for per-formance assessment is to develop confidence that thespatial variability does not include a preferential path thatwould result in the facility violating the performanceobjectives.

Data sufficiency constraints can potentially prevent thedevelopment of adequate models for performance assess-ment. The amount of data that is sufficient to build an"adequate" model cannot be identified in general, or even

I1I NUREG/CR-5927


defined unambiguously, for a specific site. A largeproportion of professional judgment is involved in theevaluation of data sufficiency. Furthermore, model"adequacy" and data "sufficiency" are likely to be deter-mined to some extent by political as well as technicalconsiderations. In addition, we note that conservativeparameter values may be used (to some extent) in placeof detailed site characterization data. In all cases, theregulatory decision will have to be made under a condi-tion of uncertainty, and data sufficiency relates to confi-dence in the regulatory decision. Consequently, conser-vative parameter values can be used to reduce uncertaintyin the regulatory decision even if they do not promote afundamental understanding of the site.

2.1.3.3 Scale Dependence of Measurements

One of the most important sources of parameter uncer-tainty is the inconsistency between the scale of parametermeasurement and the model simulation scale. It is oftenassumed that information collected at one spatial scalecan be "scaled up" to provide information of field-scalemodel parameters. Based on this assumption, we try tocollect enough data at the laboratory and intermediatescale, and subsequently use this data to develop field-scale flow and transport parameters. This problem af-fects, to some extent, our ability to model the engineeredbarriers in waste disposal systems, but it is even more ofan issue in modeling the response of the natural environ-ment.

For instance, the vadose-zone flow analyses are highlynonlinear; this fact, coupled with the extreme and com-plex spatial variability that characterizes most vadose-zone flow problems, makes identifying approaches for"averaging" local measurements to obtain field-scaleparameters a formidable obstacle. Of even more impor-tance for low-level waste performance assessment is thescale dependence of dispersion, and the uncertainty thatresults in trying to identify an appropriate dispersivity foruse in a performance assessment. Dispersion is one ofthe primary factors that can be used in reducing dosesfrom ingestion of contaminated ground water, but is diffi-cult to quantify owing to scale dependencies. Laborato-ry-scale dispersivities tend to be lower than field-scaledispersivities by orders of magnitude, but there is not anadequate way to quantify field-scale dispersivities withoutconducting field tests. Furthermore, measured dispersivi-ties in the field depend greatly on the technique used tomeasure them.

2.1.4 Treatment and Reduction of Parame-ter Uncertainty

Parameter uncertainty is treated by propagating the un-certainty through the model calculations to identify theeffects on uncertainty in model output. Extensive re-views of methods for propagating parameter uncertaintythrough models are given elsewhere [Doctor et al., 1988;Doctor, 1989; Maheras and Kotecki, 1990; Zimmermanet al., 1990]; therefore, we will not provide elaboratedetails here. Instead, we will focus on evaluating theapproaches for low-level waste performance assessment.As with other types of performance assessment uncertain-ties, the information produced in the parameter uncertain-ty analysis will be compared against a fixed, determinis-tic performance objective.

The result of accounting for input parameter uncertainty,with the exception of bounding analysis, is a distributionof doses; therefore, the issues discussed in the previoussection in relation to comparing a probabilistic answerwith a deterministic regulation become relevant. Asmentioned before, without any regulatory guidance, it isassumed that the fixed regulations cannot be exceeded.Therefore, the tail of the dose distribution obtained fromaccounting for parameter uncertainty must meet thedeterministic regulations. An alternative approach, men-tioned above, is to use some intermediate statistical mea-sure of the dose distribution. This approach is compara-ble to the EPA's guidance that the basis for comparisonbetween the deterministic Individual Protection Require-ment (which is dose based) and Ground-Water ProtectionRequirement (which is concentration based) in 40 CFRPart 191 is the greater of the mean or median of theoutput variable distribution [EPA, 1985a].

To represent the effect of input parameter uncertainty onmodeling results, the modeler must first quantify, thenpropagate the parameter uncertainty through the model toobtain model results. This may be accomplished in oneof several ways. One approach is to conduct "bounding"analyses, in which a conservative set of parameter valuesis used to produce conservative dose estimates. Themost common approach for treating parameter uncertain-ty is Monte Carlo analysis, which consists of selectingdiscrete sets of input parameter values from probabilitydistribution functions of the input variables, running eachset through the model, and constructing an output proba-bility distribution function that quantifies the uncertaintyassociated with the input. Another approach is perturba-

NUREG/CR-5927 12


tion analysis (also called analytical stochastic modeling).This approach is similar to Monte Carlo analysis, wheredistributions in input parameters are used to estimatedistributions in output parameters. However, based onsimplifying assumptions, the model equations and solu-tions are derived with the probability distribution func-tions for the input and output parameters explicitly in-cluded. Other approaches, such as the use of fuzzy settheory, are not well developed in the risk assessmentcommunity and, therefore, will not be discussed in thisreport.

2.1.4.1 Bounding Analysis

The current methodology is based on using boundingparameter values. In part, this approach was taken be-cause of the intended use of the methodology. Thepurpose of the methodology was for the NRC to conductconfirmatory analyses of a license applicant's evaluation[Starmer et al., 1988]. For this use, it may not alwaysbe necessary to conduct a full parameter uncertaintyanalysis, since the licensee should already have quantifiedthe parameter uncertainty and identified a conservativeset of model parameters.

Although adequate for NRC's purpose in some cases,generally there are several disadvantages associated withbounding analysis. First, to have confidence that a cor-rect set of conservative input parameters is chosen, itmust be compared with other likely sets of parametervalues. In treating parameter uncertainty by a boundinganalysis, it is assumed that the analyst can select theconservative combination of parameters a priori. In mostcases, particularly for nonlinear models, this prior identi-fication of the bounding parameters cannot be done.Therefore, one would have to go through some analysissimilar to Monte Carlo to estimate bounding parameters,and the advantage of bounding analysis (simplicity) islost.

A second drawback to bounding analysis is that usingonly a single realization of parameters reduces theamount of information available to the analyst and thedecision maker. This can be illustrated by consideringcalculated dose distributions from two hypothetical sites,as shown in Figure 2.1. A bounding analysis wouldsuggest that the two sites are similar: the standard isviolated, and the maximum doses are comparable. How-ever, there is clearly a distinction between the two cases.For Site A, there is a much higher probability that thestandard has been violated, which suggests that many setsof possible parameters produce a violation. In contrast,fewer sets of parameters produce the violation at Site B.

This suggests that more site characterization may be inorder to attempt to narrow the input parameter distri-butions. Further site characterization is less likely toproduce improvement in the analysis of Site A.

The goal of performance assessment should be to provideas much necessary information to the decision maker aspossible. A model prediction should be provided alongwith an estimate of the associated uncertainty in order tomaximize the information available to the decision maker[IAEA, 1989]. Furthermore, "no method based solelyon point estimates provides the decision maker with allthe available information on the nature and extent ofuncertainty, nor does it give decision makers or otheranalysts a window into the process to identify and criti-cize the assumptions made therein" [Finkel, 1990].Providing enough information so that it is easy to identifythe modeling assumptions and associated uncertainties isimportant, since the ultimate goal is public acceptance.Not only is it important to present the decision maker andthe public with as much information as possible, but thisinformation is also needed to guide data collection andvalidation efforts.

The above discussion suggests that while the use ofbounding parameter values may be appropriate in somesimple situations, in many practical cases the boundingparameter values cannot be specified a priori. Further-more, the use of a bounding analysis limits the amount ofinformation available to the analyst and decision maker.We, therefore, conclude that bounding analysis is not thebest available method for parameter uncertainty analysisin low-level waste performance assessment.

2.1.4.2 Monte Carlo Analysis

As discussed by Zimmerman et al. 11990), of the avail-able techniques for parameter uncertainty analysis, MonteCarlo analysis is the most versatile because (1) it facili-tates consistent propagation of uncertainties, (2) it can beeasily applied to a series of linked models, such as areused in low-level waste performance assessment [Kozaket al., 1990a], (3) it does not require modifications to theoriginal models; therefore, it is generally straightforwardto use, (4) it is capable of dealing with large uncertaintiesin the input variables, since it allows full stratificationover the variable ranges, and (5) it is appropriate for usewith nonlinear models [Helton et al., 1991]. The pri-mary advantage to conducting Monte Carlo analysis isthat it provides model results from a large number oflikely input parameter sets. Therefore, the output uncer-tainty is acknowledged, and there is some means foridentifying whether the output uncertainty from input

13 NUREG/CR-5927


SITE A

W00Li_0

-j

0cc

wC..00U-

.--0<0n

DOSE DOSE25 MREM 25 MREM

Figure 2.1 Comparison between two hypothetical dose distributions

parameter uncertainty has been bounded. Clearly, wehave more confidence that the output has been boundedwith increasing numbers of samples. Another advantageto this approach is that sensitivities of the model output toinput parameter variations may be identified [Zimmermanet al., 1990]; this allows the analyst to identify importantmodel parameters for future data collection efforts.

The primary disadvantages usually cited for Monte Carloanalysis are that many realizations of the data are re-quired to span the input data range, and the parametersmust be treated as uncorrelated [Harr, 1987]. However,both of these problems have been addressed to someextent. The required number of realizations can begreatly reduced by using a stratified sampling strategy,such as the Latin Hypercube Sampling method [Iman etal., 1981]. Methods are also available that allow theanalyst to introduce correlations among variables [Iman

and Conover, 1982], and these methods are included inthe computer implementation of Latin Hypercube Sam-pling [Iman et al.,1981]. Even with these techniques,Monte Carlo analysis can still require extensive computereffort.

2.1.4.3 Analytical Perturbation Methods

Another approach that has been proposed involves usingperturbation analyses (often called analytical stochasticmethods) for ground-water flow and transport calcula-tions [Polmann et al., 19881. The primary reasons forusing perturbation analyses are (1) to develop a computa-tionally efficient approach, and (2) to attempt to developimproved conceptual understanding of ground-waterprocesses by obtaining closed-form solutions to the gov-erning equations. For instance, it is well known that thedispersivities obtained from laboratory column studies

NUREG/CR-5927 14


may differ substantially (by orders of magnitude) fromthose observed in field studies. Analytical perturbationmodels have been developed that can in principle, beused to predict field-scale dispersivities from local (i.e.,laboratory scale) measurements [Gelhar and Axness,1983; Dagan, 1984; Naff et al., 1988]. Other studies(Vomvoris and Gelhar, 1990] provide information on theexpected variance of concentration at a given samplingpoint. Yet other examples relate to predictions of mois-ture-dependent hydraulic conductivity anisotropy in unsat-urated heterogeneous media [Yeh et al., 1985].

2.1.4.4 Recommended Approach for Parameter Un-certainty

We recommend that parameter uncertainty analysisshould be addressed using Monte Carlo analysis coupledwith Latin Hypercube Sampling. This approach is usedextensively, and has been recommended for high-levelwaste performance assessment [Davis et al., 1990b].Use of this approach will provide the decision makerwith considerably more information relative to boundinganalysis, and more importantly, it clearly acknowledgesand communicates the uncertainty associated with themodel output variable due to input parameter uncertainty.It also provides some basis for assessing whether or notthe model output distribution has been bounded. How-ever, as discussed in the previous section, determiningwhether the complete dose history output distributionmust fall below the regulatory performance measure, asopposed to some statistical measure of the distribution(e.g., mean, median, 95% confidence limit) is entirely aregulatory decision.

The only available method to reduce uncertainties associ-ated with data insufficiency is the collection of additionaldata about a parameter. However, since most of theparameters of interest are potentially spatially and tem-porally variable, complete elimination of this source ofuncertainty is impossible.

There is not a generally appropriate way to eliminatesystematic errors from hydrological parameters. Thebest way to verify that systematic errors are small is toattempt a validation of the overall analysis. However,this approach is hindered by (usually) ambiguous valida-tion results, by spatial variability of the parameter, andby the unavoidable convolution of potential systematicparameter errors with possible model errors. Whenmultiple measurement approaches are available for aparameter, as many as possible should be used, and thiswill provide some confidence that systematic errors areinsignificant. However, these types of potential errors

cannot be easily quantified or identified, and the analystshould recognize this additional uncertainty in parametervalues and distributions.

2.1.5 Incorporating Uncertainty Analysisinto the Methodology

We have identified the need for regulatory guidance inthe interpretation of the performance objectives in 10CFR Part 61.41, particularly for incorporating futureuncertainties in a comprehensive manner. Specifically,there needs to be a regulatory determination about wheth-er the dose limits are absolute bounds that cannot beexceeded, or if some measure of the statistical distribu-tion of doses can be used for comparison with the deter-ministic standard. The choice of uncertainty analysismethods to be incorporated into the methodology dependson this regulatory decision. Consequently, full imple-mentation of uncertainty analysis methods into the low-level waste performance assessment methodology cannotcommence until this regulatory guidance is available.

If NRC chooses to interpret the performance objectivesas relating to a statistical measure of the dose distribu-tion, the next step would be to identify that measure ofthe distribution to use. As mentioned in the previoussection, one option is to use the mean or the median,whichever is higher; this option was used by EPA forassessing compliance with the Individual Protection Re-quirement and the Ground-Water Protection Requirementin 40 CFR Part 191 [EPA, 1985a]. However, any statis-tically defined measures of the dose distribution, such asconfidence limits or percentiles, can be used.

The choice of what confidence limit to use is a regulatorydecision, but some guidance may be available based onpragmatic grounds. Analyses of ground-water traveltimes performed at SNL for the Basalt Waste IsolationProgram suggested that 85% confidence limits wererelatively robust with respect to the number of parametersets used to develop the output distribution, but that 90%and higher confidence limits required large numbers ofparameter sets to demonstrate stability of the confidencelimit value.' Similarly, Berthouex and Hau [1991] haveshown that using high percentiles can lead to high risksof making erroneous decisions in both parametric andnonparametric analyses.

In the previous sections, we have summarized approaches

for conducting uncertainty analysis for model uncertainty,

'Davis. P.A., personal communication, SNL, 1991.

15 NUREG/CR-5927


parameter uncertainty, and uncertainty about the future ofthe site and recommended methods to be used for low-level waste performance assessment. These approachesare reviewed in Table 2. 1.

Conceptual model uncertainty is addressed by identifyinga broad range of possible conceptual models, and usingeach in performance assessments. Revisions of thesemodels are made by accounting for progressive datacollected, which can be used to eliminate some modelsfrom consideration. Uncertainties about the future of thesite are addressed by acknowledging alternative futureconditions of the site. Parameter uncertainty is addressedby using Monte Carlo analysis. Latin Hypercube Sam-pling of the input parameter space is recommended to re-duce the computational effort associated with the analy-sis.

2.1.5.1 Implementation

In the following discussion, we assume that a scenarioapproach is adopted for treating future uncertainties. Asdiscussed above, it is not clear that this is the best ap-proach, nor is it clear that it will fit the regulatory philos-ophy that may be adopted by NRC. We use it here forsimplicity of demonstration of the role of future uncer-tainties on the performance assessment.

The first step in implementing the treatment of uncer-tainty in the methodology is to formulate an initial suiteof future scenarios, conceptual models, and associatedparameter distributions for use in a preliminary perfor-mance assessment. These initial postulates should beconsistent with available data, but should be as broad aspossible within the constraints set by the data. An ap-proach to broadening the interpretations is to get inputfrom a large number of independent investigators. Wemight even invite the technical advisors of intervenerorganizations to participate in the preliminary data inter-pretation for conceptual model development. Such anapproach would allow concerns of these organizations tobe addressed during the performance assessment process;this approach may be very useful in developing an even-tual political consensus about the results of the perform-ance assessment. To achieve this political consensus, itis necessary to build consensus about the decision-makingprocess beforehand. In other words, participation ofintervener organizations only makes sense if they acceptthe technical approaches in the performance assessmentprocess as the basis for the decision.

The overall approach to the uncertainty analysis is shownin Figure 2.2. A suite of future scenarios is identified by

the analysts. For each scenario, several conceptualmodels may potentially be posed. To analyze the param-eter uncertainty associated with each conceptual model, anumber of realizations are necessary from the parameterdistributions, which is depicted as the bottom tier ofblocks in Figure 2.2. These realizations are the inputrealizations to Monte Carlo analyses of the parameteruncertainty associated with each conceptual model. Oncethese realizations have been propagated through themodel, their results represent the total information avail-able to make the regulatory decision.

The approach taken to use this information for compari-son against the performance objectives is the root of theregulatory issue related to uncertainty analysis. Forinstance, NRC may choose to screen some or all alterna-tive scenarios a priori using expert judgment and sitingcriteria. In this case, the analysis would be reduced to afew (or one) scenario branches with their associatedconceptual model and parameter uncertainties. Alterna-tively, NRC may choose to evaluate all scenarios, theneliminate all or some of them a posteriori based on plau-sibility arguments, and make the regulatory decision onthe most conservative of the remaining scenario(s). Thisapproach was taken in the analysis of the Below-GroundVault Prototype License Application (PLASAR) [Rogersand Associates, 1988], but not in this formal way. In thePLASAR, a "sensitivity" analysis was performed on anincreased rainfall scenario, but the results were onlyreported in an appendix, and were not used in the mainbody of the report. This approach corresponds to aposterior screening of the analysis results. The advan-tage to this approach is that the full set of information ismade available to the decision maker, who can thereforenot be criticized for omitting some conditions. A thirdalternative is to combine all scenarios in a single proba-bility distribution by assigning likelihoods to each scenar-io. The advantage to this approach is that it combines allinformation into a single curve from which a decision canbe made. Disadvantages to this approach have beenreviewed by Andersson et al. [1989], in the high-levelwaste context; these difficulties include problems inensuring the scenarios are mutually exclusive and inde-pendent (which are requirements for combining them intoa comprehensive distribution), and time dependence andordering issues. In addition, as discussed above, thelikelihoods of the scenarios are often related more closelyto their duration than to their probability of occurrence,and this approach is not appropriate for a dose-basedstandard. A fourth alternative is to use the approach ofChhibber et al. [1991a), and combine scenarios and con-ceptual models by associating a subjective likelihood witheach. This approach has the same difficulties and

NUREG/CR-5927 16


Table 2.1 Summary of approaches to uncertainty treatment

Type of Uncertainty Proposed Approach

Model uncertainty 1. Identify multiple conceptual models2. Use all models in performance assessment3. Use performance assessment results to identify if

further data collection is necessary4. Use the most conservative model for comparison

with the performance objectives

Uncertainty about the future Dependent on regulatory decisions

Parameter uncertainty Monte Carlo analysis with Latin Hypercube Sampling

disadvantages as the third alternative approach discussedabove, but in this case the problems are much greater inassuring that conceptual models are complete, mutuallyexclusive, and independent. These four alternatives aresummarized in Figure 2.3.

We emphasize that, conceptually, the overall group ofrealizations shown in Figure 2.2 is the full amount ofinformation that the decision maker must deal with,either explicitly or implicitly. Some scenarios, such aserosion, may be potentially eliminated a priori throughsiting criteria. Others, such as climate change, may bepotentially incorporated into a base scenario with widerparameter value distributions (such as a broader infiltra-tion distribution). Still others may be eliminated by theregulator, such as a decision not to include glaciation.The decision left to the NRC is whether it is appropriateto eliminate scenarios using expert judgment (and if so,which ones). If certain scenarios are to be eliminated,then regulatory guidance will be required to identify thatconditions are to be retained for analysis.

2.1.5.2 Identification of Input Parameter Values

Once the initial set of conceptual models and parameterdistributions has been identified, each of the modelsshould be run through an initial performance assessment,accounting for parameter uncertainty by Monte Carloanalysis. From a mechanical standpoint, parameteruncertainty represents the bulk of the difficulty of imple-menting uncertainty analysis.

The analyst must identify probability distributions associ-ated with each input variable. These input distributionscan be of any form, but the form should be dictated bythe information available about the variable. As dis-cussed by Harr [19871, the amount of data available

about an input distribution, together with the principle ofmaximum entropy, may be used to dictate the form ofthe distribution. Simply stated, the principle of maxi-mum entropy identifies the least biased distribution to bethe one that maximizes entropy subject to the availabledata. The distributions that fulfill this condition areshown in Table 2.2.

A second approach used to specify input distributions forsparse data sets involves using expert opinion. Probabili-ties derived by this approach are known as "subjectivist"or "personal opinion" probabilities [Berman, 1988]. Itshould be understood that when input parameters arespecified in this manner, the output distribution also takeson the significance of a "personal opinion" probability[Vaurio, 1990]. Consequently, such an output distribu-tion should not be interpreted as an actual frequencydistribution of doses, but rather a quantification of exist-ing expert opinion about doses. We note that all uncer-tainties about the site future and about conceptual modelsare treated by subjectivist approaches, and that the over-all interpretation of the performance assessment results is(and must be) subjectivist because of the nature of theinformation available for the analysis [Atwood, 1988;Apostolakis, 1988]. Consequently, we should not beadverse to using subjectivist input parameter distribu-tions, if required; however, their use should be mini-mized, since the amount of confidence we have in adistribution is directly related to the amount of data uponwhich it is based.

Once the continuous distributions have been specified, theanalyst must sample from them to identify sets of param-eters to be used as values in the models in the methodol-ogy. Ideally, enough realizations should be taken tocompletely span the range of input values. Practically,however, this ideal cannot be met. The primary disad-

17 NUREG/CR-5927


TREATMENTOF FUTUREUNCERTAINTY

TREATMENT OFMODELUNCERTAINTY( !

I. SET 1 PARAM. SET 1 PARAM. SET 1 PARAM. SET 1 PARAM. SET 11. SET 2 PARAM. SET 2 PARAM. SET 2 PARAM. SET 2 PARAM. SET 2 TREATMENT OF1. SET 3 PARAM. SET 3 PARAM. SET 3 PARAM. SET 3 PARAM. SET 3 PARAMETER1. SET 4 PARAM. SET 4 PARAM. SET 4 PARAM. SET 4 PARAM. SET 4 UNCERTAINTY

1. SET N PARAM. SET N PARAM. SET N PARAM. SET N PARAM. SET N

FOo ECh to N ufo

Figure 2.2 Overall approach to uncertainty analysis for low-level waste performance assessment

NUREG/CR-5927 18


IDENTIFY FUTURE SCENARIOS

FOR EACH SCENARIO, IDENTIFY CONCEPTUAL MODELý

IFOR EACH CONCEPTUAL MODEL, IDENTIFY PARAMETER DISTRIBUTIONS

IPROPAGATE PARAMTE UNETAINTY FOR EACH CONCEPTUAL MODEýj

____ ICOMBINE RESULTS COMBINE RESULTSUSING PROBABIUTIES USING PROBABILITIESFOR BOTH FOR SCENARIOS ANDCONCEPTUAL MODELS ONLY THE MOSTAND SCENARIOS CONSERVATIVE

CONCEPTUAL MODELS

11!

PRESENT ALL RESULTS, PRESENT ONE RESULT,INTERPRET BY EXPERT USE THE REMAINING ASJUDGMENT JUSTIFICATION

$MAKE DECISION

Figure 2.3 Optional approaches for combining information to make the regulatory decision

Table 2.2 Least biased distributions for varying amounts of available information[adapted from Harr, 1987]

Available Data Assigned Distribution

Minimum and Maximum Values UniformExpected Value ExponentialExpected Value, Variance NormalExpected Value, Variance, Maximum and BetaMinimum Value

19 NUREG/CR-5927


vantages cited for Monte Carlo analysis are that manyrealizations of the data are required to span the input datarange, and that the parameters must be treated as uncor-related [Harr, 1987]. Even with Latin Hypercube Sam-pling, many realizations are needed to establish the tailsof the output distribution with confidence.

In the ideal situation, this initial performance assessmentanalysis may comprise a sufficient basis from which tomake a regulatory decision. The initial scenarios, con-ceptual models, and parameter ranges presumably includequite conservative conditions, owing to the inclusion ofmultiple conceptual models and broad data ranges.Consequently, if the performance measures are not ex-ceeded, the regulator can make a relatively confidentdecision. However, some site-specific validation will beneeded even in this ideal case to ensure that the initialconcepts about the site are conservative.

A more likely situation will arise when this initial per-formance assessment does not meet the performanceobjectives for some of the more conservative conceptualmodels or parameter set realizations. The task of theanalyst is then to gather more information about the siteto support or refute the conditions of concern. An appro-priate approach to complete this task is discussed inSection 2.2.1 below.

2.1.5.3 Propagating Input Parameter UncertaintyThrough the Model

Conducting parameter uncertainty analyses for eachconceptual model, and for multiple scenarios, can poten-tially consume large amounts of computer time. Ofparticular concern is the application Monte Carlo analysisto vadose-zone flow modeling, since the vadose-zoneflow equations are strongly nonlinear. In the methodolo-gy, two-dimensional vadose-zone flow modeling is usedto estimate flow into and through the disposal units.Current disposal technologies include extensive use ofconcrete structures, and multilayer "capillary break"covers; consequently, the flow analyses tend to be char-acterized by neighboring soil layers with sharply contrast-ing properties. These problems typically require manyiterations before they are solved, and consume largeamounts of time for each realization. Furthermore, it iscommon to be unable to achieve numerical convergencefor some combinations of parameters.

These issues identify a practical obstacle to using MonteCarlo analysis for vadose-zone flow modeling; the analy-sis can still be done by brute force, simply grindingthrough many realizations of a full multidimensional flow

analysis; however, this approach may require such largeamounts of computer time that it becomes impractical.In such cases, a simpler flow model can be used. Forinstance, the multidimensional flow model may be re-placed by a simple water flux through the waste that isallowed to vary over a wide enough range to representthe uncertainties. The primary disadvantage to thisapproach is the difficulty in setting "reasonable" limits onthe uncertainty in flow rate: the distribution will general-ly have to be broader than if the full-flow analysis can bedone. For instance, an arguably conservative flow ratethrough the vault might be the undisturbed infiltrationrate, and this could be used as one bound on the flowrate distribution. If this is determined to be excessivelyconservative, the analyst might then be justified in model-ing the disposal facility in more detail, and expending theextra effort to conduct the full-flow analysis.

Monte Carlo analyses are becoming increasingly moreattractive, since computing costs are currently droppingdramatically, and computing speeds are increasing equal-ly dramatically. Parallel computing is becoming avail-able for workstations, and even for personal computers,which will further improve computing speeds. Conse-quently, although these problems currently exist, theconcerns about Monte Carlo analysis for vadose-zoneflow are diminishing.

The remaining models in the performance assessmentmethodology will be much easier to use in Monte Carloanalyses, since they are linear and, consequently, do notrequire time-consuming, iterative solutions. Ground-water transport analyses are linear, assuming that onlylinear sorption terms are included in the retardation term;food-chain, dosimetry, surface-water transport, and airtransport models are all simple and linear. However,transport analyses of decay chains, even though they arelinear, are more time consuming than single-componentanalyses because of the coupling of the transport equa-tions. Nevertheless, analysis of decay chain transport isusually very fast compared to the vadose-zone flowanalysis, and should not constitute a major obstacle toimplementation of Monte Carlo analysis.

The Monte Carlo analysis should be integrated into themethodology in a user-friendly manner. In the currentstructure of the methodology, the best approach would beto implement a sampling program separately, as an addi-tion to the suite of codes in the methodology. A stand-alone code is available to implement Latin Hypercube orrandom sampling [Iman et al., 1981]. This code wouldbe run first to identify the parameter values needed in theother codes in the methodology. The performance as-

NUREG/CR-5927 20


sessment would then be run for each set of parameters.This process would have to be repeated for all conceptualmodels and future scenarios. Other options for imple-menting the uncertainty analysis are discussed in thefollowing section.

2.2 Reduction of Uncertainty

The primary purpose of low-level waste performanceassessment is to develop a confident regulatory decisionabout compliance of the disposal facility with the per-formance objectives in 10 CFR Part 61.41. Regulatoryconfidence should be understood to be different thanscientific confidence in the context of evaluating perform-ance assessment results. There may be significant uncer-tainty predicting in an absolute sense what the maximumdose will be in the future: this is scientific uncertainty.Regulatory confidence is concerned with information thatmay influence the regulatory decision. Scientific uncer-tainty may be irrelevant in making a regulatory decision,provided that there is a confident upper bound to thedoses, and that regulatory requirements are met. In sucha case, there is adequate confidence in the regulatorydecision, even if there is large scientific uncertainty aboutthe absolute predicted dose. Therefore, the performanceassessment objective is to strive for minimal likelihood ofaccepting a site that will not meet the performance objec-tives. In this and subsequent sections, a process is de-scribed that meets this objective in an efficient and defen-sible manner.

The process of performance assessment has several sa-lient features. (1) It should be iterative. During the firstiteration, the analyst begins with available informationand uses models that are conservative relative to thatinformation. Subsequent iterations serve to relax conser-vative assumptions based on new information, such asnew site characterization data, new design features, or are-evaluation of existing information. As a side benefit,this approach minimizes the resources needed to performsite characterization. More importantly, it providesregulatory defensibility where and to the extent it isneeded. The iterative nature of the process lends itself todefining endpoints. That is, performance assessment andsite characterization should only be conducted to theextent needed to meet the performance objectives. Onceadequate confidence is generated, the process may end.On the other hand, the process may end if the site devel-oper finds that it is too expensive to collect enough datato produce adequate confidence. In either case, there is afinite and well-defined endpoint to the site characteriza-tion and performance assessment. (2) It provides a

mechanism for using both generic and site-specific infor-mation in a defensible manner. (3) It integrates aspectsof the facility development that have heretofore beenconsidered separate from performance assessment. Theseaspects are site selection, facility design, site character-ization, "ability to model the site," and "validation,"which we take to mean development of regulatory confi-dence. (4) Unlike past guidance, it is a formal treatmentfor uncertainty analysis and is an intrinsic part of theprocess. Performance assessment is viewed as a methodfor quantifying the uncertainty associated with meetingthe performance objectives, and for reducing that uncer-tainty to an acceptable level. The reduction of uncertain-ty relative to the performance objectives is an intrinsicpart of the process.

The process also helps ensure that the performance as-sessment is substantially complete and defensible whenthe license application is submitted. Emphasis is placedon considering a broad range of potentially adverse con-ditions, then screening those conditions using subsequentsite investigations. Confidence in the results is intrinsi-cally built into the process, since each modification andthe reasons for it are well documented. It is highlydesirable to make the process participatory, in which caseimproved political acceptability may be generated as partof the process. Participatory performance assessment isalso expected to lead to improved technical breadth anddefensibility.

2.2.1 The Process of Performance Assess-ment

Performance assessment is a site characterization andmodeling activity directed toward evaluating complianceof a disposal facility with the performance objectives in10 CFR Part 61.41. In the following discussion, itshould be understood that the approach to site character-ization described here is directed to this end. Site char-acterization undertaken for monitoring or other activitiesis outside the scope of this discussion.

The overall process of performance assessment is shownin Figure 2.4. In this discussion, a completely participa-tory process has been assumed. A participatory processis defined as one in which all interested parties (stake-holders) participate in each step of the process as equalpartners in developing and refuting conceptual models.In the absence of this type of approach, the process isstill recommended. However, differences between par-ticipatory and nonparticipatory approaches arise at the

21 NUREG/CR-5927


START)

END

...IYES

.END)

(ACCEPT SITE*)REJECT SITE)

Figure 2.4 General approach to performance assessment

decision points. These will be discussed at the appropri-ate points in the following section.

Initial Assessment: Performance assessment begins withan initial assessment of available information about thesite, and is directed toward developing an evaluation ofthe existing knowledge of the site and facility. It isimportant to note that there is at least some informationavailable about all parts of the U.S. For instance, geo-logical maps, regional hydrology, and weather data aregenerally available. Therefore, some minimal informa-tion will be available even in the worst cases. In manycases, more detailed information will also be available.This step is intended to represent a data collection activi-

ty, and an assessment of which aspects of the data aresparse. For sparse data, conservative assumptions areappropriate.

Initial Assumptions: The initially available informationis next developed into assumptions used in specific con-ceptual models about the behavior of the system. Theconceptual models should be as broad as possible withinthe constraints of the available information. This meansthat when the initial assessment shows that only sparsedata are available, the initial conceptual models shouldinclude much more conservative conditions than if moreinformation is available. The conservatism and breadth

NUREGICR-5927 222


CONCEPTUAL CONCEPTUALMODEL 1 MODEL 2

PAAM. SETI1 PARAM. SET IPARAM. SET 2 PARAM. SET 2PARAM. SET 3 PARAM. SET 3PARAM. SET 4 PARAM. SET 4

CONCEPTUALMODEL 3

CONCEPTUALMODEL 4

PARAM. SET 1PARAM. SET 2PARAM. SET 3PARAM. SET 4

CONCEPTUAL

MODEL 5

ITREATMENT OF",( MODEL,UNCERTAINTY,,



PARAM. SET 0

/

( TREATMENT OFPARAMETERUNCERTAINTY )

PARAM. SET K PARAM. SET L PARAM. SET M PARAM. SET N

DECISI ON

Figure 2.5 Uncertainty analysis for models and parameters

of conceptual models invoked at this stage reflects thelevel of initial uncertainty in the behavior of the system.

These uncertainties result in the potential for multipleconceptual models of the site, and for broad distributionsin input parameters for the models. The structure of thisuncertainty is depicted in Figure 2.5. In this case, fivealternative conceptual models with differing parameteriza-tions have been generated by the analysts. It should benoted that in some cases the division between differentconceptual models and different parameter distributionsmay be artificial. The division between models andparameters is not always clear; however, the intent here

is to illustrate differences between fundamentally differ-ent concepts of the behavior of the system, in contrast todifferent parameterizations. These conceptual modelsshould be as conservative as possible within the con-straints imposed by the understanding of the site.

Parameter distributions for each conceptual model shouldalso be established at this point. As in the developmentof conceptual models, the goal at this stage is to retainthe broadest ranges possible within the limits of availableinformation. If no site-specific data are available for aparticular parameter, the range can be set as the maxi-mum physically possible, or the maximum ever measured

23 NUREG/CR-5927


for similar conditions. These extremes may be moderat-ed as needed as the process progresses, and more infor-mation is collected. The important point here is that if abroad range of a parameter is used and it does not influ-ence the regulatory decision, it does not need to be stud-ied further. Also, as more data are collected, it may bepossible to justify correlations among variables as need-ed. Such correlations may also depend on the modelsbeing implemented.

This stage in the process is where the participatory natureof performance assessment makes its first large impact.The breadth of conceptual models and parameters used inthe analysis is directly related to the ultimate defensibilityof the analysis. The breadth of analyses, in turn, isrelated to the number of analysts, and to the eliminationof potential bias toward nonconservatism in developingthe models. A license applicant, working independentlyat this stage, may have an inherent bias that precludesconsideration of all possible adverse conditions in devel-oping conceptual models. However, if the license appli-cant solicits conceptual models from regulators and otherinterested parties at this stage, there is much greaterassurance that the range of potential conceptual modelshas been spanned. Also, by including analysts that tendto be more risk averse, there will be a greater tendencyto capture more conservative models.

Also, it is important at this stage to emphasize the impor-tance of retaining the breadth of possible conceptualmodels. It is not typically possible to identify the relative

conservatism of conceptual models or parameter setsbefore the analysis is run. Therefore, conservatismamong analyses must be identified after the analyses areperformed.

A final consideration related to the eventual defensibilityof the analyses is that independent quality assuranceauditors should be able to trace all modeling results, thusdemonstrating that they can be reproduced.

Formulate Models: At this stage of the analysis, theanalysts formulate mathematical representations of their

conceptual models. These mathematical expressions willusually be represented and solved in particular computercodes. However, the representation of conceptual modelsshould never be constrained by the limitations of somecomputer code, simply because it is available or easy touse. Implementation of models should be done based onsite-specific physical and chemical process consider-ations, not on code capabilities. At its worst case, thismeans that the analyst must develop a computer code for

the express purpose of evaluating a particular conceptualmodel. However, it is expected that this level of effortwill rarely be necessary, since a large number of comput-er codes exist, and these can be used to represent a broadrange of potential conceptual models.

Consequence Modeling: The purpose of consequencemodeling is to propagate the parameter uncertainty asso-ciated with each conceptual model through the mathemat-ical models, in such a way that a distribution of calculat-ed doses is produced for each conceptual model. Oneapproach for propagating parameter uncertainty throughthe models is to use Monte Carlo analysis. However,this is not necessarily the only acceptable approach [Zim-merman et al., 1990].

Evaluation of Adequacy: The evaluation of adequacy issomewhat different, depending on whether a participatoryor nonparticipatory performance assessment approach isused. If a participatory approach is used, the assessmentof adequacy relates only to a simple comparison betweenthe consequence analyses and the performance objectives.The reason for this simplicity is that any technical con-cerns by any interested parties have already been ad-dressed as part of the performance assessment process.The decision to be made is between accepting the per-formance assessment if the performance objectives havebeen met, or driving further information collection if theyhave not.

By contrast, if a nonparticipatory approach is used, theperformance assessment is presented to the regulator fora regulatory decision at the end of the process. Theregulator must then review the initial conceptual modelsand parameter distributions for completeness and conser-vatism. The regulator must also evaluate the justifica-tions for changing assumptions and moderating conserva-tism as part of the process. At any stage of the processif the justification is inadequate, the regulator may eitherrequire collection of more data, or may reject the per-formance assessment.

Sensitivity Analysis: Sensitivity analysis is performedon the consequence analysis results to evaluate whichmodels and combinations of parameters were most signif-icant in producing results in excess of the regulatoryperformance objectives. The primary functions of thisstep are to (1) identify data and assumptions that affectthe regulatory decision for careful scrutiny, (2) optimizefunds by specifying the most important information to becollected to reduce regulatory uncertainty, and (3) identi-fy which assumptions and parameters do not influence the

NUREG/CR-5927 24


results. The first and third functions are important to allinterested parties in the process; the second function isimportant to the site developer.

Sensitivity analysis is not a completely necessary step,because data could be collected randomly that would stillallow the analyst to modify assumptions. However, suchan approach is not conducive to optimal use of resources,nor does it build defensibility into the process.

Sensitivity analyses for identifying important parametershave been used in related high-level waste performanceassessment problems [see, e.g., Bonano et al., 1989].However, standard techniques have not yet been identi-fied for evaluation of sensitivities relative to peak doseperformance measures, nor for the evaluation of sensitiv-ities to conceptual model (assumption) variations.

Data Worth Analysis and Evaluation of Cost Effec-tiveness: Data worth analysis has two functions in thisprocess. Both functions are performed by the site devel-oper. The first function is to identify which informationcan be produced at the least cost. Again, this function isrelated to optimal use of resources. The second functionis more serious. If the data needed to eliminate a con-ceptual model or parameter range from consideration arevery extensive, owing to site complexity or other factors,it may be more cost effective to reject the site and pro-ceed with another site.

In a completely participatory performance assessmentprocess, this is the only criterion that is used to reject asite. That is, the other participants can continue to rec-ommend new iterations indefinitely, and it is only the sitedeveloper that decides to reject the site based on econom-ic considerations. In a nonparticipatory process, this isone of two rejection points. The other point is if thelicense applicant, while working independently, has madeunacceptable conclusions.

Gather New Infornation: Once sensitivity analysis hasidentified the critical information needed to reduce regu-latory uncertainty, that information must be gathered.New information should not be gathered simply to under-stand the site without regard to reducing regulatory un-certainty. (The reader is reminded that this overall ap-proach is directed toward the needs of performance as-sessment. The needs of monitoring or other technicalareas must be addressed independently. In general,modeling to support a monitoring plan may need moreextensive data. Such data may be incorporated into theperformance assessment structure without difficulty, but

are not necessarily required by performance assessment.)Information can be one of four types: new site character-ization data, new facility design, adjunct modeling stud-ies, or new design basis information. New site charac-terization data may be generated by, for instance, drillinga new well. A new facility design might influence, forexample, how barrier degradation is modeled, or it mightalso permit geochemical modeling to be used to allowcredit for reduced solubility limitations. New designbasis information might consist of specifying inventorylimitations to reduce calculated off-site doses in the sub-sequent iteration.

Any of these sources of new information can drive thesubsequent consequence analysis iteration. Deciding thatthe previous iteration was "too conservative" without newjustification is an inadequate reason for modifying as-sumptions.

Update Assumptions: The principles of this step are thesame as the initial assumption step. In this case, howev-er, assumptions are modified based on a larger knowl-edge base. Subsequent model formulation may involveelimination of a conceptual model, modification of aconceptual model, or addition of a new model introducedby new information. However, it should be noted that ifthe initial step included a broad range of conceptualmodels, the updated models should always trend towardless conservatism. They should still always be conserva-tive relative to the information available when they areformulated.

2.2.2 Defensibility of Analyses

A common misconception about assessing low-levelwaste disposal facility performance is that performanceassessment is conducted after the conclusion of site char-acterization, and when the facility design is completed.In this approach, site investigations are vaguely directedat attempting to "scientifically understand" the site, andthere are no feedback loops to assess the progress of siteinvestigations at resolving performance issues. The risksof using this approach are (1) data needed to establishdefensibility in the performance assessment may not becollected, (2) too many data of some particular varietymay be collected, resulting in wasted resources, (3) theinvestment of large amounts of resources prior to assess-ing safety may produce a large incentive (or at least theperception of an incentive) to demonstrate facility safety;that is, the process may become biased, and (4) data maybe forced to fit the input needs on a preselected suite ofmodels, while alternative data interpretations are not

25 25 NUREGICR-5927


encouraged. Consequently, this approach is not condu-cive to building defensibility in performance assessments.

In contrast to this approach, the process described inSection 2.2.1 intrinsically builds confidence as subse-quent iterations are performed. Defensibility is producedby using formal uncertainty analysis, which quantifies theeffect of alternative conceptual models and parameterdistributions on calculated doses. Confidence is devel-oped since the range of potential adverse conditions hasbeen identified, and progressively screened based on site-specific information. Those aspects of the model thatlead to violation of the performance objectives can beidentified, and site investigations can be performed torevise assumptions about those aspects. Confidence isproduced by the visibility of this process, in that it shouldbe made apparent how alternative assumptions have beenrefuted.

The process works best to build confidence if it is per-formed in a completely participatory manner. That is, allinterested parties may propose potential adverse condi-tions that are evaluated as part of the performance assess-ment process. If those concerns are addressed as part ofthe process rather than in an adversarial manner aftercompletion of performance assessment, two importantissues may be addressed. (1) it is much more likely thatthe analysis will be technically complete and defensible,spanning a more complete range of potentially adverseconditions, and (2) if likely opponents have their con-cerns addressed during the process, there should be feweradversarial hurdles to clear at the end.

If a participatory approach is not used, this process isstill recommended for use by the license applicant.However, the role of the regulator must then be to evalu-ate in detail the span of models and parameters used ineach iteration to ensure completeness and conservatism,and to evaluate the justification for modifying assump-tions at each step of the process. That is, the licenseapplicant should be prepared to identify which adverseconditions were considered, and how those conditionswere addressed as part of the process. The risk of thisapproach, from the applicant's standpoint, is of workingin a vacuum of other ideas, and possibly not consideringsome adverse issue, which is then introduced by regula-tors or interveners at the end of the process instead of atthe beginning.

The performance assessment process discussed in Section2.2. 1 also builds confidence by combining site character-ization activities with performance assessment modeling.Performance assessments developed in this way are as-

sured to be site-specific modeling exercises, not genericmodels of dubious site-specific applicability.

The decision that the site meets the performance mea-sures is the final step in the process, and is expected tobe the result of a number of iterations undertaken torefute particular conservative conceptual models or pa-rameter distributions. As part of that process, the initial-ly conservative models have gone through successivescreening and revision. The result is still a set of conser-vative models and parameters, but they are conservativerelative to a new, broader set of site-specific information.In a participatory performance assessment, any decisionto reject a site will be made by the license applicant, aspart of a cost analysis that shows that confidence cannotbe built for a particular site within reasonable costbounds. One example of this might be a complex site inwhich an extensive site characterization would be re-quired to refute a conceptual model. This is a practicaldefinition of the requirement in 10 CFR Part 61 that thesite must be "modelable."

2.3 Summary of Uncertainty Anal-ysis Recommendations

The proposed treatment of uncertainty described in thepreceding section is the most important and far-reachingrecommendation in this report. Consequently, it isworthwhile to reiterate the most important points aboutthe proposed treatment of uncertainty.

First, the methodology would benefit from a formaltreatment of uncertainty. Such an analysis would (1)provide confidence that all important parameter rangesand models have been explored, (2) provide a wealth ofinformation to decision makers about the results of theanalysis, including the amount of uncertainty in the re-sults, and (3) potentially promote a common ground uponwhich all interested parties can agree.

The uncertainty analysis should take the following form.Parameter uncertainty should be addressed by MonteCarlo analysis. For efficiency, Latin Hypercube Sam-pling should be used to reduce the number of realizationsneeded for the analysis. Conceptual model uncertaintyshould be addressed by including a broad set of conceptu-al models in the analysis. From the perspective of thelow-level waste regulations, it is undesirable to pursuethe idea of assigning probabilities to conceptual models.The decision should be made based on the most conser-vative of the conceptual models that cannot be eliminatedusing site-specific data. Conservatism here is defined

NUREG/CR-5927 26


after all models have been carried through the analysis.The most difficult issues in low-level waste performanceassessment uncertainty analysis are related to futureuncertainties. It was concluded that a scenario approachcould be used in analyses for comparison with the per-formance objectives in 10 CFR Part 61.41, but only if aportion of the dose probability distribution is permitted toexceed the performance objectives. The scenario ap-proach was shown to be more appropriate for an integrat-ed discharge performance objective than for a dose-basedperformance objective. A possible alternative approachmight be to perform time-dependent future histories ofthe site; this approach is expected to be both self-consis-tent and appropriate, but will make the analyses morecomplex.

Prior to full implementation of uncertainty analyses inlow-level waste performance assessment, regulatoryguidance is needed from NRC in areas that are morerelated to regulatory philosophy than to technical consid-erations. Guidance is needed on the following topics:

How to treat uncertainties about the future: First,there needs to be a determination about the requiredduration of the analyses. Second, guidance is neededabout which extreme (but likely) events are to be includ-ed in the analysis. As the duration of the analysis in-creases, the need for the latter guidance becomes moreimportant.

The conditions that should be compared to the per-formance objectives of 10 CFR Part 61.41: Specifi-cally, guidance is needed about whether all calculateddoses must fall below the performance objectives, or ifsome portions of the dose distribution may exceed theperformance objectives.

2.4 User Friendliness

Methodology could be improved in the area of userfriendliness. At the present time, a significant amount ofskill in using codes is required to implement the method-ology. In addition, without automation, using the meth-odology can be a time-consuming and tedious business.Consequently, it is recommended that automated linkagesbe developed between codes in the methodology. In spiteof the need for automated links between codes, it is ex-tremely desirable to maintain the flexibility of the meth-odology. This suggests that multiple links will be neededto retain all options available to the user. For instance, itmay be necessary to develop code couplers betweenGENII and the ground-water transport codes:

NEFTRAN, PAGAN, BLT., and VS2DT. A coupling toeach of these codes will be needed to retain the currentflexibility of the methodology. The necessity of flexi-bility dictates much more complex coupling betweencodes than is necessary for a single-site methodology,such as the high-level waste methodologies for basalt[Bonano et al., 1989], tuff [Gallegos, 1991], and salt[Cranwell et al., 1987].

At the present time, one typical overall performanceassessment analysis is conducted as follows. VS2DT isrun to evaluate the flow field in the disposal unit forintact and failed conditions. The VS2DT output file isthen edited using a spreadsheet to delete all but thesource region. This file contains velocity vector com-ponents and moisture contents, which can be combined tocalculate the transport velocity in the waste region. Thisinformation can then be used in the PAGAN source termmodel. PAGAN is then used to calculate ground-waterconcentrations as a function of time. These concentra-tions are input to GENII for the exposure analysis ofinterest. The analyst must run GENII sequentially for anumber of concentrations to produce the dose history.All of these linkages are straightforward (with practice),but it is clear that the methodology would greatly benefitby having them automated.

A potential solution to the user-friendliness problemwould be to develop an integrated system for conductinglow-level waste performance assessments. In the past,integrated system codes invariably restricted the analystto a very narrow set of possible analyses. For the meth-odology, however, it is considered to be of overridingimportance to retain flexibility in the possible approach-es. In this way, a wide variety of conceptual models canbe evaluated, and the methodology can continue to beused for a broad variety of conditions. To date, thereare no existing codes with the requisite flexibility anduser friendliness.

An approach that has the potential to be flexible and userfriendly is currently being developed at SNL. The ap-proach is to integrate multiple models, an uncertaintyanalysis package, and a geostatistical analysis packageinto a single package on a single computational platform.The current system has been developed to optimize theplacement of monitoring wells for existing waste sites(Figure 2.6). The user interacts with the system througha graphical user interface that allows direct access to sitedata, which are stored in a geographical information sys-tem. The user can use the maps stored in the geograph-ical information system to develop conceptual models of

27 NUREG/CR-5927


the site; the assumptions that are made in developing theconceptual model identify which of the possible processmodels are appropriate for the analysis. The systemcurrently contains analytical models for vadose-zone flowand transport, MODFLOW and MODPATH, and anadditional analytical solution for transport in the aquifer.

Extension of the existing monitoring well optimizationpackage to low-level waste performance assessment is

quite straightforward, and consists of adding more mod-eling capabilities. We are currently adapting this systemto include the models in the low-level waste performanceassessment methodology. In addition, we will expand theuser interface to include a conceptual model manager,which will trace assumptions and ensure consistencybetween site data and conceptual models used in theanalysis. This step will use the methodology to improvethe credibility of modeling results.

-i- ---------

DECISION SUPPORT SYSTEM

.............. .....................

GRAPHICAL USER GEOGRAPHICALINTERFACE INFORMATION

INPUT PROBLEM SYSTEDEFINITION, . SIMULATION CODESASSUMPTIONS : :

UNCERTAINTYOUTPUT : ANALYSIS DRIVERRESULTS,GEOGRAPHICAL : SUPPORTDATA SOFTWARE............. ..............................'

|a o.m............ ..... .......= lm~ .=lii =.=........L CONCPTUALMODEL MANAGER

I....................................................I

Figure 2.6 The decision support system structure

NUREG/CR-5927 28

3. Pathway Assessment for Alternative Disposal Technologies

This chapter is an update of previously conducted path-way assessments [Shipers, 1989; Shipers and Harlan,1989]. The purpose of the present pathway analysis is todetermine if the current methodology can satisfactorilyanalyze the pathways and phenomena likely to be impor-tant for the full range of potential disposal options. If themethodology needs improvement in some modeling area,steps will be taken to update the methodology in thatarea. The result from this process will be assurance thatthere is adequate coverage of the types of modeling need-ed to conduct performance assessment analyses for abroad range of environmental conditions and types ofdisposal options.

We emphasize that the pathway assessment has beenperformed for a generic site. For specific sites, theremay be fewer important pathways. We have attempted tobe conservative in keeping pathways in the lists that willusually be of marginal importance. In this way we canbuild confidence that we have spanned the range of caseslikely to be encountered at a real site.

Information about the various disposal options has beenextracted from published reviews of alternative near-surface disposal options, including Bennett and Warriner[1986], Miller and Bennett [1986], Trevorrow and Schu-bert [1989], Rogers and Associates [1987], Denson et al.[1987], Denson et al. [19881, McAneny 11986], andBennett [1985], among others.

3.1 Role of Pathway Analysis inPerformance Assessment

Pathway analysis is used as a starting point to identifypotential significant exposure pathways for the low-levelwaste disposal site under consideration, so that availableresources can be concentrated on the most importantaspects of the problem. Once the key pathways havebeen determined, models are assembled that can repre-sent radionuclide transport from the disposal facility tothe environment through these pathways. Assessing theperformance of the site with respect to the relevant regu-lations is conducted based on the modeling results. Path-way analysis is qualitative and relies mainly on expertjudgment. Efforts are made to err on the side of conser-vatism; pathways that may not be significant for a specif-ic disposal site are kept for generality.

3.2 Previous Work on PathwayAnalysis

Shipers [19891 presented an exhaustive list of pathwaysbetween low-level waste and humans, considering up tofour intermediate transport media with no recyclingbetween them. This analysis of pathways consideredmedia in both environmental (non-living) media andbiological media (the food chain). The result was anenormous number of combinations of the media thatcould physically occur (Shipers and Harlan [1989] citethe number to be over 8000). Many of these combina-tions result from the inclusion of the food chain in thepathway.

Shipers and Harlan [1989] determined the key pathwaysfor shallow-land burial disposal from their list of allpotential pathways. They defined the key pathways asthose having both high likelihood and potentially highconsequence. The performance assessment methodology[Kozak et al., 1990b] subsequently developed was a suiteof mathematical models implemented in computer codesthat can simulate the important processes in all the keypathways.

3.3 General Comments on PathwayAssessment

The current chapter expands the previous work by Ship-ers and Harlan to explicitly consider significant pathwaysfor alternative low-level waste disposal methods, includ-ing belowground vault, tumulus, aboveground vault,shaft, and mine disposal. The following terms are de-fined:

Below-grade disposal: Emplacement of the waste belowthe natural surface grade.

Above-grade disposal: Emplacement of the waste abovethe natural surface grade.

Belowground disposal: Waste is covered with earthenmaterials at the time of closure.

Aboveground disposal: Vault remains as a freestandingstructure after closure.

29 NUREG/CR-5927

Pathway Assessment

The primary purpose of this chapter is to evaluate wheth-er the methodology is sufficiently general in its capabilityto model all pathways likely to be important in a per-formance assessment. In this study, we define the keypathways that the performance assessment methodologymust be capable of analyzing as ones with high signifi-cance regardless of likelihood. The basis for this choiceis that 10 CFR Part 61 has deterministic performanceobjectives. Performance assessments carried out for acomparison with these performance objectives are techni-cal analyses of the significance or consequence of thepathway. The likelihood of the pathway is not a consid-eration once the modeling has begun. Shipers and Har-lan defined the key pathways as those having both a highlikelihood of occurring and a high significance if they dooccur.

We have chosen to truncate the number of environmentalmedia to be considered in series at three. Environmentalmedia are defined as potential transport conduits betweenthe source and man. Hence, we include pathways suchas source - ground water -- surface water - land plants,but not pathways with more environmental media. (Notethat an additional implicit final link in the pathway isman.) There are two reasons for truncating the series atthis point. First, in the list of important generic path-ways generated by Shipers and Harlan [1989], the keypathways were found to consist of two environmentalmedia or less. Thus, by investigating three pathways, weare being somewhat conservative. Second, various scop-ing analyses performed using the current methodologyhave corroborated the assumption of Shipers and Harlan[1989] that secondary media (such as surface water in theabove example) are far less important than primary me-dia because of dilution effects. Media beyond the sec-ondary ones are proportionately less important. We,therefore, believe that very little information is lost byomitting pathways greater than three media. However,when assessing the pathways for a particular site, thesegeneric guidelines may not hold, and further pathwaysmay need to be assessed.

We begin the process of identifying key pathways with aninclusive list of all possible pathways, and then systemati-cally screen the unimportant pathways from consider-ation. The rules for eliminating pathways are as follows.Pathways are screened when the release is consideredunlikely or insignificant. For instance, if releases tosurface water and soil are rated as unlikely for a givendisposal method, then considering subsequent mediaalong these paths would not be appropriate. Similarly, ifthe consequence of human inhalation of gaseous radionu-clides is considered low, the effects in subsequent media

are considered very low. However, for pathways thatconsider the passage of radionuclides from a medium inwhich their significance is low to another medium inwhich bioaccumulation can occur, this screening criteriondoes not apply.

The biointrusion pathways do not enter the three-mediapathway lists because they all involve a transfer from thebiota to an environmental medium (e.g., an animal dyingor excreting into the environment) or between unlikelybiotic transfers (e.g., land animals -- aquatic animals canonly happen by an animal dying in the water, which isclearly not a primary pathway). Such pathways will tendto be of very low significance compared to other, moredirect pathways.

Keep in mind when considering the relative importanceof the pathways that drinking contaminated well wateroverwhelmingly dominates the dose in the performanceassessment. Other pathways are generally accepted to beless important. However, at a specific site, the pathwaysmay still lead to doses that contribute significantly to thetotal dose. This distinction was acknowledged in rankingthe significance of a pathway. The non-well pathwayswere ranked in importance with respect to each other.

The airborne pathway, which includes both release ofradioactive gas by the waste and direct particulate en-trainment of the waste, has been incorporated into all ofthe lists of pathways. This is contrary to the pathwaylists given in Shipers and Harlan [1989], who consideredthe likelihood of airborne releases to be high, but thesignificance to be low. We concur with this assessmentas it relates to gas production, since doses from suchreleases are generally accepted to be minor. Neverthe-less, there is uncertainty about how to model gas produc-tion [Kozak et al., 1990a]; therefore, the significance ofreleases from low-level waste facilities may be somewhatlarger than is anticipated. We have retained the airpathway for direct entrainment following exposure of thewaste at the surface. The significance of this release islikely to be high, but for most disposal options it is ex-tremely unlikely. We have also retained the air pathwaysince analysis of the air pathway is explicitly required in10 CFR Part 61.

The analysis requirement is also explicit in the rule forthe biotic intrusion pathways, hence these have also beenretained in the lists, except for the deep disposal options,for which biointrusion will clearly not play a role. Thereare two aspects to the biotic intrusion pathway. Theintrusion may be by wild animals, plants, or insects, orby domestic plants or animals. In the case of wild biota,

NUREG/CR-5927 30

Pathway Assessment

the likelihood of intrusion is high (except for deep dis-posal) but the significance is low, since wild biota areusually not major dietary items. For domestic biota, thelikelihood of intrusion is low (crops are usually shallowrooted, and domestic animals generally do not burrow),but the significance can be high. For generality, we haveranked some biotic intrusion pathways as high likelihood,high significance. However, when analyzing a specificdisposal option at a particular site, we will usually beconcerned about intrusion by either wild or domesticbiota, but not both. Therefore, in specific cases, it isunlikely that both likelihood and significance will be highfor biointrusion.

External exposures from penetrating radiation are notexplicitly included in the pathway lists, but are intendedto be implicitly included in the sense that the exposureroute between the environment and man is not specified.Concentrations in, for example, ground water can poten-tially produce both external and internal exposures.Low-level waste inventories contain nuclides that emitpenetrating radiation. If we choose as an example theinventory listed by Rogers and Associates [1988] andarbitrarily discard isotopes with half-lives less than 10years, we are left with only a handful of gamma andX-ray emitters. Of these, only Ni-59 and Mo-93 havethese as their dominant form of radiation. The rest arepure beta emitters, beta emitters with secondary gammaemission (e.g., CI-36, 1-129, Cs-137), and alpha emittersthat have secondary gamma emissions (e.g., uraniumisotopes and the transuranics). The importance of sec-ondary emissions should not be downplayed since theranking of emissions as primary or secondary is based onthe probability of the type of emission, not on the aver-age or peak energies of the emissions: from a radiologi-cal perspective, a secondary radiation may be of equal orgreater importance than the primary radiation. Forinstance, 1-129 produces more gamma energy than betaenergy, although it is commonly described as a betaemitter. Consequently, although penetrating radiation issecondary for almost all long-lived radionuclides in low-level waste, these radiations are not necessarily negligi-ble, and external exposures should be considered whenappropriate in the pathway analysis.

3.4 Alternative Disposal Technolo-gies

The experience with existing shallow-land burial facilitieshas raised questions on the effectiveness of the method.Remedial actions are underway at all three closed shal-low-land burial sites [U.S. Congress, 1989]. On the

other hand, there are many shallow-land burial sitesaround the world that are functioning satisfactorily [Kit-tel, 1989]. Furthermore, there have not been appreciableoff-site doses received from any of the closed shallow-land burial sites in the U.S. [Matuszek, 1988]. Theproblems associated with the closed sites appear to berelated more to siting problems than to any inherent flawin shallow-land burial. Nevertheless, as a result of theoperational problems at existing shallow-land burial sites,there has arisen a perception that the technology isflawed, which is not well established by experience. Theresult of this perception is that most States and Compactsare pursuing alternative disposal technologies.

The principle behind alternatives to shallow-land burial isthat it should have a lower risk/hazard ratio than shallow-land burial [Trevorrow and Schubert, 1989]. That is, fora given hazard (defined in terms of the waste as dis-posed) the risk (which accounts for the probability andconsequence of exposure) should be lower than for shal-low-land burial. Proponents of these technologies gener-ally ignore the additional uncertainties introduced whenusing untested disposal methods. It is conceivable, albeitunlikely, that the use of engineered enhancements couldworsen off-site doses compared to shallow-land burial;for example, catastrophic failure of the roof of a vault,followed by massive subsidence, which could potentiallyfunnel infiltrating water through the waste and producerelatively rapid release rates of the waste. The likelihoodof this type of failure can be minimized through appro-priate engineering designs, but this type of failure cannotbe dismissed without technical justification.

Engineered barriers can influence the performance of thefacility in several ways:

1. The barrier reduces the flux of water through thedisposal unit for some time until the structure failsand significant releases begin. Low-level waste ismade up almost entirely of short-lived (less than 30years) and very long-lived (greater than 5,000 years)isotopes. The delay until failure of the barriers canallow sufficient time for the short-lived inventory todecay, which can greatly decrease the total impact ofthe eventual release. However, the barrier will havelittle effect on the significance of releases of long-lived species unless it inhibits their release rate insome way.

2. Engineered barriers can mitigate doses to inadvertentintruders. However, intruders receive adequateprotection from the waste classification scheme in 10

31 NUREG/CR-5927

Pathway Assessment

CFR Part 61, and the engineered barriers do notsignificantly change the situation.

3. Extensive use of concrete and grout in a vault maysignificantly alter the chemistry of the pore water inthe vault. For instance, concrete-stabilized waste hasbeen shown to have pore water with a pH of 11 to13, with greater uncertainty in the Eh [Doughertyand Columbo, 1985].

These conditions tend to lead to low solubilities and highsorption of some important chemical species [Atkinsonand Nickerson, 1988; Atkinson and Marsh, 1988], whichhas a clearly beneficial influence on off-site impacts.

For the purposes of this discussion, we note that theeffect of an engineered barrier may influence the choiceof model for a particular phenomenon (e.g., leachingmodel), but will not usually change the pathways ofrelease. The only influence of engineered containmenton the pathway assessment is that the barriers may inhibitbiotic intrusion. Also, keep in mind that two disposaloptions may have identical pathway assessments, butdifferent models may be needed for them.

Rao et al. [1992a] considered belowground vaults, tumu-lus disposal, aboveground vaults, shaft disposal, andmine disposal. A belowground vault refers to placingwaste in shallow subsurface concrete structures. Below-ground vault disposal technologies essentially represent anatural evolution from shallow-land burial, and weredeveloped to address concerns about the inherent lack ofengineering control and barriers in shallow-land burialdesigns. Tumulus disposal refers to placing waste abovegrade under a mound of earth (tumulus). Tumulus dis-posal includes designs for disposal of waste entirelyabove grade and for combinations of above-grade andbelow-grade disposal. Disposal designs can include engi-neered enhancements (as in earth-mounded concretebunkers) or disposal beneath a tumulus without additionalenhancements (as in mill-tailing impoundments). In somecases, the specific design can influence the pathways ofinterest, since the presence of a concrete vault may lessenthe likelihood of biointrusion or erosive direct exposureof the waste. An aboveground vault refers to a structurethat relies on its own structural stability to provide isola-tion of the waste. This is the only commonly cited near-surface disposal option in which the waste is not coveredby geological materials. Rogers and Associates [1987]concluded that aboveground vaults may have difficultymeeting the technical requirements for stability in 10CFR Part 61. Shaft disposal facilities are holes bored

into the earth, which have a large length to diameter ratio[Trevorrow and Schubert, 1989]. Shaft disposal has alsobeen called the "augered hole" method since an auger rigis often used to construct the shafts. This method hasbeen used by several different investigators, including theU.S. Department of Energy (DOE) [Dickman and Bola-nd, 1983; Cook, 1984] and Atomic Energy of Canada,Ltd. (AECL) [Feraday, 1982]. The term mine disposalrefers to underground mines, not surface (open-pit oropen-cast) mines. The pathway analysis can be appliedto other underground excavations as well.

Evaluations of the various disposal technologies in termsof operational, cost, and safety considerations generallyrank the below-grade options (including shallow-landburial) and tumulus disposal about equally. Abovegroundvaults generally are more poorly ranked [Rogers, 1989].However, such conclusions should not be generalized,and rankings need to be performed for specific sites.Rogers and Associates [1987] concluded that earth-moun-ded bunkers, modular canisters, and belowground vaultsprovide only marginal performance improvement overshallow-land burial, at significantly larger costs andoccupational exposures. They also concluded that above-ground vault technology requires additional time andanalysis to ensure that all the requirements of 10 CFRPart 61 are met. Gershey et al. [1990] provide an infor-mative summary of the positive and negative attributes ofthe alternative disposal technologies discussed here.

Some citizen groups and state agencies have shown apreference for above-grade placement of waste [Trevorr-ow and Schubert, 1989]. Several States and Compactsare actively pursuing above-grade technologies (e.g.,Appalachian Compact, Central Interstate Compact). Ithas been suggested that above-grade facilities are pre-ferred in regions with high infiltration rates, and thatbelow-grade facilities are preferred for sites with lowinfiltration rates [U.S. Congress, 1989]. Above-gradefacilities require considerably more land area than below-grade facilities; this adds an additional siting require-ment. The ridge-trough topography of a closed above-grade facility may limit eventual unrestricted use leadingto lower likelihood of intrusion [U.S. Congress, 1989].On the other hand, the topography may provide incentivefor an intruder to level the area, thus exposing waste.

Some have suggested that above-grade structures decreasethe likelihood of inadvertent intrusion [Trevorrow andSchubert, 1989]; others suggest that above-grade facilitiesmay attract curious potential intruders [Rogers, 1989;U.S. Congress, 1989]. In either case, the potential

NUREG/CR-5927 32

Pathway Assessment

inadvertent intruder is assumed to receive adequate pro-tection through the waste classification system of 10 CFRPart 61. If an exemption from the classification schemeis requested, an intruder analysis is required to demon-strate adequate protection to the intruder. In eitherevent, the likelihood of the intrusion event is not in ques-tion, and the significance is expected to be relatively in-sensitive to the disposal system design. Therefore, al-though consideration of the probability of intrusion isimportant in choosing a disposal option, it is not relevantto the evaluation of potential pathways.

Concepts for disposal have been proposed in addition tothe ones identified above, including modular-concrete-container disposal both aboveground and belowground[Rogers, 1989], improved shallow-land burial, anddeep-trench burial [Trevorrow and Schubert, 1989]. Weconsider these options to be subsets of the technologiesdescribed in this report. For instance, aboveground-modular-canister disposal is equivalent (from a perform-ance assessment perspective) to an aboveground vaultwith the waste grouted in place. Similarly, belowground-modular-canister disposal is similar to either shallow-landburial or belowground vault disposal with the wastegrouted in place [Rogers and Associates, 1987]. Thedifferences between the technologies are in operationaland design considerations that do not affect the pathwaysor modeling requirements for postclosure performanceassessment.

3.5 Role of Temporal Progression inPathway Analysis

As the site evolves in time, the possibility exists for anumber of initiating events to alter the characteristics ofthe facility enough that alternative exposure pathwaysbecome important. Since 10 CFR Part 61 has an open-ended performance period, it could be argued that low-probability initiating events should be included in themethodology. Whether such events are included or not isa regulatory issue that we will not address here. Howev-er, we anticipate that NRC will conduct performanceassessments based on relatively minor perturbations aboutthe current state of the site. Consequently, we maydiscard low-probability initiating events such as volcanicactivity. Furthermore, we may discard processes such aslandsliding and erosion of the natural grade since they arerestricted by Subpart D of 10 CFR Part 61. We considererosion of artificial grades, such as tumulus covers, to bea separate issue from erosion of the surrounding naturalgrade since it is not covered in Subpart D.

Remaining temporal evolution problems reduce to threeprimary ones: water table changes, subsidence, andvegetative progression. Water table changes may occurbecause of crop irrigation, other man-made disturbancesto the local hydrology, or climate change. Again, weanticipate that NRC will only wish to analyze events thatcan be expected to occur at the site given the currentconditions at the site. It, therefore, seems unreasonableto include extreme (and low-probability) changes in thehydrology. There is, of course, a somewhat arbitrarydistinction between events that are anticipated to occur atthe site, and events that are only somewhat likely at thesite. The decision about the level of extremity to beincluded in the performance assessment must be made byprofessional judgment, since regulatory guidance on thisissue is not included in 10 CFR Part 61.

Subsidence can be expected to occur to some degree atall disposal facilities and, over the short term, representsan aspect of the facility that will evolve in time. Earlyon, less subsidence can be expected for facilities withengineered support for the soil overburden. However,the principal effects of subsidence on the performanceassessment are to change the flow field in the unit (e.g.,distributed infiltration becoming more intense localizedinfiltration), and to alter the values and uncertainty ininput parameters (infiltration rate). The presence andextent of subsidence will not affect the pathways that weneed to analyze, except in the extreme case in whichwaste is directly exposed by massive subsidence. Thisextreme event can be made very unlikely by appropriateengineering design of the facility. Schultz et al. 11990]have introduced the idea of using a cover that is robust tothe effects of subsidence during the period of activesubsidence, and adding a multilayer cover following thattime. This approach appears to be very promising forensuring cover performance of shallow-land burial, forwhich the period of active subsidence corresponds to theoperational and institutional control periods. However,subsidence problems associated with concrete structuresmay occur at much later times. For instance, the mostimportant subsidence in the life cycle of a belowgroundvault may be following collapse of the roof; this can beexpected to occur long after institutional control ends.

The effect of vegetative progression on low-level wasteperformance assessment has received relatively littleattention to date. Vegetative progression can be expectedto occur at all sites, to some degree, after the loss ofinstitutional control. Progression may consist, for in-stance, of the displacement of grassy cover vegetation bydeeper rooting trees or shrubs. A second example wouldbe replacement of cover vegetation by crops once the site

33 NUREG/CR-5927

Pathway Assessment

is opened to unlimited use. In either case, the progres-sion leads to two effects. The first effect is that thelikelihood of biotic intrusion may change, depending onthe root depths before and after progression. The pro-gression may be toward either deeper or shallower root-ing plants, and hence to either a greater or lesser likeli-hood of biointrusion. The second effect is a change inthe water budget. Progression toward trees, for instance,may lead to larger canopy holdup, which would alter theamount of water available for infiltration. Progressionmay tend to either increase or decrease infiltration.

Vegetative progression is also linked to climate change,since wetter or drier climates will tend to favor the incur-sion of new species into the area. The issue of climatechange will not be addressed in this chapter; however,the issues associated with climate change, as well asother possible future evolutions of the site, are discussedbelow in the section on uncertainty analysis.

3.6 Summary of Pathway Assess-ment

Details of the results of the pathway assessments forspecific disposal options may be found in Rao et al.[1992a]; we present only a brief summary here.

Given that a belowground vault facility is essentially ashallow-land burial facility augmented with engineeredstructures and barriers, the critical pathways for below-ground vault disposal are similar to those identified byShipers and Harlan [1989] for shallow-land burial. Somepathways can be eliminated for belowground vaults, sincethe engineered barriers tend to make certain pathways ofmuch lower likelihood.

Two styles of tumulus disposal may be considered. Thefirst style is a simple tumulus with neither additional con-crete support nor overpacks in the design. The secondstyle is an earth-mounded concrete bunker with a roofthat supports the tumulus. Including such a roof tempo-rarily reduces the potential for subsidence and erosion,and also reduces the likelihood of biointrusion.

Ground-water is the primary important release pathwayfor both belowground vaults and tumuli. Erosion islikely to be of little concern in belowground vault designsbecause of regulatory siting criteria; erosion of the tumu-lus can be minimized by appropriate engineering design.Incorporation of a thick rip-rap cobble layer in a coverwould inhibit biointrusion from the surface and maydiscourage potential inadvertent human intruders. Lateral

biointrusion would be impeded by the concrete wall inbelowground vault designs. Gases generated by low-le-vel waste decomposition could be easily transportedthrough earthen covers, but appropriate engineeringdesigns can be used to influence gas transport directions.

Aboveground vaults rely entirely on their concrete struc-ture to isolate the waste from the surface environment.Given the current uncertainty about the longevity ofdisposal structures, it is usual in performance assessmentanalyses to assume the vault fails completely at sometime during the performance period. For earth-moundedconcrete bunkers and belowground vaults, such failureresults in enhanced releases to ground water, but doesnot directly expose the waste at the surface. By contrast,for aboveground vaults, failure of the vault may meandirect exposure of the waste at the surface, The path-ways reflect this contrast in increased importance ofreleases to soil, surface water, and air.

The ground-water pathway is the most significant path-way for shaft disposal. Biotic intrusion is not as impor-tant because of depth of disposal. The air pathway islikely to occur since gas will be generated by the decom-position of low-level waste, which then can be transport-ed to the land surface by both diffusion and barometricpumping. However, the disposal depth increases the gastransport time, which can decrease the radioactivity ofthe gas by decay. The surface-water pathway is also notas important since direct exposure of the waste is im-probable. Although an engineered cover such as concretemay not be present, the likelihood of human intrusion issmall because of the small land surface area of the shaft.

According to McAneny [1986], at a mine disposal facility"ground water is the principal and perhaps only crediblerelease path for radionuclides." This is because mines areoften located in the saturated zone and ground-waterintrusion is probable at some point in the life of thefacility. Thus, this pathway has both a high likelihood ofoccurring as well as a large significance if the contami-nated ground water is ingested.

In summary, disposal methods can be categorized ingroups based on their depth of disposal. In deep disposaloptions (shaft and mine disposal), the key pathways areidentical, and result entirely from releases to groundwater. The shallow belowground disposal options (tumu-lus, shallow-land, and belowground vault disposal) alsomay be grouped together from a pathway analysis per-spective. For these disposal options, the ground-waterpathway remains the most important, but other pathways,

NUREG/CR-5927 34

Pathway Assessment

such as biointrusion, can become more credible than fordeep disposal options.

Aboveground vault disposal cannot be grouped with anyof the other disposal options. In the absence of geologi-cal materials covering the disposal facility, the analystcannot eliminate the potential for exposure of the waste atthe ground surface. If the waste becomes exposed, many

more pathways can become important that do not play arole in any of the below-g'0und disposal technologies.For instance, releases to surface water runoff shouldusually be considered for aboveground disposal, asshould direct exposure. Other pathways, such as bio-intrusion, become much more significant than in below-ground disposal, since the waste can become directlyaccessible at the ground surface.

35 NUREG/CR-5927

4. Ground-Water Flow and Transport Modeling

Ground-water hydrology is subdivided into two areas inthe methodology. The first area is the evaluation ofinfiltration, and the second area is the analysis of theground-water flow and transport processes. The distinc-tion between these areas is shown in Figure 4. 1, whichdepicts flow and transport processes in and near an intactnear-surface disposal facility. Processes affecting infil-tration are rainfall, evaporation, transpiration, and run-off. The resulting balance of these processes provides aboundary condition for the ground-water flow processes.

The water that percolates deeply will flow around orthrough the intact vault. Significant releases may occurfrom an intact vault if enough fractures or flaws exist inthe concrete to substantially elevate its hydraulic conduc-tivity, or if enough .joints have failed to provide signifi-cant flow paths between concrete slabs.

More importantly, however, is the time in the futurewhen the integrity of the waste containers and engineeredbarriers can no longer be relied upon. The primarydifference between performance assessment analysis ofintact and failed vaults is a change in the approach usedfor modeling ground-water flow through the waste. Theflow processes near a failed vault are depicted in Figure4.2; the vault may (or may not) retain structural stability,but it no longer provides an impediment to flow throughthe waste. The conditions usually of greatest interest inthe performance assessment are those shown in Figure4.2, since the peak dose is much more likely to be asso-ciated with failed conditions (non-design conditions) thanwith intact (design) conditions. Consequently, while thecapability to model the intact vault is important, it is evenmore important to be able to model failed and partlyfailed conditions.

steady-state flux condition is much simpler than a tran-sient condition for the long time periods of the perform-ance assessment. However, the steady-state assumptionhas not been decisively shown to be conservative com-pared to episodic infiltration.

Status

Since the publication of the methodology, Pacific North-west Laboratory (PNL) has developed the InfiltrationEvaluation Methodology (IEM) [Smyth et al., 1990].The basis of the IEM is that rainfall (in time) and soilproperties (in space) may be treated as though they arestochastically distributed, and the IEM has been devel-oped in an attempt to account for these variabilities. Inits original form, the IEM was an "approach;" we recog-nized that "because of the site-specific nature of estimat-ing drainage, the IEM does not recommend specificmodels to be used" [Smyth et al., 1990]. This approachis, therefore, in concurrence with the recommendationsof Kozak et al. [1989a], that are incorporated in themethodology. However, work at PNL is currently inprogress to recommend a group of models that will pro-vide somewhat general infiltration evaluation capabilities.

Evaluation

The primary barrier to improved infiltration modeling isthe large uncertainty in data and parameters used invalidation of the models [Olague et al., 1993]. To date,there is not a consensus about the conditions for whichparticular models are appropriate [Balek, 1988; Knuts-son, 1988]. Olague et al. [1993], recommend combiningdifferent types of experimental data to identify the condi-tions for which alternative approaches are appropriate.

Specific modeling approaches for infiltration are neededin the methodology. Currently, the experimental basisfor choosing one approach over another is not available.Consequently, we recommend that a comprehensivecomparative analysis should be conducted among themodels, as is discussed by Olague et al. [1993].

4.2 Ground-Water Flow and Trans-

port

Current Approaches

The current methodology contains the computer codesPAGAN [Kozak et al., 1990a; Chu et al., 1991] and

4.1 Infiltration Evaluation

Current Approaches

A specific method for evaluating infiltration was notincluded in the methodology; we concluded after a litera-ture review that a sufficient general method was unavail-able to evaluate the broad conditions for which the meth-odology must be capable of handling [Gee and Hillel,1988; Balek, 1988]. Instead, we suggested that thelicensee should support infiltration estimates with a com-bination of field data and site-specific modeling. Asteady-state value of infiltration derived from these ap-proaches is used in the methodology as a boundary condi-tion for vadose-zone flow modeling. In general, use of a

37 NUREG/CR-5927

Ground-Water Flow

IRUNOFF

FLOWINTOJOINTS

RAINFALL I GASES t I EVAPOTRANSPIRATION

IFRACTURE FLO

'I II -~ XT XT

-1 [Al r te

FLOW

DIVERSION

BY VAULT

AND COVER IINFILTRATION

VADOSE-ZONE FLOW III

VADOSE-ZONE TRANSPORT

__7V_AQUIFER FLOW AND TRANSPORT

Figure 4.1 Flow processes in and around an intact disposal facility

VAM2D [Huyakom et al., 1989]. PAGAN performstransport analyses of the radionuclides in the source,unsaturated zone, and saturated zone given a user-defineduniform velocity. The code's analysis of the source andunsaturated zone is usually used in a way that neglectsdispersion in these regions. This is believed to be aconservative assumption in most cases. The saturated-zone transport model in PAGAN allows for one-dimen-sional convection and three-dimensional dispersion, anduses a semianalytical solution to the convective dispersionequation. VAM2D is a fully two-dimensional finiteelement code for the analysis of both flow and transportin either saturated media, unsaturated media, or a combi-nation of the two. The flow model is based on Darcy'slaw, including the unsaturated extension of Darcy's law

(Richard's equation). Discrete fracture flow cannot beevaluated using the code. The transport model solves theconvective dispersion equation in two dimensions foreither simply decaying radionuclides or chains (up to fourmember chains, either straight or branched). VAM2D isflexible in the types of boundary conditions that can bespecified for both flow and transport. In addition,VAM2D has been shown to have robust numerical meth-ods. We chose it for use in the methodology primarilybecause of its flexibility, robustness, and consideration ofchain transport.

Kozak et al. [1990a] discussed code errors and docu-mentation discrepancies in VAM2D Version 5.0. Someof these errors were eliminated in Version 5.1, but wehave continued to find errors in Version 5.1. Several

NUREG/CR-5927 38

Ground-Water Flow

EVAPOTRANSPIRATION

RUNOFF DEGRADED74.OCNCRETEA

A A DEGR)INFILTRATION L~ AWASTE

VADOSE-ZONE FLOWIL.

AQUIFER FLOW AND TRANSPORT

.. ( )"... )

< ) )

I

ADOSE-ZONE TRANSPORT

Figure 4.2 Flow processes in and near a failed vault

versions of Version 5.1 exist (we have Version 5. 1d),and each of these informal sub-versions has been devel-oped to eliminate some error; users do not receive up-dates as they become available, nor are they even in-formed of their existence. This is a poor configurationmanagement practice that has caused us some concern.In addition, the source code is proprietary, which makesdebugging a run difficult. The most recent version of thecode, Version 5.2, is currently not available to us, andwe have not assessed its behavior. The greatest limita-tions of the code are (1) it is proprietary, and the sourcecode is not publicly available, even in hard copy form,(2) user unfriendly input structure, (3) poor configurationmanagement practice, and (4) lingering errors in thecode. These problems have not been of overriding con-

cern to date, but they may be much more important if thecode is used during the licensing process.

An updated version of PAGAN (Version 1.1) was pro-duced in late 1990 [Kozak, 1991]. The new version usesa dispersivity defined in terms of pore velocity ratherthan Darcy velocity (which was used in Version 1.0); italso uses consistently conservative values for dose con-version factors for the actinides in the code. No bugshave been reported by users of Version 1.1. Some areasof possible improvement are (1) improvement of accura-cy for long time periods, (2) elimination of some confus-ing aspects of the input shell, and (3) incorporation ofMonte Carlo analysis with either random sampling orLatin Hypercube Sampling.

39 NUREG/CR-5927

Ground-Water Flow

Status

There are many possible alternatives to using PAGANfor simple ground-water transport analyses. There arelarge numbers of analytical models for transport in theaquifer, and any of these may be appropriate at a particu-lar site. However, PAGAN combines a fairly wideamount of flexibility with user friendliness, and hasproved useful in several scoping analyses.

There are two possible alternatives to using VAM2D forcomplex ground-water modeling in the methodology:

1. Use of a different code with capabilities similar toVAM2D; and

2. Use of perturbation methods (often called analyticalstochastic models) in the methodology. Both ofthese possibilities are discussed here.

Several possible alternatives to VAM2D have recentlybecome available. Two of these are DCM3D [Updegraffet al., 1991] and FEMWATER3D/LEWASTE. 1 How-ever, DCM3D only models flow at present, and FEM-WATER3D/LEWASTE suffers from lack of adequatedocumentation and from poor quality assurance proce-dures. McCord 2 benchmarked a number of ground-waterflow codes, and identified the code VS2DT [Lappala etal., 1987; Healy, 1990] as a potential replacement forVAM2D in the methodology. Advantages of VS2DTover VAM2D are (1) better quality assurance of thecode, (2) faster convergence (in the benchmark prob-lems), (3) nonproprietary source code, (4) a more userfriendly input structure, and (5) sorption by eitherFreundlich or Langmuir isotherms. Disadvantages ofVS2DT are (1) it cannot be used to analyze chain trans-port, and (2) lack of a free drainage boundary condition,which is useful in analyzing very deep unsaturated zones.This second limitation is not considered to be a severeone.

In the original development of the methodology, prefer-ence was given to codes that were able to model eithersaturated or unsaturated media. It has since becomeclear that this constraint helped reduce the number ofcodes needed in the methodology, but also introducedlimitations. Consequently, we now consider codes thatare limited in their applicability to either saturated orunsaturated media, but which introduce some additionalneeded flexibility to the methodology.

1Yeh, G.T., personal communication, ORNL, 1991.2McCord, J.T.. personal communication, SNL. 199 1.

The first such code is MODFLOW [MacDonald andHarbaugh, 1988], developed by the U.S. GeologicalSurvey. MODFLOW is a three-dimensional block-cen-tered finite-difference code for evaluating saturated flow.It is extremely popular, widely used, and acknowledgedto have a good quality assurance background. Use ofMODFLOW allows very broad flexibility in the types ofsaturated-zone flow problems that can be modeled, fromsimple to quite complex. A particle-tracking module,called MODPATH [Pollock, 1989], has also been devel-oped, along with a packageto generate graphical output ofthe pathlines calculated in MODPATH. Another packageallows evaluation of the interaction between ground-waterand surface-water flows [Prudic, 1989]. This interactionwas treated inflexibly in the original methodology usingconservative assumptions. MODFLOW is expected to beparticularly useful for sites with complicated hydrology,and in reducing those complexities to simpler perform-ance assessment conceptual models. The biggest draw-back to the code is that it is not coupled with a transportcode.

An alternative for evaluating transport is to incorporateNEFTRAN II [Olague et al., 1991]. NEFTRAN con-tains a solution to the convective-dispersion model withone-dimensional convection and one-dimensional disper-sion, which are arranged in a network of stream tubesbetween the source and the receptor point. The primaryincentives for using NEFTRAN are (1) minimization ofnumerical dispersion for large simulation times, (2) nu-merical efficiency at very long times, and (3) the abilityto model multiple decay chains of any length. Ironically,the approaches used in NEFTRAN that make it efficientfor long simulation times also introduce difficulties in itsuse for low-level waste performance assessment. Unlikethe more common finite-element and finite-differencenumerical solution approaches to solving the governingequations, in NEFTRAN numerical dispersion is mini-mized by maximizing the time step size (within certainconstraints). This characteristic of the code is ideal foranalyzing integrated discharge for 40 CFR Part 191,which is the original intended use of the code. However,when evaluating peak ground-water concentration for 10CFR Part 61, it is often necessary to calculate manyintermediate time steps to ensure that the analysis doesnot miss the time of occurrence of the peak. This issueis particularly important for time-dependent concentra-tions that vary rapidly in time. In this case, many inter-mediate time steps may need to be calculated, and thiscan increase numerical dispersion in NEFTRAN, thusdecreasing calculated concentrations. This problem canbe overcome by careful use of the code, but the analystmust be aware of the issue.

NUREG/CR-5927 40

Ground-Water Flow

It is important to understand the limitations and uncer-tainties that can potentially be introduced by using astream tube model to evaluate concentration. As pointedout by Kozak et al. [1990a], NEFTRAN is designed tocalculate radionuclide flux through a cross-sectional area.Calculation of a concentration requires the introduction ofa dilution volume, which must be specified by assigningan effective cross-sectional area through which transportoccurs. For this reason, NEFTRAN was omitted fromthe original methodology. This limitation of the code canbe overcome in principle by using a detailed three-dimen-sional flow model to identify the boundaries of pathlinesthat can be used in assigning the stream tube volumeused in calculating concentration. This approach, whichis depicted in Figure 4.3, may be useful in some cases.However, the vadose zone-aquifer interface must beevaluated with care in the flow model, since the inter-action of the two zones will strongly affect the assignedarea of the stream tube in the aquifer. In addition, theanisotropy of the aquifer hydraulic conductivity becomesimportant, since it plays an important role in identifyinghow broadly the stream tube spreads with distance down-gradient. Other possible uses for NEFTRAN will beaddressed in Chapter 5, Source Term Modeling.

Another approach that has been proposed involves theuse of perturbation analyses (analytical stochastic meth-ods) for unsaturated ground-water flow calculations[Polmann et al., 19881. These methods reduce to anapproach to conducting parameter uncertainty analysis[Freeze et al., 1990]. Parameter uncertainty can beincorporated into the methodology using Monte Carloanalysis with codes such as VAM2D or VS2DT, or usingperturbation models (uncertainty analysis is discussedmore fully in Section 8.1). However, to use analyticalperturbation methods, the analyst is limited by a numberof restricting assumptions in the models [Freeze et al.,1990]. The drawbacks to analytical perturbation methodsmay be summarized as follows.

The purpose of the analytical perturbation models isto propagate input variable uncertainty through themodel to identify the associated uncertainty in theoutput variable. Bonano et al. [1987] and Freeze etal. [1990] have noted that analytical perturbationmodels are only one of several approaches to accountfor parameter uncertainty. However, the models arecurrently limited to normally distributed input vari-able distributions. This makes the analytical pertur-bation models much less flexible than when MonteCarlo Sampling is used in conjunction with the Mon-te Carlo analysis. Furthermore, when Latin Hyper-cube Sampling is used in conjunction with the Monte

Carlo analysis, Monte Carlo analysis may approachthe analytical perturbation methods in computationalefficiency [Bonano et al., 1987].

" The unsaturated-flow model is based on an expo-nential function for the relationship of unsaturatedconductivity to pressure head [Hills and Wierenga,19911. This function fails at low and high satura-tions, which inhibits use of these models in theseflow regimes. Including other characteristic func-tions in the model will likely eliminate the advantag-es of the analytical perturbation method, because theclosed-form analytical structure of the solution willbe lost.

" Current analytical perturbation models are based onlinear perturbation theory, with high-order terms inTaylor's series in the perturbed variable truncated atthe first-order terms. This limits the models to smallperturbations [Freeze et al., 1990]. For the flowequation under dry conditions, the perturbations areexpected to become large, and the flow model islikely to be inappropriate.

* Use of analytical perturbation analyses requires theanalyst to identify a correlation length in the fielddata. This correlation length is usually identified byevaluating a variogram of the data. Variogramanalysis has a number of limitations, among themsmoothing of data perturbations and nonrobust statis-tical behavior [Samper and Neuman, 1989]. Inaddition, interpretation of variograms is often done"by eye," and this approach has been found to bequite subjective [Samper and Neuman, 1989]. Thereis not currently a well-established method to unam-biguously identify correlation lengths from data.This means that correlation lengths cannot be speci-fied with much confidence.

* Analytical perturbation models apply continuumequations, as do more traditional Richard's equationmodels, but in the analytical perturbation approach itis assumed that the variables of interest can be treat-ed mathematically as if they are spatially random.This random field is invariably assumed to be ergo-dic and stationary, and these conditions are unlikelyto be met in the field. Dagan [19861 has identifiedstationarity as a stringent constraint on using analyti-cal perturbation ground-water models.

* The models are derived for infinite domains [Freezeet al., 1990], and the assumptions in the analyticalperturbation method about correlation lengths fail

41 NUREG/CR-5927

Ground-Water Flow

/

I

WV

K

I

I'Yr

ISTREAM TUBE VELOCITY

GENERAL AQUIFER FLOW DIRECTION

I ,"I

STREAM TUBEAREA

Figure 4.3 Defining an aquifer stream tube from a flow model. The cross-sectional area of the stream tube must bedefined from a detailed flow model to calculate a concentration using NEFTRAN II.

near boundaries. This limitation is related to the necessi-ty for stationarity, and identifies a limit to the usefulnessof the analytical perturbation models in analyzing com-plex geometries.

Evaluation

In comparing PAGAN to the other alternatives availablefor simple analyses, we conclude that it is flexible andeasy to use for scoping calculations, but the assumptionsin the model may not be appropriate for many cases.Other available solutions for simple analyses may some-times be more appropriate, but these must be chosen caseby case.

VAM2D has technical capabilities that other codes donot, but its drawbacks in quality assurance and its propri-

etary nature are continuing sources of concern. Webelieve that VS2DT has some inherent advantages overVAM2D, but at present it cannot model decay chaintransport, which is a major limitation. Analytical pertur-bation models may be useful from a purely researchstandpoint, but are neither flexible enough nor robustenough to be useful in performance assessment.

The ideal situation would be to develop a public code,such as VS2DT, to incorporate the capability to modeldecay chain transport. In this way, a code can be pro-duced that is publicly available, that is produced underappropriate quality assurance standards, and that containsall of the important technical capabilities of VAM2D. Itis recommended that VS2DT should be modified toincorporate the capability to model decay chains of arbi-trary length.

NUREG/CR-5927 42

Ground-Water Flow

MODFLOW may be useful in some cases, and introducesan additional degree of flexibility into the methodology.NEFTRAN II may also be useful in some circumstances,particularly when decay chains are important, and it isrecommended that it should be incorporated into themethodology. In principle, stream tube volumes that

identify the dilution volume used in calculating ground-water concentrations can be identified using MODFLOWand MODPATH. Caution must be used, however, indefining a stream tube volume for the calculation ofconcentration, since overestimating the stream tube vol-ume will produce lower concentrations and doses.

43 NUREG/CR-5927

5. Source-Term Modeling

Source-term models in the current methodology consist ofmodels for the breach of engineered barriers, leaching ofradionuclide chemicals into ground water, and transportof radionuclides to the boundary of the disposal unit.Current analysis methods in the methodology are theBLT code [Sullivan and Suen, 1989], and the mixing-cellcascade model [Kozak et al., 1990a], which is imple-mented in PAGAN [Chu et al., 1991]. These two ap-proaches contain advantages and limitations in how theyaddress each individual modeling area. Overall draw-backs to these methods are that neither can model chainsin the source, nor can they model gas production. Fur-thermore, in using BLT the analyst is constrained tousing FEMWATER to model the flow into and throughthe waste.

5.1 Engineered Barriers

We include in this discussion the time of failure and themechanism of failure of both concrete structures andmetal containers. These two aspects of the source-termanalysis are the only processes that can delay, the onset ofreleases from the disposal unit.

5.1.1 Concrete Structures

Current Approaches

Concrete structures are modeled in the current methodol-ogy as a step change in the concrete hydraulic permeabil-ity at some specified time. A flow analysis is performedusing VAM2D with an unsaturated hydraulic permeabilityfunction representative of intact concrete. At the time offailure, the unsaturated permeability function is replacedwith one representing a permeable soil. In essence, thisapproach assumes a rapid change between the conditionsdepicted in Figure 4.1 and those shown in Figure 4.2.There is considerable uncertainty in determining theappropriate properties for the failed vault. Less obvious,but still important, is the uncertainty in the properties ofthe intact concrete. The permeability of the concretemay vary considerably as a result of quality assurancepractices during emplacement. The only way to ensurethat the permeability of the completed concrete meetsdesign specifications is to conduct permeability tests onthe completed vault;

A step function in the flow rate through the vault isrecommended at this time in the performance assessmentanalyses. Gradual failures of the vault can also be mod-eled using the current methodology models, by using anumber of small step changes in the concrete flow prop-

erties. This approach would require the analyst to definea series of intermediate failed conditions between thoseshown in Figure 4.1 and Figure 4.2. In this approach,it is difficult to justify a gradual degradation rather thanan abrupt degradation, and to specify the flow propertiesof the partially failed concrete. The primary reason fordoing such an analysis is to take additional credit for theconcrete vault. At the present time, there are such largeuncertainties in modeling concrete failure that specifyinggradual failure is not generally defensible.

At the time the methodology was developed, an adequatemodel for concrete degradation did not exist. The primecandidate for use in modeling concrete degradation,BARRIER [Shuman et al., 1988], was proprietary andcould not be adequately assessed. Furthermore, use ofBARRIER provided an estimate of the time of failure,but the rate of failure was assumed to be an instantaneousfailure in applications of the code [Rogers and Associ-ates, 1988]. As discussed below, the performance as-sessment is relatively insensitive to the time of failure,hence it was concluded that BARRIER would not signifi-cantly enhance the capabilities of the methodology.

Status and Evaluation

Since the methodology was developed, there have beendevelopments in the area of concrete modeling. Cliftonand Knab [19891 assessed the current models for theservice life of concrete, and concluded that reasonableassurance of 500-year lifetimes should be possible.Walton et al. [1990] began developing models for thebehavior of partially failed vaults. More recent work byWalton and Seitz [1991] has provided guidelines for thedesign, construction, and operation of vaults. This workincluded models for evaluating flow and transport underpartially failed conditions. Pommersheim and Clifton[1991] identified models for evaluating the major degra-dation processes in concrete to be used in estimating theconcrete service life. In addition, Rogers and Associateswill soon produce a replacement for BARRIER, calledRAESTRICT [Shuman et al., 1991]. We do not yet fullyknow the scope or limitations of this new code.

These improvements still suffer from the limitations ofthe older models. Limitations on the current models are(1) lack of long-term experience with modem concrete,and (2) lack of an adequate validation basis for models offlow through concrete. To build confidence in thesemodels, efforts should be focused on validation experi-ments on old (50 to 100 years old) modem concrete,possibly experiments on ancient (1000 to 2000 years old)

45 NUREG/CR-5927

Source-Term Modeling

concretes, and comparisons with accelerated tests. Itwould be very useful to establish an experimental linkbetween accelerated tests and long-term analogs to pro-vide confidence that the accelerated tests represent long-term behavior [Olague et al., 1993]. Much of the issueof concrete longevity is related to quality assurance andquality control during construction rather than parametersthat are quantifiable for use in a performance assessment[MacKenzie et al., 1986]. This introduces additionaluncertainties into modeling of these structures. Olague etal. [1993] discuss a possible approach for quantifying thisuncertainty.

Another drawback to these existing models is their focuson the service life of the concrete rather than the rate ofchange of permeability. Kozak et al. [1990a] haveshown that for typical low-level waste inventories, therate of change of vault permeability of the concrete struc-ture is more important than the time at which failureoccurs. Low-level waste is made up almost entirely ofshort-lived (half life < 30 years) and very long-livedradionuclides (half life greater than 5000 years). Excep-tions to this rule are 63Ni and the actinides. However,these radionuclides will generally be present in smallconcentrations; their contributions to the total dose, whilenot negligible, will not typically dominate the perfor-mance assessment. It is relatively easy to provide aconvincing argument that a concrete structure will lastlonger than 100 to 200 years. After this time period,only the very long-lived constituents of low-level wasteremain. It would be very difficult to demonstrate that aconcrete structure will remain intact for the 2,000 to10,000 years necessary to allow decay of some of thelong-lived radionuclides. Consequently, there is verylittle change in the inventory during the crucial 200- to2000-year period, and therefore, performance assessmentresults are relatively insensitive to the time of failure ofengineered structures.

By contrast, the performance assessment is very sensitiveto the rate of change of concrete hydraulic conductivity.If the change occurs slowly, releases from the disposalunit will be spread out in time, and the peak dose will bereduced. For instance, if it could be demonstrated thatthe vault cannot reasonably fail such that from intactconditions until complete failure could not be less than100 years, this information would be very useful inperformance assessment modeling. The step functionfailure model could then be replaced by a ramp failuremodel, and it could be expected that calculated off-sitedoses would decrease as a result.

Analyzing the rate of change of the hydraulic conduc-tivity of concrete is difficult because. (1) the funda-mental mechanisms are not completely understood; (2)there is no long-term experience with modem concretefrom which to make our judgments; (3) there is uncer-tainty about the future condition of the site (see Chapter2); and (4) the most probable failure may be gradual, butthere will likely be some significant probability that thefailure will be rapid.

A second issue that needs to be addressed is how multi-ple vaults should be modeled. We could assign a proba-bility distribution of failure times to each vault, so thatall vaults do not necessarily fail at the same time. How-ever, the maximum dose from the possible sets of real-izations of these failure times would undoubtedly resultfrom a realization in which all vaults do fail at the sametime. There is not any technical justification for formallyremoving this possibility from consideration. For specif-ic sites, design considerations may suggest that the vaultfailure distributions may not overlap, in which case eachvault may be analyzed separately, and the consequencessuperposed. The approach to modeling multiple vaultswill be dictated by the overall approach to uncertaintyanalysis; optional approaches to uncertainty analysis arediscussed in Section 7.1.

A third issue that needs to be addressed is the differencebetween the idealized planned performance of a vault andits performance as emplaced. It is unreasonable to as-sume that the hydraulic conductivity of a completed vaultwill be as low as a small laboratory sample of that sameconcrete. In the concrete itself, differential settlement orstress fractures may cause the concrete permeability toincrease during the operational period. In addition, formost designs there will be enhanced degradation of theconcrete while the vault is operational, since it will beexposed to the elements above ground. The concreteslabs in vaults must be connected by joints, which aremade of materials that are not necessarily as long-lived asthe concrete itself. For instance, joints may be sealed bymetal that is subject to corrosion or by polymer materialswhose longevity is unknown. It is probable that lessconfidence can be placed in the longevity of joint materi-als than in the longevity of concrete. Another potentiallydisruptive difference between idealized behavior andactual behavior might be the obstruction of vault drainsby sediment or biological clogging, which is a frequentoccurrence in sanitary landfills [Bass, 1984].

We, therefore, conclude that the current modeling ap-proach in the methodology can be used to model eitherabrupt or gradual changes in concrete hydraulic perme-

NUREG/CR-5927 46


ability. The important issues related to this approach are(1) the hydraulic properties that are appropriate for in-tact, failed, and partially failed concrete are poorly un-derstood, and (2) it may be desirable to be able to justifygradual failure behavior for the vault. Research effortsshould be focused in these areas. The methodologycould be only marginally improved by including a modelto evaluate the service life of concrete vaults, and we donot recommend such models at this time.

5.1.2 Metal Container Degradation

Current Approaches

Degradation of metallic containers can be modeled in themethodology using either BLT or PAGAN. In PAGAN,container failures are modeled as a delay time until theonset of releases. This approach is not mechanistic, andthe time of failure must be justified by some othermeans, such as using BLT. In BLT, container degrada-tion is modeled using semiempirical models for corro-sion. The analytical structure of the model was suggest-ed by corrosion theory, and model parameters wereobtained from field data on underground corrosion. Themodel and its limitations are discussed in detail in Sulli-van et al. [1988], and in Kozak et al. [1989a]. Theprimary limitations are insufficient field data to justifysome of the parameters in the model, and insufficientdata on metals other than unpainted carbon steel.

Status

There have not been any significant changes to the corro-sion model since the development of the methodology.However, work is in progress to make BLT more user-friendly, and to extract the Breach and Leach portions ofthe code to stand alone. This improvement to the codeshould be useful as a user-friendly approach for overallperformance assessment methodology.

Evaluation

Given the increased emphasis on concrete structures inlow-level waste disposal facilities, metal degradation isless important than it would be for shallow-land burialsites. We conclude that the BLT approach for metalcorrosion is satisfactory for use in the methodology.

5.1.3 Degradation of Other Materials

A variety of other materials have been proposed to con-ýain low-level wastes in disposal. In particular, polymer-

ic materials, such as high-density polyethylene, have beenproposed. There is no long-term experience with any ofthese materials under disposal conditions, since most ofthem have only existed for (at best) a few tens of years.There are no existing models, nor is there a satisfactorydata base from which a model might be developed.Consequently, minimal credit should be given for theduration of such containers.

5.2 Leaching Processes and Near-

Field Transport

5.2.1 Leaching Processes

Current Approaches

Leaching processes are modeled in the methodologyusing either of two options in the mixing-cell cascademodel, or using the multiple options in BLT. The mix-ing-cell cascade model, depicted in Figure 5.1, wasderived for surface-wash and constant-rate leaching mod-els.

The surface-wash model was identified as being appropri-ate when the waste form is poorly understood, as inunconsolidated trash. The constant-rate model is consid-ered appropriate for some releases dominated by diffu-sion in the waste form. In particular, if the solubilitylimit is reached in the waste form, a constant internalconcentration will result in constant diffusional releases.This release model may, therefore, find its greatest use inmodeling concrete-stabilized wastes, in which the porefluid chemistry is dominated by the concrete. Pescatore[19911 has pointed out that a constant release model isnot mathematically appropriate for diffusional releaseswith a zero concentration boundary on the waste contain-er, since the mathematical solution is not convergent attime zero.

BLT is much more flexible in its treatment of the leach-ing process. The code contains models for surface-washleaching, diffusional leaching (in several geometries),congruent dissolution (appropriate for activated metals),or some combination of these mechanisms [Sullivan andSuen, 1989].

Status

Since the publication of the methodology, there havebeen several improvements to leaching models. Sullivanand Suen [1991] improved BLT to account for concentra-tions surrounding containers to accumulate, which tends

47 NUREG/CR-5927


7WaterFl77I

1

Td

-ENNOMMEContaminant Discharge I

Figure 5.1 The mixing-cell cascade model

to decrease diffusional release rates. This improvementadds an additional capability to the code. In addition, arecent extension of the mixing-cell cascade model bySullivan [1991] allows multiple simultaneous releasemechanisms. This improvement also adds a new capabil-ity to a model in the methodology. Of greatest impor-tance is the development (in progress) of a simplifiedmodel for source-term analyses, called DUST (DisposalUnit Source Term) [Sullivan, 1992]. This model isexpected to be much easier to use than BLT, because theanalyst will no longer be constrained to FEMWATER[Yeh and Ward, 1980] to generate a flow field to be usedwith BLT. Furthermore, the leaching model results willbe useable by alternative transport models, and the ana-lyst will not be constrained to FEMWASTE [Yeh andWard, 1982].

Evaluation

The mixing-cell cascade model has typically been usedwith a large number of mixing cells, which correspondsto plug flow in the disposal unit. This approach general-ly produces larger ground-water concentrations thanreleases calculated from a single mixing cell model.However, when the number of mixing cells becomeslarge, the source-term calculation becomes quite timeconsuming. This numerical inefficiency can be eliminat-ed by replacing the mixing-cell cascade model for largeN by analytical solutions for plug flow releases. Forinstance, the surface wash model in plug flow can berepresented by

NUREG/CR-5927 48


-= -Xt +- e T/I3" 0- 0'0 (1)

Q = 0, t > I/Q 0.

Here, Q0 is the initial release rate, X is the radionuclidedecay factor, and I is the inventory of the radionuclide.This simplified expression for the release rate is muchmore numerically efficient than the full mixing-cell cas-cade model.

We conclude that updated versions of the current modelsin the methodology will be useful, and should be incorpo-rated into the methodology as they become available.The mixing-cell cascade can be replaced in many casesby more efficient analytical expressions, and DUST isexpected to be much more numerically efficient andeasier to use than BLT. As in other areas of the sourceterm, the more important issue is how a validation basiscan be established that would allow the analyst to justifytaking credit for phenomena that decrease release rates.The models can account more or less rigorously forsolubility limitations, ion exchange phenomena, diffusionlimitations in waste forms, and sorption, which can allplay a significant role in reducing release rates, butwhich are all difficult to justify using because of theheterogeneities and uncertainties in low-level waste. Thistopic is discussed further in Volume 2 of this report.

5.2.2 Near-Field Transport

Current Approaches

As mentioned above, Sullivan extended the mixing-cellcascade to allow multiple simultaneous release mecha-nisms. The generalized mixing cell cascade will beuseful in any potential future developments of the model.However, incorporating more elaborate leach-modelexpressions into the model (both by us and by Brook-haven National Laboratory) has resulted in expressionsfor the global release rate so complicated that use of thismodel was unjustified. It is probable that analyticalmixing-cell cascade models may have been taken to theirpractical limit.

The alternative approach in the methodology, using a fulltransport analysis in VAM2D or BLT, can be cumber-some to apply, but it allows for flexibility in treatingnear-field transport.

Status and Evaluation

There have been no significant improvements in near-field transport modeling during the past two years. How-ever, experience in using the methodology suggests that itmay be desirable to introduce additional flexibility intothe source-term analysis. Areas in which flexibility canbe improved in straightforward fashion are (1) introduc-tion of an option to use an arbitrary user-defined releaserate (in tabular form), and (2) improved treatment ofdecay chains.

5.2.3 Decay Chains

Current Approaches

Decay chains are not explicitly considered in either themixing-cell cascade model or in the BLT. The primaryproblem with modeling decay chains in the source iswhen the parent and daughter are assigned differentretardation factors in the disposal unit. If the parent andall daughters are assigned the same retardation in thedisposal unit, the analyst can evaluate only the release ofthe parent radionuclide, and can correct for the ingrowthof daughters as the radionuclides exit the disposal unit.

If unequal retardation factors are used, an alternative toBLT or PAGAN is necessary. In using the methodology,we assume that the analyst has calculated the inventory,including decay and daughter ingrowth, at some baselineyear. The release of this inventory from particular loca-tions in the disposal unit can be modeled using leachingmodels, and the resulting time-dependent release ratescan be input to VAM2D to analyze transport in the dis-posal unit. This approach is cumbersome, because of thecomplexity of VAM2D. In addition, VAM2D is limitedto four-member chains (either straight chain orbranched), although this is not believed to be a seriouslimitation for the chains of importance in low-level wasteinventories [Kozak et al., 1990a].

Status

One alternative approach for unequal retardation factorswould be to develop analytical expressions on a case-by-case basis, assuming a mixing-cell model for the near-field transport. This analytical expression for the dis-charge rate can be used as input into a VAM2D analysisof the aquifer. This approach is straightforward and caneasily be implemented when needed for specific cases.However, it is more desirable to have the explicit capa-

49 NUREG/CR-5927


bility in the methodology to evaluate decay chains in thesource.

An existing model that uses a similar approach is theNEFTRAN II source model [Olague et al., 1991]. TheNEFTRAN II source model includes two options aboutnear-field dispersion: a mixing cell model and a "flow-through" model. The flow-through model represents aconvectively driven process, analogous to the releasesdescribed in Equation (1). Either source flow model canbe implemented with leach limits and solubility limits. Itappears that incorporating the NEFTRAN source-termmodel may be particularly useful for simple modeling ofchain members.

Evaluation

For generality and flexibility, we recommend incorpo-rating the NEFTRAN II source-term model into themethodology. In Chapter 4, we recommended incor-porating the NEFTRAN II transport model for unsatu-rated-zone transport, and potentially for some saturated-zone transport problems. The NEFTRAN II source-termand transport models represent an intermediate level ofcomplexity between very simple (PAGAN) and verycomplex (BLT or VAM2D), and additional flexibilityover the current approaches in VAM2D and PAGAN.

5.3 Gas Production

Current Approaches and Status

Issues associated with gas generation can be separatedinto two discrete subjects: generation of gases by thewaste, and migration of the gases to the surface. Thisdistinction was not made in discussions of gas evolutionin the original development of the methodology. Oncethe gases reach the surface, they form a release into theatmosphere. Depending on the exposure and pathwayassumptions, this release at the soil surface may need tobe used as an input into an air-transport model for theanalysis of off-site doses. Air-transport modeling isdiscussed in Chapter 6.

Potential gaseous radionuclides in low-level waste include3H, 14C, and 222Rn. In general, gas generation of 14C isprobably most important because of its relatively longhalf-life; therefore, the following discussion concentrateson gas generation of "4C. If the disposed waste containsnaturally occurring Th-230 or depleted uranium, thepotential exists for radon production and transport off-siteto be a significant contributor to the maximum dose. In

the case of radon, measurements are available for thedisequilibria between the radon and its daughters bothindoors and outdoors, which is important to its inhalationdosimetry for such a pathway [NCRP, 1988]. In addi-tion, although radon's half-life is relatively short, it is theparent of long-lived species (particularly Pb-210), sodaughters can potentially be transported in radiologicallysignificant amounts. The short-lived daughters of Rn-222(Po-218, Pb-214, and Bi-214) can produce significantlung doses from inhalation, since their dose-conversionfactors are large. Furthermore, the possibility exists forgaseous transport of radon to plant roots, and then decayof radon to longer lived radionuclides. This could be asignificant contributor of dose to man if bioaccumulationof the daughters in edible plant roots occurs. This sug-gests that there may be an enhanced transportation mech-anism to off-site locations for daughters of radon; thistransport pathway has been evaluated in the context ofnaturally occurring radon [NCRP, 1987], but studiesrelated to waste disposal sites are unknown to us.

The methodology does not currently contain a way toestimate the rate, volume, or radiological component ofgas production from low-level waste in a disposal facili-ty. This is an area where there has been relatively littleresearch, so data are unavailable to justify the use of anymodel at the present time. Existing approaches to model-ing gas generation have been developed for differentinventories than U.S. low-level waste streams (e.g.,Biddle, et al., 1987; Jefferies, 1990); however, gasgeneration may be important for low-level waste perfor-mance assessment because

1. Releases into the gas phase may decrease the im-pacts of the ground-water pathway. For some low-level waste inventories, "4C doses can add a substan-tial contribution to the ground-water dose. Thedoses from the air pathway and the ground-waterpathway are likely to occur at very different times,so the peak dose to the maximally exposed personfrom the ground-water pathway may potentially bereduced by accounting for gas releases.

2. The potential exists for bioaccumulation of gaseous14C0 2 in plants. This could concentrate otherwiseunimportant releases in the gas phase; consequently,this mechanism poses a possible enhanced transportpath from the disposal cells.

3. The aforementioned possibility of radon transportmay provide a preferential pathway for some of itsdaughters.

NUREG/CR-5927 50


4. Gas generation may have potentially deleteriouseffects on the integrity of vaults for some disposaldesigns [Hodgkinson et al., 1988].

Gas generation models are intimately linked to assump-tions about conditions that exist inside the low-level wastedisposal facility for the performance assessment timeperiod. Conditions inside the disposal facility have largeuncertainties associated with them because of the hetero-geneous nature of low-level radioactive waste. Theseuncertainties can be reduced to some extent for stabilizedwaste forms, but for the most part the uncertainties arelikely to remain large.

For gas production models, the mechanisms assumed tobe occurring are (1) microbial biodegradation of organicmaterials leading to releases of "4CO2 and 14CH4, and (2)production of titrated H2 gas from metal corrosion. Thelatter mechanism is believed to be less important, sincemetallic inventories are not expected to be large.

At the present time, we are unaware of any suitablemodels or experiments for gas generation that would beappropriate for evaluating U.S. low-level waste invento-ries and disposal conditions. An appropriate experimentwould consist of measurements of gas generation fromU.S. low-level waste in the physical and chemical condi-tions likely to be encountered by the waste in a disposalunit. Since most current disposal designs include mas-sive use of concrete, the experiment should be conductedfor high pH. Such an experiment may also be appropri-ate for many arid western sites, which are the only onescurrently being considered for trench burial [Olague etal., 1993].

The second aspect of the evolution of gas from the site isits transport from the disposal vault to the ground sur-face. Subsurface transport of radioactive gases has re-ceived considerable attention since the mid-1970s, be-cause of increased awareness of the potential for indoorexposure to naturally occurring Rn-222. Consequently,thete is a lot of empirical and theoretical informationavailable on subsurface transport of gases. This informa-tion will be applicable to emission of C-14, H-3, andother possible gas releases from low-level waste as wellas to radon emissions.

Current thinking about gas exhalation into houses sug-gests that it is dominated by convective gas flow in thesubsurface; measured radon concentrations in houses aretoo high to be explained by diffusion through the slab[Nazaroff, 1992]. The convective flow is the result ofbarometric pressure changes that cause transient convec-

tive transport of air into and out of the soil. However, itis not clear how the long-term average emission of gas isinfluenced by barometric pressure oscillations; moststudies in the literature have been concerned with thetemporal aspects of gas emission [NCRP, 1989; Nazar-off, 1992]. In general, at times of low pressure, gasesare "pumped" from the soil, while during times of highpressure, the emission of gas is suppressed. However,the suppression of gas emission during a high-pressureperiod can be less than the enhancement of emissionduring a low-pressure period [NCRP, 1989]. This sug-gests that barometric pumping may provide an enhancedtransport mechanism from the disposal facility to thesurface, even when averaged over long times. The long-term average emission of gas may depend on the averageduration of barometric pressure fluctuations, the radio-nuclide half life, the depth of burial, and other parame-ters. The influence of these parameters on long-termaverage gas emission from the soil surface is an aspect ofsubsurface gas transport that needs to be investigatedfrom a low-level waste performance assessment stand-point.

Evaluation

Based on the above discussion, there are not currentlyavailable adequate models for gas generation in low-levelwaste disposal facilities. Presently, the primary limita-tion is an inadequate experimental database for modeldevelopment. Some work is needed in this area.

Models are available for estimating gas transport throughsoils to the surface, which incorporate the effects ofdiffusion, convection, decay, daughter production, andspatial variability. These need to be evaluated to identifythe scope of the phenomena needed for modeling long-term average doses for performance assessment; it isrecommended that some capability is needed in the meth-odology for modeling subsurface transport of gases.

5.4 Geochemistry

Current Approaches

Geochemistry models are usually proposed for use inperformance assessment to justify taking credit for chem-ical limitations to transport. These limitations generallytake the form of either solubility limitations in the porefluid or complexation with soil minerals. To model theseeffects, geochemical models are usually developed toidentify the chemical speciation, which can then be used,together with detailed information about the chemical

51 NUREG/CR-5927


state of the ground water and soil, to evaluate the pro-cesses of interest.

Geochemistry models are likely to be most useful insource-term modeling, since the near-field chemicalenvironment in vaults may be quite well conditionedcompared to the surrounding natural soils. However,even in this well-established environment, geochemistrymodels suffer from a number of drawbacks. Olague etal. [1993] discuss the lack of an adequate validation basisfor geochemical models. This lack of adequate valida-tion, together with spatial and temporal uncertainties ingeochemistry, even in the near field, led Kozak et al.[1989a] to omit all complicated geochemical models fromthe methodology. Geochemistry is treated in the method-ology through the use of Kd sorption.

Status

The constraints on geochemistry modeling have not beenimproved since the methodology was developed. Conse-quently, there is not an impetus to change the models inthe methodology to incorporate more sophisticated geo-chemical models. Instead, we recommend that site-spe-cific geochemical data be collected and used to justifyreasonably conservative Kd values for use in performanceassessment. Detailed geochemical models may find arole in interpreting site characterization data to justifyconservative values for Kd values, but will probablycontinue to be too complicated for performance assess-inent.

A popular approach to modeling geochemical phenomenahas been to use "generic" data, usually derived fromsurrogate "conservative" data from Maxey Flats or Sa-vannah River. This approach is embodied in genericperformance assessment codes for low-level waste suchas IMPACTS [Oztunali and Roles, 1986], PRESTO[EPA, 1985b], and IMPACTS-BRC for Below-Regulato-ry Concern waste [Oztunali and Roles, 1984; O'Neal andLee, 1990]. This approach was also used recently byBaird et al. [1990] in a performance assessment of theClive, UT, low-level waste site, in spite of the differenc-es between the arid nature of Clive and the humid natureof the Savannah River site. It is often argued that Kd

values derived from humid sites are conservative, butthere is no experimental or theoretical basis for thisargument [Pescatore and Sullivan, 1991; Rao et al.,1992b]. The geochemistry of any site depends on thechemical and physical form of the waste, on the chemis-try of the waste, soil, and ground water, and on thehydrological flow regime of the site. Leaching results

from Maxey Flats depend in a complicated way on condi-tions at Maxey Flats, and other sites may bear no resem-blance to these conditions. The information produced foruse in the IMPACTS models was only intended to beused for generic modeling in support of rulemaking; itwas never intended to be used on a site-specific basis.

Evaluation

In summary, there does not appear to be a technical basisfor including any complicated geochemistry in the meth-odology. Site-specific geochemical data should be usedto identify conservative ranges for Kd to be used in per-formance assessment. We note that the new version ofNEFTRAN allows time-dependent values for Kd [Olagueet al., 1991], and these can be developed consistentlyfrom more detailed geochemistry analyses. This suggestsan additional degree of flexibility introduced when usingNEFTRAN. If site-specific data are unavailable, theminimum possible credit should be given for reduction ofimpacts as a result of geochemical effects. Surrogategeochemical data from existing low-level waste sitesshould not be used to justify Kd values when site-specificdata are absent.

5.5 Source-Term Summary

The current approach to modeling concrete degradation,which is based on the use of an unsaturated-zone flowmodel of the concrete under intact and failed conditions,is considered adequate. However, information about thehydraulic behavior of the concrete under partly failed andcompletely failed conditions is needed if any credit istaken for the behavior of the system under these condi-tions. Furthermore, detailed models for concrete degra-dation are not considered to be sufficiently advanced tojustify moderating the current approach, which assumes astep change in the flow properties of the concrete. Cur-rent concrete degradation models are considered to beuseful primarily for design analyses. The model formetal container corrosion is considered to be the bestavailable for performance assessment analyses. Some ofthe parameters required for the model have been devel-oped from a narrow experimental base, and this can beimproved. However, metal container corrosion is notimportant in most vault disposal systems, since the con-crete is generally expected to outlast the containers by awide margin. The greatest need in the area on concretedegradation is for experiments on permeability underseveral degrees of failure.

NUREG/CR-5927 52


Several improvements have been made (or are currentlybeing made) to leaching models in the methodology.These improvements generally relate to increased flexibil-ity or efficiency in the models, and should be incorporat-ed into the methodology as they become available. How-ever, they do not substantially change the existing capa-bilities of the methodology.

A significant improvement can be made to PAGAN if auser-defined source term in tabular form is used. Theuser could then model arbitrary releases by entering time-dependent values of releases. This approach would allowthe regulator to accept a license applicant's source-termoutput, and to confirm only the transport analysis inde-pendently. We recommend that this capability be devel-oped.

Adequate data to support models for gas production andgeochemistry are needed for research. Models are need-

ed, and should be developed, but such models should notbe included in the methodoiogy until there is an adequateexperimental basis for them. Treatment of subsurfacetransport of gaseous contaminants is better established,and such models will be needed for evaluation of gasreleases from a disposal facility. We recommend thatsuch models be investigated to determine which salientaspects of the processes are likely to be important from aperformance assessment perspective.

An improved treatment of decay chains in near-fieldtransport is needed. The mixing-cell cascade model andBLT are adequate for modeling decay chain transport ifall retardation factors are equal. If they are not,VAM2D or analytical solutions can be used, but are notsufficiently flexible. We, therefore, recommend that theNEFTRAN II source-term model be incorporated into themethodology. This addition will significantly enhancethe flexibility and capability of the methodology.

53 NUREG/CR-5927

6. Surface-Water Transport, Air Transport, and Exposure Pathway Modeling

The current primary computer code for all of these areasis GENII [Napier et al., 1988]. Consequently, the statusand evaluation of the code for each area has been com-bined into Section 5.4. Descriptions of the models usedin the code are given in Sections 5.1, 5.2, and 5.3.

6.1 Surface-Water Transport

Current Approaches

The current methodology uses the surface-water transportmodels recommended in NRC Regulatory Guide 1.113[NRC, 1977c]; these models are contained in the GENIIcomputer code [Napier et al., 19881. The GENII modelcan be used for a river or lake and assumes a constantflow depth, a constant convective velocity, a constantriver width, a constant lateral dispersion coefficient, astraight river channel, and a continuous point dischargeof contaminants [Kozak et al., 1990a].

The important pathways in low-level waste performanceassessment relate only to dissolved radionuclides; conse-quently, radionuclide interaction with the sediments canfrequently be neglected. For most cases, this is con-servative since adsorption of radionuclides onto the sedi-ments would cause liquid concentrations to be lower thanthose estimated neglecting sediment sorption. The possi-ble exceptions to this would be if bottom feeding fishcontributed significantly to the exposure, or if the exter-nal exposure from the contaminated surface water wasimportant in estimating total exposure. These two situa-tions are not typical; therefore, neglecting sedimentinteractions is usually acceptable, although this must bedetermined on a site-specific basis. The more usualeffect of neglecting sediment sorption is to produce con-servative estimates of exposure via the food chain[NCRP, 1984]. Nevertheless, a simple approach tosorption of radionuclides on sediments is included inGENII, and can be used if desired.

The interaction between ground water and surface water(used as an input to the surface-water models) is calculat-ed based on conservative assumptions in the methodology[Kozak et al., 1990a]. We assume that all radionuclidesdistributed in the aquifer at the location of the surface-water body are discharged into the surface water. Ingeneral, we expect that much less of the plume willactually end up in surface water. Furthermore, we as-sume that the discharge will occur at a point at the shore;a diffuse distributed plume entering the surface water isnot considered.

The methodology does not currently contain models forvarious kinds of ephemeral flows, such as surface runofffrom above-ground vaults.

6.2 Air-Transport Modeling

Current Approaches

Although air as a transport medium was not identified byShipers and Harlan [19891 as part of a significant path-way, Kozak et al. [1989a], included air-transport modelsin the methodology because of circumstances that mayarise where soils become contaminated and entrained inair (e.g., dry lake bed, intermittent stream) and becauseair pathways may become significant in intruder scenari-os. Besides simulating airborne transport of contaminat-ed particulates, models are also needed to simulate trans-port of radioactive gases that may be released from adisposal facility.

For modeling airborne transport of particulates, a mass-loading factor was recommended to estimate the amountof contaminant that is entrained in the air. The massloading model assumes that the source of airborne con-centration can be expressed as the product of the amountof soil particles suspended in the air and the radionuclideconcentration on the soil [Kozak et al., 1989a]. Thismodel assumes that materials have been mixed uniformlywith the soil and that soil and contaminants are suspendedequally. Kozak et al. [1989a], recommend this modelbecause conservative values can be identified, and be-cause it is relatively simple.

Transport of the gases or particulates is calculated usinga Gaussian plume model, which is recommended in NRCRegulatory Guide 1.111 [NRC, 1977b], and implementedin the GENII computer code [Napier et al., 1988]. TheGaussian plume model assumes one-dimensional convec-tive transport, with three-dimensional dispersive trans-port. This model has been adopted as a standard methodin regulating both radioactive [NRC, 1977a; IAEA,19801 and other [EPA, 1978] airborne species. Gaussianplume models are derived for point sources in space.For the purposes of performance assessment, area sourc-es can be treated using the conservative approaches de-scribed by Chu et al. [1991].

55 NUREG/CR-5927

Surface-Water Transport

6.3 Exposure Pathway Modeling

Current Approaches

The methodology uses the pathway models found in NRCRegulatory Guide 1.109 [NRC, 1977a] to determineradionuclide intake rates for a person from concentrationsin the environment. These models are contained in thecomputer code GENII [Napier et al.,- 1988], and accountfor bioaccumulation in plants, irrigation of various crops,inhalation, ingestion of drinking water and contaminatedfoods and external exposure [Kozak et al., 19911. Themodels can be used to account for direct biointrusion intothe waste as well as root uptake of radionuclides trans-ported to plants located off-site.

A large degree of uncertainty is associated with thesemodels because many assumptions have to be made aboutfuture conditions at the site (e.g., agricultural activity).Besides this, the most significant issue for food-chainmodels centers on radionuclide bioaccumulation, especial-ly for "4C and iodine compounds [Olague et al., 1993].The NCRP [19A] states that few efforts have been madeto validate the bioaccumulation values listed in the regu-latory guide and that validation studies are needed be-cause those in the regulatory literature are overly conser-vative. Another issue includes the lack of isotope-specif-ic transfer coefficients for specific foods, although con-servative assumptions should overestimate concentrationsin terrestrial foods and bioaccumulation factors in aquaticfood chain transport models. In addition, food consump-tion parameters are based on 1965 data; given the changein U.S. dietary habits since then, these values are quitesuspect [NCRP, 19841. Of course, these uncertaintiesare small compared to the uncertainty in U.S. dietaryhabits over the timescale of the performance assessment.

6.4 Status and Evaluation

Status

GENII was developed for the Hanford site, and maycontain some hard wired parameters that are specific tothat site. At this time, it appears likely that any issuesassociated with the model are in the exposure pathwaymodels. This has been a lingering concern about theusing GENII as a generic risk assessment tool.

A revised version of GENII, called GENII-S [Leigh etal., 1992], has recently become available. The primarydifferences between GENII and GENII-S are (1) thecapability to perform probabilistic pathway analyses as

well as deterministic ones, and (2) an improved userinterface. The user interface implemented in GENII-S isthe same one used in PAGAN [Chu et al., 1991], but hasbeen adapted to accommodate the GENII input and out-put.

Evaluation

GENII-S has attractive features, and the user interface islikely to be more user friendly than the APPRENTICEshell introduced with the original GENII code. GENII-Sshould, therefore, be used in the methodology. Themodels in GENII-S are identical to the models in GENII;we briefly evaluate these models here.

The surface-water pathway is usually much less importantthan the ground-water well pathway, because of greaterdilution and (usually) greater distances from the disposalunits. Consequently, quite conservative assumptions canoften be made without significantly affecting the decisionresulting from the analysis. However, there may beoccasions when a more elaborate method is needed forthe interaction between ground water and surface water.As discussed in Chapter 4, this need can be met by in-cluding MODFLOW [MacDonald and Harbaugh, 1988]in the methodology. There may also be occasions whenmore elaborate models are needed to evaluate surface-water transport. Such models are available [see, e.g.,Jirka et al., 1983], but are rarely used, so they are notrecommended for formal inclusion in the methodology.

Surface-water runoff models may be needed to assessaboveground vaults. As discussed in Chapter 3, when anaboveground vault fails, the waste is exposed directly atthe soil surface. The potential, therefore, exists forexposures to occur by exposure to surface-water runoffthat has contacted the waste. Contaminated runoff mightenter the food chain in several ways. First, runoff cancontaminate ground water or surface water used fordrinking, or it can contaminate nearby fields or irrigationditches that are used by crops or livestock. This pathwayis of concern only for aboveground vaults. We are cur-rently unaware of plans by any State or Compact to usean aboveground vault, so to some extent the issues asso-ciated with modeling surface runoff are of secondaryimportance. Nevertheless, the capability to conduct suchanalyses may become important in the future.

We conclude that the surface-water transport models inthe methodology are adequate for producing estimates ofthe dose that should be conservative in most cases. Themodels are very simple, and can be implemented eitheras hand calculations, or by using GENII-S. Similarly,

NUREG/CR-5927 56

Surface-Water Transport

the Regulatory Guide 1. 111 air-transport models are verysimple, and can be implemented easily, either in GENII-S or by using hand calculations.

The exposure pathway models are likely to be the onescontaining Hanford site-specific parameters. A carefulevaluation of the exposure pathway models is needed toestablish confidence in GENII results. In addition, mostlow-level waste performance assessment analyses include(at most) exposures from drinking contaminated water,and exposures to contaminated crops and livestock.These calculations are a very small subset of the overallcapabilities of GENII, so that using GENII to evaluatethese simple exposure scenarios often seems excessive.

We, therefore, propose to develop a simple generic path-way analysis that includes only the pathways used mostoften in low-level waste performance assessments. Thisanalysis might be implemented as a spreadsheet applica-tion or a simple, clear FORTRAN program. This sim-plified analysis has several advantages over GENII.First, the assumptions and limitations of the analysis canbe made more transparent to the user than they are in acomplicated code like GENII. Second, the concernsabout possible site-specific assumptions in GENII areeliminated. GENII does need to be retained in the meth-odology for its generality and flexibility, but most per-formance assessment calculations can be done on a muchsimpler level.

57 57 NUREG/CR-5927

7. Dosimetry Modeling

Current Approaches

Once the intake of radionuclides for a person have beenestablished based on exposure pathway models (seeChapter 6), dosimetry models are needed to estimate theeffect of this intake on the human body. In the method-ology, internationally accepted dosimetry models [ICRP26, 1977] based on dose-conversion factors [ICRP 30,1982-1988] are used. The dosimetry models in the meth-odology are implemented in the computer code GENII[Napier et al., 1988].

Status

The ICRP dose conversion factors are based on a modelof the human body. Critical organs are represented aswell-mixed cells, and transfer factors are specified forthe interchange of radionuclides between the organs.These transfer factors are specified by a combination ofhuman and animal radiological studies, which are blendedusing professional judgment.' These models have notbeen compared against human databases of radiologicalexposures. Furthermore, there is no attention given tothe uncertainties in the models: parameter uncertainty isacknowledged, but then neglected in the ICRP methodol-ogy. Olague et al. [1993] recommend a formal parame-ter uncertainty analysis to evaluate the uncertainty inICRP 30 dose conversion factors. In addition, the con-ceptual model uncertainty is large for low dose rates.There is statistically significant evidence that moderatelylow levels of radiation may be beneficial to humans(radiation hormesis) [Luckey, 1989; Cohen, 1990]. Forvery low levels of radiation (25 mrem), stochastic healtheffects are so small that they can only be determinedusing epidemiological studies with populations the size ofthe entire Earth's population [Gershey et al., 1990].This means that we will never be able to identify whetherstochastic health effects actually exist for very low doserates. Therefore, this represents a very large conceptualmodel uncertainty in the dosimetry models.

The ICRP has issued ICRP 60, containing updated rec-ommendations since the publication of the original meth-odology [ICRP 60, 1990]; these recommendations super-sede the recommendations of ICRP 26 [ICRP 26, 1977].There are both minor and major differences betweenICRP 26 and ICRP 60. Among the minor changes,effective dose equivalent is now merely called effectivedose; doses to individual organs are now called equiva-lent doses, a reversion to an earlier nomenclature. Com-

mitted equivalent and committed effective doses aredefined comparably to their earlier counterparts: they arethe time integral of the dose following intake over 50years for adults or 70 years for children. The term"non-stochastic" health effects has been replaced bydeterministic health effects.

Of greater importance in low-level waste performanceassessment is a change in organ and tissue weightingfactors used in calculating the effective dose. Differenc-es between tissue weighting factors between ICRP 26 andICRP 60 are shown in Table 7.1. The weighting factorshave been revised in an updated attempt to ensure thatthe effective dose would represent the same level of riskof stochastic health effects (denoted detriment in theICRP documentation) regardless of the tissue or organinvolved.

Evaluation

In spite of these changes, the ICRP 26/30 dose factorsare still widely accepted, and are still considered to be astandard. Available published guidance [Eckerman etal., 1988] does not contain enough information to calcu-late effective doses according to the ICRP 60 standard;the guidance is for the ICRP 26 standard. We, there-fore, conclude that these models are currently the bestavailable from a defensible regulatory standpoint for usein the methodology. The ICRP 60 standard should beadopted once guidance is available on values for doseconversion factors.

However, we also note that to use these models, it isnecessary to establish a regulatory position that maximumdoses will be based on the ICRP standard man. Otherassumptions about the person receiving the dose willresult in different dose calculations. In keeping with therecommendations for air transport, surface-water trans-port, and food-chain analysis, we recommend that untilthe evaluation of GENII is complete, dosimetry analysesshould be done using hand calculations. The dose con-version factors in GENII are not Hanford-specific, but ifthe other portions of the analysis are done outside ofGENII, it would not make sense to use GENII just togenerate dose conversion factors. The dose analyses canbe implemented quite easily; we recommend that a gener-ic simplified analysis that only includes ingestion dosesshould be developed for use in these analyses. The doseconversion factors can be found in Eckermann et al.[1988].

'O'Neal, B., personal communication, SNL, 1991.

59 NUREG/CR-5927

Dosimetry Modeling

Table 7.1 Differences in tissue weighting factors between ICRP 26 and ICRP 60

Tissue or Organ Tissue weighting factor Tissue weighting factor(ICRP 26) (ICRP 60)

Gonads 0.25 0.20Red Bone Marrow 0.12 0.12Colon ---- 0.12Lung 0.12 0.12Stomach ---- 0.12Bladder ---- 0.05Breast 0.15 0.05Liver ---- 0.05Oesophagus ---- 0.05Thyroid 0.03 0.05Skin ---- 0.01Bone Surface 0.03 0.01Remainder 0.30 0.05

NUREG/CR-5927 60

8. Summary

The purpose of this report is to identify models thatshould be incorporated into the low-level waste perform-ance assessment methodology. Each modeling area ofthe methodology has been reviewed in this report, andsome additional areas of concern have been discussed.

We have introduced an additional modeling area from ourearlier discussions: the area of subsurface transport ofgases, reflected in Figure 8.1, which is an updated figureshowing the modeling areas in the methodology.

The primary recommended modeling changes for themethodology (Table 8.1) are incorporating NEFTRAN IIfor the analyses of source-term and ground-water trans-port, incorporating MODFLOW to improve the flexibilityof the methodology in treating saturated-zone flow, de-veloping a simplified application to replace GENII formany applications, and replacing VAM2D by a code suchas VS2DT after adapting the latter code to handle decaychain transport. We believe that NEFTRAN will provideadditional flexibility and intermediate level of complexitythat will be useful in some circumstances. The recom-mended code development to VS2DT is intended to solvepersistent issues about using VAM2D: the code is propri-etary, and it does not have ideal quality assurance.These issues are expected to become more important if

the code is used in actual licensing analyses. Theserecommendations are shown in Figure 8.2.

Other key recommendations are as follows. A strongvalidation program is needed to begin building confidencein all of the models in the methodology as they apply totheir regulator purpose. Formal uncertainty analysismethods are recommended for model uncertainty, param-eter uncertainty, and uncertainty about the future of thesite. We recommend that model uncertainty be addressedby analyzing multiple conceptual models in parallel, anddifferentiating between them using site-specific validationand model intercomparison. The uncertainty analysis ap-proach recommended consists of allowing multiple con-ceptual models, a Monte Carlo analysis with Latin Hy-percube Sampling, coupled with a full analysis of possi-ble future states of the site. We recommend that formalapproaches are needed for acknowledging all assumptionsand their links to site-specific data. At its simplest level,this can be an explicit recognition of the assumptionsmade in the analysis. However, eventually it will bedesirable to develop a formal method for using site-spe-cific data to develop adequate (conservative) regulatoryconceptual models. Finally, we recommend that effort isneeded to improve the user friendliness of the methodolo-gy.

61 NUREG/CR-5927

Summary

Figure 8.1 Updated processes in the methodology

NUREG/CR-5927 62

Summary

Figure 8.2 Current recommendations for the methodology

63 NUREG/CR-5927

Summary

Table 8.1 Recommended changes to the models in the methodology

Modeling Area

Ground Water

Source Term

Surface Water

Recommendation

Incorporate VS2DT, NEFTRAN II, and MODFLOWAdapt VS2DT to conduct chain decay analyses

No change to concrete modelingIncorporate simplified Breach and Leach modelsIncorporate NEFTRAN IIAllow tabular inputEvaluate subsurface gas transport

Implement simplified analysisUse MODFLOW to evaluate interactions with ground water

Implement simplified analysis

Write a simplified ingestion application

Write a simplified ingestion application

Formal uncertainty analysis neededImproved user friendliness neededStrong validation program needed

Air

Food Chain

Dosimetry

Overall (All Areas)

NUREG/CR-5927 64

9. References

Andersson, J., T. Carlsson, T. Eng, F. Kautsky, E.Soderman, and S. Wingefors, "The Joint SKI/SKB Sce-nario Development Project," SKI TR 89:14, SwedishNuclear Power Inspectorate, Stockholm, Sweden, 1989.

Apostolakis, G., "The Interpretation of Probability inProbabilistic Safety Assessments," Reliability Engineeringand System Safety, Vol. 23, 247-252, 1988.

Atkinson, A. and G. P. Marsh, "Engineered Barriers:Current Status," NS/G102, Harwell Laboratory, 1988.

Atkinson, A. and A. K. Nickerson, "Diffusion and Sorp-tion of Cesium, Strontium, and Iodine in Water-SaturatedCement," Nuclear Technology, Vol. 81, 100, 1988.

Atwood, C. L., "Choose the Philosophy to Fit the Task,"Reliability Engineering and System Safety, Vol. 23, 259-262, 1988.

Baird, R. D., M. K. Bollenbacher, E. S. Murphy, R.Shuman, and R. B. Klein, "Evaluation of the PotentialPublic Health Impacts Associated with Radioactive WasteDisposal at a Site near Clive, Utah," RAE-9004/2-1,Rogers and Associates Engineering Corp., 1990.

Balek, J., "Groundwater Recharge Concepts," Estimationof Natural Groundwater Recharge, I. Simmers, ed., 3-9,D. Reidel Publishing Co., Boston, MA, 1988.

Bass, J. M., "Avoiding Failure of Leachate CollectionSystems at Hazardous Waste Landfills," EPA-600/d-84-210, U.S. Environmental Protection Agency, August1984.

Bennett, R. D., "Alternative Methods for Disposal ofLow-Level Radioactive Wastes, Task 2e: TechnicalRequirements for Shaft Disposal of Low-Level Radio-active Waste," NUREG/CR-3774, Vol. 5, U.S. NuclearRegulatory Commission, 1985.

Bennett, R. D. and J. B. Warriner, "Alternative Methodsfor Disposal of Low-Level Radioactive Wastes: Task 2bTechnical Requirements for Above-Ground VaultDisposal of Low-Level Radioactive Waste," NUREG/CR-3374, Vol. 3, U.S. Nuclear Regulatory Commission,1986.

Berthouex, P. M. and I. Hau, "Difficulties Related toUsing Extreme Percentiles for Water Quality Regula-tions," Research Journal WPCF, Vol. 63, 873-879,1991.

Biddle, P., J. H. Rees, and P. E. Rushbrook, "GasGeneration in Repositories," AERE-R-12291, September1987.

Bonano, E. J. and R. M. Cranwell, "Treatment ofUncertainties in the Performance Assessment of GeologicHigh-Level Radioactive Waste Repositories," Mathe-matical Geology, Vol. 20(5), 1988.

Bonano, E. J., P. A. Davis, L. R. Shipers, K. F. Brinst-er, W. E. Beyeler, C. D. Updegraff, E. R. Shepherd, L.M. Tilton, and K. K. Wahi, "Demonstration of a Perfor-mance Assessment Methodology for High-Level Radio-active Waste Disposal in Basalt Formations," NUREG/CR-4759, SAND86-2325, Sandia National Laboratories,Albuquerque, NM, 1989.

Bonano, E. J., S. C. Hora, R. L. Keeney, D. von Win-terfeldt, "Elicitation and Use of Expert Judgment inPerformance Assessment for High-Level RadioactiveWaste Repositories," NUREG/CR-541 1, SAND89-1821,Sandia National Laboratories, Albuquerque, NM, 1990.

Bonano, E. J., L. R. Shipers, and A. L. Gutjahr, "Sto-chastic Analysis of Contaminant Transport in PorousMedia: Analysis of a Two-Member RadionuclideChain," Water Resources Research, Vol. 2(6), 1063,1987.

Chhibber, S., G. Apostolakis, and D. Okrent, "A Prob-abilistic Framework for the Analysis of Model Uncer-tainty," Trans. L Chem. Eng., Vol. 69, Part B, 1-9,1991a.

Chhibber, S., G. Apostolakis, and D. Okrent, "On theQuantification of Model Uncertainty," in Proc. of theInternational Conf. on Probabilistic Safety Assessmentand Management (PSAM), held in Beverly Hills, CA,February 4-7, 1991b.

Chu, M. S. Y., M. W. Kozak, J. E. Campbell, and B.M. Thompson, "A Self-Teaching Curriculum for theNRC/SNL Low-Level Waste Performance AssessmentMethodology," NUREG/CR-5539, SAND90-0585,Sandia National Laboratories, Albuquerque, NM, 1991.

Clifton, J. R, and L. I. Knab, "Service Life ofConcrete," NUREG/CR-5466, NISTIR89-4086, U.S.Nuclear Regulatory Commission, 1989.

65 NUREG/CR-5927

References

Cohen, B. L., "An Experimental Test of the Linear No-Threshold Theory of Radiation Carcinogenesis," Radon,Radium, and Uranium in Drinking Water, C.R. Cothernand P. A. Rebers, eds., Lewis Publishers, Chelsea, MI,1990.

Cook, J. R., "Final Report: Greater ConfinementDisposal Technology at the Savannah River Plant,"prepared for U.S. Department of Energy by E. I. Dupontde Nemours and Co., Inc., Savannah River Laboratory,Aiken, SC, 1984.

Cranwell, R. M., J. E. Campbell, J. C. Helton, R. L.Iman, D. E. Longsine, N. R. Ortiz, G. E. Runkel, andM. J. Shortencarier, "Risk Methodology for GeologicDisposal of Radioactive Waste: Final Report," NUREG/CR-2452, SAND81-2573, Sandia National Laboratories,Albuquerque, NM, 1987.

Cranwell, R. M., R. W. Guzowski, J. E. Campbell, andN. R. Ortiz, "Risk Methodology for Geologic Disposalof Radioactive Waste: Scenario Selection Procedure,"NUREG/CR-1667, SAND80-1429, Sandia NationalLaboratories, Albuquerque, NM, 1990.

Dagan, G., "Solute Transport in Heterogeneous PorousFormations," J. Fluid Mechanics, Vol. 145, 151-177,1984.

Dagan, G., "Statistical Theory of Groundwater Flow andTransport: Pore to Laboratory, Laboratory to Form-ation, and Formation to Regional Scale," WaterResources Research, Vol. 22(9), 120S, 1986.

Davis, P. A., N. E. Olague, and M. T. Goodrich, "Ap-proaches for the Validation of Models Used for Per-formance Assessment of High-Level Nuclear WasteRepositories," NUREG/CR-5537, SAND90-0575, SandiaNational Laboratories, Albuquerque, NM, 1991.

Davis, P. A., E. J. Bonano, K. K. Wahi, and L. L.Price, "Uncertainties Associated with PerformanceAssessment of High-Level Radioactive Waste Reposi-tories," NUREG/CR-5211, SAND88-2703, SandiaNational Laboratories, Albuquerque, NM, 1990a.

Davis, P. A., L. L. Price, K. K. Wahi, M. T. Goodrich,D. P. Gallegos, E. J. Bonano, and R. V. Guzowski,"Components of an Overall Performance AssessmentMethodology," NUREG/CR-5256, SAND88-3020,Sandia National Laboratories, Albuquerque, NM, 1990b.

Davis, P. A. and N. E. Olague, "Treatment of ModelingUncertainties in Total System Safety Assessments," inProc. of the International Conference on ProbabilisticSafety Assessment and Management (PSAM), G.Apostolakis, ed., 649, held in Elsevier, 1991.

Deering, L. G. and M. W. Kozak, "A PerformanceAssessment Methodology for Low-Level RadioactiveWaste Facilities," in Proc. 12th Annual DOE Low-LevelWaste Conference, held in Chicago, IL, August 28-29,1990.

Denson, R. H., R. D. Bennett, R. M. Wamsley, D. L.Bean, and D. L. Ainsworth, "Recommendations to theNRC for Review Criteria for Alternative Methods ofLow-Level Waste Disposal: Task 2a Below-GroundVaults," NUREG/CR-5041, Vol. 1, U.S. Nuclear Regu-latory Commission, 1987.

Denson, R. H., R. D. Bennett, R. M. Wamsley, D. L.Bean, and D. L. Ainsworth, "Recommendations to theNRC for Review Criteria for Alternative Methods ofLow-Level Waste Disposal: Task 2b Earth-MoundedConcrete Bunkers," NUREG/CR-5041, Vol. 2, U.S.Nuclear Regulatory Commission, 1988.

Dickman, P. T. and J. R. Boland, "Greater ConfinementDisposal Test, Fiscal Year 1983 Progress," in Proc. ofthe 5th Annual Participants' Information Meeting, DOELow-Level Waste Management Program, DOE Con-ference Proceedings CONF-8308106, 316-323, 1983.

Doctor, P. G., "Sensitivity and Uncertainty Analyses forPerformance Assessment Modeling," EngineeringGeology, Vol. 26, 411-429, 1989.

Doctor, P. G., E. A. Jacobson, and J. A. Buchanan, "AComparison of Uncertainty Analysis Methods Using aGroundwater Flow Model," PNL-5649, Pacific North-west Laboratory, 1988.

Dougherty, D. and P. Columbo, "Leaching Mechanismsof Solidified Low-Level Waste. The Literature Survey,"BN-51899 E1.99 DE86 002088, 1985.

Eckermann, K. F., A. B. Wolbarst, and A. C. B. Rich-ardson, "Limiting Values of Radionuclide Intake and AirConcentration and Dose Conversion Factors for In-halation, Submersion, and Ingestion," EPA--520/1-88-020, DE89 011065, Oak Ridge National Laboratory, OakRidge, TN, 1988.

NUREG/CR-5927 66

References

EPA, "Guidelines on Air Quality Models," Report Nos.EPA-450/2-78-027, OAQPS No. 1.2-080, U.S, Environ-mental Protection Agency, 1978.

EPA, "PRESTO-EPA: A Low-Level Radioactive Envi-ronmental Transport And Risk Assessment Code -Methodology Manual," EPA 520/1-85-001, U.S. Envi-ronmental Protection Agency, 1985a.

EPA, "Environmental Standard for the Management andDisposal of Spent Nuclear Fuel, High-Level Waste, andTransuranic Radioactive Waste; Final Rule, 40 CFR Part191," Federal Register, Vol. 50, No. 182, Washington,DC., 1985b.

Feraday, M. A., "Canadian Experience with the Storageand Disposal of Low- and Intermediate-Level Wastes," inProc. of the Symposium on Low-Level Waste Disposal(NUREG/CP-0028), Vol. 3, U.S. Nuclear RegulatoryCommission, 1982.

Finkel, A. M., "Confronting Uncertainty in Risk Man-agement," Center for Risk Management, Resources forthe Future, Washington, DC, 1990.

Freeze, R. A., J. Massmann, L. Smith, T. Sperling, andB, James, "Hydrological Decision Analysis: 1. AFramework," Ground Water, Vol. 28, 738-766, 1990.

Gallegos, D. P., A Performance Assessment Methodologyfor High-Level Radioactive Waste Disposal in Unsat-urated, Fractured Tuff, NUREG/CR-5701, SAND91-0539, Sandia National Laboratories, Albuquerque, NM,1991.

Gee, G. W. and D. Hillel "Groundwater Recharge inArid Regions: Review and Critique of Estimation Meth-ods," Hydrological Processes, Vol. 2, 255, 1988.

Gelhar, L. W. and C. L. Axness, "Three-DimensionalStochastic Analysis of Macrodispersion in Aquifers,"Water Resources Research, Vol. 19, 161-180, 1983.

Gershey, E. L., R. C. Klein, E. Party, and A. Wilker-son, Low-Level Radioactive Waste: From Cradle toGrave, Van Nostrand, New York, 1990.

Harr, M., Reliability-Based Design in Civil Engineering,McGraw-Hill, New York, 1987.

Healy, R. W., "Simulation of Solute Transport in Vari-ably Saturated Porous Media with SupplementalInformation on Modifications to the U.S. GeologicalSurvey's Computer Program VS2D," Water ResourcesInvestigation Report 90-4025, U.S. Geological Survey,1990.

Heger, A. S., F. A. Kerl, D. P. Gallegos, and P. A.Davis, "Use of Probabilistic Networks in Quantificationof Uncertainties in Development of Conceptual Modelsfor Radioactive Waste Repositories," in Proc. ANSWinter Meeting, held in San Francisco, CA, November10-14, 1991.

Helton, J. C., J. W. Garner, R. D. McCurley, and D.K. Rudeen, "Sensitivity Analysis Techniques and Resultsfor Performance Assessment at the Waste Isolation PilotPlant," SAND90-7103, Sandia National Laboratories,Albuquerque, NM, 1991.

Hills, R. G. and P. J. Wierenga, "Model Validation atthe Las Cruces Trench Site," NUREG/CR-5716, U.S.Nuclear Regulatory Commission, 1991.

Hodgkinson, D. P., P. C. Robinson, P. W. Tasker, andD. A. Lever, "A Review of Modeling Requirements forRadiological Assessment," Radiochimica Acta, Vols.44/45, 317-324, 1988.

Huyakorn, P. S., J. B. Kool, and J. B. Robertson,"Documentation and User's Guide: VAM2D - VariablySaturated Analysis Model in Two Dimensions,"NUREG/CR-5352, HGL/89-01, April 1989.

IAEA, "Atmospheric Dispersion in Nuclear Power PlantSiting," Safety Series No. 50-SO-SG, InternationalAtomic Energy Agency, Vienna, 1980.

IAEA, "Evaluating the Reliability of Predictions MadeUsing Environmental Transfer Models," IAEA SafetySeries No. 100, International Atomic Energy Agency,1989.

ICRP 26, "Recommendations of the InternationalCommission on Radiological Protection," ICRP Publi-cation 26, Pergamon Press, New York, 1977.

ICRP 30, "Limits for Intakes of Radionuclides by Wor-kers," ICRP Publication 30, Part 1, Pergamon Press,New York, 1982.

67 NUREG/CR-5927

References

ICRP 60, "Recommendations of the InternationalCommission on Radiological Protection," ICRP Publi-cation 60, Pergamon Press, New York, 1990.

Iman, R. L., J. C. Helton, and J. E. Campbell, "AnApproach to Sensitivity Analysis of Models: Part I -Introduction, Input Variable Selection and PreliminaryVariable Assessment," Journal of Quality Technology,Vol. 13, 174, 1981.

Iman, R. L. and W. J. Conover, "A Distribution-FreeApproach to Inducing Rank Correlation Among InputVariables," Communications in Statistics, Vol. B 11(3),311-334, 1982.

Jirka, G. H., A. N. Findikakis, Y. Onishi, and P. J.Ryan, "Transport of Radionuclides in Surface Waters,"Radiological Assessment, J. E. Till and H. R. Meyer,eds. NUREG/CR-3332, ORNL-5968, Oak RidgeNational Laboratory, Oak Ridge, TN, 1983.

Jefferies, N. L., "The Evolution of Carbon-14 andTritium Containing Gases in a Radioactive WasteRepository," NSS/R-198, Harwell Laboratory, UKAEA,1990.

Kerl, F. A., A. S. Heger, D. P. Gallegos, and P. A.Davis, "Methodology to Obtain Expert InformationAbout the Conceptual Model Development Process Usedfor Performance Assessment of Waste ManagementSites," in Proc. ANS Winter Meeting, held in SanFrancisco, CA, November 10-14, 1991.

Kittel, J. H., "Introduction and Background," Near-Surface Land Disposal, J. H. Kittel, ed., Harwood Aca-demic Publishers, New York, 1, 1989.

Knutsson, G., "Humid and Arid Zone GroundwaterRecharge - A Comparative Analysis," Estimation ofNatural Groundwater Recharge, I. Simmers, ed., D.Reidel Publishing Co., Boston, MA, 271-282, 1988.

Kozak, M. W., C. P. Harlan, M. S. Y. Chu, B. L.O'Neal, C. D. Updegraff, and P. A. Mattingly,"Background Information for the Development of a Low-Level Waste Performance Assessment Methodology:Selection and Integration of Models," NUREG/CR-5453), SAND89-2505, Vol. 3, Sandia NationalLaboratories, Albuquerque, NM, 1989a.

Kozak, M. W., M. S. Y. Chu, C. P. Harlan, and P. A.Mattingly, "Background Information for the Developmentof a Low-Level Waste Performance AssessmentMethodology: Identification and Recommendation ofComputer Codes," NUREG/CR-5453, SAND89-2505,Vol. 4, Sandia National Laboratories, Albuquerque, NM,1989b.

Kozak, M. W., M. S. Y. Chu, P. A. Mattingly, J. D.Johnson, and J. T. McCord, "Background Informationfor the Development of a Low-Level Waste PerformanceAssessment Methodology: Implementation andAssessment of Computer Codes," NUREG/CR-5453,SAND89-2505, Vol. 5, Sandia National Laboratories,Albuquerque, NM, 1990a.

Kozak, M. W., M. S. Y. Chu, and P. A. Mattingly, "APerformance Assessment Methodology for Low-LevelWaste Facilities," NUREG/CR-5532, SAND90-0375,Sandia National Laboratories, Albuquerque, NM, 1990b.

Kozak, M. W. "A Comparison Between Simple andDetailed Methods for Analyzing Releases of Radionu-clides to Ground Water from Low-Level Waste Facili-ties," in Waste Management '91, held in Tucson, AZ,February 24-28, 1991.

Kozak, M. W., N. E. Olague, D. P. Gallegos, and R.R. Rao, "Treatment of Uncertainty in Low-Level WastePerformance Assessment," in Proc. 13th Annual DOELLW Conference, held in Atlanta, GA, November 19-21,1991.

Lappala, E. G., R. W. Healy, and E. P. Weeks, "Docu-mentation of Computer Program VS2D to Solve theEquations of Fluid Flow in Variably Saturated PorousMedia," Water Resources Investigation Report 83-4099,U.S. Geological Survey, 1987.

Leigh, C. D., B. M. Thompson, J. E. Campbell, D. E.Longsine, R. A. Kennedy, and B. A. Napier, "User'sGuide for GENII-S: A Code for Statistical and Deter-ministic Simulations of Radiation Doses to Humans fromRadionuclides in the Environment," SAND91-0561,Sandia National Laboratories, Albuquerque, NM, 1992.

Luckey, T. D., "Hormesis for Health Physicists," TheHealth Physics Society's Newsletter, Vol. 27(9), 1, 1989.

NUREG/CR-5927 68

References

MacDonald, M. G. and A. W. Harbaugh, "A ModularThree-Dimensional Finite-Difference Ground-Water FlowModel," U.S. Geological Survey Techniques of WaterResources Investigations, Book 6, Chapter Al, 1988.

MacKenzie, D. R., B. Siskind, B. S. Bowerman, and P.L. Piciulo, "Preliminary Assessment of the Performanceof Concrete as a Structural Material for Alternative Low-Level Radioactive Waste Disposal Technologies,"NUREG/CR-4714, BNL-NUREG-52016, BrookhavenNational Laboratory, Upton, NY, 1986.

Maheras, S. J. and M. R. Kotecki, "Guidelines forSensitivity and Uncertainty Analyses of Low-LevelRadioactive Waste Performance Assessment ComputerCodes," DOW/LLW-100, U.S. Department of Energy,1990.

Matuszek, J. M., "Safer Than Sleeping with YourSpouse - The West Valley Experience," Low LevelRadioactive Waste Regulation: Science, Politics, andFear, M. E. Burns, ed., 261, Lewis Publishers, Chelsea,MI, 1988.

McAneny, C. C., "Alternative Methods for Disposal ofLow-Level Radioactive Waste. Task 2d: TechnicalRequirements for Mined Cavity Disposal of Low-LevelWaste," NUREG/CR-3774, Vol. 6, U.S. Army EngineerWaterways Experiment Station, Vicksburg, MS, 1986.

Miller, W. 0. and R. D. Bennett, "Alternative Methodsfor Disposal of Low-Level Radioactive Wastes: Task 2cTechnical Requirements for Earth Mounded ConcreteBunker Disposal of Low-Level Radioactive Waste,"NUREG/CR-3774, Vol. 4, U.S. Nuclear RegulatoryCommission, 1986.

Naff, R.L., T-C. J. Yeh, and M. W. Kemblowski, "ANote of the Recent Natural Gradient Tracer Test at theBorden Site," Water Resources Research, Vol. 24, 2099-2104, 1988.

Napier, B. A., R. A. Peloquin, D. L. Strenge, and J. V.Ramsdell, "Hanford Environmental Dosimetry UpgradeProject. GENII - The Hanford Environmental RadiationDosimetry Software System," PNL-6584, Vols. 1, 2, and3, Pacific Northwest Laboratory, Richland, WA, 1988.

Nazaroff, W. W., "Radon Transport from Soil to Air,"Reviews of Geophysics, Vol. 30, 137-160, 1992.

NCRP, National Council on Radiation Protection andMeasurements, "Radiological Assessment: Predicting theTransport, Bioaccumulation, and Uptake by Man ofRadionuclides Released to the Environment," NCRPReport No. 76, NCRP Publications, Bethesda, MD,1984.

NCRP, "Exposure of the Population in the United Statesand Canada from Natural Background Radiation," NCRPReport No. 94, NCRP Publications, Bethesda, MD,1987.

NCRP, "Measurement of Radon and Radon Daughters inAir," NCRP Report No. 97, NCRP Publications,Bethesda, MD, 1988.

NCRP, "Control of Radon in Houses," NCRP ReportNo. 103, NCRP Publications, Bethesda, MD, 1989.

NRC Regulatory Guide 1.109, "Calculation of AnnualDoses to Man from Routine Releases of ReactorEffluents for the Purpose of Evaluating Compliance with10 CFR Part 50, Appendix I," U.S. Nuclear RegulatoryCommission, October 1977a.

NRC Regulatory Guide 1.111, "Methods for EstimatingAtmospheric Transport and Dispersion of GaseousEffluents in Routine Releases From Light-Water-CooledReactors," U.S. Nuclear Regulatory Commission, July1977b.

NRC Regulatory Guide 1.113, "Estimating AquaticDispersion of Effluents from Accidental and RoutineReactor Releases for the Purpose of ImplementingAppendix I," U.S. Nuclear Regulatory Commission,1977c.

NRC, "Final Environmental Impact Statement on 10 CFRPart 61 - Licensing Requirements for Land Disposal ofRadioactive Waste," NUREG-0945, U.S. NuclearRegulatory Commission, 1981.

Olague, N. E., D. E. Longsine, J. E. Campbell, and C.D. Leigh, "User's Manual for the NEFTRAN 2 Comput-er Code," NUREG/CR-5618, SAND90-2089, U.S.Nuclear Regulatory Commission, 1991.

Olague, N. E., M. W. Kozak, R. R. Rao, and J. T.McCord, "Evaluation of a Performance AssessmentMethodology for Low-Level Radioactive Waste DisposalFacilities, Volume 2: Validation Needs," NUREG/CR-5927, Vol. 2, SAND91-2802, 1993.

69 NUREG/CR-5927

References

O'Neal, B. L. and C. E. Lee, "IMPACTS-BRC, Version2.0," NUREG/CR-5517, SAND89-3060, U. S. NuclearRegulatory Commission, 1990.

Oztunali, 0. I. and G. W. Roles, "De Minimis WasteImpacts Analysis Methodology," NUREG/CR-3585, Vol.1, U.S. Nuclear Regulatory Commission, 1984.

Oztunali, 0. I. and G. W. Roles, "Update of Part 61Impacts Analysis Methodology," NUREG/CR-4370, Vol.1, 1986.

Pescatore, C. and T. M. Sullivan, "Review of the EPA'sRadionuclide Release Analyses from LLW DisposalTrenches Used in Support of Proposed Dose Limits in 40CFR 193," WM-4285-1, Brookhaven NationalLaboratory, 1991.

Pescatore, C., "Leach Rate Expressions for PerformanceAssessment of Solidified, Low-Level RadioactiveWaste," Waste Management, Vol. 11, 223-229, 1991.

Pollock, D. W., "Documentation of Computer Programsto Compute and Display Pathlines Using Results from theU.S. Geological Survey Modular Three-DimensionalFinite-Difference Ground-Water Flow Model," U.S.Geological Survey Open File Report 89-381, 1989.

Polmami, D. J., E. G. Vomvaris, D. McLaughlin, E. M.Hammick, and L. W. Gelhar, "Application of StochasticMethods to the Simulation of Large-Scale UnsaturatedFlow and Transport," NUREG/CR-5094, U.S. NuclearRegulatory Commission, 1988.

Pommersheim, J. M. and J. R. Clifton, "Models ofTransport Processes in Concrete," NUREG/CR-4268,NISTIR-4405, U.S. Nuclear Regulatory Commission,1991.

Prudic, D. E., "Documentation of a Computer Programto Simulate Stream-Aquifer Relations Using a Modular,Finite-Difference, Ground-Water Flow Model," U.S.Geological Survey, 1989.

Rao, R. R., M. W. Kozak, J. T. McCord, and N. E.Olague, "Pathway Analysis for Alternative Low-LevelWaste Disposal Methods, Waste Management '92, 1791-1798, 1992a.

Rao, R. R., M. W. Kozak, J. A. Rollstin, C. P. Harlan,S. D. Bensonhaver, and V. Austin, "IMPACTS-BRC,Version 2.1," NUREG/CR-5797, SAND91-2226, SandiaNational Laboratories, Albuquerque, NM, 1992b.

Rogers and Associates Engineering Corporation,"Conceptual Design Report Alternative Concepts forLow-Level Radioactive Waste Disposal," DOE/LLW-60T, June 1987.

Rogers and Associates Engineering Corporation,"Prototype License Application: Safety Analysis ReportBelowground Vault," EGG/LLW-72T, four volumes,October 1988.

Rogers, V. C., "Practical Evaluations of Low-LevelWaste Disposal Facilities," Near-Surface Land Disposal,J. H. Kittel, ed., Harwood Academic Publishers, NewYork, 1989.

Samper, F. J. and S. P. Neuman, "Estimation of SpatialCovariance Structures by Adjoint State MaximumLikelihood Cross Validation," Water Resources Res.,Vol. 25, 351-262, 1989.

Schulz, R. K., R. W. Ridky, and E. O'Donnell,"Control of Water Infiltration Into Near Surface LLWDisposal Units," NUREG/CR-4918, Vol. 4, U.S.Nuclear Regulatory Commission, 1990.

Shipers, L. R., "Background Information for the Devel-opment of a Low-Level Waste Performance AssessmentMethodology: Identification of Potential ExposurePathways," NUREG/CR-5453, SAND89-2509, Vol. 1,Sandia National Laboratories, Albuquerque, NM, 1989.

Shipers, L. R. and C. P. Harlan, "Background Informa-tion for the Development of a Low- Level Waste Per-formance Assessment Methodology: Assessment ofRelative Significance of Migration and Exposure Path-ways," NUREG/CR-5453, SAND89-2509, Vol. 2,Sandia National Laboratories, Albuquerque, NM, 1989.

Shuman, R., V. C. Rogers, N. Chau, G. B. Merrell, andV. Rogers, "Performance Assessment for Low-LevelWaste Disposal Facilities," Electric Power ResearchInstitute, EPRI NP-5745M, 1988.

Shuman, R., N. Chau, and V. C. Rogers, "ImprovedModeling of Engineered Barriers for Low-Level WasteDisposal," Waste Management '91, Vol. 2, 757-762,1991.

NUREG/CR-5927 70

References

Smyth, J. D., E. Bresler, G. W. Gee, and C. T. Kin-caid, "Development of an Infiltration Evaluation Method-ology for Low-Level Waste Shallow Land Burial Sites,"NUREG/CR-5523, PNL-7356, Pacific NorthwestLaboratory, Richland, WA, 1990.

Starmer, R. J., L. G. Deering, and M. F. Weber, "Per-formance Assessment Strategy for Low-Level WasteDisposal Sites," in Proc. 10th Annual DOE Low-LevelWaste Management Conf., held in Denver, CO, August30-September 1, 1988.

Sullivan, T. M., C. R. Kempf, C. J. Suen, and S. M.Mughabghab, "Low-Level Radioactive Waste SourceTerm Model Development and Testing," NUREG/CR-5204, BNL-NUREG-52160, Brookhaven NationalLaboratory, Upton, NY, 1988.

Sullivan, T. M. and C. J. Suen, "Low-Level WasteShallow Land Disposal Source Term Model: Data InputGuides," NUREG/CR-5387, BNL-NUREG-52206,Brookhaven National Laboratory, Upton, NY, 1989.

Sullivan, T. M. and C. J. Suen, "Low-Level WasteSource Term Model Development and Testing,"NUREG/CR-5681, BNL-NUREG-52280, BrookhavenNational Laboratory, Upton, NY, 1991.

Sullivan, T. M. "Selection of Models to Calculate theLLW Source Term," NUREG/CR-5773, BNL-NUREG-53395, Brookhaven National Laboratory, Upton, NY,1991.

Sullivan, T. M., "Development of DUST: A ComputerCode that Calculates Release Rates from a LLW DisposalUnit," Waste Management '92, 1429-1434, 1992.

Trevorrow, L. E. and J. P. Schubert, "Greater Confine-ment Disposal," Near-Surface Land Disposal, J. H.Kittel, ed., Harwood Academic Publishers, New York,169, 1989.

U.S. Congress, Office of Technology Assessment,"Partnerships Under Pressure: Managing Low-LevelRadioactive Waste," OTA-O-426, U.S. GovernmentPrinting Office, 1989.

Updegraff, C. D., C. E. Lee, and D. P. Gallegos, "DC-M3D: A Dual Continuum- Three-Dimensional, Ground-Water Flow Code for Unsaturated, Fractured, PorousMedia," NUREG/CR-5536, SAND90-7015, SandiaNational Laboratories, Albuquerque, NM, 1991.

Vaurio, J. K., "On the Meaning of Probability andFrequency," Reliability Engineering and System Safety,Vol. 28, 121-130, 1990.

Vomvoris, E. G. and L. W. Gelhar, "Stochastic Analysisof the Concentration Variability in Three-DimensionalHeterogeneous Aquifers," Water Resources Research,Vol. 26, 2591-2602, 1990.

Walton, J. C., L. E. Plansky, and R. W. Smith, "Mod-els for Estimation of Service Life of Concrete Barriers inLow-Level Radioactive Waste Disposal," NUREG/CR-5542), EGG-2597, U.S. Nuclear Regulatory Commis-sion, 1990.

Walton, J. C. and R. R. Seitz, "Performance of Intactand Partially Degraded Concrete Barriers in LimitingFluid Flow," NUREG/CR-5614, EGG-2614, IdahoNational Engineering Laboratory, EG&G Idaho, Inc.,Idaho Falls, ID, 1991.

Yeh, G. T. and D. S. Ward, "FEMWATER: A FiniteElement Model of Water Flow Through Saturated-Unsaturated Porous Media," ORNL-5567, Oak RidgeNational Laboratory, Oak Ridge, TN, 1980.

Yeh, G. T. and D. S. Ward, "FEMWASTE: A FiniteElement Model of Waste Transport Through Saturated-Unsaturated Porous Media," ORNL-5601, Oak RidgeNational Laboratory, Oak Ridge, TN, 1982.

Yeh, T-C., L. W. Gelhar, and A. L. Gutjahr, "Sto-chastic Analysis of Unsaturated Flow in HeterogeneousSoils," Water Resources Research, Vol. 21, 447-47 1,1985.

Zimmerman, D. A., K. K. Wahi, A. J. Gutjahr, andP. A. Davis, "A Review of Techniques for PropagatingData and Parameter Uncertainties in High-Level WasteRepository Performance Assessment Models,"NUREG/CR-5393, SAND89-1432, Sandia NationalLaboratories, Albuquerque, NM, 1990.

71 NUREG/CR-5927

DISTRIBUTION

External Distribution:U. S. Nuclear Regulatory Commission (6)Attn: A. Campbell (1)

R. Hogg (1)P. Lohaus (1)F. Ross (1)R. Shewmaker (1)M. Thaggard (1)

Low-Level Waste Management BranchDivision of Low-Level Waste Management and De-commissioningOffice of Nuclear Material Safety and SafeguardsMail Stop 5E4 WFNWashington, DC 20555

U. S. Nuclear Regulatory Commission (26)Attn: R. Cady (1)

T. J. McCartin (20)T. Nicholson (1)E. O'Donnell (1)J. Randall (1)M. Silberberg (1)P. Reed (1)

Office of Nuclear Regulatory ResearchWaste Management BranchMail Stop NLS260Washington, DC 20555

U. S. Nuclear Regulatory CommissionAttn: L. DeeringAdvisory Committee on Nuclear WasteWashington, DC 20555

T. M. SullivanBrookhaven National LaboratoryNuclear Waste and Materials Technology DivisionDepartment of Nuclear EnergyUpton, NY 11973

GRAM, Inc. (3)Attn: C. D. Updegraff

E. KalininaT. A. Feeney

1709 Moon N.E.Albuquerque, NM 87112

Pacific Northwest Laboratories (3)Attn: M. P. Bergeron

G. W. GeeC. Kincaid

Ground Water GroupGeosciences DepartmentP. 0. Box 999Richland, WA 99352

R. C. ArnoldApplied Power Associates10820 Farnam Drive, Suite 300Omaha, NE 68154-3260

D. GilbertHDR Engineering, Inc.8404 Indian Hills Dr.Omaha, NE 68114-4049

D. JacobsWeston704 S. Illinois Ave., Suite C120Oak Ridge, TN 37830

E. L. WilhiteWestinghouse Savannah River CompanyP.O. Box 616Aiken, SC 29802

B. A. NapierEnvironmental Health Physics GroupHealth Physics DepartmentPacific Northwest LaboratoriesP. 0. Box 999Richland, WA 99352

D. BullenDept. of Mechanical Engineering103 Nuclear Engineering LabAmes, Iowa 50011-2240

D. ThorneOak Ridge National Laboratory2597 B-3/4 RoadGrand Junction, CO 81503

73 NUREG/CR-5927

Distribution

R. Murray, ChairmanNorth Carolina Low-Level Radioactive

Waste Management Authority116 West Jones St.Raleigh, NC 27603-8003

J. HoffeltTennessee Division of Radiological HealthTERRA Building150 Ninth Ave. NorthNashville, TN 37247-3201

J. CliftonNational Institute of Standards and TechnologyU.S. Dept. of CommerceGaithersburg, ND 20899

R. SchulzDept. of Soil ScienceUniversity of CaliforniaBerkeley, CA 94720

D. W. MoellerOffice of Continuing EducationHarvard School of Public Health677 Huntington AvenueBoston, MASS 02115

R. HysongPennsylvania Bureau of Radiation ProtectionDivision of Nuclear SafetyP.O. Box 2063Harrisburg, PA 17105

W. JeterNorth Carolina Division of

Environmental MonitoringGroundwater SectionP.O. Box 29535Raleigh, NC 27626-0535

J. KadlecekNew York Department of

Environmental ConservationBureau of Radiation, Room 51050 Wolf RoadAlbany, NY 12233-7255

J. McConnellWaste Management DepartmentEG&G Idaho, Inc.P.O. Box 1625Idaho Falls, ID 83415

H. R. MeyerKeystone Scientific, Inc.5009 Alder CourtFort Collins, CO 80525

N. PrattArizona Radiation Regulatory Agency4814 South 40th StreetPhoenix, AZ 85040

A. RaineyWashington Department of HealthDivision of Radiation ProtectionAirdustrial Park, Bldg. 5, LE-13Olympia, WA 98504

R. RottaMichigan Department of Public HealthDivision of Radiological Health3423 North LoganP.O. Box 30195Lansing, MI 48909

J. RingenbergNebraska Department of

Environmental ControlLow-Level Radioactive Waste ProgramP.O. Box 98922Lincoln, NE 68509-8922

D. SchererIllinois Department of Nuclear Safety1035 Outer Park Dr.Springfield, IL 62704

Gary SmithTexas Department of Health1100 West 49th StreetAustin, TX 78763

B. SolomonUtah Geological and Mineral Survey606 Black Hawk WaySalt Lake City, UT 84108

NUREG/CR-5927 74

Distribution

M. WalleIllinois Department of Nuclear Safety1035 Outer Park Dr.Springfield, IL 62704

W. WatsonCalifornia Department of Health ServicesEnvironmental Management Branch1449 West Temple Street, Room 222Los Angeles, CA 90026-5698

K. WeaverColorado Department of HealthRadiation Control Division4210 East l1th AvenueDenver, CO 80220

D. WoodWestinghouse Hanford Co.B2-19, P.O. Box 1970Richland, WA 99352

C. HungCriteria and Standards Div.Office of Radiation ProgramsU.S. Environmental Protection AgencyANR-640, 401 M. Street S.W.Washington, DC 20460

R. J. StarmerERM Corporation7926 Jones Branch Dr.Suite 210McClean, VA 22102

J. R. PreisigNew Jersey Department

of Environmental ProtectionBureau of Environmental RadiationCN 415Trenton, NJ 08625

Foreign Addresses

J. AlonsoENRESAEmilio Vargas, 728043 MadridSPAIN

K. BerciEroterv - Power Station and

Network Engineering Co.Szechenyi Rkp. 3H-1054 BudapestHUNGARY

Mr. V. HristovDeputy ChairmanCommittee on the Use of Atomic Energy

for Peaceful Purposes55A Chapaev Str.1574 SofiaBULGARIA

R. CleggBritish Nuclear Fuels plcSellafield SeascaleCumbria CA20 1PGENGLAND

M. S. HossainWaste Management SectionDivision of Nuclear Fuel Cycle and Waste ManagementInternational Atomic Energy AgencyWagramerstrasse 5, P.O. Box 100A-1400 ViennaAUSTRIA

D. GibsonAustralian Nuclear Science and Technology OrganisationNew Illawarra RoadLucas Heights, NSWPrivate Mail Bag 1 Menai NSW 2234AUSTRALIA

P. LietavaNuclear Research InstituteCS - 250 68 RezCzechoslovakia

H. MatsuzuruJapan Atomic Energy Research InstituteTokai-mura, Naka-gunIbaraki-ken Pref. 319-11JAPAN

J. MitregaState Geological Institute4 Rakowiecka00-975 WarsawPOLAND

75 NUREG/CR-5927

Distribution

P. SantucciCommissariat a l'Energie AtomiqueInstitut de Protection et de Surete NucleaireF-92265 Fontenay-Aux-Roses CedexFRANCE

P. SasidharScientific OfficerCentralised Waste Management FacilityFuel Reprocessing & Nuclear Waste Management GroupBARC, Kalpakkam - 603 102INDIA

B. N. SelanderWaste Management SystemsAtomic Energy of Canada, Ltd.Chalk River, Ontario KOJ 10CANADA

B. SerebryakovInst. BiophysicsMinistry of HealthZhivopisnaya 46, 123182RUSSIA

A. SuarezInst. de Pesquisas Energeticas

E NuclearesTrav. R. Nr. 400-CID. Univers.05508 Sao Paolo SPBRAZIL

I. UsluTurkish Atomic Energy AuthorityAlacam Sokak No. 9Cankaya, AnkaraTURKEY

JI J. van BlerkAECSA, Ltd.Earth and Environmental Technology Dept.P.O. Box 582Pretoria, 0001REPUBLIC OF SOUTH AFRICA

V. VanclVancl RP ConsultingPohorelec 25118 00 Praha 1 - HradcanyCzechoslovakia

G. VolckaertSCK/CENBoeretang 200B-2400 MolBELGIUM

G. ZedeChina Institute for Radiation ProtectionP.O. Box 120Taiyuan, Shanxi 030006PEOPLE'S REPUBLIC OF CHINA

Sandia National Laboratories:6300 D. E. Ellis6900 T. 0. Hunter6312 F. W. Bingham6312 P. C. Kaplan6331 S. H. Conrad6331 P. A. Davis6331 D. P. Gallegos6331 M. W. Kozak (25)6331 N. E. Olague6331 D. Smith6331 E. K. Webb6340 W. D. Weart6342 D. R. Anderson6119 E. D. Gorham6400 N. R. Ortiz6403 W. A. von Riesemann6622 M. S. Y. Chu6622 J. T. McCord7141 Technical Library (5)7151 Technical Publications7223 R. Knowlton8524 Central Technical Files

NUREG/CR-5927 76

NRC FORM 335 U.S. NUCLEAR REGULATORY COMMISSION 1. REPORT NUMBER2-89) (Assigned by NRC. Add Vol., Supp., Rev.,

NRCM 1102, and Addendum Numbers, if any.)2201,.3202 BIBLIOGRAPHIC DATA SHEET

(See instructions on the reverse) NUREG/CR-59272. TITLE AND SUBTITLE SAND91-2802

Vol. IEvaluation of a Performance Assessment Methodology for 3. DATE REPORT PUBLISHED

Low-Level Radioactive Waste Disposal Facilities: MONTH I YEAR

Evaluation of Modeling Approaches August 19934. FIN OR GRANT NUMBER

L11535. AUTHOR(S) 6. TYPE OF REPORT

M. W. Kozak, N. E. Olague, R. R. Rao, and J. T. McCord7. PERIOD COVERED (Inclusive DOtri

8. PERFORMING ORGANIZATION - NAME AND ADDRESS (It NRC, provide Division, Office or Region, U.S. Nuclear Regulatory Commission. and mailing address, if contractor, provid,-name and mailing address. I

Sandia National LaboratoriesAlbuquerque, NM 87185-5800

9. SPONSORING ORGANIZATION - NAME AND ADDRESS (If NRC, type "'Same as above", if contractor, provide NRC Oivision, Office or Region. U.S. Nuclear Regulatory Commission,.and mailing address.)

Division of Regulatory ApplicationsOffice of Nuclear Regulatory ResearchU.S. Nuclear Regulatory CommissionWashington, DC 20555

10. SUPPLEMENTARY NOTES

11. ABSTRACT (200 words or less)

This report represents an update to our earlier reports on low-level wasteperformance assessment. This update addresses needed improvements andrecommended approaches to the existing state of the art in modeling, treatmentof uncertainty, and use of data. Greater attention is paid to developing anintegrated approach to performance assessment than was done in earlier develop-ments of the methodology. Furthermore, insights are being developed by partici-pating in validation exercises, and by evaluating which validation data areneeded to improve confidence in the methodology. It is emphasized that theperformance assessment methodology update is a work in progress; the recommendationsgiven here will form the general directions toward which the methodology isheading, but some of the specific approaches may continue to evolve as theresearch progresses.

12. KEY WORDS/DESCR!PTORS (List words or phrases that will assist researchers in locating the report.) 13. AVAILABILITY STATEMENT

radioactive waste Unlimited

low-level waste 14. SECURITY CLASSIFICATION

performance assessment (This Page)

validation Unclassified(This Report)

Unclassified15. NUMBER OF PAGES

16. PRICE

NRC FORM 335 (2-891

Federal Recycling Program

LOW-LEVEL RADIOACTIVE WASTE DISPOSAL FACILITIES

UNITED STATESNUCLEAR REGULATORY COMMISSION

WASHINGTON, D.C. 20555-0001

SPECIAL FOURTH-CLASS RATEPOSTAGE AND FEES PAID

USNRCPERMIT NO. G-67

OFFICIAL BUSINESS

PENALTY FOR PRIVATE USE, $300