Uncertainties In Performance Assessments For The Yucca Mountain Site And
The “Edge-of-Compliance”- 8432

Kenneth Czyscinski 

U.S. Environmental Protection Agency, 

1310 L Street, Washington, D.C. 20005

Harry Chmelynski, William Thurber 

SC&A Inc., 

1609 Spring Hill Road, Suite 400, Vienna, Virginia, 22182

ABSTRACT

Numerical performance assessments of deep geologic disposal systems of
long-lived radioactive wastes serve as the primary tool for
quantitatively assessing projected performance and for making regulatory
compliance decisions, both in the U.S. and disposal programs abroad. 
However there is a significant body of opinion, expressed in the
international literature, that the confidence that can be placed on such
assessments decreases significantly as the analysis time frames increase
into the many hundreds of thousands of years.  With the inclusion of
highly corrosion resistant metals in the planned waste package designs
for the Yucca Mountain (YM) system commercial spent fuel and
defense-generated high-level radioactive waste disposal, published
projections for the site show doses to the receptor at times extending
into the many hundreds of thousands of years.  The question of relative
confidence in dose projections becomes more important at very long time
frames, particularly when the time frame for significant doses increases
dramatically reflecting the effects of using highly corrosion resistant
materials.

To examine uncertainties in performance assessments of the YM disposal
system over very long time frames, a site model developed by the U.S.
Department of Energy to assess sensitivities in peak dose performance
was modified and used to examine the propagation of uncertainties for a
hypothetical disposal system.  These analyses start with a hypothetical
disposal system at the “edge-of-compliance” at 10,000 years,
reflecting the generic repository standard in 40 CFR Part 191.  The
hypothetical system was poised to give a mean dose of  0.15
millisievert/year (15 mrem/yr.) at 10,000 years, by allowing a fixed
number of waste packages to “fail” within the first 5,000 years
after closure.  By maintaining the number failed waste packages constant
over time, the spread in dose estimates over the time to peak dose was
calculated for this system perched at the “edge-of-compliance” at
10,000 years.  This hypothetical construct removes waste package
performance from the analyses and allows the site model to explore the
effects of uncertainties in the natural barrier and site conditions on
the dose projections out to the time of peak dose.  The sensitivities of
these projections to various “driver” parameters were examined,
including infiltration rates, solubility constraints, water chemistry
and roof collapse assumptions.  

Overall, the initial construct showed a one and one-half order of
magnitude spread between the 5th and 95th percentiles at 10,000 years,
the initial state for the analyses.  This spread increased to
approximately three and one-half orders of magnitude at peak dose,
reflecting the effects of transport and retardation mechanisms on the
fixed source term. The model eliminated portions of the transport path
from the repository where relatively little retardation would be
expected.  Travel times were therefore significantly reduced thereby
eliminating radioactive decay as an important mechanism in reducing
projected doses.  Solubility controls appear to be a major source of
uncertainties in the dose projections for the model used. These results
support the general international consensus that confidence in dose
projections over very long time frames does decrease, and illustrates
that this conclusion also applies to assessments of the YM projected
performance as well.

INTRODUCTION

To assess potential health and safety impacts of a candidate disposal
system for used nuclear fuel and high-level radioactive waste, numerical
performance assessments are the only available quantitative tool
available to analyze the complex processes involved in the post-closure
performance period.  These assessments will be a prime focus of the
regulatory process.  The issue of uncertainty propagation in repository
performance assessments is an important component in making compliance
decisions and also in structuring national standards, particularly
tiered standards.  The “sharpness” of the performance assessment
tool determines how well alternate conceptualizations of the disposal
system can be distinguished from each other, and has a direct bearing on
the implementability of standards, particularly for tiered standards. 
To gain some insight on these issues, a site-specific modeling effort
was performed to examine the propagation of uncertainties for the Yucca
Mountain commercial spent fuel and defense-generated high-level
radioactive waste (defense waste) disposal system over extended time
periods.

STANDARD SETTING CHALLENGE

 	

National standards for the acceptable performance of a deep geologic
repository have been in place since 1985, as contained in Code of
Federal Regulations, Title 40, Part 191 (40 CFR Part 191).  The Energy
Policy Act of 1992 however directed that the National Academy of
Sciences (NAS) to prepare recommendations for the U.S. Environmental
Protection Agency (EPA) at the development of site-specific standards
for the safe management and disposal of used nuclear fuel and high-level
radioactive waste at the Yucca Mountain (YM) site.  The most important
finding by the NAS [1] was to state that they “believe that there is
no scientific basis for limiting the time period of the individual- risk
standard to 10,000 years or any other value” [1, pg 55].  They then
recommended individual protection standards be developed for the period
when risks were highest whenever it occurs, within a period of geologic
stability for the site, which could be on “the order of one million
years”.   The credibility of dose projections made for such extremely
long time periods raises the question of how much confidence can be
placed in the projections as reasonable forecasts of repository
performance.  

MODELING APPROACH

The model used for the assessments presented here was developed by the
U.S. Department of Energy ((DOE) for the purpose of doing sensitivity
studies for the most important parameters contributing to peak dose
projections for the YM site [2], referred to here as the DOE Peak Dose
Model (PDM).  It differs from the Total System Performance Assessment
(TSPA) models used for more detailed assessments, being less detailed,
but was built by the same extraction process as the various TSPA models.
 Parameter values for the approximately 100 variables in the model were
developed from numerous reports documenting the results of site
characterization studies at the site for many years, as documented in
the report cited above and within the DOE-PDM.  No changes were made to
this extensive data base to preserve the integrity of the DOE_PDM
site-specific data base.  Both models operate under the GoldSim
probabilistic simulation software (version 8.02) with the Contaminant
Transport Module.  

The EPA modified this model as necessary to perform its analyses.  The
modified model will be referred to here as the EPA Uncertainty Model
(EPA-UM).  A detailed description of these modifications relative to the
DOE-PDM capabilities is described elsewhere [3].  To establish the base
case reference system, iodine (I) and technetium (Tc) had to be added to
the radionuclides already in the DOE-PDM, since we were interested in
performance within 10,000 years.  We also added the consideration of
climate variation within 10,000 years, using the timing for various
climate states given in the DOE-PDM documentation (2).   In addition,
the EPA-UM was modified to allow selection of parameter values at fixed
levels (mean values for example) so that sensitivity studies could be
performed.  The relatively simple EPA-UM is used here for several
reasons.  It can make 1,000 realizations in the course less than two
hours, as opposed to the more elaborate TSPA models which would take
weeks to perform the same number of realizations.  We do not have the
resources or time to work with the more complex TSPA codes.  We were
interested in looking at the most important “driver” parameters in
terms of their effects on peak dose projections.  These are contained in
the DOE-PDM without the complicating effects of parameters that have
less impact on dose projections, which would make the results more
difficult to interpret.

To examine the propagation of uncertainty over very long time frames, a
base case reference provides a fixed point for contrast with the forward
modeling.  Without a fixed base case for reference, it is difficult to
understand what drives the long-term modeling since all the parameters
can vary.  We were interested in examining the performance of the
natural barrier system at the site rather than the engineered barrier
system (EBS) components and our approach takes much of the EBS out of
the assessments.  The base case consists of a hypothetical disposal
system with a predetermined number of failed waste packages, set so that
the mean dose to the receptor is 0.15 millisievert per year (mSv/yr) (15
millirem/yr.) at 10,000 years.  This hypothetical disposal system is
then poised at the “edge-of-compliance” for the 10,000 year
standard.  The modeling then addresses the question, “what would the
dose variations be in the very long-term for a disposal system that
performs at the 10,000 year limit”.  Since the calculations are done
stochastically, there is a spread of dose estimates at 10,000 years as
part of the initial conditions for the reference case.  This spread is
taken as the difference in dose values between the 5th and 95th
percentiles of the dose distribution.  The spread is limited to this
range to eliminate extreme high or low values that might arise from
statistically low-probability parameter values.  The number of failed
waste packages necessary was determined iteratively with the EPA-UM,
beginning with failure of all the packages and working backward to
derive a mean dose of  0.15 mSv/yr. (15 mrem/yr.), from 1,000
realizations with all the parameters varying stochastically as in the
base case presented in the DOE-PDM documentation [2].   One thousand
realizations were used in the analyses presented her to preserve
consistency with the DOE-PDM analyses.  The number of failed packages
needed to make the reference case was found to be 520 (divided into 362
commercial spent fuel packages and 158 defense waste packages in
proportion to the planned repository loading).  The packages were
allowed to fail uniformly over the first 5,000 years of the simulations.
 This was done because the travel times through the ground water travel
path is on the order of 5-6000 years [4].  We also examined the effect
of allowing all the packages to fail at 5,000 years, but the results
were not markedly different. 

MODELING RESULTS - EPA-UM BASE CASE

The dose projections for the EPA-UM base case (“edge-of-compliance”
disposal system) are shown on Fig. 1, for runs consisting of 1,000
realizations, with all parameters varying stochastically and a fixed
number of failed waste packages.  For the 10,000 - year time line, the
dose estimates vary approximately one and one-half orders of magnitude
(difference between the 5th and 95th percentiles of the distribution)
around the mean value of 0.15mSv/yr. (15 mrem/yr.).  As the dose
projections proceed out to the peak dose, the distribution widens to
approximately three and one-half orders of magnitude surrounding the
mean value of 3.42 mSv/yr. (342 mrem/yr.).  This illustrates an
important point that the uncertainties in projecting the doses at longer
time frames increase, and gives an indication of the range of potential
doses that would occur if the disposal system were functioning at the
edge of compliance at 10,000 years.  The dose projections did not stay
in the range of tenths of a mSv/yr. (tens of mrem/yr.) at longer time
frames and did not go into the range of many mSv/yr,  (hundreds to
thousands of mrem/yr.), illustrating that the operative portions of the
disposal system functioned to keep doses relatively constrained.  These
results should not be interpreted to suggest that the actual Yucca
Mountain disposal system would yield a closely similar dose history
profile.  For the actual repository, waste packages would continue to
fail over time contributing to the inventory of radionuclides in the
repository available for removal by infiltrating ground waters and
subsequent migration from the repository. 

SENSITIVITY STUDIES WITH THE EPA-UNCERTAINTY MODEL

To examine the effect of various choices in executing modeling runs,
these calculations were repeated for differing numbers of realizations
since a decision was made initially to retain the 1,000 realizations per
model run for consistency with the DOE-PDM analyses [2].  Results with
higher numbers of realizations gave higher mean values at peak dose but
were within the 95% confidence interval of the 1,000 realization run. 
These results further illustrate the uncertainty in projecting long-term
performance even when the initial conditions of the disposal system are
tightly constrained, i.e., the failed waste packages are fixed in
number.

Sensitivity analyses were also performed to examine the effects of fixed
vs. random waste package failure times, fixed versus variable parameter
variations and Monte Carlo vs. Latin hypercube sampling for the
parameter distributions.  These variations are simply modeling choices
for the analysis of the reference base case scenario and do not measure
the effects of the site parameters, which are the main point of interest
for the sensitivity analyses.  Rather, they reflect the effects of
fundamental aspects of performance assessment in a broader sense.  The
results of these exercises are given in Table I.  Mean doses varied from
a low of 1.62 mSv/yr. (162 mrem/yr.) to a high of 4.08 mSv/yr. (408
mrem/yr.).  The low-end value corresponds to a single dose projection
within the reference base case (d in Table I).  The timing of waste
package failures made essentially no difference in the mean dose
projections (c and d in Table I), indicating that the site parameter
variations are acting in a similar way in both alternatives.  

Fig. 1.  Total Annual Dose, Mean and Selected Percentiles of
1,000 Realizations for EPA Uncertainty Model

Sensitivity studies with the EPA-UM were also performed to examine the
effects of variations in infiltration, seepage (the amounts of water
actually entering the emplacement drifts), and solubility variations
(varied by allowing pH and pCO2 to vary).  These sensitivity runs were
done by setting the stochastic parameters to fixed values and allowing
the parameters for the individual processes under consideration to vary
across their allowable range in the data base.  To interpret these
results, the mean value of dose at 10,000 years was noted along with the
spread of the 5th and 95th percentiles for these analyses.  These values
are different than the reference base case since the reference case was
calculated allowing all the parameters to vary stochastically whereas
for the sensitivity studies most parameters were fixed (see below). 
Results of these analyses are summarized in Table II.  The calculated
mean, and median values are shown, along with the values for the 5th and
95th percentiles to illustrate the variations observed.  Table II also
shows the results for variations of all the parameters to illustrate the
variations observed for the reference base case with the EPA-UM.

For infiltration sensitivity, the infiltration rate was varied from
highest to lowest rather than the blended treatment used to develop the
reference base case.  Results showed, not unexpectedly, that higher
infiltration caused the peak dose to occur earlier in time than in the
base case.  The lowest infiltration produced a mean peak dose of only
0.2 mSv/yr. (20 mrem/yr.). Higher variations in dose were calculated
with the EPA-UM in contrast to sensitivity runs presented by DOE for the
DOE-PDM results [2], probably due to the addition of I and Tc to the
EPA-UM. and the fact that these radionuclides are quickly removed from
the waste packages in the EPA-UM reference base case scenario.

Table I   Mean and Median Annual Dose Forecasts at 10,000 Years and at
Year of Peak Mean Dose in mSv/yr. (mrem/yr)

	Forecast of Annual Dose at 10,000 Years	Forecast of Peak

Annual Dose(1)

Model Specification	Mean	Median	Mean	Median

a) Fixed Parameters with fixed WP failure time (n=1)	0.05 (4.83)	0.05
(4.83)	1.63 (162.83)

@100,000 yr	1.63 (162.83)

@100,000 yr

b) Fixed Parameters with Random WP failure times (n=1000)	0.05 (4.79)
0.05 (4.78)	1.63 (162.65) @100,000 yr	1.63 (162.83)

@100,000 yr

c) Random Parameters with Fixed WP failure time (n=1000)	0.15 (14.7)
0.03 (3.03)	3.41 (341.24) @60,000 yr	0.27 (26.78)

@112,000 yr

d) Random Parameters with Random WP failure times (n=1000)
0.15{0.09,0.22}

(15.0,{8.5, 21.6})(2)	0.03 (2.98)	3.42{2.69, 4.15} 

(342.20,

{269, 415})(1) @60,000 yr	0.27 (26.59)

@112,000 yr

e) All Random Parameters with no LHS Sampling (n=1000)	0.20 (19.7)	0.03
(3.17)	4.08 (408.14)

@52,000 yr	0.31 (31.12)

@116,000 yr

(1) 	The model uses 2,000-year time steps from 10,000 to 52,000 years
and 4,000-year time steps from 52,000 to 1,000,000 years.

(2) 	An approximate 95% confidence interval for the estimated mean is
shown in parentheses for n = 1,000 realizations.

 

For seepage sensitivity studies, four parameters are varied to handle
the variation in seepage rates in the site database while the other
stochastic parameters are held constant.  Results showed that seepage
has, not unexpectedly also, a strong influence on peak dose projections.
 As more ground water enters the drifts, higher amounts of radionucides
would migrate out into the natural barrier.  In addition to the seepage
variation runs, an additional run was performed utilizing the model’s
capability to simulate collapsed and non-collapsed drifts [2].  For the
case where the drifts are not collapsed, there is a marked decrease in
peak dose, with the peak doses of less than one tenth of a mSv/yr (only
a few mrem/yr.).  Some realizations for this scenario show no releases
within one million years.

For solubility sensitivity runs, the pH and pCO2 in the model data base
were varied.  The model assumes solubility of the radionuclides (with
the exception of I and Tc) are controlled by the appropriate
thermodynamically stable phases for the repository ground water
chemistry.  For the actinides, these are carbonate and hydroxycarbonate
phases.  Solubility variations produced the largest variations in peak
dose projections, but not dramatically different than the seepage
variations.   

Table II	Statistics Measuring Uncertainty in Various Sensitivity Cases

[ Doses in mSv/yr. (mrem/yr.)]

Statistic	Vary Solubility	Vary Seepage	Vary Seepage, Solubility, and
Infiltration Rate	Vary All Parameters

5th percentile	Peak Dose	0.14(13.7)	0.07(6.5)	0.02(2.4)	0.01(1.0)

	Year of Peak Dose	96,000	144,000	144,000	22,000

Mean	Peak Dose	4.42(442)	1.83(183)	4.91(491)	3.42(342)

	Year of Peak Dose	92,000	92,000	80,000	60,000

Median	Peak Dose	1.61(161)	1.60(160)	0.87(87.1)	0.27(26.6)

	Year of Peak Dose	100,000	104,000	112,000	112,000

95th percentile	Peak Dose	22.82(2,288)	4.85(485)	30.89(3,089)
19.72(1,972)

	Year of Peak Dose	84,000	84,000	80,000	64,000

5th percentile.  (At year of peak of 95th percentile)	12.6	3.0	1.3	0.6

Range Ratio (95th/5th)	181	162	2,387	3,049

Orders of Magnitude Spread in Range Ratio a	2.3	2.2	3.4	3.5

 a - log base 10 of range ratio

The case where seepage parameters, solubility parameters, and
infiltration rates are allowed to vary (Column 4) versus the case where
all parameters are allowed to vary (Column 5) shows a slightly lower
Range Ratio for the former.  However, the 5th percentile, mean, and 95th
percentile are lower for the case where all parameters are allowed to
vary.  The fact that the two cases have similar range ratios indicates
that most of the uncertainty is captured by these three groups of
parameters.  It must be remembered that, with the EPA Uncertainty Model,
uncertainties associated with waste package and drip shield behavior are
eliminated. 

By comparing the range ratio (95th/5th) data at the time of the peak
dose from Table II with the equivalent range ratios at 10,000 years, one
can obtain an estimate of the extent that uncertainty increases with
time.  The following tabulation provides such a comparison:

	Sensitivity Case		Range Ratio at Peak/Range Ratio at 10,000 yrs.

	Vary Solubility				181/3.35 = 54.0

	Vary Seepage					162/2.32 = 23.3

	Vary Seepage, Solubility & Infiltration	2,387/7.52 = 317

	Vary All Parameters				3,049/34.0 = 89.7

The increase in the range ratio over the period from 10,000 years to the
time of the peak dose clearly demonstrates the large temporal increase
in uncertainty.  For the case where all parameters are varied, the
increase is about two orders of magnitude from 10,000 years to the time
of the peak dose. The increase in the range ratio over the period from
10,000 years to the time of the peak dose clearly demonstrates the large
temporal increase in uncertainty (the approximate range shown in Fig.1).
 

In the EPA-UM sensitivity runs, 22 parameters were varied to assess the
effects of infiltration, seepage and solubility on the dose projections.
 A step-wise regression analysis was performed (using the SPSS package)
to determine the contribution of these parameters to the total variation
seen in the results.  The results of this analysis are shown in Fig. 2. 
 

Fig. 2.  Increase in R-Squared Achieved by Stepwise Addition of the 22
Selected Parameters in Regression of Annual Dose Maxima versus All 76
Stochastic Parameters Used in EPA Uncertainty Model

(N=1000 realizations, R2 = 0.801)

The largest contributor to the variation is the solubility data for
plutonium, followed by parameters that determine seepage into the
emplacement drifts and infiltration rate [3].   These results correlate
with the results of DOE’s analyses of peak dose sensitivity [2],
indicating that our modifications to the DOE-PDM did not alter the
fundamental structure or functions contained in the model.

SENSITIVITY ANALYSES WITH THE DOE-PDM

The EPA analyses presented here removed the containment function of the
metal barriers in the EBS.  To create the reference base case, a fixed
number of waste packages were allowed to fail within 10,000 years, the
drip shields were removed and no containment credit was taken for spent
fuel cladding.  The results then reflect only the effects of site
parameters on the dose projections.  However, the performance of the EBS
components is in reality an important contributor to the disposal system
performance.  To examine the contribution of the waste package and drip
shields, the DOE-PDM was exercised to examine this aspect.  Twelve
scenarios were examined using the “switches” provided in the model
to activate or deactivate various components of the model.  The results
are shown in Table III, and Fig. 3.

Table III.  Peak Mean Dose and Selected Percentiles in Year of Peak Dose
for Base Case and 12 Alternative Scenarios – Doses in mSv/yr.
(mrem/yr.)





ID	

Scenario	Peak Dose	Year of Peak Dose	5th

	25th

	50th

	75th

	95th

	Ratio

95th/50th	Ratio

Mean/75th

0	Base Case	1.25 (125)	730,000	0	0	0.05 (5)	1.26 (126)	5.70 (570)

1.0

1	Low Infiltration Case	0.8 (84)	870,000	0	0	0.31 (31)	1.04 (1.04)	3.39
(339)	11	0.8

2	Medium Infiltration Case	1.36 (136)	690,000	0	0	0	1.24 (124)	6.39
(6.39)	-	1.1

3	High Infiltration Case	1.62 (162)	690,000	0	0	0	1.51 (151)	7.70 (770)
-	1.1

4	WP Corrosion Rate times 5	2.53 (253)	225,000	0	0	0	1.30 (130)	11.44
(1144)	-	1.9

5	Full Temperature Dependence	0.03 (3)(a)	1,000,000	0	0	0	0	0	-	-

6	DS Corrosion Rate times 5	1.51 (151)	690,000	0	0	0.07 (7)	1.56 (156)
6.64 (664)	92	1.0

7	Remove DS Functionality	1.54 (154)	690,000	0	0	0.08 (8)	1.57 (156)
6.74 (674)	80	1.0

8	2nd Phase Np Solubility Control	1.11 (1.11)	865,000	0	0	0.36 (36)	1.26
(126)	4.84 (484)	14	0.9

9	SZ Transport Length Set to 0	24.97 (2497)	775,000	0	0	2.88 (288)	27.19
(2719)	111.39 (11,139)	39	0.9

10	Non-Collapsed Drifts	0.37 (37)	860,000	0	0	0.01 (0.7)	0.28 (28)	1.83
(183)	250	1.3

11	WP & DS Corrosion Rate times 5	7.68 (768)	180,000	0	0.1	1.06 (106)
4.86 (486)	38.19 (3819)	36	1.6

12	Vary Pu242 BDCF +20%	1.38 (1.38)	730,000	0	0	0.05 (5)	1.40 (140)	6.48
(648)	133	1.0

(a)  A higher peak is reached after the end of the one-million-year
timeframe used in the current model.

Fig.  3.  Comparison of 12 Peak Dose Model Sensitivity Scenarios with
DOE-PDM Base Case (#0, Table III)

The results of particular interest are scenarios # 4, 5 & 6 (Table III),
which involve differing assumptions about the performance of the waste
package metal and drip shields.  For the case where very low corrosion
rates are assumed (# 5), which track the thermal profile of the
repository over time, failure of a large portion of the waste packages
is delayed past one million years.  For the base case scenario (#0),
where higher corrosion rates are assumed, the peak dose occurs in the
range of 700,000 - 800,000 years.  These results illustrate the
overwhelming influence of the corrosion resistance of the waste packages
to total system performance.  

Corrosion rates are measured by laboratory testing of relatively short
duration in comparison with the in-service performance period for the
repository, stretching into the hundreds of thousands of years.  The
extrapolation of the laboratory data to such extremely long time frames
is at best optimistic and assumes that all possible corrosion mechanisms
and their rates under changing repository conditions can be quantified
confidently.   Confirming these assumptions, as well as confirming the
performance of the waste package metals in their full-size waste package
configurations, is not possible in any real sense because of the
extremely long time frames involved.  Experience in the industrial
sector on the long-term performance of the corrosion resistant alloy
used in the waste packages is also not available.  If a more skeptical
approach is adopted toward such extrapolations, it may be within reason
to ask what the performance might be if higher corrosion rates then
those used in the base case and full-temperature dependence scenarios,
were assumed.  This possibility is examined in scenarios # 4, 6 and 11,
and shown in Fig. 4.

Figure 4.  Mean Annual Dose, Base Case and 12 Alternative Sensitivity
Scenarios

 

The striking shift of the peak dose forward in time from
700,000-800,000years to under 200,000   years illustrates the
significance of this variable.  The peak dose also increases from the
base case level of 1.25 mSv/yr (125mrem/yr.) to almost 8.00 mSv/yr.
(800mrem/yr.).  

While the time to peak dose predicted by the modeling should not be
taken as a realistic projection because some elements of the site system
were eliminated, the relative shift in time of peak dose from over one
million years to between 100,000 -200,000 years illustrates the
importance of corrosion rate assumptions on long-term performance.  The
omission in the models of portions of the travel path from the
repository to the down gradient controlled zone boundary (approx. 18 km)
only contribute to the ground water travel time by amounts in the tens
of thousands of years.  Their inclusion in the models would not lengthen
the time to peak dose back to in excess of one million years.  Using
corrosion resistant metals in the EBS design greatly extends the
containment time for disposal system.  The confidence that can be placed
in very long-term dose projections is strongly tied to the level of
confidence that can be placed in the extrapolation of laboratory
corrosion testing results to in-service assessments for the metallic
barriers in the EBS.   The most recent published performance assessments
for the  YM site published by DOE [5] used the temperature dependent
assumption for the selection of corrosion rates and calculated results
were similar to those for scenario #5 in table III.

Two other significant results can be seen in the DOE-PDM analyses above.
For the scenario where limited drift collapse occurs, peak doses are
very low (# 10 in Table III), consistent with the results of the EPA-UM
results, illustrating again that if little ground water seeps into the
drifts radionuclide transport is low.  Another interesting result is the
major increase in peak dose for the scenario where the saturated zone
(SZ) is eliminated from the model.  This scenario corresponds to
delivering the contaminated ground waters from directly below the floor
of the repository to the receptor’s drinking water well.  Although
this is an unrealistic scenario, it illustrates the containment and
isolation contribution of the natural barrier surrounding the repository
to the total system performance.

INSIGHTS RELATIVE TO STANDARD DEVELOPMENT

From the results of these modeling efforts, some insights pertinent to
standard development become evident.  For the highly corrosion resistant
metals in the EBS design, releases within the 10,000 year period under
undisturbed conditions will be extremely low because of the limited
amounts of ground water able to enter the emplacement drifts (UZ setting
and limited drift collapse effects) and contact the waste packages.  The
protection offered by drip shields and the waste package metals delays
the significant release of radionuclides well past the 10,000- year time
line.  While the 10,000-year standard is aimed at providing protection
for that time period, a peak dose limit beyond 10,000 years would extend
the protection to even longer time frames.  From the results of the
modeling, it appears that the period of very low doses could extend from
100,000-200,000 to over a million years as a function of the corrosion
rates assumed in performance assessments.  Other assumptions for site
parameters and processes would amplify or reduce peak dose projections
for the more complex TSPA models, but the larger trends in dose
projections are evident in the modeling presented here.  Peak dose
limits remaining in the low  single digit mSv/yr. range (low hundreds of
mrem/yr.) would suggest that the period of very low doses would extend
easily into the time frame of many hundreds of thousands of years to
potentially in excess of one million years.  Establishing a peak dose
standard in addition to the 10,000 year standard increases the period of
very low doses to time frames well beyond that involved in 40 CFR Part
191.

The increasing uncertainties in performance assessments as time frames
increase dramatically also affect the development of any form of tiered
standards. The ability of the assessment tool to distinguish between
alternative performance scenarios plays a significant role in framing
dose limits and time frames for any form of tiered standards.   Setting
dose limits for performance periods ultimately is a societal decision
about what exposure levels are acceptable, but the limits of the
assessment tool determines the regulatory challenge for the applicant in
making a credible safety case and the regulatory decision maker to come
to a compliance decision.

CONCLUSIONS

Results of peak dose modeling using both the DOE-PDM and the EPA-UM
provide some insights and implications for understanding the performance
of the Yucca Mountain disposal system at the time of peak dose and the
behavior of important uncertainties through the geologic stability
period.  By using a base case where some components of the disposal
system are eliminated, uncertainty in projecting doses could be
examined.

Removing EBS metallic components from the EPA-UM analyses removes a
major source of uncertainties and allows the uncertainties in site
parameters and processes to be examined 

 Modeling the disposal system with the EPA-UM showed that uncertainty
increases over time relative to a fixed reference base case under the
site conditions of the Yucca Mountain disposal system.  These results
support a general intuitive conclusion that uncertainties in projecting
dose for the complex natural barrier system should increase as time
frames for the projections extend into the many tens to hundreds of
thousands of years.   

The most important parameters affecting dose projections are those
involved with estimating ground water movement through the repository
(infiltration and seepage), radionuclide mobility (release from the
waste packages, solubility and ground water chemistry) and stability of
the emplacement drifts to collapse after the thermal period has passed.

Corrosion rates assumed for long-term assessments have the most dramatic
effects on dose projections, both in contributing to the time frame and
magnitude of peak dose estimates.  Confidence in corrosion rate
assumptions and dose assessments leans heavily on the extrapolation of
laboratory corrosion rates over time and scale.

Setting a peak dose limit in the range within a few mSv/yr. at most (low
hundreds of mrem/yr.) would assure that the period of very low doses for
the Yucca Mountain disposal system is extended in time significantly
beyond the 10,000 - year period embodied in 40 CFR Part 191.

REFERENCES

National Academy of Sciences, “Technical Bases for Yucca Mountain
Standards”.  Committee on Technical Bases for Yucca Mountain
Standards, Board of Radioactive Waste Management, Commission on
Geosciences, Environment and Resources, National Research Council,
National Academy of Sciences Press, Washington, DC., 1995.

2.	Office of Civilian Radioactive Waste Management, “Management and
Technical Support Peak Sensitivity Analyses, Las Vegas Nevada: Office of
Civilian Radioactive Waste Management”, November 2005, EPA Yucca
Mountain Docket, OAR-2005-0083 # 0352.

3.	SC & A, Inc., “Support to the Revision of 40 CFR Part 197, task 4
– Modeling Uncertainty Effects on a Reference Dose Level” EPA Yucca
Mountain Docket, OAR-2005-0083 -# 0385.

A  Al. EDDERBARH,  “What is the Mean and Variance Transport Time of a
Conservative Species in the SZ”,  handout for presentation to the
Nuclear Waste Technical Review Board meeting, January 30-31, 2001,
Amargosa Valley, Nevada.

Office of Civilian Radioactive Waste Management, U.S. Department of
Energy, “Draft Supplemental Environmental Impact Statement for a
Geologic Repository for the Disposal of Spent Nuclear Fuel and
High-Level radioactive Waste at Yucca Mountain, Nye County, Nevada”,
DOE/EIS-0250F-SID, October 2007.

WM2008 Conference, February 24-28, 2008, Phoenix, AZ

