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.
4 ADEQUACY OF TREATMENT OF RADIONUCLIDE MOVEMENT The primary barrier against radionuclide release in the KBS plan is the copper canister enclosed in its nest of compacted bentonite. If this barrier functions as planned, safe isolation of waste for a million years and more could be guaranteed in any kind of rock, at any depth where groundwater is not actively flowing or markedly corrosive. But some unexpected, very low probability event is always possible, and the KBS authors devote much effort to showing that the kind of rock they plan to use for a repository, at a depth of 500 m, aided by the bentonite buffer, will constitute an adequate secondary barrier for controlling radionuclide release even if something goes radically wrong. Suppose that a canister is breached so effectively that all of its contained fuel is at once in contact with groundwater. This is the extreme situation that the KBS authors postulate in calculating rates of radionuclide migration. The extreme nature of the assumption should be kept in mind. Certainly it is plausible to expect that a few canisters will fail prematurely, because of flaws in manufacture or some very unusual natural dis- turbance; but the failure in all probability would be only a minor crack permitting very limited access of groundwater to the fuel. Dissolution of radionuclides would be slow and localized, and the quantity being added to groundwater at any time would be small. But the extreme assumption of instantaneous exposure of much fuel to groundwater is useful to explore as a means of assess- ing just how conservative the KBS design is. Nuclides dissolved from fuel elements exposed in a breached canister will move first by slow diffusion through the buffer (provided the buffer is still reason- ably intact), and then into the surrounding rock, where 42
43 motion will be in large part by diffusion through water contained in narrow cracks and to a lesser extent by diffusion into the rock matrix. For discussing the motion, a distinction is made in KBS-3 between the "near- field" (the buffer and the rock within a few meters of it, where the major questions concern the rate of dissolu- tion and then diffusion in clay and small fissures) , and the "far-field" (bedrock between the near-field and the point where groundwater enters a major fracture). An additional consideration is a possible short-circuit path to the biosphere by flow of groundwater through tunnels and shafts, either because the backfill is an inadequate barrier or because rock adjacent to the tunnels and shafts has been rendered more permeable by blasting during repository construction. These three itemsâmigration of radionuclides in the near-field, migration in the far- field, and the effectiveness of planned seals in con- trolling migration along tunnels and shaftsâare discussed in the following paragraphs. A final section presents a brief analysis of the estimates of doses to humans that would result from radionuclide transport. The treatment in KBS-3 is similar to that in KBS-2, but additional experiments, field tests, and mathematical analyses are adduced to support the conclusions. RETARDATION IN THE NEAR-FIELD To what extent will slow dissolution and diffusion through bentonite and adjacent small fractures control the migration of radionuclides away from a breached canister in the near-field? "~ Dissolution. In estimating the rate of dissolution, the KBS-3 authors note that the radionuclides formed from uranium atoms by fission or neutron capture will in large part remain dispersed within the uranium oxide pellets, and hence will dissolve only as fast as the crystal struc- ture of UOj is destroyed by the dissolving of uranium. The isotopes of Cs and I are treated as an exception, because they or their precursors are sufficiently volatile to move through the crystal structure to the outer part of the pellets. As long as conditions remain reducing, dissolution is slow because it is controlled by the very low solubility of U02 (1 to 2 yg/1). The rate, however, may be affected by radiolytic decomposition of water, and this possibility is examined in detail in
44 KBS-3. Irradiation of the water in contact with the fuel pellets will produce hydrogen and peroxy free-radicals; the escape of hydrogen gas may make the groundwater in close proximity to the canister highly oxidative. In this oxidizing water, uranium would be much more soluble, as would be Tc, Np, and Pu. The oxidized forms of these elements could diffuse into the buffer, but within a short distance would encounter the prevailing reducing conditions and would tend to precipitate. The calcula- tions indicate that even with the most generous assump- tions about the efficiency of radiolysis and escape of hydrogen, the effects of oxidation would be limited to a zone within a few meters of the fuel (TR 83-66, TR 83-68) . Thus, in most of the near-field, the amounts of radio- activity moving through the buffer and adjacent rock would be limited by the low solubility of many of the elements in a reducing environment and, for the more soluble elements, by the rate of dissolution of the uranium oxide. As a general confirmation of the low solubility of uranium in reducing groundwater (and hence of the low leachability of nuclides dispersed through UO2 pellets), uranium concentrations were measured in many samples of water obtained from the recently studied field areas. These concentrations, presumably representing equilibrium with uraninite (UC>2) known to be present in traces in much of the Swedish bedrock, are all below 10 vg/l, and mostly below 1 ug/l, in agreement with values calculated from thermochemical solubility data (TR 83-40). Migration. The dissolved nuclides will diffuse through the buffer away from the fuel rods at widely different rates. Some will be greatly retarded by sorption, others scarcely at all. For nuclides that diffuse slowly and are strongly sorbed (most of the actinides, for example), the delay in moving through 0.38 m of compacted bentonite could amount to thousands of years (TR 82-27). Eventu- ally, however, a steady state will exist in which about as many nuclides are escaping from the buffer's outer surface as are entering its inner surface from the dis- solving fuel pellets. Thus the migration of nuclides in the near-field can be broken down into a transient phase and a steady-state phase. The transient phase is important only for those nuclides that will be retarded for lengths of time many times greater than their half-lives. Calculations show that a few nuclides will decay significantly as they
45 migrate through the bentonite; e.g., 241Am would decay to innocuous levels in the buffer, and the activity of "*Pu would be reduced about tenfold (TR 82-27). For many radionuclides, however, their transit time in the buffer is not long enough to permit significant decay. For this reason the KBS authors, in calculating release times into the far-field and into the biosphere, assign no credit to retardation in the buffer. The steady-state phase is more important than the transient phase in assessing migration through the near- field. In the steady-state phase, the concentration of a nuclide moving into the far-field depends on the reductive precipitation and the diffusive resistances of the buffer and of the water in small bedrock fractures that makes contact with the buffer. Calculations indicate that the latter is more important than the former, if it is assumed that, on average, one fracture of width 0.1 nun intersects each meter of bentonite and hence that the cross-sectional area for flow is much less in the rock than it is in the clay (TR 82-24, Table 2.2). The diffusive resistance of the buffer is small enough to be neglected, and the quantity of any nuclide transported to the far-field can be calculated from the diffusive resistance in the water- bearing fractures, expressed as "equivalent water flow." This quantity is calculated from an assumed groundwater flow rate at repository depth. Thus the calculated migration rates through the near- field depend on a series of assumptions, which lead to the conclusion that the buffer plays only a minor role in both transient and steady-state phases. The panel was concerned about the degree of conservatism involved in the basic assumptions, given the uncertainties in estimates of groundwater flow rates. When asked about the conservatism, Ivars Neretnieks (Royal Institute of Technology, personal communication, 1983) pointed out that, even with much different assumptions, the integrity of the bentonite as a retardant for groundwater has little importance in the near-field calculations, and also that the "equivalent water flow" is not very sensitive to the assumed Darcian velocity (a tenfold increase of the latter increases the former only threefold). Even in the absence of any bentonite buffer, radionuclide concentrations in the geosphere and biosphere would increase only twenty- fivefold for the most pessimistic water flow rates and at most threefold for more representative flows (Ivars Neretnieks, Royal Institute of Technology, private communication, 1983). The panel concludes that the
46 assumptions are adequately conservative for estimates of radionuclide concentrations delivered to the far-field. This does not mean that the important diffusion processes have been completely and rigorously studied. The Swedish scientists recognize this lack ("Diffusion through compacted bentonite is governed by complex mech- anisms and cannot be accommodated by a simple pore- diffusion model." TR 83-37, p. 17) and plan additional research on diffusion rates. The Soret effect and surface-migration mechanisms (versus pore diffusion) are examples of topics that are not yet completely docu- mented. But in the panel's opinion, the refinements to be gained by additional research are not likely to change greatly the current estimates of migration rates in the near-field. RETARDATION IN THE FAR-FIELD If radionuclides escape from the near-field into the groundwater of .the far-field, will retardation, dis- persion, and dilution in the bedrock be sufficient to keep concentrations acceptably low? General. Movement of radionuclides in the far-field of a Swedish repositoryâthat is, movement through granitic or metamorphic rock between the immediate vicinity of breached canisters and the biosphereâis one aspect of the general problem of how effectively natural rock environments can be expected to control nuclide migration. Debate on this subject has gone on for many years, and there is still no prospect of a definitive answer. The difficulty is that radionuclide movement depends on many variables, not all of which are easily measurable. Values for some variables can be obtained in laboratory experiments, but then questions arise as to how closely laboratory conditions simulate those in nature. Experiments in the field are often useful, but they are limited by the long times commonly required and by uncertainties about rock properties in the inaccessible parts of a rock mass. Research in this area is neces- sarily a multiyear and multicountry enterprise, and the KBS authors have been active participants since their project was started. Recent progress in Sweden has been especially notable along two general lines, dispersion phenomena and sorption, and the panel's review focuses primarily on these subjects.
47 It should be noted to begin with that one basic premise has changed. In KBS-2 the movement of radionuclides was assumed to take place from a repository at a depth of 500 m through partly filled fractures and crush zones to the land surface; hence the radionuclides would be subject to retardation by sorption both in the fracture fillings and in the rock matrix throughout the entire distance. The more conservative KBS-3 model pictures a repository shielded from major fractures zones on all sides by at least 100 m of relatively impermeable rock, and the only effects of dispersion and sorption considered are those that occur during movement through this 100-m thickness. Once in a fracture zone, radionuclides are assumed to have free access to the biosphere, where radionuclide concentrations are affected only by dilution or by chemical processes at the receptor location. Dispersion. In a porous medium, dispersion can be adequately described by a convective-diffusion equation, using a dispersion coefficient that is fixed for a given flow field (Bear, 1972; Anderson, 1979). In fractured rock, the dispersion phenomenon is much more complex. Over the past several years, Neretnieks and his co-workers have advanced considerably our quantitative understanding, making it possible to reproduce reasonably well with mathematical models many of the salient features observed in experiments with tracers in fractured rocks. One modification of the dispersion theory for porous media that must be made for application to fractured material is to allow for diffusion from fractures into the adjacent relatively unfractured rock, referred to in KBS-3 as the "matrix" or "rock mass" (Neretnieks, 1980). A second modification is to introduce "channeling," the variation in size of fractures that leads to faster flow in some than in others. Dissolved material is thereby dispersed because of the differences in velocity among channels as well as through other mechanisms, e.g., hydrodynamic dispersion. Neretnieks (TR 82-03, TR 83-69) has developed a model, called the "stratified-flow model," that takes both channeling and diffusion into rock adjacent to fractures into account. This is a new concept in the modeling of dispersion in fractured rocks; the only additional information required in this model, over and above that in standard dispersion models, is the size-frequency distribution of the fractures. One pre- diction from the stratified-flow model, i.e., that dispersion increases with distance of travel, has been
48 verified in field tests at both Finnsjon (TR 81-07) and Studsvik (TR ll0, TR 82-10). Other models can be used to calculate dispersion effects in fractured media, for example, a hydrodynamic dispersion model in which the dispersion coefficient is made to increase with spatial scale (thereby mimicking the observed behavior at FinnsjSn and Studsvik). This can be done by requiring that the Peclet number, rather than the dispersion coefficient, remain constant. Reason- able agreement with experimental results can be obtained with this model, using a wide range of values for the Peclet number. Neretnieks is aware, of course, that much additional work is needed to perfect the dispersion models for frac- tured rocks, but a good start has been made toward under- standing the complex phenomena involved. Sorption. The extent to which sorption can be expected to retard the movement of radionuclides dissolved in groundwater is particularly difficult to estimate. A common procedure is to measure sorption in the laboratory, either by letting a radionuclide solution stand in contact with crushed rock or by arranging a system of constant flow, and then to assume that the laboratory results will apply to rocks in-situ. The method is subject to criti- cism on several grounds: laboratory conditions may differ markedly from those in nature; the radionuclides in a natural environment may be in a different oxidation state or in the form of complexes with very different sorption properties; and the nuclides may be carried in groundwater as colloids, and for that reason may be more mobile than laboratory results would indicate. Despite these cogent objections, laboratory experiments seem the only feasible way to get rough estimates of the role of sorption in nature, and much effort has been concentrated on improving the experiments so as to counter the objections. In this effort the KBS scientists have played a major part, especially in the last few years. Basic laboratory data on sorption coefficients (Kd's) have been generated in abundance by Allard, Rydberg, and their colleagues at Chalmers University (e.g., TR 82-2l, TR 83-07, TR 83-61). A substantial part of the tabulated data now used worldwide comes from this laboratory, and of course the Swedish workers are continually in touch with similar work in progress elsewhere. The tabulated values for sorption coefficients in KBS-3 (Table 12-7) , changed slightly from those in KBS-2, are generally
49 similar to values in recent American publications (Moody, 1982; NRC, 1983) although somewhat larger for the actinide elements. In the KBS work, laboratory methods for determining sorption coefficients were refined long ago in the more obvious ways of ensuring similarity with natural environ- ments: use of groundwater samples in the experiments; excluding oxygen and carbon dioxide; control of Eh, pH, and temperature; and exposure of both fresh and altered mineral surfaces. Recent results from the continuing laboratory effort have established (1) that sorption of a given nuclide is a maximum in the pH range where neutral hydroxide complexes of the nuclide are dominant in solu- tion, (2) that the sorption of actinides in their lower oxidation states is independent of ionic strength over much of the pH range, and (3) that the amount of sorption on granitic rock can be approximated by a formula using Kd's for individual minerals (TR 82-21). A literature survey of work on organic complexes leads KBS scientists to conclude that increased mobility of radionuclides from this cause is not a serious problem in Sweden, because the content of organic matter in the groundwater is low and because many organic complexes are markedly sorbed on mineral surfaces (TR 83-09). The possible flooding of sorption sites by lead ions dissolved from the canister filling was considered in the review of KBS-2; if KBS-3 adopts the alternative of an all-copper canister, as seems likely from conversations with Swedish scientists, a flood of lead ions (and also possible galvanic reaction with copper) is no longer a concern. In addition to the continuing work on sorption coef- ficients and complexes of various sorts, much recent effort has been devoted to the question of the role of colloids, especially for the actinide elements (TR 83-08) . Experiments in the presence of air showed, for example, that Am and Pu form colloidal particles as the pH is raised from low values; but Np and U do not, presumably because they are complexed by carbonate. Precipitation of the Am colloid increased, as would be expected, with increasing ionic strength and increasing temperature. Comparison of solutions of colloidal and noncolloidal Am passed through a column of crushed granite showed little increased mobility for the colloid; evidently these colloidal particles were sorbed as readily as the dis- solved ions. Certainly not all the nagging questions about colloids and complexes are answered by the new experimental work, but a good start has been made.
50 In a different kind of experiment, the movement of radionuclides dissolved in artificial groundwater was followed along a single fissure in a cylindrical specimen cut from a drill core (TR 83-01). Three nuclides were used: Eu111, Npv, and Puiv. As would be expected from batch experiments, Np showed little retardation, Pu a great deal, and Eu intermediate values. K^'s calculated from the flow experiment were somewhat larger than those obtained from static batch experiments, but the agreement was fairly good. Clearly, it would be desirable to check laboratory results with sorption measurements made in the field, and in this difficult enterprise the KBS scientists have been especially active. At Studsvik, where solutions could be injected in three holes spaced at various distances (up to 30 m) from a central borehole that could be pumped, the movement of Sr permitted a calculation of K^ giving values in the range 4 to 6 m3/t, which checked well with numbers obtained from laboratory batch experiments using the same kind of rock (6 to 8 m3/t) (cubic meters per tonne, an SI unit numerically equivalent to milli- liters per gram) (TR 83-18). These are unusually low values for Sr, but the check is satisfying. Cs migrated so slowly that K<j could only be estimated as greater than 30âat least an order of magnitude greater than for Sr, in agreement with most laboratory results. A similar experiment with Sr at FinnsjBn showed less satisfactory agreement with laboratory values, probably because frac- tures here are lined with calcite, which has poor sorptive properties (TR 81-07). In the experimental facility at Stripa, a study is under way to follow details of the diffusion of solutes into the rock matrix as solutions move along a small fissure. Preliminary results show satisfactory agreement with diffusivities previously obtained in the laboratory (TR 82-08, TR 83-39). The above discussion is a brief and incomplete sampling of recent Swedish work on retardation. Despite these efforts, the proof is not completeâand probably never will beâthat laboratory data on retardation by sorption are adequate for completely reliable predictions of radio- nuclide movement from a breached canister to the bio- sphere. The number of variables in laboratory work, plus the much larger number in natural environments, is simply too great for complete control to be achievedâas is witnessed by the wide disparity in values often reported for experiments conducted under seemingly identical con- ditions. The KBS scientists, nevertheless, have succeeded
51 in gaining much understanding of the effects of the more important variables, such as pH, Eh, temperature, ionic strength, complexing, and colloidal behavior. In the panel's opinion, recent work has added greatly to this understanding and created increased confidence that the Swedish estimates of rates of radionuclide transport are soundly based. The estimates will be even better when some of the research in progress is completed, but the additional refinements are hardly needed as long as con- servative values are used in the calculations. Despite the reasons for skepticism alluded to above, the panel thinks that the KBS authors have developed a sound scien- tific basis for their conclusion that the barrier to radionuclide movement provided, to some extent, by the bentonite buffer, but mainly by the bedrock will be adequate insurance against unacceptable releases to the biosphere even in the very unlikely event of large-scale canister failure. Dilution. When contaminated water moving along a fissure zone from the repository comes in contact with the biosphere, for example when it flows into a well or lake, it will be greatly diluted by uncontaminated water from near-surface sources. The radiation dose to users of the water depends on the amount of dilution. In KBS-3 the contaminated water is assumed to be diluted 10,000 times before it is consumed by humans, but the basis for this assumption is questionable. According to Tbnis Papp (KBS, personal communication, 1983) , 104 was chosen as a "reasonably conservative" value after review of available information. Quantitative estimates of the dilution factor for wells range very widely, from 102 to 107. The lower part of this range (102 to 105) is advocated by Thunvik (TR 83-50) , who uses a finite-element model to calculate water flux through the repository and into a well for several combinations of boundary conditions and well-repository configurations. The model rests on the assumption that flow in fractured granite can be treated as porous-medium flow. The lowest figures from the model are about 102 for a well 200 m deep (much deeper than most domestic wells in Sweden) and 103 for a well 60 m deep. Other KBS authors (T6nis Papp, KBS, personal communication, 1983) think that Thunvik's assumptions for deriving the lowest factors (high withdrawal rate and little infiltra- tion from the land surface) are unrealistic. Carlsson (TR 83-45), using a different model, calculates factors
52 ranging from about 105 to 107, based on assumptions that other KBS authors regard as somewhat optimistic. The chosen figure, 104, is a compromise among these widely diverse estimates. The panel has little basis for judging the suitability of 104 as a dilution factor for the well scenario. It is indeed "reasonably conservative," but perhaps not conservative enough for a safety analysis that claims to be based on pessimistic choices throughout. There seems no reason to dismiss out-of-hand the assumptions behind Thunvik's lower estimates. If a factor of 103 were used instead of 104, calculated radiation doses would still be acceptably low except for the most extreme scenarios of early canister failure. The panel suggests that model-dependent uncertainties in estimating the factor may be reduced by incorporating fracture-flow considerations into the calculation of dilution at various receptor locations. In the case of a lake, dilution is determined by the lake volume and the turnover rate, with the assumption that all of the contaminated water from the repository enters the lake. Lakes with low inflows and turnover rates represent unfavorable conditions for dose cal- culations, and so Morpa Lake in the Fjallveden area was chosen for the scenario calculations (TR 83-49) ; for this lake a dilution factor of about 10^ was estimated. TUNNEL SHAFT, AND BOREHOLE SEALING Can tunnels, shafts, and boreholes be sealed effectively enough to keep movement of water no faster than through adjacent undisturbed rock? Shafts and tunnels used in constructing a repository, as well as boreholes that may have been drilled during preliminary exploration, must be filled and sealed when the repository is ready for closure, to prevent their becoming channels of easy groundwater flow that could bypass the normal slow movement through relatively impermeable rock to major fissure zones. Methods of filling and sealing described in KBS-3 are similar to those in KBS-2: boreholes will be plugged with bentonite, and tunnels and shafts with a bentonite-sand mixture; to prevent movement of fluids through the disturbed zone adjacent to tunnels and shafts produced by blasting, seals will be constructed at intervals by sawing slots through the disturbed zone into sound rock and filling the slots with compacted bentonite blocks.
53 The NRC subcommittee reviewing KBS-2 thought that the planned sealing procedures were a weak spot in the KBS proposal, especially since their practicability and effectiveness had not been demonstrated in the field. To some extent this deficiency has been remedied by recent work. Regarding emplacement of backfill, Pusch (TR 82-07 and 1983 personal communication, accompanied by recent photographs) reported that he and his colleagues, using an experimental tunnel in the granite at Stripa, have shown the feasibility of emplacing backfill by mech- anically compacting layers of sand-plus-bentonite on the floor and then filling the upper part of the tunnels by use of shotcrete equipment. The shotcreted material remained homogeneous and had a density only slightly less than that of the compacted layers below. An experiment is under way at Stripa to demonstrate the sealing of a tunnel by filling a slot sawed through the disturbed zone with compacted bentonite blocks and concrete (Roland Pusch, University of Lulea, personal communication, 1983) . To ensure the complete filling of boreholes with bentonite, Pusch (TR 81-09) has devised an ingenious method of placing cylinders of compacted bentonite in perforated copper tubing that is then inserted into the borehole. As the bentonite absorbs water, it swells through the perforations and ultimately fills the entire hole. The method has proved successful in both labora- tory and field experiments. In the field experiment (at Stripa), overcoring of the filled borehole and slicing of the extracted blocks for examination showed that the bentonite had expanded uniformly and formed a tight seal against the rock walls. The experiments are impressive, but whether they are sufficient to silence all doubts remains unclear. A recent American technical report (Meyer and Howard, 1983) on the use of clays for repository sealing emphasized the need for research into the long-term stability and even solubility of clays, their possible reactions with adjacent rock, and methods of emplacement so as to ensure a tight seal. Pusch and his colleagues at Lulea have made a good start at this kind of research, but addi- tional in-situ experiments would make the demonstration more convincing. In the panel's opinion, the KBS conclusion that open- ings into a repository can be adequately sealed has much better support than it did five years ago. Continuing research may take care of remaining uncertainties long before the actual sealing of a filled repository is
54 necessary. It should be kept in mind also that the adequacy of seals in a wet-rock repository is important only in the unlikely event of early failure of a large number of canisters. CALCULATIONS ON RELEASES TO THE BIOSPHERE Are the calculated radiation exposures to present and future populations based on adequate data and analyses? Evaluation of the feasibility of safe disposal of spent nuclear fuel depends ultimately on the calculated radiation doses to this and future generations from the radionuclides that may escape from a repository. The dose calculations for KBS-3 (TR 83-49) are made using a compartment model to simulate the various pathways by which radionuclides can move to human beings from assumed contaminated water sources. The KBS-3 authors make dose assessments primarily for two pessimistic receptor locations in waterâa well drawing from a contaminated fracture zone in bedrock and a lake with very slow turnoverâas well as for a special scenario where a lake eutrophies to become a peat bog that is used 10,000 years later as a soil conditioner. Concentrations in the water sources are estimated by using some of the models des- cribed in preceding sections, starting with dissolution of spent fuel pellets, tracing the movement of leached nuclides through the near-field and far-field, and finally postulating substantial dilution as contaminated ground- water approaches the places where water is obtained for human use. Different concentrations are obtained at the receptor locations depending on various postulated sequences of events ("scenarios") by which canister failure and exposure of fuel to groundwater might take placeâdifferent assumed times at which canisters are breached, different numbers of canisters affected, different chemistries of the invading and downstream groundwater. The worst-case scenarios described in TR 83-49 assume transport chemistries that lead to high solubility (oxidizing conditions throughoutâScenario C) or little sorption (colloid transportâScenario D). Even such unlikely radionuclide transport conditions produce radiation exposures that are at least 100-fold less than International Commission on Radiological Protection (ICRP) limits. If a single canister somehow fails completely within the first century after repository closure (Scenario B) ,
55 exposing all of its contained fuel to oxidizing ground- water, the derived doses are surprisingly small, roughly 100,000-fold less than ICRP limits. Even with a far more extreme assumption, involving the highly improbable simul- taneous early failure of all the canisters rather than a single one, the resulting calculated dose (4400 times the dose in Scenario B) to a maximally exposed individual would be approximately 4 mrem/yr (0.04 mSv/yr) (milli- sieverts per year; 1 mSv - 0.01 millirem). This level of exposure is about 4 percent of the ICRP limit for an average individual in a nearby population and 1 percent of the ICRP limit for the maximally exposed individual. Since the postulated sequence of events is hardly credible even as a worst-case scenario, this result shows that requirements of the "stipulation law" are certainly satisfiedâprovided that the calculations leading to the dose estimate are reliable. How secure is the basis for these calculations? Reliability of the calculations depends on (1) the validity of the models and (2) the assumptions made as to the values of parameters used in the models. The effort by the KBS-3 group in the development of models (espe- cially the fracture flow work of Neretnieks) is impres- sive. A complex box (compartment) model has been used to describe the dynamic processes that distribute both radioactive and nonradioactive substances throughout the biosphere, with simplifications tailored to the variables important to a specific site. The BIOPATH model (Bergstrbm and Rftjder, n.d.; TR 83-40, TR 83-28), selected for dose assessment in KBS-3, calculates internal and external radiation doses to indi- viduals living in four ecosystems, each of expanding breadth. Thus, the accumulation of radiation doses by an individual has been evaluated from sources local (e.g., well-water and/or crops), regional (e.g., meats, fish and dairy products), intermediate (e.g., shellfish from the Baltic), and remote (e.g., ocean fish) to his immediate environment. Collective (population) dose is calculated by summing the annual individual doses. BIOPATH is a system made up of a finite number of reservoirs (at this time, up to 20 compartments), each of which is homogeneous and well-mixed, and the compartments interact by exchanging elements. The dispersion rates and deposition sites are mainly governed by the calculated turnover rates of the elements and the compartment size chosen (either volumetric or gravimetric). The exchange of each element and corresponding radionuclides between
56 reservoirs is described in the model by first-order rate equations that depend on transfer coefficients expressed as turnover rates. All of the interactions between the roost important reservoirs that may contribute to radiation exposures in the near and distant future appear to have been adequately accounted for in BIOPATH. Compartment models of this type have been used extensively to evaluate biospheric transport of radionuclides produced in nuclear weapons tests and by natural processes. The parameters used in the BIOPATH calculations include nuclide-independent and nuclide-dependent transfer coef- ficients, retention and weathering fractions, terrestrial and aquatic yield values (productivity), and diet and consumption rates. Bergstrom provided an extensive eval- uation (TR 83-28) of these parameters and the values chosen for them. Where a range of values is available for any parameter, the geometric mean is generally selected as input, although pessimistic values were selected in a few instances where the range was extreme. The values for the parameters appear to be well-documented and to have been selected from an abundant source of references of international scope; we judge them to be sufficient for dose assessment. The BIOPATH code also makes use of dose conversion factors (in sieverts per becquerel (Sv/Bq)) as revised by Johansson (TR 82-14) from values originally developed by the ICRP (ICRP-30). For the most part, the revisions consist of estimating the dose commitment for a 70-year integration period (full lifetime) rather than a 50-year period (occupational lifetime) as used by ICRP. The change in integration period causes only a small increase (less than twofold) in the dose commitment for most radio- nuclides. For three radionuclides, 239Pu, 231Pa, and 237Np, new uptake data suggest a fivefold increase, an eightfold increase, and a ninefold decrease, respectively, from the ICRP-30 values. These departures from ICRP recommendations appear to be justified adequately by appropriate references. A sensitivity analysis has been performed (TR 81-03) for an earlier version of the BIOPATH code (TR-100). Thus, the panel believes that the KBS-3 authors have developed in the BIOPATH code a model that adequately represents the complex biospheric and dietary processes that contribute to radiation exposures. The panel found it difficult, however, to follow just how the KBS authors use each model and how the models and input data are fitted together to arrive at a dose esti-
57 mate. For example, to calculate dose for the postulated pathway in which water is obtained from a contaminated well, it appears from pages 92 to 93 of TR 83-45 that only a simple analytic solution to the pore-flow model is used to obtain a dilution factor as input to the calcula- tions in TR 83-49. Yet page 10 of the latter document states that the dilution factor comes from a different calculation in TR 83-50. The use of different models may not lead to great differences in calculated dose. How- ever, the uncertainty in the method of calculation is disconcerting, because it cannot be tested readily within the framework of the safety analysis as performed for the KBS-3 report. In response to the panel's request for clarification, Toms Papp (KBS, private communication, 1983) provided a calculation scheme specifying the model used at each step of the safety evaluation, as well as the points at which judgmental factors influenced the input data. That diagram is reproduced here (Figure 4-1) as a description of the panel's understanding of the actual sequence of calculations and data input used for the KBS-3 safety analysis for the well recipient. In the diagram, model calculations or analytic solutions are represented by rectangles with the appropriate supporting report refer- enced, whereas each rhombus represents a judgmental decision for which an explanation can be found in the referenced KBS-3 chapter, and input-output data are represented by arrows. As shown on Papp's diagram, the "hydromodel" of TR 83-45 is used to obtain an estimate of flux, U0, through the bedrock surrounding a repository. The single-value output from the hydromodel is used directly as input to the near and far-field calculations (TR 82-24, TR 83-48) to obtain an estimate of the equivalent water flow past each canister, Qeg. Although the Canister Corrosion calculations indicate canister breaching will not occur before 106 years and perhaps not until 108 years, the KBS authors pessimistically assume as input to the calculation of Fuel Dissolution that canister failure will begin at 105 years with the last canister failing at 106 years. As noted in the section in Chapter 4 entitled "Retardation in the Far-Field," evaluation of dilution factors for the well recipient ranging from a worst-case value of 102 (TR 83-50) to a highly optimistic value of 107 (TR 83-45) led to selection (rhombus indicating KBS-3, Chapter 15.2.1) of a dilution factor of 10,000 as input to
58 HYDRO- MODEL CALCULATIONS TR 83-46 O^ CALCULATION TB 82-24 DOSES FIGURE 4-1 Calculation sequences in KBS-3. Biosphere Calculations (TR 83-49) in combination with output from Far-Field Calculations. The safety analysis for a lake recipient is performed in an analogous manner, except for substitution of turnover rates in Morpa Lake (TR 83-52) to estimate a dilution factor for the lake recipient (KBS-3, Chapter 18.2.5). For parameter values, the KBS authors maintain that "pessimistic" or "conservative" assumptions are con- sistently used. That is, if a choice or a range of values is available, the number chosen is one that would lead to overestimation rather than underestimation of the resulting dose. In general, the choices do indeed tend to be pessimistic, but in a few instances the pessimism can be questioned. The aforementioned selection of a 10,000-fold dilution from the repository to a well is probably the most important one. As another example, the
59 main scenario assumes failure of the first canister 105 years after repository closure and failure of the others at regular intervals of 250 years thereafter, with fuel dissolution occurring over five million years; a higher rate of release may occur should many canisters fail at intervals grouped more closely around the time of average canister life (Chu et al., 1983), or should dissolution of the fuel occur at a greater rate (NRC, 1983). Also, the retardation factors used for KBS-3 (TR 83-48) are somewhat larger (therefore leading to smaller or more "optimistic" doses) than the factors used in some recent U.S. reports (NRC, 1983; Moody, 1982). Failure to use pessimistic values for every parameter, however, may not seriously affect the estimated doses, especially when the uncertainty in some of the values is very large. For each of the above examples, the KBS-3 authors have provided to the panel persuasive arguments in support of the parameters selected (Tonis Papp, KBS; Ivars Neretnieks, Royal Institute of Technology; Leif Carlsson, Swedish Geological; and B. Allard, Chalmers University of Technology; private communications, 1983). They state, for example, that according to studies by the Swedish Corrosion Institute (TR 83-24), canister failure should be assumed to occur at times 100- to 1000-fold longer than those assumed for the KBS-3 calculations, with a resultant spreading of the failure times (contrary to Chu et al.). Furthermore, based on equivalent water flow and groundwater composition, they believe that chemical solubility limits will restrict the fractional dissolution rate for the fuel to at most 10~^ per year, with more likely rates extending to 10~8 per year or less. The factors selected for radionuclide retardation are justified by the KBS-3 authors on the basis of extensive laboratory work, especially by Allard and Rydberg (TR 83-6l, TR 79-26). Even in the absence of bentonite as a buffer surrounding the canister, KBS estimates that doses would increase, at most, threefold to twenty-five-foldâvalues that meet ICRP limits. Finally, the conditions assumed by Thunvik to obtain the worst-case dilution factors (TR 83-50), are thought by KBS to be so improbable (especially the assumption that there would be a complete lack of infiltrating precipita- tion from ground surface to the well recipient) that pessimism beyond the assumption of 10,000-fold dilution is not warranted. One consequence of the insertion of arbitrary param- eters into the modeling procedure and the persistent
60 choice of pessimistic values for parameters is an inabil- ity to make systematic sensitivity and uncertainty analyses. Such analyses would be helpful in assessing which parameters and which variants of modeling methods have the greatest effect on calculated radiation expo- sures, and in calculating the range of uncertainty in the dose estimates. The panel requested clarification of the criterion that KBS would apply to a decision for sealing off a drift, or section of drift, should a major fracture be encountered after disposal operations begin. Papp (KBS, private communication, 1983) stated that KBS would wish to maintain at least the safety margin provided within Scenario C (approximately 100-fold less than the ICRP dose limit). Since dose to an individual is proportional to the amount (in becquerels per year) of radioactivity delivered to the geosphere (TR 83-49), it is also roughly proportional to Qgg (TR 83-48). Thus, using the dose obtained in Scenario C as a design objective, a large fracture must deliver 20 to 30 1/m2 yr to all of the 4400 canisters in the repository before it, or the drift it intercepts, would be considered a candidate for avoidance or sealing. However, only at most a few canisters are likely to be affected by such a large fracture, because canisters more than 100 m away from the fracture have no influence on the safety calculations, whereas those at lesser distances have decreasing influence with distance due to design procedures for repository closure. Therefore, it appears unlikely that one or two fractures with excessively large hydraulic conductivity will be inadvertently, or by choice, left in contact with sufficient buried waste to cause any concern as to the safety of the repository. The panel notes that the KBS-3 plan did not address the matter of canister recovery in the event of extremely unlikely catastrophic events. Despite these questions concerning the calculations, it appears that only under the most unlikely and extreme combination of worst-case conditions (e.g., improbably early canister failure, rapid fuel dissolution, poor radionuclide retardation, and a complete absence of surface water infiltration to a well) would the dose to the maximally exposed individual approach or exceed ICRP limits. The panel concludes that the KBS-3 authors have provided sufficient evidence that radiation doses to a maximally exposed individual, to an individual in a nearby populace, and to a large population (collective dose) will be acceptably small.