Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
s Review of Methodology for Dose Analysis In the United States and internationally, there have been several attempts to provide technical guidance concerning the doses that might be associated with various clearance policies for slightly radioactive solid material (SRSM). As part of its charge, the study committee has reviewed the relevant public reports, as well as various commentaries and critiques of those reports. In addition, the committee met with knowledgeable experts involved in preparing the reports, to clarify specific issues, particularly the reasons why dose factors differ between reports. Because one of the reports, the draft report NUREG-1640, Radiological Assessments for Clearance of Equipment and Materials from Nuclear Facilities (USNRC, 1998b), was prepared for the U.S. Nuclear Regulatory Commission, the committee gave it particular attention. The committee has been able to delve sufficiently deeply into the report to form an overall judgment about its useful- ness and to make recommendations for next steps. Based on its review of technical documents from around the world, the committee has drawn a number of conclusions on technical issues. These find- ings are collected at the end of this chapter. The body of the chapter supports the findings. Most of the technical material in this field falls under the rubric of risk assessment, which means it inherits both the strengths and the limitations of this discipline. In particular, although risk estimates can provide useful guidance, they do not substitute for policy decisions on what risks are acceptable. Further- more, "although the conduct of a risk assessment involves research of a kind, it is 80
REVIEW OF METHODOLOGY FOR DOSE ANALYSIS 81 primarily a process of gathering and evaluating extant data and imposing science- policy choices" (NRC, 1994~. One of the science policy choices to be imposed involves setting boundaries on the scope of the analysis. In this case, the boundaries involve using radiation dose as a surrogate for health impacts and ignoring other consequences consid- ered to be of lesser significance, such as psychological impacts. When it comes to assigning risk to dose, analysts generally accept standard estimates of dose-risk coefficients established by scientific bodies such as the Committee on Biological Effects of Ionizing Radiation (BEIR) of the National Research Council and the United Nations Scientific Committee on the Effects of Atomic Radiation (NRC, 1990; UNSCEAR, 1988~. Nevertheless, the technical reports in this field can assist the USNRC and interested parties in making policy judgments about the clearance of SRSM, as long as the following three conditions are met: (1) the boundaries of the relevant risk assessments must be kept in mind; (2) policy decisions about acceptable risk must be separated from technical issues; and (3) the major limitations of the technical reports, as identified in this chapter, must be addressed. The flow chart in Figure 5-1 shows points at which technical informa- tion can inform decision makers about clearance of SRSM, if a rulemaking pro- cess one day advances to a decision point about clearance. KEY TECHNICAL ASSESSMENTS OF ANNUAL DOSES ASSOCIATED WITH CLEARANCE OF SOLID MATERIALS A great deal of effort in a number of countries over the last 20 years has gone into developing the numerical coefficients, also called dose factors, needed by policy makers to (1) understand the dose commitment implied by various clear- ance concentrations and (2) convert a primary dose standard into secondary activ- ity standards that can be used by licensees to ensure compliance with the primary standard (see Box 5-1~. The major compilations of these dose factors are listed in Table 5-1, along with the scientific bodies that have reviewed the underlying technical analyses. (See Appendix D for a summary of efforts here and abroad on SRSM clearance standards.) All of the reports in Table 5-1 estimate doses to classes of persons, such as SRSM transport workers, or consumers, and focus on the group that is estimated to have the highest dose under all the scenarios considered (the critical group); see Figure 5-2. The principle is that if the most exposed group of individuals is identified correctly and the dose to that group is shown to fall below the primary standard, then the dose to any other member of the general public will fall below the standard as well. Thus, the dose to the critical group (for a unit release) determines the dose factor. Note that the critical group can differ for different radionuclides, which complicates implementation of any clearance standard that relies on dose factors. Also, the critical group, and thus the secondary standard, may change when the allowed clearance categories are restricted, as in condi-
on on ~ F _ .____ FICURE 5-1 Pouts ~ ~bicb ~cbn~1 in~rm~on Ed judgment can idiom ~lem~- ing decisions related to cle~=ce of sligbOy radioactive solid mogul NO~: Circles indict policy decisions. Rectangles indict ~cbuic~1 con~ibuUons. donut clc~=cc. To dam, most of the abandon to dosc Actors bas assumed tab they would be used in scHing ~=d~ds ~r (uncon~hon~) clc~=cc. Bcc~sc pdm~ dosc studs ~r clc~=cc or ~spo~hon of soNd m~cd- ~s ~ usually divan in dosc par yam, the dosc Actors arc pcncr~ly c~prcssod in
REVIEW OF METHODOLOGY FOR DOSE ANALYSIS 83 units of dose per year per unit of activity released.] Cumulative total doses can be obtained by multiplying the estimated dose by an assumed duration of expo- sure for instance, a person's remaining years of life. Presumably, the total dose per year (i.e., the committed dose per year). In some cases, effective dose equivalent is used, which accounts for both the relative biological effectiveness of different types of radioactivity and the differing sensitivity of organs to cancer mortality. In some studies, only effective doses are used, without the weighting by cancer mortality that produces dose equiva- lents.
84 THE DISPOSITION DILEMMA TABLE 5-1 Technical Analyses Supporting Numerical Coefficients for Deriving Secondary Activity Standards from Primary Dose Standards Study Status Reviewer Reference USNRC NUREG- 1640a NUREG-0518 EPA TSD 97a Draft CNWRA USNRC, 1998b Draft USNRC, 1980 Draft NCRP EPA, 1997a TSD gob In progress ANSI/HPS N13.12-1999 Final ANSI/HPS, 1999 IAEA Safety Practice No. 111-P-l.1 Final IAEA, 1992 Technical Document 855 Interim IAEA, 1996 European Commission Radiation Protection-89 Final EC, 1998b Radiation Protection-114 Final EC, 2000 NOTE: ANSI/HPS = American National Standards Institute/Health Physics Society; CNWRA = Center for Nuclear Waste Regulatory Analyses; EC = European Commission; EPA = Environmental Protection Agency; IAEA = International Atomic Energy Agency; ICRP = International Commission on Radiological Protection; NCRP = National Council on Radiological Protection and Measure- ments; TSD = Technical Support Document. aThe coefficients given in the USNRC and EPA source documents have built into them, or the opportunity to use, an explicit margin to account for uncertainty. In EPA TSD 97 a margin was built into the dose coefficients. Specifically, the semiquantitative uncertainty analysis described in Chap- ter 10 showed that, depending on choice of input parameters, normalized doses could be higher by a factor of 5-50 or lower by a factor of 100-500, i.e., they favored more protective levels. The NUREG- 1640 draft shows a distribution of dose factors based on Monte Carlo simulations of the aggregate uncertainty resulting from uncertainties in the component estimates. Both the mean values and the 95th percentile given in NUREG-1640 for the dose coefficients lie above the median, 50th percentile value. If either of these properties of the distribution were chosen to define the regulatory dose coefficients, a margin above the best estimate (median) would automatically be included. bThe study committee has seen the first EPA report, TSD 97. A new report, TSD 99, was prepared and given very limited distribution, presumably in 1999. Proposals were due to the EPA on April 13, 2001, for final revision of TSD 99, with submittal of the draft to the EPA by May 31, 2001. Both TSD 99 and this new revision will supplant Chapters 1-7 of TSD 97, using ICRP-68 guidance. By this revision the EPA will be reacting to comments from the NCRP review and others. The remaining chapters of TSD 97 after Chapter 7 will apparently stand without revision. (Information on status of TSD 99 and revision efforts was received in a personal communication from Debbie Kopsick, EPA, to Robert Bernero, Board on Radioactive Waste Management, National Research Council, April 11, 2001.)
REVIEW OF METHODOLOGY FOR DOSE ANALYSIS Population Group I and / Critical Group (drivers) Population groups - 85 \ ~opulatiop/G roup I I $\~:X / (workers) \ \ /opulati r group (golfers \ / playing on't~op of postclosure jpulation grouggenera;\ / public) / 1 ~ \ landfill) \` by / Populatior drinkers) ~ group (we: \/ r ~ ! FIGURE 5-2 Illustration of scenario pathways following SRSM clearance and hypothet- ical affected critical groups.
86 THE DISPOSITION DILEMMA The committee's assessments of the individual technical sources listed in Table 5-1 are presented in the next five sections. Then the committee compares the methodologies used across these studies, including comments on the useful- ness and quality of the dose factors they contain, general limitations that should be corrected, and potential inconsistencies in the dose factors used by different countries. Before concluding with the summary statement of the findings for the chapter, the committee explores in further detail specific issues that should be addressed in subsequent work on the draft NUREG-1640. USNRC STUDIES The committee reviewed two technical documents on clearance standards developed for the USNRC. Draft NUREG-1640, which has been mentioned in earlier chapters of this report, is particularly relevant to the new rulemaking on clearance standards for SRSM, which the Commission is contemplating. The second document, NUREG-0518, represents an earlier effort at analysis to sup- port clearance standards for SRSM. Draft NUREG-1640 Draft NUREG-1640 (USNRC, 1998b) contains estimates of the total effec- tive dose equivalent to an average individual in a critical group from direct reuse of equipment, recycling, or disposal of materials, for a wide range of radionu- clides that may be present in solid materials from decommissioning of nuclear facilities. The risk assessment methodology is largely state of the art. Critical groups are chosen by assuming a policy of clearance, although information in the appendixes may be sufficient to allow choices of other critical groups to support derivation of dose factors for possible conditional clearance policies. The draft does not discuss implementation issues. Although NUREG-1640 is a draft for review and comment, it is a sophisti- cated product and does many things well. The various scenarios considered for clearance of materials with surface or volume contamination are well docu- mented and easy to understand. The major analytical effort is for recycling steel (31 scenarios), with less analysis for recycling copper (23 scenarios), aluminum (17 scenarios), and concrete (7 scenarios). There is an in-depth analysis of current recycling practices and how the inclusion of SRSM would show up as exposure to humans. In addition, the study does a good job of documenting the impact of equipment reuse. The chemistry, metallurgy, geology, and physics used in the report seem reasonably sound. Considerable information is provided on the dose factors re- sulting from external exposure, inhalation, and ingestion of radioisotopes from recycled material, waste, and release of effluents to air or water. Most of the
REVIEW OF METHODOLOGY FOR DOSE ANALYSIS 87 critical groups turn out to be workers, not the public at large. The report does not discuss whether this pattern would change for conditional clearance. The committee found the overall conceptual plan of draft NUREG-1640 to be the best of all of the studies that it reviewed. It is closest in spirit to recommen- dations on risk assessment that have been made by expert bodies, including committees of the National Research Council (NRC, 1994~. For instance, the estimates in draft NUREG-1640 are traceable, and a formal uncertainty analysis has been performed for each dose factor.2 The study presents the mean and the 5th, median, and 95th percentile values for each dose factor, derived from Monte Carlo uncertainty analyses. The authors of draft NUREG-1640 use the range from the 5th to the 95th percentile to define a "90 percent confidence interval" (about the median) (USNRC, 1998b, Tables 4.10 and 4.11~. The result of a Monte Carlo calculation, such as carried out by NUREG- 1640, is a distribution of doses for each scenario delivered to the representative member of a critical group for a particular radionuclide. There is no single dose estimate to a critical group, and hence no single dose factor for that critical group. Nevertheless, a decision must be made, if NUREG-1640 is to be used to support clearance or conditional clearance, about which dose factor should be used to assign secondary activity standards. If one takes the median of the distribution, then 50 percent of the dose factors are below and 50 percent above. Choosing the median as the de minims value for use in clearance or conditional clearance standards, however, would leave the decision maker without the higher degree of assurance that the dose to the critical group is below the ordinary dose standard as when higher percentile-valued dose factors are chosen (e.g., 90th percentile). This additional assurance is above and beyond the conservatism that applies to individuals within population groups that receive less exposure than the critical group. Table 5-2 averages the uncertainty factors computed for NUREG-1640 across radionuclides. In this table, uncertainty is represented by the geometric standard deviation (GSD), which is appropriate for quantifying the spread in variables with large variations. Except for concrete recycle, the GSDs are small.3 2For the uncertainty analysis, NUREG-1640 works with the individual steps involved in making a dose estimate. The analyst gathers data from the literature on the ranges that individual parameters required for the estimate might take and then propagates the individual uncertainties to the final coefficient using techniques such as Monte Carlo simulation (EPA, 1996). There is a subjective element in choosing the parameter distributions used to fit the literature data, but these choices are one or more steps removed from the final uncertainty estimate for the dose factor. Also, guidelines exist for selecting the functional form for a parameter distribution (Seller and Alvarez, 1996). 3A sampling of papers published in Health Physics showed GSDs ranging from 1.7 to 20, with most in the range 2-4. Thus, values of GSD below 2 can be considered small, and values above 4 considered high (sreshears, 1989; Johnston, 1991; Till, 1995; Sheppard, 1997; Belch, 2001).
88 TABLE 5-2 NUREG-1640 Uncertainty Factors Averaged Across Radionuclides THE DISPOSITION DILEMMA Average Geometric Standard Deviation (GSD)a Volume Contamination Surface Contamination Steel recycled 1.8 2.0 Concrete recycles 3.0d 3.4 Copper recyclee 1.5 1.6 Aluminum recyclef 1.4 1.7 Reuse of large piece of equipment" NAh 1.9 aOne standard deviation is equal to the product of the median times the GSD. Two standard deviations (~95th percentile limits) equal the square of the GSD. For the table, GSDs were approxi- mated by computing the square root of the ratio of the 95th percentile dose factors to the 50th percentile results, as presented in tables in NUREG-1640 (USNRC, 1998b). bFrom Tables 4.1, 4.2 (USNRC, 1998b). CFrom Tables 7.2, 7.3 (USNRC 1998b). The distribution is bimodal, with one group of radionuclides having a GSD around 1 and another group having a GSD around 6. eTables 5.5, 5.6 (USNRC, 1998b). Gables 6.4, 6.5 (USNRC, 1998b). "Tables 3.2, 3.3 (USNRC, 1998b). hNA = not applicable. Formal uncertainty analysis can be an important tool for building confidence in the use of dose estimates for policy decisions. It addresses the reported ten- dency of even experts in a field to underestimate uncertainty bands when profes- sional judgment alone is used (Cooke, 1991~. This tendency exists even in the physical sciences (Shlyakhter and Valverde, 1995~. It is therefore not wise to rely on professional judgments of estimates of overall uncertainty because of the subjective bias found in such estimates. Of the studies listed in Table 5-1, only draft NUREG-1640 includes a formal uncertainty analysis that reduces the amount of professional judgment required in assigning uncertainty bands to dose factors. Excellent discussions of formal uncertainty analysis can be found in other USNRC documents (e.g., USNRC, 1995) and in Morgan and Henrion (1990~. The authors and planners of draft NUREG-1640 are to be commended for developing an excellent approach. The execution of draft NUREG-1640's con- ceptual plan, however, has been clouded by questions of contractor conflict of interest concerning the recycle option (see discussion in Chapter 2~. One question is how the USNRC could have failed to identify the conflict of interest. These questions highlight the need to include the possibility of organizational failure when assessing overall system uncertainty. After the conflict of interest was
REVIEW OF METHODOLOGY FOR DOSE ANALYSIS 89 identified, the matter was investigated by USNRC counsel and the contract in question was terminated. The USNRC engaged another contractor to complete the work on draft NUREG-1640. Meanwhile, the USNRC asked the Center for Nuclear Waste Regulatory Analyses (CNWRA) to perform an independent technical review of the draft NUREG-1640. The CNWRA, located at the Southwest Research Institute in San Antonio, Texas, is a dedicated contractor providing technical support to the USNRC on waste management matters. The committee has studied the CNWRA review (CNWRA, 2001), which is actually an audit of the mathematics and completeness of scenarios considered in draft NUREG-1640. CNWRA recom- mended that some additional scenarios be added to the mix considered in draft NUREG-1640 but otherwise found the mathematics to be correct. Although this CNWRA review is comforting and is confirmed by the committee' s spot check of some of the scenarios, there has not yet been a similarly thorough review of the choice of parameters and parameter ranges, term by term, for the component estimates in deriving the dose factors. (The choice of parameters and parameter ranges are listed in Appendix B of draft NUREG-1640, Tables B.1, B.2, etc. Although the committee generally confirmed the reasonableness of many of these choices, it was able to review only a sample of the dose factors, given its full set of tasks.) In addition to any lingering questions about the choice of parameters, whether due to a potential bias or for other reasons, there are a number of other limitations in draft NUREG-1640. These limitations have to be addressed before the docu- ment will be fully usable by the USNRC and interested parties in reaching valid conclusions about related SRSM policy issues. The limitations are discussed in the penultimate section of this chapter. One option for the USNRC, faced with any lingering concerns over charges of conflict of interest, is to start all over again. However, it is likely that any new contractor would simply repeat the work in NUREG-1640 as far as it goes and build upon it in the way the committee recommends. Therefore, from a scientific perspective, the committee does not believe it is cost-effective to repeat the work done in draft NUREG-1640. The committee believes that once the remaining questions about and limi- tations in draft NUREG-1640 are addressed, either in the final version of the report or in follow-up reports, the USNRC and interested parties will have a sound technical basis for evaluating the health impacts, measurement issues, and implementability of various primary dose standards and the unavoidable uncertainties involved in risk estimates. However, the committee notes that the dose factors developed through the NUREG-1640 process cannot be adopted for use with Department of Energy (DOE) or other SRSM without further analysis. Changes are likely to be needed to some of the dose factors and/or their uncertainties because the quantity and types of DOE SRSM, as well as
9o THE DISPOSITION DILEMMA some potential release scenarios, differ from wastes generated by USNRC-li- censed facilities.4 NUREG-0518 Prior to NUREG-1640, the USNRC published a risk assessment in 1980 for the release of SRSM. In response to a 1974 amendment to the Atomic Energy Act (AEA), which authorized release of de minimis quantities of special nuclear material if justified, the Atomic Energy Commission (AEC) began developing a de minimis standard for enriched uranium and the attendant fission product tech- netium. The development side of the AEC (later the Energy Research and Devel- opment Administration, ERDA) requested guidance from the regulatory side of the AEC (later the USNRC). The ERDA developed data for the quantities of scrap steel, copper, and nickel that would become available from the 1976-1982 cascade improvements at the gaseous diffusion plants used for enriching ura- nium. (See also "DOE Facilities" in Chapter 3.) Data were provided on the extent of decontamination that could be achieved by smelting. These data showed that smelting could not be relied upon to reduce the contaminant content to less than 17.5 parts per million (ppm) uranium and 5 ppm technetium. The USNRC staff prepared and issued NUREG-0518, Draft Environmental Statement Concerning Proposed Exemption from Licensing Requirements for Smelted Alloys Containing Residual Technetium-99 and Low-Enriched Uranium (USNRC, 1980~. NUREG-0518 contained analyses of the expected scrap metal inventories from the gaseous diffusion plants and of scrap metal from other sources. Using several important assumptions, the study estimated both doses to an individual member and collective doses to the entire group for several critical groups. The most important assumption was that the proposed exemption from regulatory control would apply only to scrap metal ingots coming out of a li- censed smelter, thus ensuring a radionuclide content in the scrap of no more than 17.5 ppm uranium and 5 ppm technetium. Conservative assumptions were ap- 4DoE may have disposition opportunities, and therefore clearance scenarios, that are not available to USNRC licensees. As for dose calculations, the committee notes that uncertainties about migra- tion of transuranics have become important for DOE SRSM, whereas they are far less important for SRSM from USNRC licensees. For example, the transuranic radionuclides in the SRSM stream from USNRC-licensed facilities constitute a relatively minor component of the radioactive contaminants. USNRC analysts therefore do not need to delve too deeply into chemical and biological processes in landfills that might speed up migration of transuranic radionuclides, which are thought to migrate at a slow rate under usual subsurface conditions. By contrast, DOE has a great deal of material poten- tially contaminated with transuranics at substantially higher concentrations than occur in nuclear power plant wastes. An analysis by DOE to support conditional clearance standards for DOE SRSM may have to consider in some detail the chemical or biochemical processes in Subtitle C or other landfills.
REVIEW OF METHODOLOGY FOR DOSE ANALYSIS 91 plied to scenarios for possible uses of the ingots. For example, all of the steel scrap released was assumed to be used in a continuous 80-day run at the exempt steel plant and made into products of the reference content. Steel plate, iron tonic, and even a production run of 9 million cast iron frying pans, were considered as possible products from the steel. Jewelry, coins, and prostheses were considered as possible products from the other metals. The estimated doses listed in NUREG-0518 include a 10 mrem/yr whole- body external dose for one exposed group (workers spending 1,000 hours per year in a steel vault), a 2 mrem total-body dose commitment for another group (1 year of iron tonic ingestion), and a 20 rem contact bone dose for a third group (prosthesis pins implanted for 50 years). The collective dose for the worst-case scenario was estimated to be 80 person-rem. NUREG-0518 does not contain any uncertainty analysis as such. Instead, it invokes conservative bounding conditions to make the point estimates of dose usable for regulatory purposes. In NUREG-0518 the USNRC staff concluded that the proposed exemption, as qualified, was acceptable for consideration by the Commission for amendment of its regulations. There was substantial negative public reaction to NUREG- 0518, and the proposed exemption process was suspended (51 Federal Register 8842; March 14, 1986~. ENVIRONMENTAL PROTECTION AGENCY DOCUMENTS ON DOSE FACTORS The Environmental Protection Agency (EPA) Technical Support Document (TSD) Evaluation of the Potential for Recycling of Scrap Metals from Nuclear Facilities ("TSD 97") contains numerous tables of background information on the sources and inventories of radioactively contaminated metal scrap from vari- ous government and commercial sources (EPA, 1997a). The document develops various normalized individual doses, collective doses, and collective risks, nor- malized to curie-per-gram concentrations in the scrap metal streams. It also con- tains valuable information, compiled in an insightful way, about detection limits for contamination as a function of various parameters and about various scrap metal processes, including how radionuclides partition in these processes. This is all useful information. The methodology employed and the handling of uncer- tainties helped the study committee understand the relevant issues. TSD 97 also contains useful discussions about possible pathways from con- taminated metal (sources) to humans (receptors). These pathways are sorted into a few important pathways and a much larger number that were judged to be less important. The basis for the sorting is explained well. Another useful element is the discussion of an estimated "timetable," cover- ing the next few decades, indicating when the waste streams might become available for potential commercial recycle (or other disposition alternatives).
92 THE DISPOSITION DILEMMA This discussion, although inexact in detail because of some assumptions that cannot be verified, succeeds in putting the issues in context. To assess the uncertainty associated with doses to the critical group, TSD 97 performed a semiquantitative uncertainty analysis that "evaluated the uncertainty/ variability in the dose evaluation results due to uncertainty/variability in the calculational parameters and assumptions."5 Although not a formal uncertainty analysis, the analysts used their inspection of these results and professional judg- ment to conclude that the dose factors they calculate represent a 90th percentile. (That is, in 90 percent of cases, use of the calculated dose factor will result in a dose to a member of the critical group that is at or below the primary dose standard.6) The National Council on Radiation Protection and Measurements (NCRP, 1998) has produced a detailed critique of TSD 97. Among its major findings and recommendations are the following (NCRP, 1998, pp. 9, 11~: 1. The NCRP task group concluded that, "as it now stands, [TSD 971 over- emphasizes the evaluation of a limited number of scenarios with data that are incomplete and/or unsupported." 2. The NCRP task group recommended "the use of a probabilistic risk as- sessment model, such as the Monte Carlo method (as recommended by [the EPA's] established policy relative to the conduct of [probabilistic risk assessments]), for analyzing the potential uncertainties and for iden- tifying areas for improvements in the input data." 3. The NCRP task group recommended that the EPA evaluate the feasibility for implementation, stating, "Standard development cannot be devoid of information regarding implementation." These comments from the NCRP task group are apparently being taken into account by the EPA as it works on a revision of TSD 97 (EPA, in progress). The committee has not seen the revision, which was still in progress when the various technical documents on dose factors were being reviewed for this report. AMERICAN NATIONAL STANDARDS INSTITUTE AND HEALTH PHYSICS SOCIETY STANDARD N13.12-1999 The Health Physics Society (HPS) Standards Working Group developed this standard.7 The document defines primary (dose) and secondary screening (activity level) criteria (ANSI/HPS, 1999~. The primary dose standard is 5TSD 97 (EPA, ~997a, p ES-8, see also, Ch. lo, p. ~2). 6TSD 97 (EPA, ~997a, Ch. 3, p 3). 7The standard was consensus balloted and approved by the ANS~-accredi~ed HPS NO committee on October 19, 1998. It was approved by ANSI, Inc., on August 31, 1999.
REVIEW OF METHODOLOGY FOR DOSE ANALYSIS 93 10 pSv/yr (1 mremlyr), which is consistent with international values. The docu- ment tabulates derived screening levels, above background, for the clearance of SRSM or items containing surface or volume activity concentrations of radioac- tive materials. These screening levels are derived by applying dose factors to the primary dose standard. The ANSI/HPS document contains a great deal of useful information on uncertainties in dose factors. Furthermore, the working group took on the diffi- cult task of developing an implementation protocol, which specifies areas over which measurement averages should be taken. It also groups radionuclides based on similarity of dose factors and assigns group-level screening levels ranging from 0.1 to 100 Bq/cm2 or Bq/g, depending on the group considered. The dose factors chosen are quite similar to the International Atomic Energy Agency (IAEA) values. To derive dose factors, the working group reviewed a range of dose estimates produced by different analysts for different activities, such as landfill disposal and steel recycling. It also used reports that examined exposures for different forms of contamination (either volume or surface contamination). In contrast to other reports the committee has reviewed, the working group did not use the range of dose estimates across categories to define a critical group in a docu- mented manner. As a result, the method for deriving the screening levels is not traceable by independent reviewers.8 Although the ANSI/HPS working group was composed of analysts of great skill and experience, only a traceable approach could be judged and ranked by the committee. INTERNATIONAL ATOMIC ENERGY AGENCY DOCUMENTS The committee reviewed two documents developed by the International Atomic Energy Agency: Safety Practice No. 111-P-l.l, Application of Exemp- tion Principles to the Recycle and Reuse of Materials from Nuclear Facilities (IAEA, 1992), and a more recent interim document, IAEA-TECDOC-855, Clear- 8Based on a discussion with a working group member, it appears that the working group used professional judgment to discount or reduce dose values from scenarios if the group believed the value to be unreasonably conservative. It then picked the highest remaining value to use in setting screening levels (personal communications from William Kennedy, HPS Standards Working Group, to Jan Beyea, committee member, April 20, 2001). Had the working group included a table in the standard with the discounted factors, the methodology would have been traceable. As part of the working group's analysis, it concluded that dose factors appeared to be similar for surface and volume contamination, when units were expressed in becquerel per gram or becquerel per centimeter squared. (IAEA, 1996, came to a similar conclusion.) Consequently, the group chose the same "derived screening levels" to apply to both surface and volume contamination in the imple- mentation protocol. Again, no summary of the values from which the group drew its conclusions was included in the report, making its analysis untraceable. The study committee recognizes that a volun- teer group, such as the HPS Standards Working Group, can include only a limited amount of detail in its reports.
94 THE DISPOSITION DILEMMA ance Levels for Radionuclides in Solid Materials: Applications of Exemption Principles (IAEA, 1996~. Comments on each document are presented below. Safety Practice No. 111-P-1.1 In Safety Series No. 89, Principles for the Exemption of Radiation Sources and Practices from Regulatory Control, the IAEA established the principles that underlie its technical estimates of dose factors (IAEA, 1988~. The USNRC has produced no similar generic document. The IAEA dose factors are contained in Safety Practice No. 111-P-l.l, Application of Exemption Principles to the Re- cycle and Reuse of Materials from Nuclear Facilities (IAEA, 1992~. Two of the IAEA recommendations from these documents may differ from the concept of clearance of SRSM under discussion in the United States: 1. "The dose to the individual of the critical groupers) and the dose to the whole population exposed by the practice should not be significantly affected by other similar (or identical) practices (e.g., several waste dis- posal sites in the same region)" (IAEA, 1988, p. 6~. 2. "The formulation of an exemption should not allow the circumvention of controls that would otherwise be applicable, by such means as deliberate dilution of material or fractionation of the practice" (IAEA, 1992, p. 4~. The technical calculations for Safety Practice No.111-P-l.1 were completed in 1993. The authors considered recycle of steel, aluminum, and concrete. They also analyzed reuse of surface-contaminated rooms in buildings and reuse of tools and equipment. The report contains no uncertainty analysis. Instead, a con- servative approach was taken to deterministic calculations. Parameters were as- signed values from the upper end of their observed or expected ranges. This approach produces results that "are likely to overpredict doses which will be received in practice (if they are received); however, it is difficult to say by how much they are higher than the 'real' values" (IAEA, 1992, p. 49~. A Monte Carlo analysis was carried out for 60Co in asphalt, which confirmed that the base case estimate produced an overestimate of exposure (IAEA, 1992, pp. 104-105~. In addition, a limited sensitivity analysis was undertaken for steel recycling to study the effects of three basic assumptions on the partition, dilution, and quantity of contaminated steel (IAEA, 1992, p. 49~. A limitation in the report is the use of values for some parameters without citing sources,9 which makes it difficult for independent reviewers to trace the analysis. 9See, for example, Appendix IL p. 97, of IAEA (1992), where values for a resuspension factor, the fraction of surface contamination available for resuspension, the rate of secondary ingestion of re- movable surface contamination, and the transfer factor for secondary ingestion are given without citation.
REVIEW OF METHODOLOGY FOR DOSE ANALYSIS Interim Report IAEA-TECDOC-855 95 In 1996 the IAEA prepared an interim report Clearance Levels for Radionu- clides in Solid Materials: Application of Exemption Principles, in which it re- viewed a set of studies, including its own, to pick a set of dose factors to use in deriving secondary activity standards for clearance (IAEA, 1996~. The secondary standards are derived by dividing the primary standard recommended by the IAEA (10 mSv/yr) by the dose factor that the authors decided on for each radio- nuclide. A similar approach was later used by the HPS Standards Working Group to prepare the ANSI/HPS clearance standard. However, unlike ANSI/HPS, the IAEA study includes the steps the authors took to discount various studies, so the work is traceable. To simplify implementation, the authors grouped radioncuclides with similar clearance levels by rounding values. No uncertainty analysis is presented in the report. EUROPEAN COMMISSION DOCUMENTS The European Commission (EC) has produced a number of technical and policy documents that deal with clearance issues. The two main technical reports are EC-RP-89, Recommended Radiological Protection Criteria for the Recycling of Metals from the Dismantling of Nuclear Facilities (EC, 1998b), and EC-RP- 114, Definition of Clearance Levels for the Release of Radioactively Contami- nated Buildings and Building Rubble (EC, 2000~. These reports address metals recycling, equipment and building reuse, and building demolition. For buildings and building rubble, the analysts used a few scenarios that are assumed to be representative of the many others that have been studied by other analysts. An analysis assuming homogeneous volume contamination produced "nuclide specific clearance levels" (i.e., secondary standards) that were prohibi- tively restrictive for large buildings, so the authors took into account the likeli- hood of inhomogeneous contamination and other factors to reduce the clearance levels by a factor of 10 (EC, 2000~. An explicit assumption in the EC analyses, which is built into the EC recommendations, is that it is forbidden to mix highly contaminated surfaces or rubble with the uncontaminated bulk of the structure. Apparently, no uncertainty analysis was carried outfit Presumably the under- lying doses were calculated with a tendency to choose individual parameters that produced an overestimate in dose, but no statement to that effect is included in the reports. However, the study committee has not reviewed the full consultant's report for EC-RP-114, only what is included in the report itself. In deriving nuclide specific clearance levels, the EC reports use a collective dose standard of 1 person-sievert (person-Sv) per year and a derived dose stan- {OThe authors did consider what they called "pessimistic" assessments in developing dose factors and clearance values.
96 THE DISPOSITION DILEMMA card for individuals of either 10,uSv/yr (1 mrem/yr) or a skin dose of 50 mSv/yr (5 rem/yr) (EC, 1998b, p. 4~. If the collective dose exceeds the 1 person-Sv/yr standard, a decision must be made on whether the activity has been optimally reduced, (i.e., is as low as reasonably achievable [ALARAj). This approach suggests a refinement that the USNRC should consider as it deliberates over clearance standards. Suppose that the variations in contaminant levels of a material were so large that the highest values surveyed exceeded the allowed dose to a member of the public, even though the average value was at or below the USNRC clearance standard. It might be desirable to require reduction of the activity level to the point that the dose standard was not exceeded by the highest survey reading. COMPARISON OF CLEARANCE STUDIES Table 5-3 compares specific features of the general methodologies used in the studies reviewed by the committee. Not surprisingly, the studies do not al- ways agree on the numerical values for best estimate. To capture the rough magnitude of these differences, Table 5-4 shows the average of the ratios of the NUREG-1640 dose factors to the dose factors presented in other studies. Note that Table 5-4 uses the mean NUREG-1640 dose factor coefficient, which lies somewhere between the 50th and 95th percentile values for the dose factor, depending on the radionuclide. On average, the dose factors for metals in the draft NUREG-1640 and the EPA study are in relatively good agreement. Using the computation explained in Table 5-4, the NUREG-1640 values are lower but on average are within a factor of two of the EPA values. With respect to the dose factors selected in the IAEA and EC reports however, the NUREG-1640 values are on average about 5 to 14 times higher and hence would allow less activity to be released on average given the same primary dose standard. For particular radionuclides and particular criti- cal groups, the disagreement between the U.S. dose factors (NUREG-1640 or EPA TSD 97) and those from the EC studies can be much greater than a factor of 10. For instance, the draft NUREG-1640 dose factor for 60Co is 200 times more restrictive than the EC value for clearing surface-contaminated metals (USNRC, 1998b, Table 2.5). One reason that dose factors computed for different studies vary is that different simplifying approximations are used. Another reason is that different critical groups and different exposure scenarios for those groups are selected to model doses. In some cases, heterogeneity of contamination was assumed, from which one could derive a lower dose in a given exposure scenario than if uniform contamination were assumed, and therefore increase the activity level allowed for clearance. For example, the EC studies estimate that "the mass specific activity averaged over the total quantity of building rubble (105 metric tons) will be around one order of magnitude less than the clearance level" (EC, 2000~.
REVIEW OF METHODOLOGY FOR DOSE ANALYSIS 97 Similar assumptions, which have the effect of reducing the dose factor (and therefore allowing a higher secondary standard [see Box 5-11) have not been introduced into the analyses from which either the EPA or the USNRC dose factors were estimated. These and other differences in methodology explain some of the difference between the European dose factors and those from the EPA or USNRC studies. Finally, different degrees of conservatism may have been built into the estimates. Large differences do not necessarily imply that one approach or the other is objectively mistaken, although that is possible. Another way to look at the uncertainty in dose factors other than simply computing ratios of dose factors is to look at the variability around the ratios. To this end, we use the geometric standard deviation as a measure of variability, which can provide an estimate of the confidence that can be placed in any particu- lar coefficient. The GSD of the ratios between draft NUREG-1640 and other studies amounts to a factor of 6 to 12,ll which is a much larger range than the GSDs computed by draft NUREG-1640 based on its internal analysis of uncer- tainty (see Table 5-2~. Although some difference would be expected, such a large discrepancy raises questions as to whether or not draft NUREG-1640's uncer- tainty bands are sufficiently wide to incorporate the range in which experts may reasonably disagree and therefore the bands might need rechecking. At the very least, the USNRC should understand and be able to explain the reasons for the discrepancy. Given the complexity of the scenarios, the committee believes that an order of magnitude difference in dose estimates is reasonable for risk estimates of this type. With so much effort having gone into these studies over the past 20 years, it seems unlikely that additional, reasonable effort will be able to reduce dramati- cally the uncertainty in the coefficients that differ by less than a factor of 10 at least until there is real-world experience that can be used for benchmarking purposes. On the other hand, for the dose factors that show unusually large differences it would make sense to mount an international benchmarking exer- cise, with the goal of trying to understand the technical reasons for the major disagreements. On average, the dose factors in draft NUREG-1640 and EPA TSD 97 will yield more restrictive secondary standards (i.e., the derived allowable activity level for release of a contaminated material will be lower) for the same primary dose standard than will the dose factors from the IAEA and EC studies. In other 1lFor instance, the committee looked at the GSD of the ratio of NUREG mean dose factors to those computed by the EPA and the EC (volume-contaminated metals), using data combined from Table 2.4 and Table 2.5 of NUREG-1640. The GSD was 6. A similar analysis was done for the ratio of NUREG-1640's mean dose factors to those computed by the IAEA (all materials), this time using Table 2.6 of NUREG-1640. The GSD was 8 for volume contamination and 12 for surface contamina- tion. Note that 1 standard deviation is equal to the product of the median times the GSD; 2 standard deviations (~95th percentile for a log-normal distribution) equal the square of the GSD.
98 THE DISPOSITION DILEMMA TABLE 5-3 Comparison of Dose Factor Estimates Made to Support Clearance Proposals Category USNRC NUREG- 1640 EPA TSD 97 Nuclides Scenarios Approach Materials Dose criteria Exposed population Conversion coefficients Collective dose considered Comparison to fluxes from NORM or NARM, case-by- case clearances Dose uncertainty Level of conservatism in dose calculationsd Measurement uncertainty Human error Sensitivity studies Benchmarking or validation 85 79 Generic geometries Fe, Al, Cu metals; concrete, equipment None established, estimates included for 10 ,uSv/yr when comparing results of other studies Member of critical group Traceable No No Monte Carlo, traceable Can be determined by policy maker Not considered Not considerede None None 40 37 Specific situations Fe, Al metals (copper in preparation) None established Reasonable maximally exposed individual Traceable Yes No Sensitivity studies and judgment Implicit, thought to represent 90th percentile (e.g., 90% of members of critical group get lower doses) Considered in part Not considered To determine which parameters contribute most to uncertainty None NOTE: NA = not applicable; NARM = naturally occurring and accelerator-produced radioactive material; NORM = naturally occurring radioactive material. aIAEA (1988, p. 10). bIAEA (1996, p. 47). CTo provide perspective.
REVIEW OF METHODOLOGY FOR DOSE ANALYSIS 99 EC-89, EC-113, EC-114 IAEA TECDOC-855 ANSI/HPS 104 Limiting pathway Specific situations Metals for recycle, buildings or rubble, all solids, equipment reuse 10 ~Sv/yr; 1 person-Sv collective dose per year or, if higher, optimization (ALARA); skin dose of 50 ,uSv/yr Member of critical group Traceable Yes No Not formally analyzed Implicit Not considered Not considered None None 56 NA Most conservative of dose factors from range of studies considered reasonable All solids 10 ,uSv/yr; 1 person-Sv per year or optimization (ALARA)a Member of critical group Traceable for volume- contamination factors Yesb In part None Implicit Not considered Not considered None None 52 NA Most conservative of dose factors from range of studies considered reasonable All solids 10 ,uSv/yr; higher on a case- by-case basis. Unspecified Not traceable Qualitative discussion No Assessed on an overall basis, not nuclide by nuclide Implicit Not considered Not considered None None dDose calculations that result in higher percentile-valued dose factors are more conservative. NUREG-1640 reports a distribution of values and hence the selection is at the discretion of the policy maker. eThe USNRC has commissioned a separate study dealing with accidents.
100 THE DISPOSITION DILEMMA TABLE 5-4 Ratio of NUREG-1640 Dose Factors to Other Estimates, Averaged Across Radionuclides c`Mean,, Ra~Oa Volume Contamination Surface Contamination EPA metalsb o.64c NAd EC metalse 5.4c lOf IAEA all materials" 1 4h 4.5c aComputed as the exponential of the average of logarithms of ratios. The values from NUREG- 1640 are all mean values that lie between the 50th and the 95th percentiles for all radionuclides. bDenved from Table 2.4, draft NUREG-1640. C+~26 percent. Standard deviation for art individual radionuclide, however, is approximately a factor of 5. dNot applicable. eDenved from Table 2.5, draft NUREG-1640. f+37 percent. Standard deviation for an individual radionuclide is a factor of 12. gDenved from Table 2.6, draft NUREG-1640. h+36 percent. Standard deviation for an individual radionuclide is a factor of 8. words, the draft NUREG-1640 and EPA dose factors are more protective. The committee has not been able to determine the precise reason for the differences from other estimates. The question of whether the total uncertainty could be greater on average than a factor of 10 is discussed in the next section. Usefulness and Quality of Dose Factors The committee's review of the studies listed in Table 5-1 found that some of the dose factors estimated in these studies, particularly those for radionuclides causing external gamma radiation doses to workers, can easily be shown to be reliable. Other dose factors require the use of parameters that are highly uncer- tain. One way to compensate for uncertainty in setting a protective standard is to set the dose factor for each radionuclide at a fixed margin above the best estimate for the dose factor. This allows the decision maker to compensate for the lack of complete knowledge in the dose analysis and thus increase confidence that the dose to the critical group will be below the primary dose standard. For example, the value for the dose factor can be set to the 95th percentile in the distribution of values for that dose factor rather than the median. Taking the mean value of the distribution, in almost all complex dose analyses (i.e., for right-skewed distribu- tions), will increase the value of the dose factor over the median or 50th percen- tile result of the Monte Carlo calculation. The mean value has the property, in most calculations of this type, that its distance above the median automatically increases when uncertainty is large and decreases when uncertainty is small. (Although NUREG-1640 gives explicit values for the 5th, 50th, and 95th percen- tiles, it would be possible for the authors to extract other values e.g., the 85th
REVIEW OF METHODOLOGY FOR DOSE ANALYSIS 101 percentile from the computed Monte Carlo distributions that would exceed the median by varying amounts.) However, the choice of any percentile level (and its corresponding dose factor), like the choice of a primary dose standard, is a matter of policy that cannot be decided by scientists through analysis or facts alone. For instance, policy makers could decide to choose dose factors closer to the median of the distribution of dose factors forgoing the additional margin of protection afforded when a higher percentile-valued dose factor is selected because they consider a 1 mrem/yr dose to be too far below background to be of concern. Conversely, they could pick a higher percentile-valued dose factor (e.g., the 95th) to assure the public that doses are very unlikely to exceed 1 mrem/yr. If this additional margin of protection (which is implicit in the choice of higher percentiles) is not used in setting a dose factor, one must either pay close attention to the uncertainty in the estimate for each dose factor or fall back on assurances that analysts tended to be protective of public health (i.e., they picked parameter values e.g., landfill leaching rates, resuspension coefficients from the range of uncertain values that would end up being restrictive on the amounts of radioactivity that could be released to produce a given dose). However, the committee is reluctant to recommend reliance on statements by experts about the protectiveness of calculations. This is just the area in which experts have been found to perform poorly (Cooke, 1991; Shlyakhter and Valverde, 1995~. Although picking the percentile value appropriate for selection of dose fac- tors is a policy choice, decision makers need to be informed about the quality of the supporting information. Over time, risk analysts have devised ingenious ways to reduce what at first glance appear to be unavoidable uncertainties in an analy- sis. For instance, it is often not necessary to know the amount of radioactivity released by a licensee in order to make use of a dose factor; often knowing the mass concentration is enough (i.e., the activity per gram). Analysts often simply consider releases that are large enough to saturate the doses to members of a candidate critical group, such as an entire truckload or industry-wide totals.l2 In general, bounding assumptions are made to eliminate the need to consider the total quantity of material released. Although this tends to overestimate dose factors and reduce allowed release concentrations, such as when concentrations are kinetically limited, it simplifies regulatory considerations. However, there are exceptions,l3 and some residual assumptions may still be necessary, such as the amount of mixing that takes place with nonradioactive material; see Box 5-2 for 12For example, once the volume of cleared material exceeds a truckload, the dose to the truck driver during one trip cannot go higher, which allows the number of trips one driver can make before receiving the allowed dose to be computed. The number of drivers needed to move the cleared material will increase as the quantity of cleared material increases, but this affects only the number of drivers who receive the dose. The collective dose increases, but not the dose to an individual driver. 13Exposure of workers in a steel plant may depend on the total quantity recycled (IAEA, 1992, p. 54), although even there, the dependence is limited. In the IAEA study, a hundredfold increase in the total amount of contaminated steel being handled produced an eightfold increase in individual dose (IAEA, 1992, p. 58).
102 THE DISPOSITION DILEMMA an illustration. In contrast to individual doses, collective doses under a clearance standard are directly related to the total amount of radioactivity released. Despite the use of bounding assumptions, considerable uncertainty remains in some sce- narios, particularly when it comes to predicting the behavior of radioactive mate- rials leaching from landfills. Analysts often add margins of protection to components of a dose factor calculation because information about a parameter is lacking or because the analyst is seeking greater generality for the analysis. Because different analysts may not use the same margins in their computation, the various studies listed in Table 5-1 are difficult to compare. The numbers are neither pure "best estimates" (i.e., estimates of central tendency) nor pure bounding estimates (estimates of the upper and lower bounds of a percentile range). It is particularly difficult to esti- mate how the dose factor calculated for one study would change if an assumed margin of protection were changed to improve its agreement with other studies.
REVIEW OF METHODOLOGY FOR DOSE ANALYSIS 103 Until a clearance system is implemented and concentrations of radioactivity in key scenarios are measured, one cannot be certain that assumptions made to provide margins of protection or other safety-enhancing factors have been ad- equate or are unrealistically restrictive. One way to deal with hypothetical model error is to adopt a policy of "adaptive management" in which real-world perfor- mance is monitored through validation that is possible only after implementation, or through retrospective analysis of selected case studies. For example, leachate can be sampled from representative landfills, or con- centrations of radioactivity in sample pieces of recycled steel can be checked, to ensure that the model assumed in calculating dose factors reasonably represents reality, with an adequate margin of protection. The model, and the dose factors calculated from it, should be updated if the primary dose standard is being ex- ceeded or even if key assumptions in the model are clearly inadequate.l4 The IAEA encourages this type of retrospective review, including the "testing of radioactive consumer products on the market" (IAEA, 1988, p. 14~. Reaching most of the limiting conditions that were assumed in estimating dose factors, such as truck drivers handling slightly contaminated truckloads every work day or concentrations in landfills reaching the maximum capacity, will sometimes take considerable time (typically, years of activity after clearance standards are imple- mented). If a validation program is in place soon after a standard is implemented, there will be sufficient time to adjust dose factors (and the clearance standards derived from them) if corrections are needed. Based on Table 3-7, it seems unlikely that SRSM from USNRC-licensed facilities cleared under a dose-based standard will come close to matching the 14A validation program might also include measuring the distribution of radioactivity, or limits on the amount of radioactivity, that arrives at monitored landfills. Even data from portal monitors placed at both the sending and the receiving facility would be useful, particularly in assessing how often human error leads to gross errors in maintaining transport constraints. Other ideas for useful data collection can be gleaned from the EC guidelines, which require licensees to track the total amount of material cleared for disposition (EC, 2001). If the amount of material per shipment was recorded, as well as the activity measurements made to check compliance with the secondary standard, then uncertainty margins relative to the assumption of clearance at the activity level of the secondary standard could be computed. This analysis would also aid in determining if significant mixing of waste was occurring. The USNRC may not find it justifiable to require this degree of data gathering and reporting by licensees, but it might fund a program of research-oriented activities. During the 1980s, when the Low-Level Radioactive Waste Policy Amendments Act was passed, the USNRC considered includ- ing a requirement in 10 CFR Part 61 for reporting data on the radioactive content of low-level radioactive waste shipments to disposal sites. This requirement was not included in the rule. It was believed that the data would be useful only as a broad check of assumptions made in the environmen- tal impact analysis for disposal, not for material balance. For some years, such data were obtained by contract for such a broad check (personal communication from Robert Bernero, Board on Radioac- tive Waste Management, National Research Council, July 17, 2001). However a detailed material balance would not be necessary for the validation activities discussed.
104 THE DISPOSITION DILEMMA concentration and total amounts of naturally occurring and radioactive material (NORM), naturally occurring and accelerator-produced radioactive material (NARM), and comparable materials that are cleared today under a case-by-case approach. Consequently, field data will probably prove useful only in assessing how well the clearance models have bounded the concentrations and thus esti- mated the doses. Nevertheless, a modest monitoring effort would boost confi- dence in the dose factors, particularly for those who are skeptical of the models being used. It may also provide useful incidental information on where NORM and NARM are ending up. General Limitations of the Reviewed Studies Failure to Consider Uncertainties Associated with Implementation of a Primary Dose Standard Dose factors as estimated to date are useful theoretical tools. However, they have practical value only within a specific implementation protocol, where such a protocol can introduce uncertainties into dose estimates tied to primary dose standards. Only a few studies (e.g., EPA, 1997a) appear to have explicitly consid- ered any implementation issues in assigning uncertainties to the estimated dose factors. Among these sources of added uncertainty are averaging error, sampling error, rounding error, and treatment of multiple radionuclides: Averaging error. The area or volume over which one averages radioactiv- ity can introduce errors (EC, 2000, p. 20~. This will increase the uncer- tainty associated with dose estimates. Sampling error. Guidance for a volume contamination standard would probably include acceptable sampling and modeling methods, which would allow some level of sampling error. Sampling error, in turn, could add to overall dose uncertainty. To a degree, any error incurred from a finite number of samples might be offset by the fact that not all of the cleared material will have an activity level exactly matched to the second- ary standard. On the other hand, there is also the possibility that hot spots may have been missed. Rounding error. For practical reasons, regulatory authorities may decide to round secondary activity standards to a few convenient values for instance, 0.1, 1, 10, and 100 Bq/g, and so forth. This can result in an error of a factor of three or so in dose factors. This practice, which has been adopted by the European Union (EC 2001, Table 1) and is used in the ANSI/HPS standard (ANSI/HPS, 1999), is equivalent in effect to choos- ing higher or lower percentile-valued dose factors. The possibility of rounding the derived secondary standards to integral powers of 10 should be considered when assessing uncertainties and selecting the percentile
REVIEW OF METHODOLOGY FOR DOSE ANALYSIS 105 value corresponding to the dose factors. The percentile level implicit in a rounded activity standard should be roughly the same as the percentile level sought in a dose factor that will not be subject to rounding. For example, if the policy choice for selecting dose factors is to maintain a 95 percentile level, then the implicit percentile level of a rounded activity standard should be at least 95 percent. Alternatively, information such as the implied confidence level after rounding should be presented with the proposed activity standards so that policy makers understand the implica- tions of adopting a policy of rounding the activity standards. · Multiple radionuclides. If a dose-based clearance standard was chosen, a decision would have to be made on whether its implementation for mul- tiple radionuclides should apply a sum-of-the-fractionsl5 computation or apply the individual clearance levels for any nuclides detected. The sum- of-the-fractions method is used routinely for control of radioactive efflu- ents (10 CFR Part 20, Appendix B) and is recommended by the EC for clearing solid material (EC, 2001, p. 14~. For a given protocol, an analyst can estimate the uncertainty that may result from using it with contamina- tion from multiple radionuclides and include the estimated uncertainty in the dose factors. Without the specification of a protocol for treating mul- tiple nuclides, it is difficult to assess whether any changes need be made, up or down, to the uncertainty estimates for dose factors. Lack of Validation of Model Estimates Validation against field data provides the best way to check for model error, as well as unexpected problems with parameter assignments. As noted in the previous section, a validation program should be used to correct and refine a system of dose-based clearance standards, given the inevitable uncertainties in the process of estimating dose factors. Furthermore, the confidence of policy makers, licensees, the public, and skeptics in the predictions from risk assess- ments can be increased by undertaking validation exercises. The committee heard only one presentation about a study in which clearance model estimates have been field-tested. 16 In that case, an international group led by the Swedish Radia- tion Protection Institute attempted to check predictions of model estimates against 15A sum-of-the-fractions computation is used when the governing standard sets the amount of each isotope that, if alone, would reach the dose limit of the standard. When materials containing many isotopes are analyzed for compliance with the dose-based standard, a fraction is calculated for each isotope present (the amount detected divided by the dose limit amount set for that isotope). The sum of all these fractions must be less than or equal to 1 if compliance with the dose limit is to be ensured. 16Shankar Menon, program co-ordinator, OECD/NEA Co-operative Program on Decommission- ing, presentation to the committee, June 13, 2001.
106 THE DISPOSITION DILEMMA results of actual recycling of SRSM. The committee did not review this work but wishes to encourage that such studies be undertaken. Lack of Inclusion of Accidents and Human Errors in the Dose Factors The IAEA recommends consideration of accidents in estimating exposures of the public from disposal exemptions (EC, 2000, p. 20~. Examples of human error that can initiate or contribute to accidents involving error in clearance of materials at a nuclear power plant include failure to monitor properly, failure to properly handle and contain loose contamination, and delivery of material to the wrong recipient. Specifically, a facility that was routinely required to screen all scrap material for radioactivity, but rarely encountered any contamination, might disable the radiation alarms, fail to keep them in working order, and/or ignore them when they actually went off. Human error was not explicitly addressed in the analyses supporting dose factor estimates in any of the studies reviewed. However, the USNRC has carried out an (as yet unpublished) analysis of one form of human error (accidents), which suggests that this type of human error is not likely to have a significant impact on dose factor estimates. USNRC staff were not able to provide the study committee with the frequency at which exit monitors at licensed facilities were triggered by shipments on their way to final disposition, following clearance based on Regulatory Guide 1.86, a license provi- sion, or approved by case-by-case review. However, a health physicist from the steel recycling industry told the committee that shipments from USNRC-licensed facilities have been sent back from recycling facilities because the shipments triggered portal monitors. Although alarm events could be false alarms since the portal monitors are set as close as possible to background radioactivity levels, they may also indicate that human errors were made in the release of material from the source facility. Consequently, it must be presumed at this time that some shipments will leave licensed facilities with contamination in excess of a clear- ance threshold level. Clearance coefficients that are estimated using a probabilis- tic approach, such as draft NUREG-1640, can account for this possibility.l7 Human error may have only limited impacts on dose factor estimates, espe- cially for those coefficients where simplifying methods have been used to make 17If human error is not correctly accounted for in the dose rate coefficients themselves, other methods can be used to handle it in the system itself. For instance, portal monitors can be placed not just at the exit of licensed facilities, but at recipient sites, such as landfills or recycling facilities, if release to these facilities is allowed under the standard adopted. In many cases, steel mills have such portal monitors (and, in some cases, monitors in other portions of the facility), as do landfills and licensees that generate wastes. Pennsylvania already has a requirement that all landfills be outfitted with portal monitors to catch orphan sources, along with a formal plan for dealing with radiation sources that trigger the monitors. As one example: if landfill disposition of SRSM were restricted to landfills that installed portal monitors, one protection against human errors made at licensed facilities might be institutionalized.
REVIEW OF METHODOLOGY FOR DOSE ANALYSIS 107 estimation easier and more robust. However, human error can also be embedded in the larger framework of system failure, which includes the following interre- lated sources of failure: (1) hardware, (2) software, (3) organizational, and (4) human (Haimes, 1991~. A follow-up study might take such a systems approach. Potential Inconsistencies in Dose Factors Between Countries As noted above, analysts from different countries have estimated different dose factors. These differences can lead to inconsistencies between clearance policies adopted in different countries. However, in discussing transnational con- sistency of dose factors and derived secondary clearance standards, two types of consistency must be distinguished. If countries agree on the same primary dose standard, they have agreed on the level of risk that sets the ceiling on clearable SRSM. For instance, there is widespread agreement on a 10 mSv/yr primary dose standard in the European Union. If countries disagree on which sets of dose factors are appropriate the second type of inconsistency they are differing over technical calculations, possibly differing only over degrees of conservatism that are needed to simplify the estimates. Consistency of clearance standards across national boundaries is clearly de- sirable, particularly for materials that might find their way into international commerce. However, it would be inappropriate for one country to change its view of the supporting scientific evidence simply to achieve consistency with the standards in effect in other countries. Such an approach would not be conducive to building confidence in the scientific and engineering foundations for clearance standards. Even the appearance of making changes in technical documents to make policy choices easier could undermine public confidence in the overall results. If rationalization of standards across borders becomes paramount, after attempts at technical rationalization have failed, the effort should be separated from the scientific deliberations by which dose factors are estimated. The decisions to rationalize for reasons beyond those supported by technical studies should be made as a clear policy choice (e.g., accepting more or less conservatism in the adopted dose factors). DETAILED COMMENTS ON NUREG-1640 As noted, the committee paid particular attention to the draft NUREG-1640 because it was prepared for the USNRC in preparation for reconsideration of clearance standards. The discussions in this section supplement the earlier gen- eral discussion of analytical limitations in the draft document. Many of the issues raised here may have been considered intuitively by the analysts and staff that prepared the draft and judged to be of little consequence. Some may be currently under study at the USNRC. In any case, the committee believes that all of the
108 THE DISPOSITION DILEMMA following issues have to be considered explicitly, at some point, in the technical support process. Issue 1: Landfill Disposal Scenarios Landfill issues in the draft NUREG-1640 were difficult to understand. They require clarification and justification. The following are examples: . . · Fraction of material that goes to landfill. The justification for the as- sumed 0.15 percent fraction of volume of material that ends up in a landfill is weak (USNRC, 1998b, p. 4-98~. The +50 percent uncertainty assigned to the fraction seems small. Alternative economic models for landfill deposits. Draft NUREG-1640 does not consider the situation in which only a small number of facilities are willing to take cleared material. Neither does the EPA, although TSD 97 does mention this possibility. If the postclearance landfill industry splits this way, the net result would be to increase the fraction of released material in the few facilities that would take contaminated material, thereby increasing the dose to landfill workers and nearby residents. This possibility is sufficiently realistic that it deserves assessment. It can prob- ably be handled in draft NUREG- 1640 by changing the uncertainty distri- bution currently assigned to landfill clearance calculations. Uncertainties. Landfill scenarios in draft NUREG-1640 did not have de- fined critical groups, so they did not get the consideration they might have if conditional clearance had been under consideration. Leaching rates, liner failure, and long-range transport are possible issues that should be addressed more carefully as part of the technical support process. Issue 2: Incineration Pathway Once material is released into general commerce, it may one day enter the municipal waste stream. Since a certain percentage of trash is incinerated to reduce volume, one possible immediate or delayed-clearance pathway would be incineration; yet this pathway was not addressed. Even though this pathway is unlikely to be significant, it should be explicitly considered. Issue 3: Sensitivity Analysis The uncertainty analysis was reasonable, but since the study uses a Monte Carlo analysis, the committee wondered why a set of sensitivity analyses was not carried out. Sensitivity analyses can be misconstrued as uncertainty ranges, but the committee believes that they can be constructive. Sensitivity studies yield important information about the significance of an input parameter's value to the
REVIEW OF METHODOLOGY FOR DOSE ANALYSIS 109 output value predicted by the model. In this case, such a study would allow a better assessment of the effect of the parameter's uncertainty on the calculated dose factors. (See also discussion of resuspension of contamination below.) Issue 4: Validation There is no benchmarking or validation provided in the appendix material to draft NUREG-1640. Benchmarking or validation exercises would be appropriate to demonstrate the validity of the modeling technique. Issue 5: Sample Calculations There was a dearth of sample calculations that could have provided clarity for readers as to the overall method. Issue 6: Multiple Pathways The draft report does not consider multiple pathways. The committee notes that when exemption from regulatory control is considered, the IAEA (1988) recommended as follows: "The dose to the individuals of the critical groups and the dose to the whole population exposed by the practice should not be significantly affected by other similar (or identical) practices (e.g. several waste disposal sites in the same region)." Issue 7: Resuspension of Contamination The draft document has only limited consideration of resuspension of sur- face contamination into the air. Of all the factors that can play a role in exposure to toxic substances, resuspension is probably the most difficult to address (IAEA 1992, p. 66; USNRC,1998b, p.3-8~. Even after loose material is removed during cleaning, some residual radioactive material can be available for resuspension over a longer time. Resuspension rates, which generally affect only inhalation exposures, can span many orders of magnitude, as the authors of draft NUREG- 1640 acknowledge: "The resuspension factor, RFSc, is the most poorly known parameter in the inhalation pathway analysis . . ." (USNRC, 1998b, p. 3-8~. The method of uncertainty analysis adopted by NUREG-1640, which the committee applauds, can nevertheless be disconcerting when applied to param- eters with large uncertainty ranges. The 95th percentile can end up being many times greater than the highest value measured to date. There is a tendency for analysts to disbelieve such numbers and make some form of downward adjust- ment. This is a potential form of downward bias that bears watching, given the known problem of expert overconfidence (Cooke, 1991), which leads to underes- timation of uncertainty ranges when subjective judgments are made.
110 THE DISPOSITION DILEMMA For example, in estimating doses to workers in reused trucks, the draft NUREG-1640 analysts selected the bottom of the range of resuspension values available to represent the median of the distribution, with little justification. The choice of geometric standard deviation was also made with little justification. With measured resuspension rates varying by many orders of magnitude, it is difficult to determine how to handle this problem. At a minimum, a sensitivity analysis should be performed to inform readers as to how the dose factor would vary with a change in the resuspension coefficient. A sufficient technical basis may not yet exist for assigning a credible uncer- tainty factor to certain types of releases that are sensitive to resuspension. If so, such clearance categories could be excluded by regulation until a sufficient tech- nical basis is developed. Issue X: Collective Dose Draft NUREG-1640 has no consideration of collective dose. The EC and the IAEA have a two-part primary dose standard, 10 ,uSv/yr for an individual and 1 person-Sv/yr for the collective dose to the population. Specifically, IAEA recom- mends that regulatory authorities conduct a generic study in the early stages of regulatory development to determine whether the annual dose from exempt prac- tices will exceed one man-Sv. If not, then further optimization of the regulatory option being proposed is not needed (IAEA, 1988, pp. 10-11~. Unlike practices from which individual doses may vary over a wide range and be a significant fraction of background radiation, doses from activities that result in low indi- vidual doses result in doses and therefore risks that are individually and collec- tively very small both in absolute value and in comparison to natural background and man made exposures levels at which the significance of collective dose has been controversial. (To exceed the collective dose requirement, more than 100,000 persons would have to receive the allowed individual dose.) The EPA has also examined collective dose (EPA, 1997a). However, technical analysis by the USNRC has focused only on the individual dose. Although the dose to an individual in a critical group, and thus the secondary activity standard, does not ultimately depend on the total radioactivity released, the collective dose does increase with total radioactivity. For example, if more than one truckload of material is shipped at a given concentration from a licensed facility using differ- ent drivers for each truck, the dose to an individual driver from a full load does not increase, only the number of exposed truck drivers increases. Even if the same driver makes multiple trips, the dose will be limited by the total number of trips that can be made in one year. Consequently, it may be of interest in shaping policy to have some idea of collective dose, recognizing that such estimates may carry much greater uncer- tainty than will the dose to an individual in the critical group. Given a collective primary dose standard in the range of 1 person-Sv/yr, the individual dose estimate
REVIEW OF METHODOLOGY FOR DOSE ANALYSIS 111 for material from USNRC-licensed facilities is likely to be more restrictive than the collective dose (Clarke, 2001~. Issue 9: Size of Critical Groups The draft NUREG-1640 does not discuss the total number of people exposed in any critical group. Although most critical groups will include a relatively small number of persons, other critical groups may include greater numbers of people. When groups are large, it is easy to think of smaller subgroups that could get higher doses. For instance, iron workers, train conductors, and elevator operators could receive higher doses from slightly radioactive steel than would users of common consumer objects. Knowledge of the approximate size of critical groups assists in building confidence that a more important subgroup has not been over- looked. Issue 10: Total Activity Buildup and Mass Balance The draft NUREG-1640 contains limited information on total activity buildup and mass balance. The methodology chosen to estimate doses for draft NUREG- 1640 largely eliminates the need to know the total inventory of curies released. The authors consider (justifiably) that the total amount of curies released and stored affects the estimation of cumulative doses more than the estimates of critical doses (i.e., the individual doses on which dose factor selection is based). Nevertheless, the committee is uncomfortable with the lack of activity balance estimates. Given that a material flow model has already been developed for the analy- ses, it should be straightforward to account for approximately how much of the radioactivity released each year is removed from the commerce "pool" as it enters landfills, how much will build up in the steel content, and how much would end up stored in structures. Since 85 percent of the steel cleared from USNRC facilities is likely to end up in landfills, steel made from cleared scrap will constitute only a tiny fraction of the total recycle in the United States. With radioactive decay, the committee does not believe that the buildup is likely to be significant, but without supporting estimates there is no explicit basis for this pOSition.l8 18It would also be useful to compare the amount of radioactivity in material projected to enter commerce and landfills from various proposed clearance policies with the amounts entering now from both the USNRC's case-by-case clearance policy and NORM or NARM sources. To aid in estimating the quantities entering commerce and landfills now from USNRC-licensed facilities, ana- lysts could collect a random sample of case-by-case decisions from each USNRC region and analyze the dose implications using NUREG-1640 coefficients.
2 THE DISPOSITION DILEMMA Issue 11: Accounting for Human Error Accounting for human error is good risk assessment practice. Draft NUREG- 1640 does not consider human error and specifically assumes that there is none. Although USNRC staff has already taken steps to analyze the impacts of acci- dents on dose factor estimates, more of this type of analysis should have been done in the draft document. For instance, in one case, the analysis assumes that loose surface contamina- tion is always removed according to good health physics practice (USNRC, 1998b, p. 3-2~.19 Yet inclusion of a modest human error rate could end up domi- nating the dose estimate. It is inconceivable that all loose surface contamination will always be removed prior to clearance. The probability that loose material may be overlooked may be low, but the downstream dose from loose contamina- tion could in principle be sufficiently high to overcome the low probability that an error will occur. Issue 12: Uncertainty in Conversion between Intake and Dose The authors of draft NUREG-1640 did not consider uncertainties in the coefficients that convert inhalation and ingestion to dose,20 relying instead on coefficients developed by the EPA. Although the uncertainty in these coefficients may not be significant compared to other uncertainties that enter the estimate of dose factors, especially for USNRC-licensed facilities,21 this contribution to un- certainty should be explicitly considered. FINDINGS Finding 5.1. Analytical work in the United States and abroad over the past two decades is useful in understanding the likely doses associated with exposure scenarios that might occur under various clearance standards. Much of the techni- cal analysis in this field has the objective of understanding "dose factors," which to date have been analyzed in depth only for (unconditional) clearance scenarios. A dose factor is used to convert a concentration of radioactivity that is about to be released, whether it be confined to a surface or contained within a volume, to a primary dose level (measured in microsieverts per year or millirems per year). With such a dose factor in hand, a primary dose standard can be converted to obtain a secondary clearance standard in terms of radionuclide activity, which 19''Based on assumed good health physics practices at NRC licensed facilities, removable surface contamination has been removed during decontamination procedures prior to final survey and clear- ance" (USNRC, 1998b, p. 3-2). 20Constant values taken by draft NUREG-1640 included "the dose equivalent due to radionuclide intake." 21The radionuclides of significance at USNRC-licensed facilities are generally not transuranics.
REVIEW OF METHODOLOGY FOR DOSE ANALYSIS 113 could then be used at USNRC-licensed facilities. A dose factor can be used with any choice of primary dose standard. Finding 5.2. Selecting a primary dose standard is a policy choice, albeit one informed by scientific estimates of the health risk associated with various doses. For instance, as shown in Table 1-2, a lifetime dose rate of 10,uSv/yr (1 mrem/yr) equates to an estimated increased lifetime cancer risk of 5 x 10-5, which falls within the range of acceptable lifetime risks of 5 x 10= to 10-6 used in developing health-based radiation standards other than radon in the United States (NRC, 1995, p. 50~. When setting primary dose standards, regulators can make a policy decision to include a level of conservatism such that the final standard is in excess of the best-estimate dose factor and in this way account for uncertainty (e.g., selecting the 90th, 95th, or other percentile in the distribution for the dose factor, instead of the best-estimate value). Finding 5.3. The uncertainty in dose factor estimates is a key technical issue. When an uncertainty has been estimated, a quantitative determination can be made of the likelihood that the dose to an individual in the critical group will be below the primary dose standard. Quantitative uncertainty estimates can also assist regulators in assigning a level of conservatism to dose factors in excess of the best estimate. Dose factors developed by analysts from different countries show wide variation, which highlights the need for careful consideration of un- certainties. Finding 5.4. The committee concludes from its review that of the various reports, draft NUREG-1640 (USNRC,1998b) provides a conceptualframework that best represents the current state of the art in risk assessment, particularly with regard to its incorporation of formal uncertainty, as judged using recommendations of this committee and other committees of the National Research Council. Once the limitations in draft NUREG-1640 have been resolved (see Findings 5.5 and 5.6) and the results are used in conjunction with appropriate dose-risk estimates in the final version of the report or in follow-up reports the USNRC will have a sound basis for considering the risks associated with any proposed clearance standards and for assessing the uncertainty attached to these dose estimates. Finding 5.5. The development of the NUREG-1640 draft has been clouded by questions of contractor conflict of interest. The mathematics and completeness of scenarios considered in draft NUREG-1640 have been verified through an audit carried out by another USNRC contractor. The committee also carried out its own review that generally confirmed the reasonableness of several dose factor analy- ses. However, a thorough review of the choice of parameters and parameter ranges, term by term, is needed to complete the reassessment of draft NUREG- 1640.
114 THE DISPOSITION DILEMMA Finding 5.6. Draft NUREG-1640 did not consider human error and its possible effect on dose factor predictions, nor did it consider scenarios involving multiple exposure pathways. In addition, draft NUREG- 1640 does not provide a sufficient basis to analyze conditional clearance options, such as disposal in a Subtitle D landfill. Finding 5.7. The dose factors developed in draft NUREG-1640 should not be used to derive clearance standards for categories of SRSM other than those con- sidered in the draft NUREG-1640, without first assessing the appropriateness of the underlying scenarios. Some of the dose factors developed in draft NUREG- 1640 are likely to require modification when applied to other mixtures of radio- nuclides (e.g., mixtures in which transuranics dominate) and other clearance scenarios, such as may be relevant to DOE material and technologically en- hanced naturally occurring radioactive material (TENORM).
