DRC-2018-003164 - 0901a068807df804

RECEIVED
APR 0 2 2018

ENERGYSOLUTIONS

III- ENVIRONMENTAL QUALM'
Div of Waste Managernent
and Radiation Control
APR 0 2 2018

April 2, 2018 CD18-0036
Mr. Scott T. Anderson D R C - ZO t 6 - 0031 6 1H
Director
Utah Division of Waste Management and Radiation Control
195 North 1950 West
Salt Lake City, Utah 84114-4880
Subject: Radioactive Material License UT2300249: Responses to the Amended and
New Interrogatories Related to Clive DU PA Modeling Report Version 1.4
Dated November 2015
Dear Mr. Anderson:
EnergySolutions herein submits responses to the Utah Division of Waste Management
and Radiation Control's Amended and New Interrogatories Related to Clive Depleted
Uranium Performance Assessment Modeligg Report Version 1.4 from November 2015
(dated May 11, 2017). With submission of this final set of responses, EnergySolutions
understands that "the Division intends to move forward with a second version of the
[Safety Evaluation Report] for public comment, after which the Director will be prepared
to make a determination." EnergySolutions herein repeats its request made with the 2015
submission of version 1.4 of the Depleted Uranium Performance Assessment,
"Prior to releasing the next [Safety Evaluation Report] draft, EnergySolutions
requests that the comments contained in [Clive Depleted Uranium Modeling
Report Version 1.4] be addressed and that they have opportunity to consult with
the Division on any further issues identified"2
1 Anderson, Scott. "Depleted Uranium Performance Assessment (DUPA) Clive Facility; Model Version
1.4 interrogatories; Radioactive Material License 2300249" Letter from the Division of Waste
Management and Radiation Control to Vern Rogers of EnergySolutions, May 11, 2017.
2 Rogers, Vern. "Radioactive Material License UT2300249: Safety Evaluation Report for Condition 35.B
Performance Assessment; Response to Issues Raised in the April 2015 Draft Safety Evaluation Report
(CD16-0266)." Letter from EnergySolutions to Mr. Scott Anderson of the Utah Division of Waste
Management and Radiation Control, November 25, 2015.
299 South Main Street, Suite 1700 • Salt Lake City, Utah 84111
[phone] • Fax: [phone] • www.energysolutions.com
ENERGYSOLUTIONS
Mr. Scott T. Anderson
CD18-0036
April 2, 2018
Page 2 of 3
As originally reported with EnergySolutions ' submission of version 1.4 of the Depleted
Uranium Performance Assessment, the five categories of deficiencies with the 2015
Safety Evaluation Report included,
1) The April 2015 draft Safety Evaluation Report goes beyond the scope
outlined by the Nuclear Regulatory Commission (NRC) and inappropriately
provides policy and legal direction to the Division Director.
2) The April 2015 draft Safety Evaluation Report is in conflict with Utah Code
Section 19-3-104(7)(a) and Utah Code Ann. 19-5-102(5) as it applies a more
stringent standard than R313-25-9(5)(a) requires. Additionally R313-25-
9(5)(a) is "more stringent" than the draft corresponding federal regulation
10 C.F.R. 61.13(e).
3) The April 2015 draft Safety Evaluation Report inappropriately addresses
issues with general data rather than using available site-specific analysis.
4) The April 2015 draft Safety Evaluation Report includes several factual
errors that require correction.
5) The "Unresolved Issues" in the April 2015 draft Safety Evaluation Report
have been addressed in version 1.4 of the DU PA Modeling Report.
EnergySolutions recognizes that the 2015 Safety Evaluation Report's first condition
prerequisite to any Director approval requires an enforceable agreement be executed
between EnergySolutions, the State of Utah and the U.S. Department of Energy for
transfer, after decommissioning, of ownership of that portion of EnergySolutions ' Clive
facility on which depleted uranium has been land disposed. Legal justification for this
requirement was provided in a letter dated August 9, 2017.3 The justification questions
the applicability of EnergySolutions ' existing land ownership exemption gyanted by the
former Radiation Control Board to disposal of depleted uranium in significant quantities.
Holland and Hart's response to the Division's justification, prepared on behalf of
Energysolutions ', has been submitted separately.4
3 Anderson, Scott. "Response to Question Regarding Tri-Party Agreement for Depleted Uranium
Disposal." Letter from the Division of Waste Management and Radiation Control to Vern Rogers of
EnergySolutions, August 9, 2017.
4 Smith, Amanda. "Tri-Party Agreement Requirement Response." Letter from Holland and Hart to Scott
Anderson of the Division of Waste Division of Waste Management and Radiation Control, April 2, 2018.
299 South Main Street, Suite 1700 • Salt Lake City, Utah 84111
[phone] • Fax: [phone] • www.energysolutions.com
________,----•-----.--,:-
EN ERGYSOLUTIONS
Mr. Scott T. Anderson
CD18-0036
April 2, 2018
Page 3 of 3
Thank you for your thoughtful review of these responses. Please let me know when the
Division is available to discuss preparation of the second revision to the Safety
Evaluation Report.
Sincerely,
Vern C. Rogers tit e clro Mar 23 2018 9:56 AM
Sian
Vern C. Rogers
Manager, Compliance and Permitting
CC Don Verbica, DWMRC
Helge Galbert, DWMRC
enclosure
I certify under penalty of law that this document and all attachments were prepared under my direction or supervision in accordance with a system designed
to assure that qualified personnel properly gather and evaluate the information submitted. Based on my inquiry of the person or persons who manage the
system, or those persons directly responsible for gathermg the information, the information submitted is, to the best of my knowledge and behef, true,
accurate, and complete I am aware that there are significant penalties for submitting false information, including the possibility of fine and imprisonment
for knowing violations
299 South Main Street, Suite 1700 • Salt Lake City, Utah 84111
[phone] • Fax: [phone] • www.energysolutions.com

RML UT2300249 – CONDITION 35.B: RESPONSES TO
INTERROGATORIES RAISED WITH VERSION 1.4 OF THE
DEPLETED URANIUM PERFORMANCE ASSESSMENT

MARCH 23, 2018

NAC-0115_R0

Introduction to DU PA Model
Version 1.4 Interrogatory
Responses
23 February 2018

Prepared by

NEPTUNE AND COMPANY, INC.
1435 Garrison St, Suite 201, Lakewood, CO 80215
Introduction to DU PA Model Version 1.4 Interrogatory Responses
23 Feb 2018 ii
1. Title: Introduction to DU PA Model Version 1.4 Interrogatory Responses
2. Filename: Introduction to DU PA Model Version 1.4 Interrogatory Responses.docx
3. Description: Introductory document providing context for seven associated topic-based
response documents.

Name Date
4. Originator Sean McCandless 12 February 2018
5. Reviewer Mike Sully 12 February 2018
6. Remarks

Introduction to DU PA Model Version 1.4 Interrogatory Responses
23 Feb 2018 iii
CONTENTS
CONTENTS ............................................................................................................................. iii
ACRONYMS AND ABBREVIATIONS ................................................................................... iv
1.0 Introduction ........................................................................................................................ 1
2.0 Introduction to Probabilistic Modeling ............................................................................... 2
3.0 References .......................................................................................................................... 4

Introduction to DU PA Model Version 1.4 Interrogatory Responses
23 Feb 2018 iv
ACRONYMS AND ABBREVIATIONS
CSM Conceptual Site Model
DEQ (Utah) Department of Environmental Quality
DU depleted uranium
ET evapotranspiration
GIGO garbage in, garbage out
LLRW low-level radioactive waste
PA performance assessment
PPA probabilistic performance assessment
SER Safety Evaluation Report
UDEQ Utah Department of Environmental Quality

Introduction to DU PA Model Version 1.4 Interrogatory Responses
23 Feb 2018 1
1.0 Introduction
Beginning in 2009, EnergySolutions contracted Neptune and Company, Inc. (Neptune) to create
a probabilistic performance assessment (PPA) for the disposal of large quantities of depleted
uranium (DU) at their Clive, Utah low-level radioactive waste (LLRW) disposal facility.
The initial model was submitted as version 1.0 on June 1, 2011 (Neptune 2011) and was revised
to version 1.2 on June 5, 2014 (Neptune 2014). A Safety Evaluation Report (SER) based on
review of version 1.2 was issued by the Utah Department of Environmental Quality (UDEQ) in
April 2015 (SC&A 2015).
On November 25, 2015, EnergySolutions submitted Radioactive Material License UT2300249:
Safety Evaluation Report for Condition 35.B Performance Assessment; Response to Issues
Raised in the April 2015 Draft Safety Evaluation Report (EnergySolutions 2015). This document
included version 1.4 of the DU PA (Neptune 2015), prepared in response to open primary and
supplemental interrogatories included in Appendix C and Appendix B, respectively, of the SER.
On May 11, 2017, UDEQ provided Amended and New Interrogatories Related to Clive DU PA
Modeling Report Version 1.4 Dated November 2015 (Utah DEQ 2017). This document contains
revised and new interrogatories regarding version 1.4 of the DU PA.
UDEQ has arranged the interrogatories and SER comments in the Amended and New
Interrogatories Related to Clive DU PA Modeling Report Version 1.4 (Utah DEQ 2017) into
seven general categories. In order to retain focus on each subject area, responses are grouped into
individual papers by these seven categories. Each category is addressed within a standalone
Neptune response document as follows:
1. NAC_0106: Evapotranspiration (ET) Cover Design
2. NAC_0108: Erosion
3. NAC_0105: Deep Time Supplemental Analysis
4. NAC_0102: Other Wastes
5. NAC_0104: Groundwater Exposure
6. NAC_0103: Recycled Uranium
7. NAC_0101: Federal Cell Design
Note that, in an attempt to improve readability, the complete interrogatory number is cited as
each is first introduced and in headings, while the abbreviated number is used during discussion.
For example, Interrogatory CR R313-25-7(2)-05/2 is introduced by its full number, then is
abbreviated to Interrogatory 05/2, since the interrogatory numbering system employed by UDEQ
applies a unique number after the last hyphen in the sequence.
Note also that some interrogatories are linked to more than one section of the Utah
Administrative Code; thus, a single interrogatory can appear at first glance to be two. For
example, “Interrogatory CR R313-25-3 and R313-25-8-195/1” is a single interrogatory
referencing two parts of the Utah Administrative Code related to the subject at hand. The
shorthand for this interrogatory becomes “195/1.”
Introduction to DU PA Model Version 1.4 Interrogatory Responses
23 Feb 2018 2
Full text of each interrogatory, with UDEQ images, tables, and references, is available in Utah
DEQ (2017). Within these responses, the interrogatory is briefly quoted or paraphrased to
identify what is understood to be the issue of concern. When quoting longer passages of
interrogatory text, blue font in Arial size 10.5 is used and indented to visually distinguish the
interrogatory from the response. An example is shown below:
Sample format for quoting interrogatory text.
When quoting shorter passages within a paragraph, italics are used to distinguish the
interrogatory text being quoted.
The attached responses address issues raised in the interrogatories point by point, providing a
defensible basis for closing all open review items.
2.0 Introduction to Probabilistic Modeling
The Clive DU PA Model is a probabilistic model of the potential long-term risk consequences
from disposing DU in the Clive facility. This means that the inputs to the Model are specified as
probability distributions. The PA results depend not only on the model structure, but also on the
model specification (input probability distributions). The Clive DU PA Model structure is a
complex fully coupled organization of features, events, processes, and scenarios as described in
the Conceptual Site Model (CSM). Each of the inputs to this Model needs to be specified with
the same degree of care and attention that is given to the model structure. The PA Model results
depend on the probabilistic inputs; consequently, it is critical that the specification addresses
underlying statistical principles that relate to such long-term dynamic modeling. Note also that
the sensitivity analysis for the PA Model depends on the specific probabilistic input, in which
case identification of the appropriate important, or sensitive, parameters also requires correct
specification of the input probabilistic distributions.
The input distributions must address uncertainty in the data/information available for
specification, and must address spatial and temporal scaling. A well-formed probabilistic model
includes distributions that: 1) are based on what is thought to be known (expectation) and the
uncertainty about that expectation; 2) address spatial and temporal scaling, and 3) address
correlation between inputs where possible.
There is a tendency in PA modeling to ignore all three facets and instead to perform modeling
that is based on notions associated with conservatism (often in a deterministic model instead of a
probabilistic model), to ignore spatial and temporal scaling, and to ignore correlation. This leads
to projected human health risks or doses that are biased high and are far more uncertain than
appropriate, which in turn leads to poor decision making.
It is fine to make conservative decisions, but it is not fine to make important decisions based on
purposefully erroneous models. This is the same basic message that was attributed to Charles
Babbage after his invention of a “difference machine” in the late 1800s. Babbage wrote, “On two
occasions I have been asked ‘Pray, Mr. Babbage, if you put into the machine the wrong figures,
will the right answers come out’ I am not able rightly to comprehend the kind of confusion of
ideas that could provoke such a question.” This is the genesis of the term “garbage in, garbage
Introduction to DU PA Model Version 1.4 Interrogatory Responses
23 Feb 2018 3
out” (GIGO), which also applies to today’s modeling paradigm if the “wrong figures” are used.
It is crucial to pay attention to the statistical details of specifying a probabilistic model if
reasonable output is desired.
The Clive DU PA Model has been built to the extent possible based on expectation and
uncertainty. Correlation is included when and if supporting data/information are available. And,
considerable effort has been put into ensuring that the spatial and temporal scales of the input
distributions used in the PA Model are appropriate. There are several aspects of the supporting
data/information, supporting models (process level models such as Hydrus, SIBERIA, and
AERMOD, which were used to support the PA Model), and the PA Model itself that must be
considered when establishing the correct spatial and temporal domain of the PA Model and the
supporting data. Often data that are useful for specifying input distributions for a PA are
available at a very small spatial and temporal scale. Process level models need to address scaling
at a refined level, and the systems level PA model must address scaling at yet another level.
Statistical scaling must also consider the endpoint of interest. For PA models, an intermediate
endpoint is concentration of radionuclides in some media thousands of years (or in the case of
the Clive DU PA, millions of years) into the future. Some assumptions are critical, such as that a
PA model will project current knowledge into the future, so that changes to many aspects of the
environmental system are considered unlikely (for example, the basin contents are silt/clay, and
this is assumed to not change throughout time). Scaling can address temporal (or spatial)
changes, but the point here is to explain why scaling is needed, and relatively simple examples
are used for demonstration purposes. Consequently, for the purpose of this explication, the
parameters of interest are assumed stable in time and space (e.g., chemical/physical
characteristics).
Other considerations relate to the computational set up and the computational complexity of a
PA model. Suppose, for simplicity, that the model looks forwards 1,000 years, and that each time
step is one year. In principle a new random number could be drawn from an input distribution
every year; however, in practice this adds greatly to the computational complexity. Instead, an
approach is taken that draws a random number at the beginning of time, and applies that random
number in each of the 1,000 time steps. The trick is to find the appropriate distribution for this
approach to mimic pulling a random number every time step. If the response to the input
parameter is linear and the system with respect to that input parameter is stationary (or steady
state), then the input distribution that is needed at the beginning of time to mimic the effect of
drawing new random numbers every year (time step) is the distribution of the average of the
distribution that would be used for 1,000 random draws over time. For non-linear responses, and
non-stationary systems, the statistical scaling issues are more complex, but they are conceptually
the same.
In effect, the spatial and temporal scaling that is needed must address the differences in scale
between the data/information available and the models that are used, including the difference in
scale of the supporting process level models and the systems level PA model. Scaling in this
context is largely an averaging process. If the response is linear and stationary, then pure
averaging works directly, and this is the approach that has been taken in the Clive DU PA Model
for projecting future concentrations to which human receptors are exposed.
Introduction to DU PA Model Version 1.4 Interrogatory Responses
23 Feb 2018 4
A simple example of scaling considers the proportion of green M&Ms in a packet as a predictor
of the proportion of green M&Ms that are produced. The distribution of green M&Ms from
many packets might range from near 0% up to about 33%, but that is not the correct distribution
to use to predict the proportion of green M&Ms in 1,000 packets. Scaling takes over, and a better
estimate for 1,000 packets would be very close to 16% overall (the production rate of green
M&Ms is about 16%). One could estimate the number of green M&Ms in 1,000 packets by
adding up how many are in each packet (similar to choosing a new random number each time
step), or by taking the distribution of the average of the packet specific distribution. To make this
more concrete, suppose the proportion of green M&Ms in a packet follows a normal distribution
with mean 16% and standard deviation 5%; then, simulating by drawing a new random number
from that distribution for each one of the 1,000 packets gives the same result (distributional
estimate of overall proportion) as that obtained from simulating directly from a normal
distribution with mean 16% and standard deviation 5%/sqrt(1,000), where the 1,000 represents
the 1,000 packets. However, the wrong thing to do would be to simulate from the N(16%, 5%)
distribution and apply that to the overall proportion of green M&Ms in 1,000 packets. This
describes the essence of the scaling problem, which depends on the endpoint of interest for the
problem, and the scale of both the data and the model.
The basic approach to developing input probability distributions is described in Appendix 14 of
the initial Clive DU PA submittal. Further explanation of some of the details is provided in
response to interrogatories concerning the ET Cover, because there are more interrogatories
about input distributions for the ET Cover than for other aspects of the Clive DU PA Model. The
critical issue is the importance of spatial and temporal scaling so that parameter uncertainties are
appropriately captured in the Model.
3.0 References
EnergySolutions, 2015. Radioactive Material License UT2300249: Safety Evaluation Report for
Condition 35.B Performance Assessment; Response to Issues Raised in the April 2015
Draft Safety Evaluation Report, EnergySolutions LLC, Salt Lake City UT, November
2015
Neptune, 2011. Final Report for the Clive DU PA Model version 1.0, Neptune and Company
Inc., Los Alamos NM, June 2011
Neptune, 2014. Final Report for the Clive DU PA Model, Clive DU PA Model v1.2, NAC-
0024_R2, Neptune and Company, Inc., Los Alamos NM, August 2014
Neptune, 2015. Final Report for the Clive DU PA Model, Clive DU PA Model v1.4, NAC-
0024_R4, Neptune and Company Inc., Los Alamos NM, November 2015
SC&A, 2015. Utah Division of Radiation Control, EnergySolutions Clive LLRW Disposal
Facility, License No: UT2300249; RML #UT 2300249, Condition 35 Compliance Report;
Appendix A: Final Report for the Clive DU PA Model, Safety Evaluation Report, Volume
1, SC&A Inc., Vienna VA, April 2015
Introduction to DU PA Model Version 1.4 Interrogatory Responses
23 Feb 2018 5
Utah DEQ, 2017. Division of Waste Management and Radiation Control, EnergySolutions Clive
LLRW Disposal Facility License No: UT2300249; RML #UT 2300249, Amended and
New Interrogatories Related to Clive DU PA Modeling Report Version 1.4 Dated
November 2015, Utah Department of Environmental Quality (DEQ), Salt Lake City UT,
May 2017

NAC-0106_R0

ET Cover Design Responses for the
Clive DU PA Model
23 February 2018

Prepared by

NEPTUNE AND COMPANY, INC.
1435 Garrison St, Suite 201, Lakewood, CO 80215
ET Cover Design Responses for the Clive DU PA Model
23 Feb 2018 ii
1. Title: ET Cover Design Responses for the Clive DU PA Model
2. Filename: ET Cover Design Responses for the Clive DU PA Model.docx
3. Description: Responses to UDEQ Interrogatories and Safety Evaluation Report Comments
received May 11, 2017.

Name Date
4. Originator Mike Sully and Sean McCandless 12 Feb 2018
5. Reviewer Dan Levitt
6. Remarks

ET Cover Design Responses for the Clive DU PA Model
23 Feb 2018 iii
CONTENTS
CONTENTS ............................................................................................................................. iii
FIGURES ................................................................................................................................... v
TABLES ..................................................................................................................................viii
ACRONYMS AND ABBREVIATIONS ................................................................................... ix
1.0 Overview and Conceptual Model........................................................................................ 1
1.1 Modeling for Probabilistic Performance Assessments ................................................... 1
1.2 Setting .......................................................................................................................... 9
1.2.1 Disposal Cell Design ............................................................................................... 9
1.2.2 Unsaturated Zone and Shallow Aquifer ................................................................. 11
1.2.3 Climate .................................................................................................................. 14
1.2.4 Vegetation ............................................................................................................. 14
1.3 HYDRUS-1D Software Package ................................................................................. 15
2.0 UDEQ Interrogatory Responses........................................................................................ 17
2.1 Interrogatory CR R313-25-7(2)-05/2: Radon Barriers ................................................. 17
2.1.1 Interrogatory Response .......................................................................................... 19
2.2 Interrogatory CR R313-22-32(2)-10/3: Effect of Biologicals on Radionuclide
Transport .................................................................................................................... 37
2.2.1 Interrogatory Response .......................................................................................... 38
2.3 Interrogatory CR R317-6-2.1-20/2: Groundwater Concentrations ............................... 39
2.3.1 Interrogatory Response .......................................................................................... 41
2.4 Interrogatory CR R313-25-8(4)(d)-21/2: Infiltration Rates .......................................... 42
2.4.1 Interrogatory Response .......................................................................................... 47
2.5 Interrogatory CR R313-25-8(4)(a)-28/3: Bioturbation Effects and Consequences ....... 61
2.5.1 Interrogatory Response .......................................................................................... 62
2.6 Interrogatory CR R313-25-7(2)-59/2: Bathtub Effect .................................................. 63
2.6.1 Interrogatory Response .......................................................................................... 64
2.7 Interrogatory CR R313-25-7(3)-60/2: Modeled Radon Barriers .................................. 64
2.7.1 Interrogatory Response .......................................................................................... 64
2.8 Interrogatory CR R313-25-7(1–2)-90/2: Calibration of Infiltration Rates .................... 64
2.8.1 Interrogatory Response .......................................................................................... 65
2.9 Interrogatory CR R313-25-7(2)-150/3: Plant Growth and Cover Performance ............ 65
2.9.1 Interrogatory Response .......................................................................................... 66
2.10 Interrogatory CR R313-25-8(4)(d)-153/2: Impact of Pedogenic Processes on the
Radon Barrier ............................................................................................................. 66
2.10.1 Interrogatory Response .......................................................................................... 66
2.11 Interrogatory CR R313-25-7(2)-175/1: Infiltration Rates for the Federal Cell Versus
the Class A West Cell ................................................................................................. 67
2.11.1 Interrogatory Response .......................................................................................... 68
2.12 Interrogatory CR R313-25-8(5)(a)-176/1: Representative Hydraulic Conductivity
Rates........................................................................................................................... 68
2.12.1 Interrogatory Response .......................................................................................... 68
ET Cover Design Responses for the Clive DU PA Model
23 Feb 2018 iv
2.13 Interrogatory CR R313-25-7(2)-189/3: Modeling Impacts of Changes in Federal Cell
Cover-System Soil Hydraulic Conductivity and Alpha Values .................................... 70
2.13.1 Interrogatory Response .......................................................................................... 71
2.14 Interrogatory CR R313-25-7(2)-192/3: Implications of Great Salt Lake Freezing on
Federal Cell Performance............................................................................................ 72
2.14.1 Interrogatory Response .......................................................................................... 72
2.15 SER B.1 Supplemental Interrogatory Comment 1 ....................................................... 73
2.15.1 Interrogatory Response .......................................................................................... 74
2.16 SER B.2 Supplemental Interrogatory Comment 2 ....................................................... 77
2.16.1 Interrogatory Response .......................................................................................... 78
2.17 SER B.3 Supplemental Interrogatory Comment 3 ....................................................... 82
2.17.1 Interrogatory Response .......................................................................................... 83
2.18 SER B.4 Supplemental Interrogatory Comment 4 ....................................................... 86
2.18.1 Interrogatory Response .......................................................................................... 88
2.19 SER B.5 Supplemental Interrogatory Comment 5 ....................................................... 90
2.19.1 Interrogatory Response .......................................................................................... 92
2.20 SER B.6 Supplemental Interrogatory Comment 6 ....................................................... 97
2.20.1 Interrogatory Response .......................................................................................... 97
2.21 SER B.7 Supplemental Interrogatory Comment 7 ..................................................... 100
2.21.1 Interrogatory Response ........................................................................................ 101
2.22 SER B.8 Supplemental Interrogatory Comment 8 ..................................................... 106
2.22.1 Interrogatory Response ........................................................................................ 107
2.23 SER B.9 Supplemental Interrogatory Comment 9 ..................................................... 108
2.23.1 Interrogatory Response ........................................................................................ 108
2.24 SER B.11 Supplemental Interrogatory Comment 11 ................................................. 110
2.24.1 Interrogatory Response ........................................................................................ 111
3.0 Conclusion ..................................................................................................................... 116
4.0 References ...................................................................................................................... 119

ET Cover Design Responses for the Clive DU PA Model
23 Feb 2018 v
FIGURES
Figure 1. Evapotranspiration (ET) cover profile showing materials, observation nodes, and root
distribution used in the HYDRUS-1D models. ......................................................... 11
Figure 2. Stratigraphic profile showing ET cover, waste zone, and stratigraphy below the Federal
DU Cell. .................................................................................................................. 12
Figure 3. Eolian silt in trench located at Clive Pit 29 overlying Lake Bonneville sedimentary
deposits (Neptune 2015b). ....................................................................................... 21
Figure 4. An example of upper soil-modified eolian silt in Pit 29. Basal contact of the silt is
approximately located at the middle of the pick handle. It is a gradational contact
between eolian silt intermixed with regressive Lake Bonneville marl (bottom of the
pick handle). ............................................................................................................ 22
Figure 5. Figure 3 from Benson and Gurdal (2013) showing the data requested by
EnergySolutions. ...................................................................................................... 24
Figure 6. Estimated linear relationships between α and Ks for all observations (solid line) and
without the high-leverage points making up the clusters of points in the upper right
half of the plot with α greater than 0.10 kPa-1 (dotted line). The estimated correlation
changes from 0.627 for all the data to 0.384 for the restricted range. ........................ 25
Figure 7. Estimated linear relationships within each soil type for the data provided by Benson
(2017). Pearson’s correlation coefficients, and 95% confidence intervals, are shown
in Table 2 for the soil textural classes, and the individual soil classes are shown in
panels in Figure 8. ................................................................................................... 26
Figure 8. Estimated linear relationships by soil textural class for the data plotted in Figure 6 and
Figure 7. Note the x and y axes are allowed to change among panels and are on the
log10 scale. .............................................................................................................. 29
Figure 9. Estimated linear relationship on the log scale between α and Ks for Silty Clay.
Pearson’s correlation coefficient is -0.2, with a very wide 95% confidence interval of
(-0.97, 0.94). ............................................................................................................ 30
Figure 10. Estimated linear relationships based on 16 realizations of six random pairs of
observations that come from statistically independent variables (x and y were drawn
independently from two standard normal distributions). ........................................... 31
Figure 11. Borrow soil cross-section below a greasewood plant shows the compacted clay layer
at approximately 60-cm depth. Roots extend laterally and do not penetrate the
compacted layer (SWCA 2011)................................................................................ 35
Figure 12. Table 16 from the Clive PA Model Parameters v1.4 document providing the
distribution used. A description of the method used to select the geometric mean,
geometric standard deviation, and minimum reported in this table, and the associated
parameterization of the log normal were also provided in Appendix 14 of the Clive
DU PA Model Final Report for v1.4. ....................................................................... 48
Figure 13. R Statistical Software (R Core Team 2017) code and output for getting quantiles from
the distribution of Ks used in v1.4. ........................................................................... 50
ET Cover Design Responses for the Clive DU PA Model
23 Feb 2018 vi
Figure 14. The lognormal distribution used for the Ks in v1.4. It is parameterized by a geometric
mean of 3.37 (log-scale mean 1.215), a geometric standard deviation of 3.23 (log-
scale standard deviation of 1.17), and a minimum of 0.00432 implemented through a
shift of the distribution. The target 1st, 50th, and 99th percentiles are shown by the
vertical dotted lines. ................................................................................................. 51
Figure 15. The average annual net infiltration values obtained from HYDRUS compared to the
predictions from the linear regression model abstraction. The results from the 50
HYDRUS realizations were used to develop the regression model abstraction, and
therefore this plot depicts in-sample predictive performance. The one-to-one line is
shown for reference. ................................................................................................ 57
Figure 16. The average annual net infiltration values obtained from HYDRUS compared to the
predictions obtained from test datasets via 3-fold cross validation. The original
HYDRUS values predicted were not used to fit the regression equations used to
obtain the predictions. This approximates out-of-sample predictive performance of
the linear regression model abstraction. The one-to-one line is shown on each plot for
reference. ................................................................................................................. 60
Figure 17. Comparison of 1,000 realizations of net infiltration using the linear model in GoldSim
with the results of the 50 HYDRUS simulations of infiltration. ................................ 61
Figure 18. Relationship between hydraulic conductivity and water content used for Unit 4
material.................................................................................................................... 69
Figure 19. Comparison of 1,000 realizations of net infiltration using the linear model in GoldSim
with the results of the 50 HYDRUS simulations of infiltration. ................................ 76
Figure 20. A comparison of predictions (fitted values) from the linear and quadratic regression
models, against the HYDRUS results for net infiltration used in the model fitting. ... 94
Figure 21. A comparison of predictions from the linear and quadratic regression models based on
the input values used for the 50 HYDRUS runs. ...................................................... 95
Figure 22. Comparison of predicted net infiltration rates at the inputs used with the 50 HYDRUS
runs for the linear regression model and the “exponential” model suggested by
UDEQ. .................................................................................................................... 96
Figure 23. Re-creation of information in UDEQ Figure B-2 showing complementary cumulative
distribution functions (CDFs) for the 50 α values used in the HYDRUS realizations
and 1000 values drawn from the distribution used in GoldSim for the Clive DU PA
Model v1.4. ........................................................................................................... 103
Figure 24. Re-creation of information in UDEQ Figure B-3 showing complementary cumulative
distribution functions (CDFs) for the 50 n values used in the HYDRUS realizations
and 1000 values drawn from the distribution used in GoldSim for the Clive DU PA
Model v1.4. ........................................................................................................... 103
Figure 25. Re-creation of information in UDEQ’s Figure B-4 showing complementary
cumulative distribution functions (CDFs) for the 50 net infiltration values from the
HYDRUS realizations and 1000 values drawn from the distribution used in GoldSim
for the Clive DU PA Model v1.4. ........................................................................... 104
ET Cover Design Responses for the Clive DU PA Model
23 Feb 2018 vii
Figure 26. Comparison of Bingham Environmental (1991) water content data with water content
calculated using the regression equation for the DU PA GoldSim model and with the
results of the 20 HYDRUS simulations. ................................................................. 110
Figure 27. Time series of infiltration into the waste zone for one of the 20 HYDRUS-1D
simulations. ........................................................................................................... 115

ET Cover Design Responses for the Clive DU PA Model
23 Feb 2018 viii
TABLES
Table 1. Texture class, thickness range, and average thickness for the hydrostratigraphic units
underlying the Clive site. ......................................................................................... 12
Table 2. Estimated Pearson’s correlation coefficients between Ks and α on the log scale by soil
textural class, and associated 95% confidence intervals, calculated using the data
provided by Benson in 2017 data. ............................................................................ 28
Table 3. Percentiles associated with elicited information for the Ks distribution (cm/day), and the
distribution actually used. These are based on a lognormal distribution with geometric
mean of 3.37 cm/day and a geometric sd of 3.23 cm/day, with shifts for the minimum
associated with each row. See Figure 13 for example R code to get the percentiles. . 50
Table 4. Minerals in Unit 4 soil clays. ....................................................................................... 67
Table 5. Coefficients calculated from multiple linear regression models. ................................... 75
Table 6. Parameter sets of van Genuchten α, n, and Ks used for HYDRUS modeling. ............... 80
Table 7. Results of 50 flow realizations described in Appendix 5 of DU PA Model v1.4. .......... 98
Table 8. Water Content Data from Table 6 of Bingham Environmental (1991). ....................... 109

ET Cover Design Responses for the Clive DU PA Model
23 Feb 2018 ix
ACRONYMS AND ABBREVIATIONS
ACAP Alternative Cover Assessment Program
bgs below ground surface
BSC biological soil crust
CCD complementary cumulative distribution
CSU Colorado State University
DEQ (Utah) Department of Environmental Quality
DOE (United States) Department of Energy
DU depleted uranium
DWMRC Division of Waste Management and Radiation Control
DWR (Utah) Division of Water Rights
ET evapotranspiration
HAL Hansen, Allen, and Luce
HSU hydrostratigraphic unit
LLRW low-level radioactive waste
LM DOE’s Office of Legacy Management
MAUP modifiable areal unit problem
NOAA National Oceanographic and Atmospheric Administration
NRC (United States) Nuclear Regulatory Commission
NRCS Natural Resources Conservation Service
PA performance assessment
PPA probabilistic performance assessment
PDF probability density function
PET potential evapotranspiration
RML Radioactive Material License
RMSE root mean squared error
sd standard deviation
SER Safety Evaluation Report
SWCC soil water characteristic curve
UDEQ Utah Department of Environmental Quality
USDA United States Department of Agriculture
WSS (NRCS) Web Soil Survey

ET Cover Design Responses for the Clive DU PA Model
23 Feb 2018 1
1.0 Overview and Conceptual Model
This document begins with three sections that provide necessary background information for the
interrogatories and responses on the subject of the proposed evapotranspiration (ET) cover
design. Section 1.1 offers a discussion of the basis for the statistical approaches applied to
probabilistic PA modeling. This includes fitting distributions to data or elicited information,
scaling, and model abstraction. Section 1.2 is a summary of the setting of the Clive Site that
includes a description of the cell design, local hydrogeology, climate, and vegetation. Section 1.3
is a brief summary of the features of the variably saturated flow model HYDRUS-1D used to
develop models for net infiltration and water content in the GoldSim DU PA model. The
implementation of the software for modeling flow in the cover is also summarized in this section.
Recharge is an important process in controlling the release of contaminants to the groundwater
pathway. Site characteristics influencing movement of water from precipitation through the
vadose zone to the water table at the Clive Site include climate, soil characteristics, and native
vegetation. Engineered barriers are used at the Clive Site to control the flow of water into the
waste.
A hydrologic model of the waste disposal system must realistically represent precipitation, the
source of water to the system, runoff, evaporation, transpiration, and changes in storage to
estimate the flow through the system. Under natural conditions, plants remove water from the
upper soil zone through root uptake and transpiration, reducing the water available for seepage
deeper into the profile. The same processes occur in an engineered cover layer that has been
revegetated. Seepage through a cover system can occur when soils become wet enough to
increase their conductivity to water. Cover surface layers with adequate storage capacity can
hold the water in the near surface until it can move back into the atmosphere through
evaporation, reducing the seepage of water to the waste (net infiltration).
These processes would be expected to show temporal variability at the Clive Site on the time
scale of minutes to hours in the near surface and days to years deeper in the disposal cell.
Processes that tend to change cover properties such as plant and animal activity and climate
influences (e.g. frost heave, erosion) are expected to be slowed by the effects of eolian
deposition.
1.1 Modeling for Probabilistic Performance Assessments
Introduction
The Clive DU PA Model is a probabilistic model; consequently, its inputs are specified as
probability distributions. The PA Model results depend on both the model structure and the
model specification through the input probability distributions; consequently, it is critical that the
specification addresses underlying statistical principles that relate to this type of long-term
dynamic modeling. The sensitivity analysis for the PA Model also depends on the specific
probabilistic input, in which case identification of the appropriate important, or sensitive,
parameters also requires correct specification of the input probabilistic distributions. The
ET Cover Design Responses for the Clive DU PA Model
23 Feb 2018 2
statistical aspects of this type of dynamic probabilistic modeling are critical to the success and
defensibility of the model results.
A crucial piece of developing and evaluating distributions is understanding what the values from
the distributions should, or do, represent. This understanding is necessarily linked to how values
of the input parameters are used in the PA Model, and to the assumptions needed to use available
data/information to inform the distributions. A challenge for evaluating and developing
distributions is identifying the appropriate temporal and spatial scales that need to be
represented. The data/information to support specification are typically available at different
scales than those required by the model. The following provides a more detailed discussion of
the implications of this “scaling” on the development of input distributions for the Clive DU PA
Model, and is intended to supplement the information provided in Appendix 14 of the Clive DU
PA Final Report (particularly Section 6.0, titled Scaling and Model Abstraction). Appendix 14 of
the Clive DU PA Final Report contains supporting information for the probability distributions—
and methods to obtain those distributions—used to develop the distributions for the Clive PA.
Probabilistic PA
The probabilistic Clive DU PA Model uses probability distributions to represent the current state
of knowledge in the PA Model inputs, with the goal of running the PA Model at many sets of
plausible inputs. The approach is designed to provide information regarding the expected
performance of the Site. In general, the modeling effort should be focused on running the PA
Model under different realistic scenarios defined by different combinations of values of the input
parameters. Judging what is realistic is the fundamental challenge, and it depends on the spatial
and temporal scale represented by the values of the input in the PA Model.
This is in contrast to testing the potential performance of the Site “deterministically” by using a
single value to represent each input parameter. In addition, probabilistic modeling is aimed at
modeling what is thought to be known and the uncertainty in that knowledge, whereas
deterministic models are often run using extreme, and likely unrealistic, input parameter values
that are usually described as a conservative or protective. A problem with so-called conservatism
in PA models is that the direction of conservatism is often difficult to justify. Conservative
models are also difficult to explain because they are knowingly wrong, and thus use of
conservatism leads to poor decision making. If there is a desire to make a conservative
(protective) decision, then those value judgments should be separated from the PA model, and
should be based on a model that explains the system response as best possible.
The PA model uses time steps to propagate movement of radionuclides through the engineered
and environmental system to project concentrations in various media at various future points in
time. However, for a single simulation, random numbers are drawn from the input distributions
at the beginning of time, and are used throughout time (thousands of years, or more). This means
that the input distributions must represent long time periods, and the data/information available
must be scaled appropriately to represent those long time periods. The same is the case for the
spatial scale of the model; the available data/information often represent small spatial scales, and
must be adjusted to represent the large areas or volumes that underlie the PA model. Each
simulation is, essentially, a deterministic run of the PA model. Uncertainty in the output is
ET Cover Design Responses for the Clive DU PA Model
23 Feb 2018 3
obtained by running many (thousands of) simulations, each one of which has a distinct set of
input values drawn at random from the underlying distributions of the input parameters.
For example, using a value that is extreme at a small temporal scale (such as a daily soil
temperature in the summer) implies that it is realistic to hold that input constant at such an
extreme for the entire time period represented by a run of the PA model (thousands of years or
more). Therefore, it is crucial to question whether a value is “plausible” or “realistic” to
represent a constant condition over a long period of time or a large spatial area. A probabilistic
PA provides a natural framework for running the model at sets of values that are deemed realistic
for representing long-term conditions over the spatial extents represented in the model. If this is
accomplished, the model outputs can be interpreted as indicative of expected performance of the
entire site under realistic long-term conditions, given the current state of knowledge in input
parameters and the assumptions underlying the PA model.
The Importance of Scaling and Its Challenges
All models are approximations of reality, and the assumptions should always be evaluated and
discussed. However, given the assumptions of a PA model, the distributions should be developed
to explicitly match, as much as possible, the assumptions and design of the model. The use of
distributions representing values at a different scale than that represented by the model is not
consistent with the fundamental goals underlying probabilistic PA modeling. Therefore, model
developers and reviewers must ensure they are considering the problem relative to the same
spatial and temporal scales in order to have a meaningful and productive conversation and to
ultimately improve the distributions, and hence models, used in a probabilistic PA.
Developing an appropriate distribution for a PA model input parameter is not simply an exercise
in approximating a histogram of available data thought to be relevant to the site. It is unlikely
that available data, information from the literature, and/or expert knowledge exist at spatial and
temporal scales consistent with the design of the PA model. Therefore, an additional step in
distribution development is taking the available information and “scaling” it appropriately; this
typically means taking available values that are representative of relatively small scales (e.g.,
locations in space and/or points in time) and combining them to represent larger scales, typically
through averaging.
Identifying a scale consistent with the PA model requires in-depth understanding of how the
values from the distribution will ultimately be used in the PA model. That is, there should be an
explicit connection between what the values are assumed to represent within the PA model and
the distribution developed to produce those values. To accomplish this, the values coming from
the distribution should be aligned with the spatial and temporal scales upon which the PA model
is built; that is, they should represent the spatial and temporal scales represented by a value of the
input. Tangible examples are provided later in this discussion. If the scale represented in the
distribution does not match that used for the model, then the results from the model cannot be
interpreted as representing the performance of the site under realistic conditions consistent with
the model assumptions.
ET Cover Design Responses for the Clive DU PA Model
23 Feb 2018 4
The Change-in-Scale Problem
A mismatch in spatial and/or temporal scale between available data and what is needed to inform
a model, or to make predictions from a model, is a problem that arises in many disciplines. It is
not a new problem or a problem unique to PA work, and it has different names in different
disciplines such as change-of-support, change-of-scale, downscaling and/or upscaling, and the
modifiable areal unit problem (MAUP) in the geography literature (see Cressie and Wikle
(2011)). The need to match the scale of available data to that needed for the model or predictions
is fundamental to problems with temporal and/or spatial components. Higgs et al. (2017) offer a
general description of the problem within the context of PA work.
Understanding the Scales Represented in the PA Model
The first step in understanding what spatial and temporal scales a distribution should represent is
asking questions about how a value drawn from that distribution will ultimately be used in the
PA model. In the current implementation of the Clive DU PA Model in GoldSim, volumes of
material at the Site are represented as a network of “cells,” and the material represented by a
single cell is assumed to be completely mixed (homogeneous). For a single realization of the
Model, a property of the entire volume of material/soil associated with a cell is represented by
the value of input parameter drawn from a distribution for that property. Therefore, a value from
a distribution used to specify a property of material within a cell should represent the entire
volume of completely mixed materials within the cell. If the value is used to represent properties
of multiple cells, then the value should represent the larger spatial scale associated with the
volume of the combined cells. What values are appropriate to represent the entire volume of
material? How should available information, collected at various scales from the Site or
elsewhere, be used to inform the distribution of such values?
A value drawn from a distribution is typically held constant over the entire time period of a PA
model run or realization; it is usually not changed over time within a particular realization.
Therefore, the distribution should provide plausible values for summarizing the input parameter
over thousands of years (or longer), such as a long-term average. For example, even if values of
the input parameter may vary seasonally, that does not mean it is reasonable to run the model as
if a seasonal extreme is held constant for 1000 years. It is most realistic, and thus informative of
expected performance, to run the model at plausible long-term averages, not at short-term
extremes. Running the model at values representative of smaller time scales (such as individual
seasons) results in many unrealistically extreme realizations that will not adequately characterize
expected site performance over the period of interest. Over a long period of time, the input
variable is expected to be below average and above average during small periods of time, but the
performance is assessed over such large time scales that the extremes should “balance each other
out” given the model structure. Data may only be available over short time periods, such as
months, but distributions for longer-term averages can be developed using the short-term data
and assumptions regarding the future.
The relatively large spatial and temporal scales of input parameters represented in the PA model
result in challenges for development and evaluation of distributions. It is often not intuitive how
to use the information from available data, or from subject matter knowledge of “possible
ET Cover Design Responses for the Clive DU PA Model
23 Feb 2018 5
values” for an input parameter, to develop a distribution representing the scale consistent with
the PA model. For example, it is tempting to assume that any value that could possibly be
obtained from a single sample from one location at one point in time should be represented in the
distribution. However, such an approach is generally misaligned with the scale of the PA model,
as described previously. This idea is further explored in the following discussion, with spatial
and temporal dimensions considered separately, and in more detail.
Spatial Scaling Example
Suppose data are collected on material properties from 20 soil cores taken at random locations
from around the actual site. The variability in the resulting values from the samples will reflect
the heterogeneity across the site. An individual sample represents the spatial scale of a point
location (or small core), and a distribution fit directly to the observed values from all the cores
could be used to represent possible values that could be obtained from other potential cores taken
from the same site. Now, suppose values that reflect the entire volume of soil of the site are of
interest, and need to be obtained from the 20 soil cores. Is it reasonable to use a value
representing a single core as if it represents the larger volume of soil? In general, it is not
reasonable for a value from an individual sample to represent the larger volume; instead, an
average (or estimated mean) would typically be used as a representative value of the larger
volume. The idea can be thought of in terms of prediction; what value is reasonable to use to
predict the same property of a different site of the same size: the value from an individual core or
the average over all the cores sampled? Choosing the estimated mean will clearly give better
predictions and should have smaller errors on average (uncertainty) and less variability in errors.
A single data set has an associated estimate of the mean, but different estimates would be
obtained from different sets of 20 locations sampled. That is, there is uncertainty in the value of
the mean because the entire volume of material represented in the PA model will never be
sampled. Therefore, a probability distribution is used to incorporate uncertainty in the mean
rather than variability among values from different point locations; the distribution should
represent plausible means for the entire volume of material, rather than variability among
measurements from cores taken from individual locations within a site. Using values from a
distribution describing variability among individual small-scale locations to represent the entire
volume of material leads to extreme, and unrealistic, model runs. Practically, this often translates
into using standard errors for the estimated mean over a larger scale to define distributions, rather
than standard deviation of available small-scale measurements.
Another way to think about the problem is through the assumption of complete mixing used in
the conceptual construction of the PA model cells. First, envision the 20 samples described in the
previous paragraph. Now, suppose a huge mixer is employed to completely mix the entire
volume of material/soil before the 20 samples are taken. What would variability among the
values from these new 20 samples look like compared to variability among the first 20 samples
from the heterogeneous material? Because the soil is thoroughly mixed, the relatively large
values and small values are now dispersed through the whole volume so that the samples will
now all have values relatively close together (the more completely the material is mixed, the
closer they will be). The average value of the 20 samples in both cases (heterogeneous vs.
mixed) should be the same (or close to the same), but the variability may be drastically different.
ET Cover Design Responses for the Clive DU PA Model
23 Feb 2018 6
The point estimate of the mean from the original 20 samples can be used as the center for the
desired distribution, with the variance adjusted to reflect uncertainty in that mean because it is
based on a limited number of samples.
Temporal Scaling Example
The time scale associated with the input parameters in a PA model is the length of time over
which a value from a distribution is held constant in a run of the model. The Clive DU PA Model
most often uses a single value, drawn from an input distribution at the beginning of a model
realization, over the whole time period represented by the Model. Therefore, it must represent
the spatial volume (described in the previous section) over thousands of years (or longer). This is
obviously challenging because available data exist at much, much smaller temporal scales (less
than a day to perhaps 50 years) and an assumption must be made that current knowledge is
projected into the future. The lack of knowledge about the future requires making assumptions
about future conditions, and typically the assumption is made that the conditions under which
data were collected will hold into the future (i.e., stationarity). This implies the mean and
associated variance will stay constant into the future represented by the PA Model. Under this
assumption, information in the available data and expert knowledge can be used to directly
inform the distribution; uncertainty and lack of trust in assumptions can also be incorporated.
As discussed for the spatial scale, the distribution should represent the current state of knowledge
about the input at the PA model scale, meaning random draws from the distribution should
capture variability among plausible long-term means for the volume of material represented by
the PA model cell(s). Using soil temperature as an example, consider the variability among daily
average soil temperatures in a region with seasonal variation, with extremes captured by the
daily average soil temperatures during the coldest time of the year, and the daily average soil
temperatures during the warmest part of the year. Now, contrast that with the extremes expected
across different annual average soil temperatures; one extreme is a year that is colder on average
over all seasons and the other extreme is a year that is warmer on average over all seasons. The
difference between these two annual extremes is expected to be far less than the difference
between the two extremes defined on the daily scale. Is it reasonable to use an extreme daily
average to represent an extreme annual average soil temperature? Using a daily average as an
annual average is equivalent to representing incredibly extreme conditions that are not supported
by available data—it amounts to assuming that conditions in the system are going to change so
drastically that daily averages become representative of annual averages.
These ideas continue to larger scales, where, at each larger scale, less variability is expected
among averages at that scale. In other words, we expect less variability among decadal averages
than annual averages, and less variability among 50-year averages than decadal averages, and so
on. The degree to which the distributions representing the current state of knowledge in larger
and larger scales get narrower does depend on the trust in the assumption of stationarity of
conditions into the future. However, regardless of the proposed difference in variance between a
distribution of annual averages and a distribution of 100- or 1000-yr averages, it is clear that
seasonal variation occurs at a scale much smaller than that represented by the PA model. It is not
appropriate to assume that a value measured at a location at one point in time should be
represented in the collection of draws from a distribution for that input used for the PA model
ET Cover Design Responses for the Clive DU PA Model
23 Feb 2018 7
when the value drawn from the distribution is held constant over thousands of years (or longer).
When evaluating the reasonableness of an input distribution, given the modeling conditions, one
should ask “Are these values plausible given they will be held constant in the model over
hundreds or thousands of years?”
Combining Spatial and Temporal Scaling
As described above, from a temporal perspective the values captured in a distribution should be
judged reasonable in the context of being held constant over long time periods, and therefore
should not reflect extremes that are only realistic at smaller time scales. Thinking about plausible
long-term averages for the values representing the spatial volumes of the PA model is the easiest
way to judge reasonableness of the distribution relative to time. However, this is challenging
because it is clearly impossible to collect data over hundreds or thousands of years into the future
to inform the distribution of a long-term average.
Therefore, temporal scaling is often more abstract than the methods discussed for spatial scaling,
and often must actually use mainly spatial variability as a surrogate for variability that might be
observed over time. For example, using data from different locations and regions captures
conditions in different materials that may be observed at one location over time. Therefore, the
heterogeneity over space is used as a surrogate for heterogeneity over time for the volumes
represented by the PA model.
This is necessary if distributions are to be based on empirical information because it is not
possible to collect the data needed to capture the heterogeneity over time into the future. While
the scaling does appear to be simply “spatial scaling” on the surface, it is not a misnomer to call
it spatio-temporal scaling because both scales are always simultaneously considered even if most
of the information available is spatial. The resulting distributions should be evaluated relative to
both dimensions by asking “does the distribution provide values that are plausible when applied
to the spatial volumes of PA cells over long periods of time?”
Choice of Statistical Methods
There are many statistical methods that can be used to estimate a mean, along with associated
uncertainty; the choice depends on the method of data collection (e.g., sampling design) and
characteristics of the data such as dependence (e.g., correlation) among values from different
samples. The average (arithmetic mean) is the simplest, and is often appropriate, as long as
samples do not severely violate underlying linearity and stationarity assumptions.
If direct averaging applies, then the scaled distribution can often be represented by a normal
distribution (depending on the nature of the underlying data). This is because sampling
distributions of averages tend to be approximately normally distributed, even when the
population of values from which the samples are taken are not normal, and normality improves
with increasing sampled sizes. The t-distribution is also an option to account for uncertainty in
estimating the standard deviation of the population, but the t-distribution is barely
distinguishable from the normal distribution for sample sizes greater than 30, and the normal
distribution is simpler to implement and communicate about for a PA model.
ET Cover Design Responses for the Clive DU PA Model
23 Feb 2018 8
The data distribution has a mean and a standard deviation, whereas the scaled distribution
obtained as the distribution of the average has the same mean and a standard deviation of the
mean, which is usually called the standard error of the mean.
Scaling Challenges Related to Sources of Information
If samples are available directly from the volume of material of interest (represented by a PA
model cell), then the development of the distribution at the appropriate scale based on smaller-
scale measurements is a straightforward statistical exercise. However, if the information
available to inform the distribution is composed of measurements from other sites, under
different experimental conditions, representing predictions from other estimated relationships,
from expert judgment, etc., then distribution development becomes much more challenging and
must rely on professional and expert judgment decisions to factor in uncertainty regarding what
measurements might be from the site of interest and to gauge how different sources of
information should be weighted against each other in the distribution development process.
In such cases, assumptions and decisions should be explicitly stated, and it is not reasonable that
every rational expert would provide the same distribution. It is rare that many samples are taken
from the site to inform a distribution, and the variety of types of information and “data” available
certainly add to the challenges of representing current state of knowledge in the mean of a spatial
volume at the site. Different values and summaries presented in the literature may represent
different spatial scales, and this information should be used, if possible and if available.
Additionally, development of distributions is often complicated by limited site-specific
information over space and over time. This limited information must be transformed to represent
the scales consistent with the model, after taking into account the quality of the different data
sources and their relevancy to the site, and the assumptions of the PA model.
Conclusions
While all rational experts may not develop the exact same distribution, they should all be
critically examining the distributions with the same spatial and temporal scales in mind, given
the PA model in question. The relevant question to ask is not “Does the distribution cover all
plausible values of the input variable that could be observed from a single sample?” but instead
“Does the distribution cover all plausible values of the input variable that could be representative
of the large volume of material represented by the PA model cell over long periods of time?”
These questions have different answers because they are tied to different spatial scales, and
answers depend on the current state of knowledge regarding site conditions, including linearity
and stationarity.
Development of distributions for PA model inputs would certainly be easier if each distribution
could simply represent possible values that have been observed in data collected from single
locations and single points in time. However, running the PA model with such distributions
would result in many of the realizations of a PA model capturing unrealistically extreme
scenarios (e.g., an extreme daily average temperature of 90 degrees assumed to be held constant
over not only a year, but thousands of years). The problem as described for a single distribution
magnifies as the number of distributions with spatial and temporal scales misaligned with the PA
ET Cover Design Responses for the Clive DU PA Model
23 Feb 2018 9
model increases, possibly to the point where only a small proportion of the PA model
realizations represent realistic and useful outcomes.
The goal of a probabilistic PA is to represent expected outcomes of the site. Using distributions
at temporal and spatial scales misaligned with the representations built into the PA model makes
this goal unattainable.
If there is interest in running the model at extreme, unrealistic scenarios to test hypothetical
limits of the model and the associated performance of the site, then this can be done in a separate
investigation, but this should not be built into the distributions meant to capture the current state
of knowledge at a scale consistent with the current PA model. Distributional goals are
necessarily tied to spatial and temporal scale, and therefore developers and reviewers must
ensure they are considering the problem relative to the same spatial and temporal scales in order
to have a meaningful and productive conversation about the input distributions. This leads to a
defensible model, and a sensitivity analysis that yields meaningful results that can ultimately be
used to direct further data/information collection to improve, or refine, the distributions used in a
probabilistic PA.
Further examples of scaling in the Clive DU PA are provided in Appendix 14 to the Clive DU
PA Final Report.
1.2 Setting
1.2.1 Disposal Cell Design
The design of the Federal DU Cell is a covered embankment, with relatively steeper sloping
sides nearer the edges. The upper part of the embankment, known as the top slope, has a
moderate slope (2.4%), while the side slope is markedly steeper (20%). The embankment is also
constructed such that a portion of it lies below-grade. The overall length of the embankment is
1317.8 ft and the overall width is 1775.0 ft. A detailed description of embankment dimensions
and a discussion of the representation of the Federal DU Cell in the GoldSim model are provided
in Embankment Modeling for the Clive DU PA Model (Neptune 2015g).
Disposal involves placing waste on a prepared clay liner that is approximately 8 ft below the
ground surface. For the Federal DU Cell design, the depth of the waste below the top slope is a
maximum of 47.5 ft (14.5 m). A cover system is constructed above the waste. The objective of
the cover system is to limit contact of water with the waste, limit biointrusion, and protect the
lower layers of the cover from freezing. The cover is sloped to promote runoff and is designed to
limit water flow by increasing evapotranspiration (ET). The arrangement of the layers used for
the ET cover design is shown in Figure 1. Beginning at the top of the cover, the layers above the
waste used for the ET cover design are:
• Surface Layer: This layer is composed of native vegetated Unit 4 silty clay material with
15 percent gravel mixture on the top slope and 50 percent gravel mixture for the side
slope. This layer is 6 inches thick. The functions of this layer are to control runoff,
minimize erosion, and maximize water loss from ET. This layer provides storage for
water accumulating from precipitation events, enhances losses due to evaporation, and
ET Cover Design Responses for the Clive DU PA Model
23 Feb 2018 10
provides a rooting zone for plants that will further decrease the water available for
downward movement.
• Evaporative Zone Layer: This layer is composed of Unit 4 material. The thickness of
this layer is 12 inches. The purpose of this layer is to provide additional storage for
precipitation and additional depth for the plant rooting zone to maximize ET.
• Frost Protection Layer: This material ranges in size from 16-inch cobbles to clay-size
particles. This layer is 18 inches thick. The purpose of this layer is to protect layers below
it from freeze/thaw cycles and wetting/drying cycles, and to inhibit plant, animal, or
human intrusion.
• Upper Radon Barrier Layer: This layer consists of 12 inches of compacted clay with a
low hydraulic conductivity. This layer has the lowest conductivity of any layer in the
cover system. This is a barrier layer that reduces the downward movement of water to the
waste and the upward movement of gas out of the disposal cell. The design specification
for saturated hydraulic conductivity (Ks) of this layer is 5.00 × 10-8 cm/s (Whetstone
Associates (2011), Table 15).
• Lower Radon Barrier Layer: This layer consists of 12 inches of compacted clay with a
low hydraulic conductivity. This is a barrier layer placed directly above the waste that
reduces the downward movement of water and the upward movement of gas out of the
disposal cell. The design specification for Ks of this layer is 1.00 × 10-6 cm/s, from Table
15 of Whetstone Associates (2011).
ET Cover Design Responses for the Clive DU PA Model
23 Feb 2018 11

Figure 1. Evapotranspiration (ET) cover profile showing materials, observation nodes, and
root distribution used in the HYDRUS-1D models.
1.2.2 Unsaturated Zone and Shallow Aquifer
The following description of the Clive Site hydrology is taken from the review prepared by
Envirocare (2004). The Site is described as being located on lacustrine (lake bed) deposits
associated with the former Lake Bonneville. The sediments underlying the facility are principally
interbedded silt, sand, and clay. Sediments at the Site are described by Bingham Environmental
(1991, 1994) and Envirocare (2000, 2004) as being classified into four hydrostratigraphic units
(HSU). Predominant sediment textural class, layer thickness range, and average layer thickness
for each unit are listed in Table 1. A diagram of the unsaturated zone is shown in Figure 2.
Unit 4: This unit begins at the ground surface and extends to between 6 ft and 16.5 ft below the
ground surface (bgs). The average thickness of this unit is 10 ft. This unit is composed of finer
grained low permeability silty clay and clay silt.
Unit 3: Unit 3 underlies Unit 4 and ranges from 7 ft to 25 ft in thickness. The average thickness
of this unit is 15 ft. Unit 3 is described as consisting of silty sand with occasional lenses of silty
to sandy clay.
Unit 2: Unit 2 underlies Unit 3 and ranges from 2.5 ft to 25 ft in thickness. The average
thickness of this unit is 15 ft. Unit 2 is described as being composed of clay with occasional silty
ET Cover Design Responses for the Clive DU PA Model
23 Feb 2018 12
sand interbeds. A structure map was prepared by Envirocare (2004) (their Figure 5) with
contours representing the elevations of the top of the unit. This map shows that the top surface of
Unit 2 slopes downward gradually from east to west in the vicinity of the Class A South cell.
Unit 1: Unit 1 underlies Unit 2 and is saturated beneath the facility, containing a locally confined
aquifer. Unit 1 extends from approximately 45 ft bgs and contains the deep aquifer. The deep
aquifer is reported to be made up of lacustrine deposits consisting of deposits of silty sand with
some silty clay layers. One or possibly more silty clay layers overlie the aquifer (Bingham
Environmental 1994).
Table 1. Texture class, thickness range, and average thickness for the hydrostratigraphic
units underlying the Clive site.
Unit Sediment Texture Class Thickness
Range (ft)
Average
Thickness (ft)
4 silt and clay 6–16.5 10
3 silty sand with interbedded silt and clay layers 7–25 15
2 clay with occasional silty sand interbeds 2.5–25 15
1 silty sand with interbedded clay and silt layers >620 >620

Figure 2. Stratigraphic profile showing ET cover, waste zone, and stratigraphy below the
Federal DU Cell.
ET Cover Design Responses for the Clive DU PA Model
23 Feb 2018 13
The aquifer system in the vicinity of the Clive Site is described by Bingham Environmental
(1991, 1994) and Envirocare (2000, 2004) as consisting of unconsolidated basin-fill and alluvial
fan aquifers. Characterization of the aquifer system is based on subsurface stratigraphy
observations from borehole logs and from potentiometric measurements. The aquifer system is
described as being composed of two aquifers: a shallow, unconfined aquifer and a deep confined
aquifer. The shallow unconfined aquifer extends from the water table to a depth of
approximately 40 ft to 45 ft bgs. The water table in the shallow aquifer is reported to be located
in Unit 3 on the west side of the Site and in Unit 2 on the east side.
The deep confined aquifer is encountered at approximately 45 ft bgs and extends through the
valley fill (Bingham Environmental 1994). The boring log from a water supply well drilled in
adjoining Section 29 indicates continuous sediments to a depth of 620 ft bgs as described in Utah
Division of Water Rights water right number 16-816 and associated well log 11293 (DWR
2014).The deepest portion of the basin in the Clive area is believed to be north of Clive in Ripple
Valley, where the basin fill was estimated to be 3,000 ft thick (Baer and Benson as cited in Black
et al. (1999)).
Deeper saturated zones in Unit 1 below approximately 45 ft bgs are reported to show higher
potentiometric levels than the shallow unconfined aquifer. Differences in potentiometric levels
are attributed to the presence of the Unit 2 clays. These observations are interpreted as indicating
that the shallow unconfined aquifer below the Site does not extend into Unit 1 but is contained
within Units 2 and 3 (Bingham Environmental 1994). The aquifer systems are described in more
detail in the Saturated Zone Modeling white paper (Neptune 2015a).
Recharge to the shallow aquifer in the vicinity of Clive is thought to consist of three components:
a small amount is due to vertical infiltration from the surface, some small amount of lateral flow
is from recharge areas to the east of the Site, and the majority of recharge is believed to be from
upward vertical leakage from the deeper confined aquifer (Bingham Environmental 1994).
Average annual groundwater recharge from the surface in the southern Great Salt Lake Desert in
the precipitation zone typical of Clive was estimated by Gates and Kruer (1981). An estimated
300 acre-feet per year were recharged to lacustrine deposits and other unconsolidated sediments
over an area of 47,100 acres. This is a recharge rate of approximately 0.08 in/yr (2 mm/yr).
Groundwater recharge from lateral flow occurs due to infiltration at bedrock and alluvial fan
deposits away from the Site, which moves laterally through the unconfined and confined aquifers
(Bingham Environmental 1994). This is evidenced by the increasing salinity of the groundwater
due to dissolution of evaporate minerals as water moves from the recharge area to the aquifers
below the Site facility (Bingham Environmental 1994). The majority of recharge to the shallow
aquifer is believed by Bingham Environmental (1994) to be due to vertical leakage upward from
the deep confined aquifer due to the presence of upward hydraulic gradients.
The higher potentiometric levels in the deeper saturated zones in Unit 1 described previously are
attributed to the presence of the Unit 2 clays (Bingham Environmental 1994). Vertical gradients
between shallow and deeper screened intervals in the monitor well clusters were calculated by
Bingham Environmental (1994). An upward vertical gradient was observed ranging in magnitude
from 0.02 to 0.04 based on the distance between the screen centers. For a vertical hydraulic
ET Cover Design Responses for the Clive DU PA Model
23 Feb 2018 14
conductivity of 1 × 10-6 cm/s (Bingham Environmental 1994), this corresponds to a recharge
range from 0.25 in/yr to 0.5 in/yr (6.35 mm/yr to 12.7 mm/yr).
1.2.3 Climate
Precipitation measurements taken at the Site over the 17-year period 1992 to 2009 show a mean
annual value of 8.53 inches (21.7 cm) (Whetstone Associates 2011). Precipitation exceeds the
annual average from January through June and again in October and is below average for the
remaining months. The nearest National Oceanographic and Atmospheric Administration
(NOAA) station with a long-term record is located in Dugway, Utah, approximately 40 miles to
the south. The mean annual precipitation for the same 17-year period measured at the Dugway
station is 8.24 inches (20.9 cm). A comparison of the Dugway precipitation data for the 17-year
period 1992 to 2009 with the long-term average for Dugway was made by Whetstone Associates
(2011). This comparison indicated that annual average precipitation at the Site during this 17-
year period has been greater than the long-term average at Dugway by 8 percent. Whetstone
Associates (2011) concluded that simulations of cover performance using precipitation data from
this 17-year period might be overestimating this component of the site water balance.
The HYDRUS-1D modeling performed is based on the 17-year record for consistency with the
modeling results reported in Whetstone Associates (2011). However, an additional 2 years of
monthly precipitation data are available from MSI (2012). The 19-year average precipitation is
8.62 inches (21.9 cm). This difference is driven primarily by the 4.28 inches of rainfall in May
2011. The small change in the overall average suggests that the modeling results presented for
this analysis would not change significantly if the 19-year precipitation record had been used
instead of the 17-year record.
The close correspondence between mean monthly temperatures measured at the Clive Site and
the Dugway NOAA station was demonstrated by Whetstone Associates (2011). Average
monthly temperatures measured at the Clive Site over the 17-year period 1992–2009 ranged
from 27.7 °F in December to 79.5 °F in July.
1.2.4 Vegetation
Actual transpiration is dependent on the characteristics of the plant communities at the Site.
Vegetation cover at the Site is less than 20 percent, with soils supporting a range of native and
invasive shrubs. Excavations at the Site have shown plant rooting depths extending to
approximately 31 inches (80 cm) below the ground surface, with root density decreasing with
depth (SWCA 2011).
Vegetation surveys of three field plots on or adjacent to the Clive Site were conducted by SWCA
(2011). The three low desert vegetation associations were characterized as: black greasewood,
Plot 3; halogeton-disturbed, Plot 4; and shadscale-gray-molly, Plot 5. The dominant shrub in Plot
3 was black greasewood with a percent cover of 4.5% and the dominant forb was halogeton with
a percent cover of 0.7%. In Plot 4, the dominant shrub was shadscale saltbush with a percent
cover of 2.3%, and the dominant forb was halogeton with a percent cover of 3.3%. In Plot 5, the
ET Cover Design Responses for the Clive DU PA Model
23 Feb 2018 15
dominant shrub was shadscale saltbush with a percent cover of 12.5%, and the dominant forb
was halogeton with percent cover of 0.9%.
Black greasewood, shadscale saltbush, and halogeton are all classified as facultative halophytes
(Anderson 2004; Pavek 1992; Simonin 2001). Facultative halophytes are known to benefit from
high salt concentrations in their growth media (Shabala 2013). Halophytes are able to adjust to
saline environments through various physiological adaptations such as compartmentalization of
ions in cell vacuoles, succulence, and the elimination of salt through salt-secreting glands and
bladders (Shabala 2013). Optimal growth for halophytes has been demonstrated by Shabala
(2013) to occur in media with a concentration of approximately 50 mM NaCl for monocots, and
between 100 and 200 mM for dicots. For the optimum range for dicots of 100 to 200 millimoles
per liter (mM), the corresponding range of electrical conductivity for a NaCl solution is 9.5 to
18.4 mmho/cm (https://www.wolframalpha.com/).
Depending on the extent of the area defined on and adjacent to the Clive Site, approximately 80
to 90 percent of the soils are mapped by the Natural Resources Conservation Service (NRCS) as
the Skumpah on 0 to 2 percent slopes (NRCS 2016). This Unit is characterized as having
maximum salinity ranging from 8.0 to 16.0 mmhos/cm. The top end of this range of maximum
salinity does not exceed the maximum of the range of salinity considered optimum for halophyte
growth of 18.4 mmho/cm. Given the similarity in ranges of salinity in the surface soils at the
Clive Site and that needed for optimum halophyte growth, the influence of the osmotic head
reduction in the root-water uptake water stress response function is considered negligible and
was, consequently, not included in the model.
1.3 HYDRUS-1D Software Package
HYDRUS-1D (Šimůnek et al. 2013) was selected for simulating the performance of the ET
cover proposed for the DU waste cell. The HYDRUS-1D platform was selected for this project
because of its ability to simulate processes known to have a significant role in water flow in
landfill covers in arid regions. HYDRUS includes the capabilities to simulate:
• water flow in variably saturated porous media,
• material hydraulic property functions,
• atmospheric surface boundary conditions including precipitation and evapotranspiration,
• root water uptake, and
• free-drainage boundary conditions.
The flow component of unsaturated flow and transport software packages with atmospheric
boundary conditions such as HYDRUS solves modified forms of the Richards equation for
variably saturated water flow. The flow equation incorporates a sink term to account for water
uptake by plant roots. HYDRUS can be applied to one-, two-, and three-dimensional problems.
The HYDRUS software includes grid generators for structured and unstructured finite element
meshes. Programs such as HYDRUS require detailed data to represent the atmospheric boundary
conditions and plant responses that are the dominant influences on flow in the cover in arid and
semi-arid conditions. These programs use the infiltration capacity of the soil at any time as
calculated in the model to partition precipitation into infiltration and overland flow. HYDRUS
ET Cover Design Responses for the Clive DU PA Model
23 Feb 2018 16
has been used for many applications for unsaturated zone modeling and has received numerous
favorable reviews such as Scanlon’s (Scanlon et al. 2002) review of HYDRUS-1D, Diodato’s
(Diodato 2000) review of HYDRUS-2D, and McCray’s (McCray 2007) review of the most
recent program, HYDRUS (2D/3D).
HYDRUS-1D was selected for simulating flow in the Federal DU Cell ET cover since previous
numerical modeling of flow in the similar ET cover design for the Class A West cover
demonstrated that subsurface lateral flow was not significant (EnergySolutions 2012). To test the
importance of 2-D flow effects in the ET cover design, 2-D transient flow simulations were
conducted for representative sections of the cover. The approach taken was to model a section of
the side slope in two dimensions. Representative hydraulic properties were assigned to the ET
cover layers and the models were run with daily atmospheric boundary conditions for 100 years.
The results of these 2-D simulations demonstrated that water flow in the cover system for both
designs is predominantly vertical with no significant horizontal component. These results
demonstrate that 1-D models can be used to provide a defensible analysis of cover performance
for the ET cover design due to the lack of lateral flow.
HYDRUS-1D models were developed for the evapotranspiration cover design for the Federal
DU Cell (Figure 1). Model development requires construction of a computational grid based on
the geometry of the model domain. Hydraulic properties for each layer required for the model are
available from previous studies at the Site or can be estimated from site-specific measurements
such as particle size distributions. HYDRUS requires daily values of precipitation, potential
evaporation, and potential transpiration to represent the time-variable boundary conditions on the
upper surface of the cover. Representative boundary conditions were developed from records of
nearby meteorological observations. Parameters for describing root water uptake were available
from the literature.
HYDRUS implements the soil-hydraulic functions of van Genuchten (1980), who used the
statistical pore-size distribution model of Mualem (1976) to obtain a predictive equation for the
unsaturated hydraulic conductivity function in terms of soil water retention parameters. The
expressions of van Genuchten (1980) are given by
𝜃(ℎ)=&𝜃'+𝜃)−𝜃'
[1 +|𝛼ℎ|/]1 ℎ<0

𝜃)ℎ≥0
(1)
𝐾(ℎ)=𝐾7𝑆9:[1 −;1 −𝑆9
<
1=
1
]> (2)
𝑆9 = 𝜃−𝜃'
𝜃7 −𝜃'
(3)
ET Cover Design Responses for the Clive DU PA Model
23 Feb 2018 17
where
θ is volumetric water content (–),
h is the matric potential (l),
θr is residual volumetric water content (–),
θs is saturated volumetric water content (–),
α is a van Genuchten fitting parameter (l-1),
n is a van Genuchten fitting parameter (–),
m is 1-1/n with n > 1,
K is hydraulic conductivity (l/t), and
Ks is the saturated hydraulic conductivity l/t.

The above equations contain five independent parameters: θr, θs, α, n, and Ks. The pore-
connectivity parameter “l” (lower-case L) in the hydraulic conductivity function was estimated
(Mualem, 1976) to be about 0.5 as an average for many soils. The value for l is commonly taken
to be 0.5, and this value was used for all simulations for all soil types.
2.0 UDEQ Interrogatory Responses
UDEQ has arranged the interrogatories and SER comments in the Amended and New
Interrogatories; Clive DU PA Modeling Report Version 1.4 (Utah DEQ 2017) into seven groups.
The first group named by UDEQ as ET Cover Design will be addressed in this document. This
includes Open Interrogatories CR R313-25-7(2)-05/2, CR R313-22-32(2)-10/3, CR R317-6-2.1-
20/2, CR R313-25-8(4)(d)-21/2, CR R313-25-8(4)(a)-28/3, CR R313-25-7(2)-59/2, CR R313-
25-7(3)-60/2, CR R313-25-7(1–2)-90/2, CR R313-25-7(2)-150/3, CR R313-25-8(4)(d)-153/2,
CR R313-25-7(2)-175/1, CR R313-25-8(5)(a)-176/1, CR R313-25-7(2)-189/3, and CR R313-25-
7(2)-192/3, and Supplemental Comments from Appendix B of the SER (SC&A 2015) 1, 2, 3, 4,
5, 6, 7, 8, 9, and 11.
2.1 Interrogatory CR R313-25-7(2)-05/2: Radon Barriers
DEQ Conclusion from April 2015 SER, Appendix C:
Based on several unresolved issues related to the evapotranspiration (ET) cover, DEQ indicated
in the DU PA SER that the cover design was deficient. Therefore, this interrogatory remains
open. The unresolved issues are as follows:
Evapotranspiration Cover—There are still a number of unresolved issues with respect to the
selection of parameter ranges, distributions, and correlations, as well as the modeling approach
and predicted sensitivities. These concerns are detailed in Appendix B. Further, because the
model-predicted infiltration rates will be sensitive to the hydraulic properties assigned to each ET
layer, DEQ recommends that EnergySolutions develop hydraulic properties for the cover system
based on the approach outlined by Dr. Craig H. Benson in Appendix F to this SER. Issues
related to this portion of the performance assessment cannot be closed until these concerns
have been resolved. (All references in prior interrogatories to Appendices of “the SER” refer to
the April 2015 SER (SC&A 2015)).
ET Cover Design Responses for the Clive DU PA Model
23 Feb 2018 18
Clay Liner—As with the ET cover, there is still an unresolved concern that Ksat values will
increase greatly over time, and that the α and Ksat values assumed for modeling flow through
the liner must either be correlated or a sensitivity analysis be conducted to demonstrate that the
lack of correlation assumed does not adversely affect the modeling results. In addition, there are
problems with assumed liner hydraulic conductivity values. Furthermore, the DU PA Model v1.2
does not account for liner degradation over time. These issues must be resolved before DEQ
can determine the adequacy of this portion of the DU PA.
DEQ Critique of DU PA v1.4, Appendix 21:
Modeling conducted for the clay liner beneath the waste should employ hydraulic parameters
representative of a compacted clay liner. Typical α, n, and Ɵs for compacted clays can be found
in Tinjum et al. (1997). Typical saturated hydraulic conductivities for clay liners can be found in
Benson et al. (1994).
Infiltration—Before the adequacy of the DU PA can be determined, additional modeling of the
ET cover infiltration rates must be conducted based on in-service hydraulic properties and
correlated log(α) and log(Ksat) values as described in Appendix E. Without this information, DEQ
is unable to conclude if the infiltration rates predicted by the DU GoldSim model are reliable or
representative of future conditions (i.e., ≥ 10,000 years).
DEQ Critique of DU PA v1.4, Appendix 21: See Interrogatories 21, 175, 176, and 189 for
discussions regarding the relationship between infiltration and the in-service hydraulic properties.
Erosion of Cover—Before the adequacy of the DU PA can be determined, EnergySolutions
needs to clarify certain issues relating to Appendix 10 to the DU PA Model v1.2 (June 5, 2014;
Neptune 2014g) as described in Section 4.4.2 of the SER. The Division of Waste Management
and Radiation Control (DWMRC) is currently reviewing a proposed ET cover test request as part
of a Stipulation and Consent Agreement to use a cover of similar design to that proposed for the
Federal Cell in the DU PA. Any recommendations and conclusions from that review will need to
be applied to the proposed Federal Cell as well.
DEQ Critique, v1.4, Appendix 21: See Interrogatories 20, 28, 160, and 191 for discussions
regarding cover erosion.
Effect of Biologicals on Radionuclide Transport—EnergySolutions has not shown that the
cover system is sufficiently thick or designed with adequate materials to protect the cover system
or the underlying bulk waste in the embankments against deep rooting by indigenous
greasewood (a species known to penetrate soils at other sites down to 60 feet) or other plants, or
against biointrusion by indigenous ants or mammals (e.g., with maximum documented burrowing
depths greater than the proposed cover thickness). Higher rates of infiltration are typically
associated with higher contaminant transport rates. Under Utah rules, infiltration should be
minimized [see UAC Rule R313-25-25(3) and (4)]. DEQ cannot determine the adequacy of the
DU PA until EnergySolutions accounts for greater infiltration through the cover system at the
proposed Federal Cell embankment due to biointrusion by plant roots and by animals.
DEQ Critique of DU PA v1.4, Appendix 21: See Interrogatories 10, 20, 28, and 71 for discussions
regarding enhanced transport due to biological processes.
Frost Damage—With the current proposed Federal Cell design, EnergySolutions should account
in modeling for substantial disruption of near-surface layers above and within the radon barriers
by frost, with accompanying decreases in ET and increases for initially low-permeability soil in
ET Cover Design Responses for the Clive DU PA Model
23 Feb 2018 19
both hydraulic conductivity and correlated α values, which could affect modeled infiltration rates
and radon release rates. UAC R313-25-25(3) and (4) require a licensee to minimize infiltration;
therefore, EnergySolutions must model infiltration under realistic long-term assumed site
conditions before DEQ can determine that this requirement has been met.
DEQ Critique of DU PA v1.4, Appendix 21: See Interrogatory 192 for discussions regarding
depth of potential frost impacts.
2.1.1 Interrogatory Response
Evapotranspiration Cover (ET Cover)
UDEQ argues that modeling of infiltration rates “must be conducted based on in-service
hydraulic properties and correlated log(α) and log(Ksat) values” as described by Dr. Craig Benson
in Volume 2, Appendix E, of the safety evaluation report (SER) prepared by SC&A (SC&A
2015a) [see interrogatories and comments 05/2, 21/2, 60/2, 90/2, 176/1 and B.4].
However, the hydraulic property recommendations and cover material naturalization presented in
Benson et al. (2011) and in Appendix E of SC&A (2015a) are inappropriate for the Clive Site.
When included in the model, they produce a model inconsistent with the observed site conditions
of Clive. This model can be considered “conservative” in terms of modeling groundwater
concentrations but dose results are lower for this model implementation than for the Clive DU
PA Model v1.4 (Neptune 2015d), which does not imply “conservative.” The rationale for not
using these naturalized or in-service cover properties in the Clive DU PA Model v1.4 (Neptune
2015d) are presented in this section.
The hydraulic property recommendations provided in Benson et al. (2011) are based on
measurements for samples from in-service covers made at 12 sites throughout the continental
United States. One element of the characterization of a site’s climate is the ratio of mean annual
precipitation to mean annual potential evapotranspiration. The magnitude of this ratio is
estimated to be 0.17 for Clive. Only one of the sites sampled by Benson et al. (2011) was
considered to be arid, having a ratio of 0.06. The mean value of this ratio for all sites sampled
was 0.51, with a highest value of 1.10. At two of the sites rainfall exceeded potential
evaporation, which is completely inappropriate for the arid conditions at Clive. All but one of the
sites that form the basis for the hydraulic property recommendations have much wetter
conditions than Clive.
The conceptual model of cover material “naturalization” for Clive based on the work of Benson et
al. (2011) is described in Appendix E of SC&A (2015a) as including changes in the hydraulic
behavior of the material following construction. These changes are characterized by increasing
values of hydraulic properties such as Ks and the hydraulic function α parameter (see Equation
(1)) that begin soon after cover completion. These changes are commonly attributed to soil
forming processes including wet-dry and freeze-thaw cycles, activity of roots and soil animals,
decomposition of organic matter by microbes producing compounds that tend to bind soil
particles into aggregates, and changes in cations adsorbed onto soil particle surfaces. In this
conceptual model these processes lead to the development of soil structure but not soil horizons.
ET Cover Design Responses for the Clive DU PA Model
23 Feb 2018 20
Under the wetter conditions considered by Benson et al. (2011), plant and animal activity are
greater than in an arid setting. These wetter conditions promote a faster rate of disruptive
processes due to plant and animal activity and in some cases freeze-thaw activity that were
shown by Benson et al. (2011) to lead to formation of an aggregated soil structure and natural
mixing of soil layers within a 5 to 10 year period at their study sites. Most importantly, the sites
considered by Benson et al. (2011) also lack significant eolian deposition.
This is not the case for a site like Clive. Recent field studies (Neptune 2015b) provide evidence
for a site-specific conceptual model of weak development of soil profiles (limited soil formation)
in a setting influenced by deposition of eolian silt in Holocene history. The Site is within a region
of significant eolian activity evidenced by locally thick accumulation of gypsum dunes west and
southwest of the Site and a laterally continuous layer of suspension fallout silts preserved
beneath the modern surface throughout the Clive Site. Clive quarry exposures examined in a
field study (Neptune 2015b) showed sections of eolian silts immediately below a modern
vegetated surface (Figure 3). The bottom of the eolian silt formed a gradational but definable
contact with the lake muds and marl below. The upper vegetated surface at the top of the eolian
section was distinct and noted as being partially indurated. In addition, buried soils were found in
the eolian and lake sediments below the Lake Bonneville lacustrine sequence.
The eolian deposits in the upper part of the stratigraphic section shown in Figure 3 represent a
10,000-year-old record of deposition and soil formation (Neptune 2015b). Primary soil features
developed over this time interval include an indurated Av-zone, and slight reddening of the silt
profile with local platy structure from formation of clays (Figure 4). These observations are
consistent with slow processes of soil formation in a high elevation semi-arid setting and
continuing suppression and burial of developing soils by a relatively low rate of deposition of
eolian silt. There is no evidence of soil structure development extensive enough to influence soil
hydraulic properties.
Observations of Holocene eolian silt throughout the Clive Site support a conceptual model of
long-term eolian deposition on a stable surface that promotes and preserves concurrent eolian
deposits which are only slightly modified by slow processes of soil formation. The past
Holocene depositional conditions at the Clive Site are promoted by a combination of extensive
wet playa sources of eolian source material to the west and southwest of the Clive Site and the
extremely low gradient paleo-Lake Bonneville surface surrounding the Site with sparse surface
vegetation and limited surface erosion. These conditions will persist at the Clive Site as long as
the lake levels remain below the site elevation. Rates of eolian deposition would be expected to
increase as future lakes approach the Site with increased formation of dunes (deposition of eolian
sands). Recurring lakes during ice ages (climate cycles) will rework and mix the eolian deposits
with aggrading clastic lake sediments. The expectation is that eolian deposits will drape and
slightly stabilize closure covers until future lakes return to the Clive Site. These studies at the
Clive Site were conducted by two highly experienced Ph.D. geologists, one of whom, Dr. Jack
Oviatt (https://www.k-state.edu/geology/faculty-staff/Oviatt.html), has decades of experience
working in the Lake Bonneville region.
Climate conditions and soil formation processes at the Clive Site contradict the assumptions of
rapid soil structure formation in the cover layers observed by Benson et al. (2011) at other sites
ET Cover Design Responses for the Clive DU PA Model
23 Feb 2018 21
and demonstrate the inapplicability of the conceptual model to the Clive Site. The one-size-fits-
all generalization of hydraulic properties does not appropriately represent the unique conditions
at the Clive Site. As described by Neuman et al. (2003) in NUREG/CR-6805:
Hydrogeologic models are by nature site-specific. Though there is an established (and
evolving) set of general hydrogeologic principles that apply to many sites on many scales,
they are insufficient to either describe (conceptualize) or quantify (model) the hydrogeology
of a particular site over a given range of scales. Because each site is unique, general
principles must always be supplemented by regional and site-specific data to be useful for
conceptualizing and modeling subsurface flow and transport at a site, regardless of purpose.
The soils at the Clive Site used for construction of the cover have been mapped, described,
sampled, and their properties measured. These soils have hydraulic properties that derive from
their origin and their environment, which are unique to the Clive Site. These properties are
therefore already “naturalized.”

Figure 3. Eolian silt in trench located at Clive Pit 29 overlying Lake Bonneville
sedimentary deposits (Neptune 2015b).
ET Cover Design Responses for the Clive DU PA Model
23 Feb 2018 22

Figure 4. An example of upper soil-modified eolian silt in Pit 29. Basal contact of the silt is
approximately located at the middle of the pick handle. It is a gradational contact
between eolian silt intermixed with regressive Lake Bonneville marl (bottom of the
pick handle).
UDEQ asserts that “the photographs in Figures 4 and 5 of Appendix 21 [reproduced herein as
Figure 3 and Figure 4] are inconclusive and provide no quantitative basis to support inferences
that structural development and alterations in hydraulic properties do not occur at Clive.
Structural development that occurs in covers due to pedogenesis generally is not visible at the
scale represented in these photographs. Moreover, the smearing that occurs in test pits can
obscure structure that is present.” The interpretation of the expert geologists that recent soil
formation is minimal at the Site was based on their direct field observation of the exposures, not
on examination of the photographs.
UDEQ states further, “If EnergySolutions wishes to use these analogs as evidence to support
hydraulic properties representing long-term conditions significantly different from NUREG/CR-
7028 (Benson et al. 2011), EnergySolutions should conduct appropriate measurements on these
in-place materials to demonstrate that the hydraulic properties are indeed different from the
abundance of data in NUREG/CR-7028.” The profiles excavated and examined by the geologists
were not analogs but observations of the Unit 4 material at the Clive Site while the “abundance
of data” referred to by UDEQ was acquired from other (nearly all wetter) sites throughout the
continental United States.
ET Cover Design Responses for the Clive DU PA Model
23 Feb 2018 23
Correlation Between Saturated Hydraulic Conductivity, Ks, and the van Genuchten α
Parameter
The hydraulic parameters α and Ks used in the Clive DU PA Model v1.4 are considered to be
statistically independent, implying they are theoretically uncorrelated and distributions can be
developed separately for each parameter. UDEQ believes this assumption is not valid based on
summaries of data presented by Benson and Gurdal (2013) as presented in Appendix E of the
Safety Evaluation Report (SER) (SC&A 2015a), Sections 4.1.1.1 and 4.4.1. Both references
examine data aggregated over soil textural classes, while the Clive DU PA Model v1.4 (Neptune
2015c) focuses on a single primary texture class (silty clay), and distributions for α and Ks are
developed specifically for that soil textural class. In this response, EnergySolutions provides new
analysis of data associated with Benson and Gurdal (2013) to demonstrate the importance of
considering soil textural class when attempting to quantify correlation between α and Ks.
In addition to providing detail on the different patterns seen in aggregated data versus
subpopulations defined by soil textural classes, several other fundamental statistical concepts
support a lack of evidence for building strong positive correlation into the distribution
development for α and Ks. There are several fundamental statistical concepts that should be taken
into account when considering generalization of a correlation observed in a particular data set to
a larger population or context: (1) correlation is designed as a summary of the linear association
between two random variables, and therefore linearity should be assessed before relying too
heavily on correlation as a summary of a relationship, (2) data aggregated over many
subpopulations with different ranges for the variables may exhibit a drastically different
correlation from the correlations observed within the subpopulations, (3) even when random
variables are statistically independent, random realizations of data will have estimated
correlations that are different from zero, and (4) points with extreme values for both variables
have high leverage and can be very influential on estimated correlations and estimated slopes.
Investigations into observed correlations between α and Ks are typically carried out on the
logarithmic scale because the relationship is expected to be more linear on the log scale than the
original scale. All plots and correlations provided in this response are based on the log10 scale for
both α and Ks.
A closer look at the Benson and Gurdal (2013) data
The data from the Benson and Gurdal (2013) paper are used to investigate the evidence of a
correlation. Benson (C. Benson, personal communication, October 11, 2017) supplied
EnergySolutions with the 253 pairs of Ks and α values presented in the 2013 paper, along with
information regarding the percent sand/clay/silt, as measured for the material for each sample.
The plot provided in Benson and Gurdal (2013), and included here as Figure 5, does not display
or take into account the soil texture associated with the measurements (i.e., data are aggregated
over soil textural classes). They reported a sample correlation coefficient for the aggregated data
of 0.462 and did not provide a measure of uncertainty or information about what correlation type
was calculated.
ET Cover Design Responses for the Clive DU PA Model
23 Feb 2018 24

Figure 5. Figure 3 from Benson and Gurdal (2013) showing the data requested by
EnergySolutions.
Using the USDA soil classification system (Jury and Horton 2004), each data point in the data
provided by Benson (2017) was assigned a soil textural class. This information is then used to re-
visualize the data (see Figure 6 and Figure 7) and to further investigate estimated correlation
coefficients. Pearson’s correlation coefficients are used unless otherwise stated.
ET Cover Design Responses for the Clive DU PA Model
23 Feb 2018 25

Figure 6. Estimated linear relationships between α and Ks for all observations (solid line)
and without the high-leverage points making up the clusters of points in the upper
right half of the plot with α greater than 0.10 kPa-1 (dotted line). The estimated
correlation changes from 0.627 for all the data to 0.384 for the restricted range.
●●●●
●
●
●
●●
●
●
●
●
●
●
●
●
●●
●●●
●
●
●●
●●
●
●●
●
●●
●
●●●●●●
●●
●
●●
●
●●
●●
●●●●
●
●●
●
●
●
●
●●
●●●●●●
●
●
●●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●●
●
●
●●
●
●●
●
●
●
●●
●
●
●
●
●
●●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●●
●●
●
●●●●
●
●
●
●
●●
●●●
●
●●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●●
●
●
●
●
●
●
●
●
●●
●
●●
●
●
●
●
●
●●●●●●●
●
●
●
●
●
●
●
●
●
●
●
●●●
●
●
●
●
●●
●
●
●●
●
●
●
●
●
●
r = 0.627
0.001
0.010
0.100
1.000
1e−07 1e−05 1e−03
Ks (cm/sec)
a (
1
/
k
P
a
)
SoilType
●
●
●
●
●
●
●
●
●
●
ClLo
Lo
LoSa
Sa
SaCl
SaClLo
SaLo
SiCl
SiClLo
SiLo
ET Cover Design Responses for the Clive DU PA Model
23 Feb 2018 26

Figure 7. Estimated linear relationships within each soil type for the data provided by
Benson (2017). Pearson’s correlation coefficients, and 95% confidence intervals, are
shown in Table 2 for the soil textural classes, and the individual soil classes are shown
in panels in Figure 8.
A few important observations from Figure 6 are:
1. The clusters of observations in the upper right with large value for Ks and α belong to
three distinct soil textural classes (sand, sandy-loam, and loamy-sand); they do not
represent pairs of Ks and α across a range of soil types. The estimated linear relationship
and correlation are heavily influenced by the location of the clusters relative to the cloud
of other points because they have high leverage on the estimated relationship (see the
difference in slope between the solid and dotted lines in Figure 6).
2. The relationship displayed in Figure 5 and the reported correlation coefficient based on
the aggregated soil textural classes are not indicative of the relationships observed within
individual soil textural classes. In fact, the observed relationships between Ks and α vary
widely across soil textural classes, as seen in Figure 7. The estimated correlation for the
aggregated data does not appear to be useful for individual soil textural classes and,
therefore, should not be generally applied to distribution development for PA models
meant to apply to specific soil textural classes.
3. The Pearson’s correlation coefficient and the estimated relationship shown in the plot in
Figure 5 do not match those for the data provided to EnergySolutions by Benson in 2017
●●●●
●
●
●
●●
●
●
●
●
●
●
●
●
●●
●●●
●
●
●●
●●
●
●●
●
●●
●
●●●●●●
●●
●
●●
●
●●
●●
●●●●
●
●●
●
●
●
●
●●
●●●●●●
●
●
●●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●●
●
●
●●
●
●●
●
●
●
●●
●
●
●
●
●
●●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●●
●●
●
●●●●
●
●
●
●
●●
●●●
●
●●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●●
●
●
●
●
●
●
●
●
●●
●
●●
●
●
●
●
●
●●●●●●●
●
●
●
●
●
●
●
●
●
●
●
●●●
●
●
●
●
●●
●
●
●●
●
●
●
●
●
●
0.001
0.010
0.100
1.000
1e−07 1e−05 1e−03
Ks (cm/sec)
a (
1
/
k
P
a
)
SoilType
●
●
●
●
●
●
●
●
●
●
ClLo
Lo
LoSa
Sa
SaCl
SaClLo
SaLo
SiCl
SiClLo
SiLo
ET Cover Design Responses for the Clive DU PA Model
23 Feb 2018 27
(compare Figure 5 to Figure 6). The data presented in Figure 6 are based on 253 pairs of
Ks and α, which matches the number of pairs expected based on Table 2 in Benson &
Gurdal (2013). However, it is clear the data do differ slightly (e.g., there are two points
near α of 0.1 kPa-1 in Figure 5 that do not appear in Figure 6). This discrepancy could not
be reconciled in this investigation. Regardless, the general conclusions and arguments do
not depend on the value of the correlation coefficient reported or small differences
between the two data sets.
Patterns in aggregated data do not apply to individual soil textural classes
For the correlation based on data aggregated over all soil textural classes to be an appropriate
description for individual soil textural classes, the relationships within soil textural classes
should look similar to the pattern observed in the aggregated data. The existence of a different
relationship for aggregated data (or a population) as compared to data separated by subgroups (or
subpopulations) is a common and well-studied statistical phenomenon known by names such as
Simpson’s Paradox and ecological fallacy (e.g., Wagner (1982) and Ch. 9 of Ramsey and
Schafer (2013)). This observation is often expected when, as shown in Figure 6 and Figure 7, the
ranges of values for each variable vary by group, sometimes greatly. This creates a situation
where subgroups with little correlation, or even negative correlation, can stack together to induce
positive correlation over the population. Potentially misleading conclusions and predictions can
result through the assumption that subgroups behave similarly to the aggregated population with
respect to the association between the two variables. The assumed correlation within a
population need not be the same as the correlations within its subpopulations.
The observed relationships between Ks and α within the data provided by Benson in 2017 clearly
vary in magnitude and sign for different soil textural classes. The data for each soil textural class
are plotted in their own panels (with different x and y axis scales) in Figure 8 to supplement
information in Figure 7. Table 2 provides the estimated correlation coefficients and their
associated 95% confidence intervals for each soil textural class. There is no evidence in these
data that the pattern in the aggregated data across soil types should apply at the scale of a
particular waste cover primarily made up of one soil textural class. Benson and Gurdal (2013)
suggest a large positive correlation exist for data aggregated across soil textural classes, but, as
shown above and below, it is important to consider the soil textural class when introducing the
assumption of correlation into distribution development for the PA model.
Correlation coefficients can be useful statistics to provide the direction and strength of a linear
relationship between two random variables (Ramsey and Schafer 2013). Few of the soil textural
classes in the data from Benson (2017) seem to exhibit clear linear associations, thus limiting the
usefulness of the correlation coefficient as a summary measure of the relationship. Statistical
hypothesis tests can be helpful if evidence against a particular null hypothesis is of interest, but
confidence intervals provide more information about uncertainty and are not tied to a particular
null hypothesis. The 95% confidence interval can be interpreted as a range of plausible
correlations that could be obtained from different random datasets of the same size from the
same population.
ET Cover Design Responses for the Clive DU PA Model
23 Feb 2018 28
Table 2. Estimated Pearson’s correlation coefficients between Ks and α on the log scale by
soil textural class, and associated 95% confidence intervals, calculated using the data
provided by Benson in 2017 data.
Soil Textural
Class
Estimated
correlation
95% lower
limit
95% upper
limit
Clay Loam -0.01 -0.64 0.62
Loam -0.15 -0.49 0.23
Loamy Sand 0.79 -0.06 0.98
Sand -0.28 -0.79 0.48
Sandy Clay 0.96 -0.02 1.00
Sandy Clay Loam 0.25 -0.03 0.49
Sandy Loam 0.80 0.63 0.90
Silty Clay -0.20 -0.97 0.94
Silty Clay Loam 0.39 0.10 0.62
Silty Loam 0.36 0.12 0.55

On the log scale, seven of the ten soil textural classes represented in the data from Benson and
Gurdal (2013) have negative estimated correlations or have 95% confidence intervals that
overlap zero. The widths of the confidence intervals demonstrate the substantial uncertainty in
the correlations of the underlying populations, given the available data.
It is also important to combine the information in Table 2 with that in Figure 7 and Figure 8. For
example, the sandy loam soil class (SaLo) has a point estimate with a large magnitude (0.80) and
a confidence interval suggesting a positive correlation between Ks and α (0.63, 0.90), but a closer
look at the plot of the data for that soil textural class in Figure 8 calls into question the validity of
that conclusion based on the cluster of points with high leverage driving the correlation.
For the Clive DU PA Model v1.4, the objective is to develop distributions for the Unit 4 soil
described as Silty Clay, which shows no evidence of a positive correlation in the four samples in
the Benson and Gurdal (2013) data (see Figure 8, Figure 9, and Table 2).
ET Cover Design Responses for the Clive DU PA Model
23 Feb 2018 29

Figure 8. Estimated linear relationships by soil textural class for the data plotted in Figure
6 and Figure 7. Note the x and y axes are allowed to change among panels and are on
the log10 scale.
●
●
●●●
●
●
●●●
●
●
●●
●
●
●
●
●●
●●
●
●
●
●
●
●●
●
●
●●
●
●
●
●
●
●
●
●
●●
●●
●
●
●
●
●●
●
●
●
●
●
●●●●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●●●●
●
●●
●
●●
●
●
●●
●
●
●
●
●●
●
●●
●
●●
●
●
●
●
●
●
●
●
●
●
●●●
●●
●●
●●
●●
●●
●
●
●
●
●●●
●
●●
●●
●
●●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●●●
●●
●
●
●
●●
●
●
●●
●●●●
●
●●●●
●
●
●
●
●●●
●
●●
●●
●
●
●
●
●
●
●
●
●
●●
●●●●●●●●●●●●●●
●
●●●●●
●●
●●●●
●
●●
●
●
●
●●●●
●●
●
●
●●
SiClLo SiLo
SaCl SaClLo SaLo SiCl
ClLo Lo LoSa Sa
1e −06
1e −04
1e −07
1e −06
1e −05
1e −07 1e −07 1e −05 1e −06 1e −04 1e −08 1e −07 1e −06
1e −07 1e −05 1e −06 1e −04 1e −05 1e −04 0.001 0.010
0.001
0.001
0.010
0.100
1.000
0.001
0.010
0.001
0.010
0.001
0.010
0.001
0.010
0.100
0.001
0.001
0.010
Ks (cm/sec)
a (
1
/
k
P
a
)
ET Cover Design Responses for the Clive DU PA Model
23 Feb 2018 30

Figure 9. Estimated linear relationship on the log scale between α and Ks for Silty Clay.
Pearson’s correlation coefficient is -0.2, with a very wide 95% confidence interval of
(-0.97, 0.94).
Observed sample correlations from independent random variables
Assessing evidence of a theoretical correlation between two variables based on relatively few
samples, as is the case for the separate soil textural classes, is very difficult. Figure 10 illustrates
estimated linear relationships that could be observed even when the two variables are truly
independent. The panels are created by taking random draws from independent standard normal
distributions with sample sizes of 6. Note the different conclusions one might draw about
correlation between the two random variables just from looking at one of the sixteen panels. The
difficulty of assessing evidence for a theoretical correlation due to small sample sizes can lead to
the desire to increase the number of samples used by aggregating the data of immediate interest
with that from related groups. As discussed above, this should be done with caution.
●
●
●●
r = −0.2
0.001
1e−08 1e−07 1e−06
Ks (cm/sec)
a (
1
/
k
P
a
)
ET Cover Design Responses for the Clive DU PA Model
23 Feb 2018 31

Figure 10. Estimated linear relationships based on 16 realizations of six random pairs of
observations that come from statistically independent variables (x and y were drawn
independently from two standard normal distributions).
Summary
In summary, there is not evidence of a strong linear relationship between α and Ks that should be
incorporated into PA distributions, particularly for the Unit 4 soil within the Silty Clay soil
textural class found at the Clive Site. Additionally, incorporating correlation into distribution
development without adequate empirical and theoretical evidence should be done with care to
avoid excluding combinations of values of the two variables that are actually plausible (i.e., to
avoid inadvertently excluding realistic scenarios). The argument for correlation between Ks and
α suffers from some common statistical misconceptions and misuses of correlation, as described
in detail in the response. It is clear from the exploratory analysis conducted here that soil textural
classes should be taken into account when assessing possible correlation between Ks and α for a
particular site. The two variables exhibit very different relationships within different soil textural
classes—a clear indication that the aggregated correlation should not be applied to the individual
soil textural classes. In the particular soil textural class used in the Clive DU PA Model v1.4,
Silty Clay, there are very few samples available in Benson and Gurdal (2013) and no evidence of
a positive linear relationship between the Ks and α parameters.
●
●
●●●
●
●
●
●●●
●
●
●●●●
●
●●
●●
●
●
●
●●
●
●●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●●●
●●●●
●●
●
●
●●●
●
●
●●
●●●
●●●●●●
●
●●●
●
●
●
●●
●●●
●
●
●●
●●
●
●●
●●●
13 14 15 16
9 10 11 12
5 6 7 8
1 2 3 4
−2 −1 0 1 2 −2 −1 0 1 2 −2 −1 0 1 2 −2 −1 0 1 2
−2.5
0.0
2.5
5.0
−2.5
0.0
2.5
5.0
−2.5
0.0
2.5
5.0
−2.5
0.0
2.5
5.0
X
Y
ET Cover Design Responses for the Clive DU PA Model
23 Feb 2018 32
Clay Liner
The Federal Cell is constructed on top of a compacted clay liner. Disposal involves placing waste
on a prepared clay liner that is approximately 2.5 m (8 ft) below the ground surface. The liner is
constructed of compacted local clay, i.e., Unit 4 material. Porosity and bulk density values for
the clay liner are assumed to be the same as for the Radon Barrier Clays, as these clays are all
compacted, unlike the surface and ET layer Unit 4 material. Properties of nine Unit 4 cores
acquired at the Clive Site were determined by Bingham Environmental (1991, 1994). Grain size
distributions were determined for nine cores and saturated hydraulic conductivities and water
retention relations determined for two of the cores.
The GoldSim software platform cannot calculate flow of water from the cover surface to the
water table using the unsaturated flow equation the way that variably saturated numerical flow
models (e.g. HYDRUS) can. GoldSim includes flow by assigning a flow rate or flux for a
realization from a statistical distribution of fluxes. In this context a GoldSim realization is “a
single model run within a Monte Carlo simulation. It represents one possible path the system
could follow through time” (GTG 2014). This water flux distribution is developed from a
collection of results obtained from simulations using more complex models that incorporate daily
changing weather conditions at the surface and that allow for many flow models to be run
encompassing a range of hydraulic input parameter values. The resulting statistical distribution
represents the net infiltration or water flux at the top of the waste. For any given realization, the
same value of the flux drawn from the distribution is assigned to each cell representing the
pathway through the waste as inflow to the cell and outflow from the cell. The clay liner is
represented in the model by four cells below the waste. The outflow from the bottom cell
representing the waste is connected to the inflow of the top cell of the clay liner. Each of the clay
liner cells are assigned the same flux as the waste.
The layer above the clay liner with the lowest hydraulic conductivity is the Upper Radon Barrier.
As described in Section 2.4.1, the distribution for the hydraulic conductivity of this layer has a
minimum value of 4.32 × 10-3 cm/day, which is the engineering design specification. The
distribution has 1st, 50th, and 99th percentile values of 0.65 cm/day, 3.8 cm/day, and 52 cm/day,
respectively, which are from a range of in-service (“naturalized”) clay barrier Ks values
described in Section 6.4, p. 6-12 of Benson et al. (2011). Given this distribution, any Ks value
chosen from the distribution other than the minimum is a value greater than the initial
construction specification. As described above, each cell from the bottom of the waste to the top
of the upper aquifer is assigned the same water flux drawn from the distribution for that
realization, including the clay liner cells. As larger fluxes are drawn representing the variability
due to naturalization of the cover materials, those same fluxes are applied to the clay liner
representing naturalization of that layer.
UDEQ has argued that the modeling approach described above does not account for liner
degradation over time and that there are problems with the assumed liner hydraulic conductivity
values (see the responses to UDEQ Interrogatories 05/2 and 90/2). With respect to the first issue,
liner degradation is accounted for by using a statistical distribution of water fluxes for the
modeling as described above. This distribution takes into account a range of possible values for
saturated hydraulic conductivities greater than the engineering specification. The next issue is
ET Cover Design Responses for the Clive DU PA Model
23 Feb 2018 33
immaterial because the water flux through the clay liner cells depends only on the value of the
water flux at the top of the waste drawn from the water flux distribution. This means that liner
hydraulic properties are assumed to be the same as the radon barriers.
UDEQ states that modeling of the clay liner “should employ hydraulic parameters representative
of a clay liner.” They go further to recommend a reference for clay hydraulic properties not
based on samples from the Clive Site. As described above, hydraulic properties for the clay liner
were obtained from testing of two cores acquired from the Clive Site of Unit 4 silty clay
(Bingham Environmental (1991, 1994)).
The last issue raised by UDEQ with respect to the clay liner is that van Genuchten hydraulic
function α parameter and the saturated hydraulic conductivity values must be correlated (also
Interrogatories 05/2, 21/2, 60/2, 90/2, 153/2, and Comment B.4). See the response to
Interrogatory 05/2.
Infiltration
See the response to Interrogatory 05/2 Evapotranspiration Cover (ET Cover).
Erosion of Cover
UDEQ points to issues on erosion raised in Section 4.4.2 of the SER (SC&A 2015b). These
issues are addressed in the response to Interrogatory 71/1.
Calculations to evaluate the stability of the design with respect to gully erosion for the ET cover
of the Class A West cell were provided in Appendix D of EnergySolutions (2015). Similar
calculations for the Federal Cell are presented in the response to Interrogatory 71/1.
Effect of Biologicals on Radionuclide Transport
SWCA Environmental Consultants (SWCA) was contracted by EnergySolutions to acquire field
data at the Clive Site to identify representative study areas and to collect data to “document the
diversity and composition of plant and animal species and to quantify soil mixing by burrowing
mammals and ants associated with each vegetation community” (SWCA 2013). In addition, six
excavations were completed and surveyed to quantify the aboveground and belowground size of
the dominant plant species, and the maximum rooting depth and width of root of the dominant
species. This work was focused on characterizing the biotic aspects of the ET cover system at the
Clive Site for the DU PA Model (SWCA 2011). SWCA is a nationwide firm of environmental
professionals operating since 1981 and is well known for their work in multiple environmental
sectors, including ecology.
Field studies and analyses were conducted by SWCA at the Clive Site including quantifying
vegetation, small mammal distributions, and mammal burrow and ant mound size and densities
at the Clive Site and nearby ecologically analogous sites. Additionally, several excavations in the
Unit 4 soils were completed to estimate ant mound depth and soil volume transported to the
surface by ants for the DU PA Model (SWCA 2011).
ET Cover Design Responses for the Clive DU PA Model
23 Feb 2018 34
SWCA describes the ecological context for the ET cover design at the Clive Site in their
responses to Round 1 Interrogatories from UDEQ. They note that the design for Clive is
consistent with both published ET cover-system recommendations (ITRC 2003; Peace et al.
2004; Scanlon et al. 2005) and with the site-specific climate and ecology documented through
the studies conducted by SWCA (2011, 2012). As summarized from SWCA (2013), the ecology
at the Clive Site can be characterized by the following:
• The dominant shrub species are small in stature compared to those of the same species in
deeper more fertile soils and/or in areas of greater annual precipitation.
• The distributions, densities, and stature of plant species at the Site are determined by high
soil salinity, soil pH, low fertility, and aridity.
Based on data collection and analysis and the proposed cover design, SWCA evaluated the
potential for disturbance by plant roots, mammals, and ants to result in increased infiltration
(SWCA 2013). The Surface and Evaporative Zone layers are composed of native Unit 4 material
and revegetated with native, locally adapted plant species. The target vegetation community is
designed to represent the diversity and density of the native vegetation (SWCA 2013). With this
combination of local materials and plants, the hydrology and ecology of the upper layers of the
cover will be similar to native undisturbed areas at the Site.
Based on evidence from analogous sites detailed in SWCA (2013), the densities of plant roots
and burrows will be sparse in these layers and infiltration would be minimally affected. Total soil
disturbance due to mammal burrowing was estimated by SWCA (2013) for two cases of
vegetation conditions. These cases are considered because these authors noted an association
between vegetation conditions and mammal activity. Under expected vegetation conditions, soil
volume disturbed by mammal activity was less than 1/100th of a percent of the total soil volume.
Under a worst-case scenario of vegetation dominated by greasewood, the soil volume disturbed
was 1/10th of a percent of the total soil volume of the ET cover. Their conclusions, supported by
detailed evidence presented in SWCA (2013), were that the Surface and Evaporative Zone layers
would be deep enough to allow for ant activity and some mammal burrowing and soil movement
without compromising the functions of the lower layers.
SWCA concludes that, despite the low density and small size of plants and mammals at the Site,
local plants and animals could penetrate the Frost Protection layer but “the amount of soil
disturbance that could potentially occur on the ET cover is minute, compared to the total soil
volume on the ET cover. Similarly, the amount of water infiltration that could occur in
association with biointrusion of the ET cover is also minute.”
ET Cover Design Responses for the Clive DU PA Model
23 Feb 2018 35

Figure 11. Borrow soil cross-section below a greasewood plant shows the compacted clay
layer at approximately 60-cm depth. Roots extend laterally and do not penetrate the
compacted layer (SWCA 2011).
The potential for greasewood roots to extend to depths greater than the cover layers is recognized
by SWCA. SWCA (2013) notes, however, that “the aboveground mass of greasewood plants on
and near Clive is consistent with low water availability.” Roots are expected to follow water
availability and not penetrate clay layers. This behavior was documented by SWCA (2013) in an
excavation at the Clive Site showing lateral root growth of a greasewood plant at the Site. Roots
were seen to extend vertically downward to the top of a clay layer at approximately 60 cm depth
(Figure 11). At the clay layer roots extended only laterally and did not penetrate the layer. The
explanation for this rooting behavior from the observation of SWCA, consistent with other
research, is that infiltrating water would tend to perch on lower permeability clay layers causing
plant roots to grow laterally following water availability (Groeneveld 1989).
SWCA (2013) conclude from their observations at the Clive Site and ecological analog sites that
the low fertility and alkalinity of the soils and the aridity of the climate limit the growth of plants
at the Site and would inhibit the development of large deep-rooted plants in the cover. Their
opinion is that, while greasewood will eventually become established at the Site, roots will
follow available water and remain within the Surface, Evaporative Zone, and Frost Protection
ET Cover Design Responses for the Clive DU PA Model
23 Feb 2018 36
Layers. Additional evidence supporting the projection of limited greasewood root penetration at
the Site presented by SWCA (2013) includes:
• The water table depth exceeds maximum levels accessible by greasewood.
• There is little or no capillary rise above the water table.
• The root to shoot stature of greasewood at Clive is not sufficient to root deeply.
• Regardless of plant stature, the majority of root biomass is concentrated in the upper soil.
• Nutrients are concentrated near the surface in desert soils.
Local Unit 4 material is used for the upper layers of the cover, and these layers will be vegetated
with native, locally adapted plant species. The hydraulic properties of these layers are expected
to be similar to those of undisturbed areas of the Site, which will experience the same biotic
activity as the upper layers of the cover. Using hydraulic conductivity measurements of native
Unit 4 samples from the Site as inputs to the model accounts for the influence of biotic activity
on infiltration.
Frost Damage
Calculations of frost depth at the Clive Site were made by Hansen, Allen, and Luce (HAL), a
professional engineering firm. These calculations were provided to UDEQ in Appendix E of the
Updated Site-Specific Performance Assessment (Revision 2) (EnergySolutions 2015).
HAL chose to use the modified Berggren equation to calculate frost depth because of its long-
established use and acceptability by the engineering community. This method, first presented by
Berggren in 1943 and further refined by Aldrich and Paynter in 1953, was later adopted by the
US Army Corps of Engineers and other agencies as their preferred method for frost depth
determination (Departments of the Army and the Air Force 1988).
The frost depths calculated as part of this analysis give results that are in line with the depths of
cover and frost protection proposed in the EnergySolutions ET Cover system design. The
proposed radon barrier begins at depths ranging from 30 inches to 42 inches, which provides
frost protection for the calculated 100-year frost penetration depth of 22.4 inches to 27.8 inches
for the top slope and side slope, respectively.
A statistical distribution for saturated hydraulic conductivity of the radon barriers used for
modeling flow in the cover was developed from in-service properties based on the results of
Benson et al. (2011). Use of these properties as inputs to the flow models takes into account
changes in hydraulic properties due to freeze/thaw cycles assumed by Benson et al. (2011). See
the response to Interrogatory 21/2 for a discussion of the development of the saturated hydraulic
conductivity distribution for the radon barriers (Section 2.4.1).
Early degradation of cover layer properties described by Benson et al. (2011) is not consistent
with soil formation observations at the Clive Site. Recent field studies (Neptune 2015b) provide
evidence for a site-specific conceptual model of weak development of soil profiles (limited soil
formation) in a setting influenced by low rates of deposition of eolian silt in the Holocene
history. See the response to Interrogatory 05/2 Evapotranspiration Cover (ET Cover) (Section
2.1.1) for a detailed discussion of soil formation at the Clive Site.
ET Cover Design Responses for the Clive DU PA Model
23 Feb 2018 37
2.2 Interrogatory CR R313-22-32(2)-10/3: Effect of Biologicals on
Radionuclide Transport
DEQ Conclusion from April 2015 SER, Appendix C:
As discussed in the DU PA SER (Section 4.4.3), EnergySolutions has not shown that the cover
system is sufficiently thick or designed with adequate materials to protect the cover system or the
underlying bulk waste in the embankments against deep rooting by indigenous greasewood (a
species known to penetrate soils at other sites down to 60 feet) or other plants, or against
biointrusion by indigenous ants or mammals (e.g., with maximum documented burrowing depths
greater than the proposed cover thickness). Higher rates of infiltration are typically associated
with higher contaminant transport rates. Under Utah rules, infiltration should be minimized [see
UAC Rule R313-25-25(3) and (4)]. DEQ cannot determine the adequacy of the DU PA until
EnergySolutions accounts for greater infiltration through the cover system at the Federal Cell
embankment due to biointrusion by plant roots and by animals. Therefore, this interrogatory
remains open.
DEQ Critique of DU PA v1.4, Appendix 5: EnergySolutions/Neptune retain the same
assumptions with respect to biointrusion depths and potential impact on infiltration in v1.4 as
were provided in v1.2.
In v1.4 Appendix 5 (p. 33), EnergySolutions indicates that root water uptake was modeled
assuming the roots extended to the bottom of the evaporative zone layer and that rooting density
decreased with depth. This text seems to contradict the statement in v1.4 Appendix 5 (p. 33) that
root distribution was modeled as extending into the frost protection layer with a maximum depth
of 31 inches (80 cm). The base of the evaporative zone would be at 18 inches. Figure 1 indicates
that the roots cease within the frost protection layer. The impact of the rooting depth in v1.4 is to
remove water from the system and thereby reduce the infiltration rates. The concern raised by
the interrogatory is related to the roots creating preferential pathways and thereby increasing the
infiltration.
DEQ Critique of DU PA Appendix 21: EnergySolutions/Neptune state (p. 15):
It is important to recognize how the range of rooting depths discussed in the comment
actually relates to what was used as a maximum rooting depth in GoldSim Models v1.2
and v1.4. A maximum root depth of 5.7 meters (18.7 ft) (Robertson 1983) is used in the
Model, so the Model already assumes that roots extend beyond the radon barrier. In
addition, v1.4 of the GoldSim Model assumes increased permeability, correlation
between saturated hydraulic conductivity and the hydraulic function alpha parameter,
and homogenization of the cover materials, with no physical barriers to either plant roots
or infiltration.
It is unclear how the specification of the rooting depth in GoldSim is particularly relevant to the
concern expressed in the comment pertaining to potential increased infiltration rates due to
biointrusion of plants and animals. The rooting depth in GoldSim is related to the depth of
contaminant uptake, redistribution of contamination, and assimilation of contaminants once the
plant dies rather than changes to the hydraulic properties that would allow greater infiltration.
Plant roots will almost certainly extend downward and into the radon barrier. These roots will
then penetrate into the underlying waste if water is available in the waste. As described Benson
et al. (2008), roots were observed in the radon barrier in the caisson lysimeters exhumed at
ET Cover Design Responses for the Clive DU PA Model
23 Feb 2018 38
Monticello in 2008. These were at depths of 1.6–1.9 m bgs (see Figure 10-1 below). The roots
desiccated the radon barrier, causing large cracks and an increase in Ks.
Furthermore, EnergySolutions has used a homogeneous cover profile in the most recent
simulations. This was not the intent of the previous comments and approach outlined in
Appendix E to the April 2015 SER and was misconstrued from the parameter recommendations
provided in Appendix E. The cover profile should retain a layered structure representative of the
materials planned for each layer, but with the hydraulic properties of each layer adjusted to
reflect pedogenesis. The parameters in the 2015 recommendations were presented as a guide
for reasonable ranges consistent with the recommendations in NUREG/CR-7028 (Benson et al.
2011).
2.2.1 Interrogatory Response
As described in the response to Interrogatory 05/2, SWCA acquired field data at the Site and
conducted extensive literature reviews to “document the diversity and composition of plant and
animal species and to quantify soil mixing by burrowing mammals and ants associated with each
vegetation community” (SWCA 2013). Based on data collection and analysis and the proposed
cover design, SWCA evaluated the potential for disturbance by plant roots, mammals, and ants
to result in increased infiltration (SWCA 2013) and determined that soil disturbance and
increased infiltration due to biotic activity would be minute. Their evaluation does not support
UDEQ’s assertion that biointrusion of plants and animals will substantially increase infiltration.
See the response to Interrogatory 05/2: Radon Barriers.
UDEQ correctly points out an error in the description of the maximum modeled rooting depth in
Section 12.1, p. 33 of Appendix 5 of the Clive DU PA Model v1.4 (Neptune 2015c). The correct
description is on p. 36 of Section 12.3 of the same document: “Root distribution was modeled as
extending into the frost protection layer with a maximum depth of 31 inches (80 cm). Root
density was modeled as decreasing linearly with depth.” This description is consistent with the
maximum rooting depths and rooting density shown for the model (Figure 1 of Neptune (2015c))
and observed in excavations at the Site (Figure 9 of Neptune (2015c)).
UDEQ makes the inference that observed root penetration and cracking of clay layers at the
Monticello Site means that similar root penetration and degradation of the radon barriers will be
likely at Clive. Monticello and Clive are not comparable sites with respect to rooting depths.
SWCA (2013) describes the important ecological differences between Monticello and Clive:
• Monticello receives approximately 50% greater average annual precipitation than Clive
(15.4 in).
• The Monticello ET cover consists of clay-loam to sandy-loam soils that are less alkaline
and more fertile than the saline, alkaline silty-clay soils at Clive (Waugh et al. 2008).
• The native vegetation at Monticello is dominated by big sagebrush shrublands and
grasslands that are more diverse and of larger stature—with greater target plant densities
and cover for the ET cover—than those proposed at Clive.
The last issue of this interrogatory raised by UDEQ is that the hydraulic parameter distributions
and correlations used for modeling of flow in the cover system based on the recommendations of
ET Cover Design Responses for the Clive DU PA Model
23 Feb 2018 39
Benson et al. (2011) were not how the recommendations were intended by UDEQ to be
implemented.
Furthermore, EnergySolutions has used a homogeneous cover profile in the most recent
simulations. This was not the intent of the previous comments and approach outlined in
Appendix E to the April 2015 SER and was misconstrued from the parameter
recommendations provided in Appendix E. The cover profile should retain a layered
structure representative of the materials planned for each layer, but with the hydraulic
properties of each layer adjusted to reflect pedogenesis.
The proposed cover system shown in Figure 1 is composed of layers that either store and release
water or act as a barrier. The proposed system does not include geomembranes. UDEQ argues
that input parameter distributions for the α, n, and Ks hydraulic parameters be developed using
recommended ranges for hydraulic properties from NUREG/CR-7028 (Benson et al. 2011).
Benson et al. (2011) provide these recommendations in Section 10.2:
• The saturated hydraulic conductivity of fine-textured earthen storage and barrier layers
can be assumed to range between 1 × 10-7 m/s and 5 × 10-6 m/s.
• The porosity of earthen storage and barrier layers will likely range between 0.35 and
0.45.
• The α-parameter in the van Genuchten equation, which is used to describe the SWCC
(soil water characteristic curve) for hydrologic simulations, varies between 0.01 and
0.33 kPa-1 for field-scale barrier and storage layers.
• The n-parameter in van Genuchten’s equation, which is used to describe the SWCC for
hydrologic simulations, varies over a very small range (typically between 1.2 to 1.4).
Note that in assigning hydraulic properties no distinction is made by Benson et al. (2011)
between storage and barrier layers. Based on their conceptual model of formation of soil
structure in cover systems, there is no difference in the in-service properties of what were
constructed as storage layers or barrier layers. The only distinctions in properties between depth
intervals in the flow model used for the cover simulations are the initial conditions, which are
minimized by long simulation times, root density (which is maintained as a constant throughout
the cover depth), and the hydraulic properties of the intervals. Given their conceptual model that
makes no distinction between the properties of storage and barrier layers, the cover can no longer
be represented by a layered system. UDEQ objects to this homogeneity in hydraulic properties in
this interrogatory, but that is the only logical outcome of applying the Benson et al. (2011)
method to the ET cover.
2.3 Interrogatory CR R317-6-2.1-20/2: Groundwater Concentrations
DEQ Critique of DU PA v1.4, Appendix 10: See Interrogatory 201 for further discussion.
DEQ Critique of DU PA Appendix 21: EnergySolutions/Neptune state (p. 16):
The conceptual model of cover “naturalization” described in Appendix E of the SER
(SC&A 2015) is that plant and animal activity and freeze-thaw cycles result in
disturbance and mixing of soil layers in the upper portion of the cover system subject to
their influences. The extent of the influence of these processes decreases with depth of
ET Cover Design Responses for the Clive DU PA Model
23 Feb 2018 40
roots, animal burrowing, and frost penetration. This conceptual model does not maintain
the designed functions of store and release layers and barrier layers to reduce net
infiltration. Using this conceptual model, the upper portion of the soil profile subject to
naturalization processes is considered to be homogeneous with respect to the hydraulic
properties affecting net infiltration. For the Clive Site, the hydraulic properties of the
waste below the cover are modeled as Unit 3 material and would be subject to the same
naturalization processes as the materials used to construct the cover.
With this conceptual model, the depth to the waste would be reduced by erosion but the
net infiltration will not vary. The net infiltration is determined by climate and hydraulic
properties. If the hydraulic properties are assumed to be homogeneous and determined
by climate and biotic activity, loss of material from the surface of the cover will not
change the net infiltration.
EnergySolutions has used a homogeneous cover profile in the most recent simulations. This was
not the intent of our previous comments and approach outlined in Appendix E to the April 2015
SER and was misconstrued from the parameter recommendations provided in Appendix E. The
cover profile should retain a layered structure representative of the materials planned for each
layer, but with the hydraulic properties of each layer adjusted to reflect pedogenesis. The
parameters in the 2015 recommendations were presented as a guide for reasonable ranges
consistent with the recommendations in NUREG/CR-7028.
EnergySolutions has conducted a series of analyses to evaluate the impact of erosion on
percolation rates from the cover. In one case, the simulation included loss of 1.2 m of cover soil.
EnergySolutions reports that percolation rates obtained for the full thickness cover and a cover
eroded by 1.2 m are essentially the same.
This is not logical given that the soil in the cover is required to store the water during cooler and
wetter periods, and then release the water during drier periods. The proposed cover is 1.52 m
thick. If the cover thickness is reduced by 1.2 m via erosion, then the soil water storage capacity
of the cover will be reduced by approximately 80%, and the percolation should change
accordingly. This result without supporting analysis makes all of the HYDRUS modeling suspect.
Additional quantitative and mechanistic evidence is needed to support the outcomes in this part
of the report. Water balance graphs, which depict the temporal variation in water balance
quantities (rather than a water balance quantity chart) could be used to illustrate whether the
outcomes are reasonable. Water balance graphs typically are created using daily output
predicted from a water balance model and show the seasonal variation in each water balance
quantity. Examples of water balance graphs are shown in Figure 20-1. These graphs depict
actual water balance data; water balance graphs from a model prediction would be similar. The
soil water storage record in the water balance graph would be compared to the soil water storage
capacity of the eroded profile.
Significantly higher technetium-99 (99Tc) concentrations were obtained for percolation rates
predicted using the hydraulic properties EnergySolutions developed with the recommended
approach (Appendix E, April 2015 SER) relative to the percolation rates predicted in their
previous analyses (Figure 20-2). The differences are very large, which is difficult to understand
given that the percolation rates predicted for the cover are on the order of 1 mm/yr and are
consistent with percolation rates measured for covers placed at other sites in the region.
If the impact on groundwater concentrations is this sensitive to percolation rates on the order of
1 mm/yr, then detailed assessment and proof of the cover design should be particularly
important. EnergySolutions should consider installing a lysimeter to confirm that the cover
modeling is reliable.
ET Cover Design Responses for the Clive DU PA Model
23 Feb 2018 41
2.3.1 Interrogatory Response
In this interrogatory UDEQ proposes contaminant transport from the Federal Cell beginning with
the formation of gullies on the cover leading to increased infiltration rate and leaching of
contaminants into the upper aquifer.
The first part of this interrogatory points to soil erosion issues raised in Section 4.4.2 of the
Safety Evaluation Report (SC&A 2015a). UDEQ expresses concerns that gullies will form and
enhance radon diffusion, deep infiltration, and contaminant transport. EnergySolutions, however,
has provided plans for both ecological and engineering measures to address these concerns by
minimizing gully formation. Demonstration of the stability of the cover is addressed in the
response to interrogatory 71/1.
At the request of UDEQ, EnergySolutions has simulated the impact of erosion on net infiltration
using the conceptual model of in-service (naturalized) covers described in Appendix E of the
SER (SC&A 2015b). These simulations showed that a loss of 1.2 m would have little effect on
the net infiltration rate. UDEQ argues that “This is not logical given that the soil in the cover is
required to store the water during cooler and wetter periods, and then release the water during
drier periods.”
The conceptual model of the in-service cover (naturalization) described in Appendix E of the
SER (SC&A 2015b) and Benson et al. (2011) is that plant and animal activity and freeze-thaw
cycles result in disturbance and mixing of soil layers in the upper portion of the cover system
subject to their influences. In this conceptual model, these processes lead to the development of
soil structure but not soil horizons. The extent of the influence of these processes decreases with
depth of roots, animal burrowing, and frost penetration. This alternative conceptual model of soil
formation is discussed in the response to interrogatory 20/2.
UDEQ maintains that input parameter distributions for the α, n, and Ks hydraulic parameters for
water flow modeling be developed using recommended ranges for hydraulic properties from
NUREG/CR-7028 (Benson et al. 2011). However, these ranges are considered unrealistic. Based
on UDEQ’s conceptual model of formation of soil structure in cover systems, there is no
difference in the in-service properties of what were constructed as storage layers or barrier layers
(Benson et al. 2011). Normally, the distinction in properties between depth intervals in the flow
model that constitute layers used for the cover simulations are the hydraulic properties of the
intervals. The UDEQ conceptual model does not maintain the differences in hydraulic properties
that provide the designed functions of store and release layers and barrier layers. Thus, the cover
can no longer be represented by a layered system. UDEQ objects to this homogeneity in
hydraulic properties in this interrogatory, but that is the only logical outcome of applying the
Benson et al. (2011) method to the ET cover.
Using this conceptual model, the upper portion of the soil profile subject to naturalization
processes is considered to be homogeneous with respect to the hydraulic properties affecting net
infiltration. For the Clive Site, the hydraulic properties of the waste below the cover are modeled
as Unit 3 material and would be subject to the same naturalization processes as the materials
used to construct the cover. As soil is lost due to erosion, disturbance due to biotic activity and
ET Cover Design Responses for the Clive DU PA Model
23 Feb 2018 42
freeze-thaw extends deeper to maintain the same thickness of “naturalized” soil characterized by
the same ranges of hydraulic properties. The net infiltration is determined by climate and
hydraulic properties. If the hydraulic properties are assumed to be homogeneous and determined
by climate and biotic activity, the loss of material from the surface of the cover due to erosion is
compensated by a deepening of the naturalized soil profile. The thickness of the naturalized soil
remains the same and net infiltration will not change. Given the approximately 10-meter depth of
burial of the DU, the logical projection of this conceptual model is that this process of soil loss
from the surface and deeper extension of soil formation would continue for a very long time
before the soil above the waste was thin enough to influence the net infiltration. The hydraulic
property recommendations and cover material naturalization presented in Benson et al. (2011)
and in Appendix E of SC&A (2015a) are not reasonably applicable for the Clive Site (see the
response to Interrogatory 05/2). Further, the cover design is demonstrated to be stable with
respect to sheet erosion and gully formation. See the response to Interrogatories 71/1 and 191/3
for further discussion on erosion analyses.
UDEQ requests that daily water balance plots of the flow simulation results be provided. UDEQ
has been provided with annual averages for water balance components of precipitation, runoff,
evapotranspiration, storage, and deep drainage. Daily water balance is not the appropriate scale
to evaluate a performance assessment model. See the response to Interrogatory 21/2 for a
discussion of evaluation of flow model water balance.
UDEQ notes that higher technetium-99 concentrations resulted from the simulations using the
unrealistic ranges for hydraulic properties from NUREG/CR-7028 (Benson et al. 2011) requested
by UDEQ. It is not unexpected that unrealistic results will come from a model using unrealistic
input parameters.
2.4 Interrogatory CR R313-25-8(4)(d)-21/2: Infiltration Rates
DEQ Critique of DU PA v1.4, Appendix 5: EnergySolutions/Neptune describe their approach to
parameterizing the radon barriers for v1.4 as follows (pp. 39–40):
An expanded assessment of the performance of the radon barriers was made possible
by developing a distribution for the saturated hydraulic conductivity (Ks) of the radon
barriers to use for the modeling. The Ks values for the radon barriers were sampled from
a distribution developed from a minimum value of 4.32×10-3 cm/day corresponding to the
design specification for the upper radon barrier (Whetstone 2007, Table 8), and 1st,
50th, and 99th percentile values of 0.65 cm/day, 3.8 cm/day, and 52 cm/day,
respectively, which are from a range of in-service (“naturalized”) clay barrier Ks values
described by Benson et al. (2011, Section 6.4, p. 6-12). A shifted lognormal distribution
was fit to the 1st, 50th, and 99th percentiles, and the minimum value of 4.32E-3 cm/day
was used as a shift. The resulting distribution is:
𝐾𝑠 ~ 𝐿𝑜𝑔𝑛𝑜𝑟𝑚𝑎𝑙 𝑔𝑒𝑜𝑚. 𝑚𝑒𝑎𝑛: 3.37 𝑐𝑚/𝑑𝑎𝑦, 𝑔𝑒𝑜𝑚. 𝑠𝑑: 3.23 𝑐𝑚/𝑑𝑎𝑦, with a right shift of
0.00432 cm/day
For all HYDRUS simulations, the same Ks value was applied to both the upper and
lower radon barriers.
ET Cover Design Responses for the Clive DU PA Model
23 Feb 2018 43
Correlations between α and n were investigated by analyzing the combinations of α and
n for the 12 textural classes in Rosetta (Schaap, 2002), and no correlations were
evident. There were also no statistically significant correlations between Ks and α or n.
The developed 50 sets of uncertain parameters for α, n, and Ks were then used as
hydraulic property inputs to 50, 1000-year simulations using HYDRUS-1D.
This approach varies from that taken in DU PA v1.2 as described below (Neptune 2014,
Appendix 5, pp. 41–41):
An expanded assessment of the performance of the radon barriers was made possible
by developing a distribution for the saturated hydraulic conductivity (Ks) of the radon
barriers to use for the modeling. The Ks values for the radon barriers were sampled from
a distribution developed from a minimum value of 4 × 10-3 cm/day corresponding to the
design specification for the upper radon barrier (Whetstone 2007, Table 8), and 50th and
99th percentile values of 0.7 cm/day and 52 cm/day, respectively, which are from a
range of in-service (“naturalized”) clay barrier Ks values described by Benson et al.
(2011, Section 6.4, p. 6-12). A normal distribution was fit to the 50th and 99th
percentiles, and the minimum value of 4E-3 cm/day was used as a shift. For all
HYDRUS simulations, the same Ks value was applied to both the upper and lower radon
barriers.
Correlations between α and n were investigated by analyzing the combinations of α and
n for the 12 textural classes in Rosetta (Schaap, 2002), and no correlations were
evident. There were also no correlations between Ks and α or n.
The developed 20 sets of uncertain parameters for α, n, and Ks were then used as
hydraulic property inputs to 20 1000 year simulations using HYDRUS-1D.
The infiltration results for v1.4 are presented on p. 45 of Appendix 5:
The 50 HYDRUS-1D simulations resulted in a distribution of average annual infiltration
into the waste zone, and average volumetric water contents for each ET cover layer.
Infiltration flux into the waste zone ranged from 0.0067 to 0.1 mm/yr, with an average of
0.024 mm/yr, and a log mean of 0.01 mm/yr for the 50 replicates.” These fluxes are
significantly lower than those calculated in v1.2 and provided on p.45 (Appendix 5)
“Infiltration flux into the waste zone ranged from 0.007 to 2.9 mm/yr, with an average of
0.42 mm/yr, and a log mean of 0.076 mm/yr for the 20 replicates.
Since it appears that the greatest change between v1.2 and v1.4 is that the Ksat values were
increased in v1.4, it is not clear why the infiltration rates would decrease since increasing Ksat
values are typically accompanied by increasing infiltration rates. However, deciphering why the
predictions differ is nearly impossible with the output provided. Understanding the outcome
requires water balance graphs showing the seasonal hydrologic cycle and the dynamics of water
throughout the year. The difference in the predictions may have to do with the shape of the
normal distributions that were used. They are similar, but as described below using the lower
bound constraint may have affected the distribution of K values that are predicted.
Probability density functions (PDFs) are shown in Figure 21-1 that were used to describe
uncertainty and spatial variability in the saturated hydraulic conductivity of the radon barrier in the
Unsaturated Zone Modeling reports submitted in June 2014 and October 2015. A PDF is
analogous to a histogram, describing the probability density associated with a particular value of
the random variable for a defined probability distribution (in this case, the three-parameter log-
ET Cover Design Responses for the Clive DU PA Model
23 Feb 2018 44
normal distribution). The distributions for 2014 and 2015 were parameterized to the extent
practical using the methodology described in the 2014 and 2015 Unsaturated Zone Modeling
reports. A three-parameter log-normal distribution was used given that the 2014 and 2015
reports indicate that a lower bound > 0 was stipulated in the 2014 and 2015 reports. A
description of the three-parameter log-normal distribution can be found in Zhai and Benson
(2006).
For 2014, the distribution was parameterized using a lower bound (x) = 0.004 cm/d, a log-mean
(µ) of -0.357 corresponding to a 50thpercentile of 0.7 cm/d, and a log-standard deviation (s) of
1.85. The lower bound and log-mean are equal to the values stipulated in the 2014 report. The
log-standard deviation was obtained iteratively by ensuring that the 99th percentile equaled
52 cm/d, as described in the 2014 report.
Two PDFs are shown for 2015 in Figure 21-1 below because the fitting methodology and
parameters cited in the 2015 report lead to ambiguity. The PDF marked “2015 reported”
corresponds to x = 0.00432 cm/d (lower bound indicated in 2015 report), µ = 1.215
(corresponding to geometric mean of 3.37 cm/d indicated in 2015 report), and s = 1.17
(corresponding to 3.23 cm/d referred to in 2015 report as the “geom. sd”). These parameters
(“2015 reported”), however, do not yield a 1st percentile of 0.65 cm/d and a 99th percentile of
52 cm/d as indicated in the report (mathematically impossible). Thus, a second parameter set
was selected (referred to as “2015 reported and fit”). This parameter set has x = 0.00432 cm/d
(lower bound indicated in 2015 report), µ = 1.215 (corresponding to geometric mean of 3.37 cm/d
indicated in 2015 report), and s = 1.17. The log-standard deviation (s) was selected by iteration
so that the 99th percentile equaled 52 cm/d, as indicated in the report. However, the 1st percentile
could not be matched along with the 99th percentile (mathematically impossible). The 1st
percentile hydraulic conductivity for the distribution “2015 reported and fit” is 0.1 cm/d.
The PDFs in Figure 21-1 provide insight into the unexpected outcomes for the percolation rates
predicted in 2014 and 2015, the latter percolation rates being lower despite substantially higher
geometric mean saturated hydraulic conductivity. For the PDF marked “2015 reported and fit,”
which seems to be the PDF most likely used as input to the model, the upper tail of the
distribution is much lighter than for the 2014 PDF (e.g., the probability of high hydraulic
conductivities is lower in the 2015 modeling). Consequently, the percolation rates tend to be
lower in the 2015 report relative to those in the 2014 report. This would not be the case if the
parameters corresponding to “2015 reported” were used as input to the model, as the PDF for
this case generally has a heavier upper tail relative to the PDF used as input to the 2014 model.
This ambiguity highlights an important issue: reports issued by EnergySolutions should include
sufficient information for an independent party to reproduce the outcomes without ambiguity. At a
minimum, probabilistic descriptions should show a mathematical description of the distribution
employed (e.g., probably distribution and definition of parameters) and a list of the values
assigned to each parameter for each case being analyzed.
DEQ Critique of DU PA v1.4, Appendix 21:
Naturalized Cover
Significant disagreement remains regarding appropriate hydraulic properties to represent
“naturalized” conditions (EnergySolutions nomenclature) for the cover. EnergySolutions states
correctly that hydraulic properties they developed with the approach recommended in Appendix
E to the 2015 SER are significantly different from those used in their previous analyses for the
DU PA. This is not surprising, as the hydraulic properties EnergySolutions had used in previous
ET Cover Design Responses for the Clive DU PA Model
23 Feb 2018 45
analyses (Bingham Environmental 1991) were obtained nearly three decades ago using poorly
documented sampling and testing methods. Techniques for undisturbed sampling and
measurement of unsaturated hydraulic properties have improved dramatically since the Bingham
Environmental data set was created. The quality and relevancy of the Bingham Environmental
data used by EnergySolutions is suspect, and there is good reason for hydraulic properties
obtained using the approach recommended in Appendix E (April 2015 SER) to differ significantly
from those EnergySolutions has used in past analyses.
EnergySolutions also states that the parameters sets obtained with the approach recommended
in Appendix E (2015 SER) “are conservative” and “do not represent the likely evolution of the
cover system.” EnergySolutions also states that the model predictions “do not make sense.”
EnergySolutions will need to provide quantitative evidence to support these assertions. The
photographs in Figures 4 and 5 of Appendix 21 are inconclusive and provide no quantitative
basis to support inferences that structural development and alterations in hydraulic properties do
not occur at Clive. Structural development that occurs in covers due to pedogenesis generally is
not visible at the scale represented in these photographs. Moreover, the smearing that occurs in
test pits can obscure structure that is present. If EnergySolutions wishes to use these analogs as
evidence to support hydraulic properties representing long-term conditions significantly different
from NUREG/CR-7028 (Benson et al. 2011), EnergySolutions should conduct appropriate
measurements on these in-place materials to demonstrate that the hydraulic properties are
indeed different from the abundance of data in NUREG/CR-7028.
EnergySolutions goes on to argue that the Clive location is not represented properly using the
data set in NUREG/CR-7028, and indicates that less extensive pedogenic change should be
expected at Clive relative to the sites in NUREG/CR-7028. They attribute more extensive
pedogenic change to a greater abundance of biota as well as surface and subsurface biomass at
sites in humid climates, which is incorrect. Changes in hydraulic properties due to pedogenesis
are predominantly caused by cycling in state of stress due to seasonal changes in pore water
suction. Those cycles tend to be larger in arid regions than in humid regions, which promotes
greater volume change and more rapid pedogenesis. In fact, conceptually, pedogenesis should
occur more rapidly, and be more extensive, in a more arid climate such as Clive relative to a
more humid climate. However, as shown in NUREG/CR-7028, climate effects are not significant
over time, as structure develops and hydraulic properties are altered in essentially all climates.
EnergySolutions also suggests that the Clive site is outside the range of sites represented in the
data included in NUREG/CR-7028. DEQ does not agree with the suggestion that the semi-arid
climate at Clive is greatly different from the climate at sites in Apple Valley, California, Monticello,
Utah, or Boardman, Oregon. Each of these sites is semi-arid to arid and not greatly different from
Clive. To further address this issue, data from other sites in the region should be considered as
discussed in Interrogatory 189. These sites include the Monticello Uranium Mill Tailings
Repository, the Blue Water Uranium Mill Tailings Reclamation Site near Grants, New Mexico,
and the Cheney Disposal Facility near Grand Junction, Colorado. While none of these sites has
the same climate as Clive, they are sufficiently similar to be considered reasonable analogs. An
argument against the relevancy of these analogs, especially without data, is not logical.
Homogeneous Cover
EnergySolutions has used a homogeneous cover profile in the most recent simulations. This was
not the intent of our previous comments, and was misconstrued from the parameter
recommendations provided in Appendix E of the 2015 SER. The cover profile should retain a
layered structure representative of the materials planned for each layer, but with the hydraulic
properties of each layer adjusted to reflect pedogenesis. The parameters in the 2015 SER
ET Cover Design Responses for the Clive DU PA Model
23 Feb 2018 46
recommendations were presented as a guide for reasonable ranges consistent with the
recommendations in NUREG/CR-7028.
Correlation and Range of Hydraulic Properties
The hydraulic properties EnergySolutions developed based on the multivariate normal random
generator as recommended by DEQ/SC&A are consistent with those in NUREG/CR-7028 for
“naturalized” conditions. The cross-correlation structure between Ks and α, based on ln Ks and
ln alpha, is also consistent with the literature, as shown in Figure 21-2.
The scatter in this correlation is characteristic of real data, and the correlation is realistic.
However, the range is constrained for both Ks and α because EnergySolutions used the lower-
end standard deviation provided in the 2015 Appendix E SER recommendations. A broader
range would have been obtained using the typical and high-end recommendations for the
standard deviation.
EnergySolutions indicates that the lower end standard deviation was used “to keep the input
parameters within the ranges” of the 2015 Appendix E SER recommendation, which was not the
intent of the recommendation. EnergySolutions should conduct their simulation using a typical
standard deviation for each parameter. This will likely affect only the tails in the percolation data
(high and low percolation rates in Figure 2 of Appendix 21) but likely will affect the 95th percentile
doses (reported in Table 5 of Appendix 21).
Furthermore, the NUREG/CR-7028 recommended range of α values utilizes averaged values for
the entire cover system for each embankment studied in the NUREG, not individual sampling
points, or small parts of an embankment. The information is already presented at the scale
needed for application to a single cover system on a single embankment. Therefore, either
upscaling, or sub-sampling of the data, by Neptune to get a narrower range of α values for an
embankment cover system would be neither necessary nor appropriate.
For all sets of realizations, the mean and the standard deviation (or ln std deviation for Ks and
alpha) should be cited.
Unsaturated Flow Model Output
Percolation rates predicted with the hydraulic properties developed by EnergySolutions using the
procedure recommended in Appendix E to the 2015 SER are reasonable and consistent with
percolation rates measured and predicted for other final covers in regions of similar aridity, as
reported in NUREG/CR-7028. EnergySolutions predicts percolation rates ranging from 0.57 to
1.31 mm/yr using hydraulic properties developed with the procedure recommended by
DEQ/SC&A. As a comparison, percolation ranging from 0.0 to 3.8 mm/yr have been measured
using an ACAP lysimeter at the U.S. Department of Energy’s (DOE’s) Monticello U-Mill Tailings
Disposal Facility in Monticello, Utah, over the period 2000–2016. Percolation rates at other arid
or semi-arid sites described in NUREG/CR-7028 with comparable cover profiles include Apple
Valley, California (0–1.8 mm/yr), Boardman, Oregon (0 mm/yr), and Underwood, North Dakota
(1.9–9.4 mm/yr).
As in past reports from EnergySolutions, the model predictions are difficult to interpret and
evaluate with the level of detail provided. We have requested water balance graphs (see CR
R317-6-2.1-20/2, Figure 20-1), which depict the important interplay between the water balance
quantities throughout the water year. EnergySolutions has included an annualized water balance
chart (Figure 3, Appendix 21), but this chart does not provide the additional information or insight
ET Cover Design Responses for the Clive DU PA Model
23 Feb 2018 47
that is necessary for a proper evaluation of the model predictions. Water balance graphs should
be provided.
Regression Model
The regression model used in GOLDSIM was updated using predictions obtained with the
hydraulic properties EnergySolutions developed based on the method recommended in
Appendix E to the April 2015 SER. This model relates the average annual percolation rate into
the waste to the hydraulic properties of the cover soils. The regression method is not described
in Appendix 21, but is likely the same method used by EnergySolutions in the past. Appendix 21
does not include supporting statistics confirming the significance of the regression and each of
the independent variables included in the regression model. Thus, the efficacy of the regression
cannot be evaluated.
Percolation rates predicted with the regression model and obtained directly from HYDRUS show
a good comparison (see Figure 6 of Appendix 21). This is expected, because the regression
model is based on the HYDRUS output. A concern raised before, and yet unresolved, is whether
good agreement would exist between percolation rates predicted with the regression model and
an independent set of predictions from HYDRUS using the same underlying inputs (e.g., a blind
forward comparison). That type of evaluation is needed to confirm the validity of the regression
model. For example, if an analysis was conducted with the typical standard deviations to obtain a
broader range in outcomes, would the comparison between the predictions from the regression
model and predictions from HYDRUS be in comparable agreement?
At a minimum, EnergySolutions should conduct an independent set of simulations where
percolation is predicted with HYDRUS and then compared with predictions obtained with the
regression model. This is the only fair means to evaluate the efficacy of the regression model.
These predictions should be conducted with the typical standard deviations to get a realistic
representation of the tails of the distribution of percolation.
2.4.1 Interrogatory Response
Radon Barrier Ks Distribution
UDEQ raises a number of issues in this interrogatory. The first is a request for clarification of
what probability distribution was used in v1.4 for the saturated hydraulic conductivity, Ksat, of
the radon barriers. The distribution is clearly provided in Table 16 of the Clive PA Model
Parameters v1.4 document (shown below in Figure 12), and is stated correctly in the
interrogatory as a three-parameter log-normal distribution with “x = 0.00432 cm/d (lower bound
indicated in 2015 report), µ = 1.215 (corresponding to geometric mean of 3.37 cm/d indicated in
2015 report), and s = 1.17 (corresponding to 3.23 cm/d referred to in 2015 report as the “geom.
sd”).” It appears the confusion regarding the distribution stems from the difference between
actual quantiles (percentiles) of the final distribution and the target quantiles that were used as a
basis to inform the development of the distribution. Therefore, this response describes why this
difference exists and provides more detail regarding how the final distribution was selected
based on the available information.
ET Cover Design Responses for the Clive DU PA Model
23 Feb 2018 48

Figure 12. Table 16 from the Clive PA Model Parameters v1.4 document providing the
distribution used. A description of the method used to select the geometric mean,
geometric standard deviation, and minimum reported in this table, and the associated
parameterization of the log normal were also provided in Appendix 14 of the Clive
DU PA Model Final Report for v1.4.
For DU PA v1.4, the Ks distribution was based on 1st, 50th, and 99th target percentiles elicited
from the relevant literature (Benson et al. 2011) to capture plausible values of Ks that could apply
to the entire cover described by the distribution. A lognormal distribution was chosen as the
distributional form to capture skew toward larger values. Parameters of the lognormal
distribution were chosen such that the percentiles of the final distribution were close to the target
percentiles.
In this case, the following information was used to obtain target quantiles for the distribution of
Ks, where the values included in the distribution are meant to represent possible values for the
entire cover (as opposed to small cores from different locations within the cover):
• Per Section 6.4, p. 6-12 (Benson et al. 2011), “For all but one site, Ks falls within the
range of 7.5 × 10-8 and 6.0 × 10-6 m/s regardless of cover type…”. The lower value of
7.5 × 10-8 m/s is 0.648 cm/day after conversion. This was rounded to 0.65 cm/day, and is
used as the elicited target for the 1st percentile.
• The upper value quoted above, 6.0 × 10-6 m/s, is 51.84 cm/day after conversion. This is
rounded to 52 cm/day, and was used as the elicited target for the 99th percentile.
• Per Section 6.4, p. 6-12 (Benson et al. 2011), “The geometric mean in-service hydraulic
conductivity is 4.4 × 10-7 m/s.” This geometric mean is 3.8016 cm/day after conversion
from m/s to cm/day, and this was rounded to 3.8 cm/day. The geometric mean of a
lognormal distribution is equal to the median, so this value was used as the elicited target
for the 50th percentile of the distribution.
• A minimum value of 0.00432 cm/day was also specified to correspond to the design
specification for the layers (Whetstone Associates 2011). This minimum value was used
for the final distribution, but not to define a target percentile.
Neptune uses the open-source statistical software package R (R Core Team 2017), and functions
to facilitate fitting of distributions to target quantiles, or percentiles. The function implements an
optimization algorithm to find the log scale mean, log scale standard deviation, and, for a three-
parameter distribution, a minimum value that together provide a lognormal distribution with
ET Cover Design Responses for the Clive DU PA Model
23 Feb 2018 49
quantiles close to the targets. Due to the restrictions on the shape of the lognormal distribution, it
is not possible to exactly match percentiles using only two parameters and the usual lower bound
of 0; the goal is finding a distribution with percentiles close to, but not exactly equal to, the
targets.
As a starting point, a three-parameter lognormal distribution was fit to provide a close match to
the target percentiles. With three parameters to vary, the target quantiles were achieved, but this
corresponded to a minimum value (implemented as a shift) of 0.43 to go with log-scale mean and
standard deviation of 1.215 and 1.17, respectively. That is, the minimum value chosen shifts the
lognormal distribution defined by a lower bound of zero by 0.43, and after shifting the
distribution quantiles are very close to the targets (see Table 3). However, the minimum value
needed to match the target quantiles is greater than the 0.00432 minimum deemed reasonable
from expert knowledge of the site design, and, therefore, the smaller minimum was used for the
final distribution for the PPA Model to ensure that is it possible to obtain random draws between
0.00432 and 0.42. The mean and standard deviation for the pre-shifted distribution were not
changed. The change in minimum from 0.43 to 0.00432 changes the percentiles of the final
distribution, but these were judged to be close enough to the target quantiles to be reasonable
(see Table 3); the 1st percentile of the final distribution is 0.22 instead of 0.65 and the 99th
percentile is 51.5 instead of 52. Note that a shift of 0.00432 is so small relative to the spread of
the entire distribution that it is not discernible from a lower bound of zero in a plot of the
probability density function (Figure 14) and it only changes the percentiles in the third decimal
place.
Table 3 provides the percentiles from the distributions discussed above, compared to the target
percentiles used to inform the distribution, and Figure 14 shows the associated probability
density function (PDF) for the distribution used for v1.4 and reported in Figure 12. As is clear
from Table 3, percentiles of the selected lognormal distribution are close to the elicited
percentiles, while also adhering to the minimum chosen based on design knowledge of the Site.
The interrogatory presents Figure 21-1, along with curves meant to represent PDFs associated
with different possible distributions used. It is not clear from the description how the curves were
created for the “2015 reported” and “2015 reported and fit.” By definition, the area under a PDF
must integrate to 1, and this is not the case for the curves displayed in Figure 21-1. It appears
from the text that the “2015 reported” curve should match the PDF displayed in Figure 14, but it
does not (the y-axis values between the two plots do not match). Because they are not truly
probability density functions, it is difficult to evaluate them and to attempt to match them to
actual probability distributions. For additional clarification of the percentiles associated with the
final distribution, the code to get the percentiles from the final lognormal distribution using R
statistical software is provided in a screenshot in Figure 13; the code also explains how the
percentiles reported in Table 3 were obtained.
ET Cover Design Responses for the Clive DU PA Model
23 Feb 2018 50

Figure 13. R Statistical Software (R Core Team 2017) code and output for getting quantiles
from the distribution of Ks used in v1.4.

Table 3. Percentiles associated with elicited information for the Ks distribution (cm/day),
and the distribution actually used. These are based on a lognormal distribution with
geometric mean of 3.37 cm/day and a geometric sd of 3.23 cm/day, with shifts for the
minimum associated with each row. See Figure 13 for example R code to get the
percentiles.
1st Percentile 50th Percentile 99th Percentile
Elicited Percentiles (Target) 0.65 3.8 52
Percentiles from distribution
(minimum=0, i.e., no shift)
0.220 3.370 51.548
**Percentiles from v1.4
distribution
(minimum = 0.00432)
0.225 3.374 51.552
Percentiles from distribution
with minimum allowed to
vary to match quantiles
(minimum = 0.43)
0.650 3.800 51.978

ET Cover Design Responses for the Clive DU PA Model
23 Feb 2018 51
Figure 14. The lognormal distribution used for the Ks in v1.4. It is parameterized by a
geometric mean of 3.37 (log-scale mean 1.215), a geometric standard deviation of 3.23
(log-scale standard deviation of 1.17), and a minimum of 0.00432 implemented
through a shift of the distribution. The target 1st, 50th, and 99th percentiles are shown
by the vertical dotted lines.
A log-triangular distribution was not used for this fit. Please see the response to Comment B.3
for a description of problems with using a log-triangular distribution.
UDEQ requests that “At a minimum, probabilistic descriptions should show a mathematical
description of the distribution employed (e.g., probably [sic] distribution and definition of
parameters) and a list of the values assigned to each parameter for each case being analyzed.”
The description of hydraulic parameter distributions and the values of each of the parameters
used for each flow simulation case were provided in Appendix 5 of the Clive DU PA v1.4
(Neptune 2015c). Values for parameters that were held constant were provided in Table 8 of
Neptune (2015c). Development of a probability distribution for Ks of the radon barriers is
provided here again in the response to Interrogatory 21/2. Values of each parameter varied for
the 50 simulations are provided again in the response to Comment B.2.
ET Cover Design Responses for the Clive DU PA Model
23 Feb 2018 52
Naturalized Cover
The next issue raised by UDEQ in this interrogatory is objection to the use of site-specific
hydraulic property data provided in a report by Bingham Environmental (1991). UDEQ has
criticized the use of hydraulic property results from testing of samples of Unit 3 and Unit 4
materials at the Clive Site by Colorado State University that were published in a report prepared
by Bingham Environmental (see also UDEQ Comment B.4). UDEQ states that “Techniques for
undisturbed sampling and measurement of unsaturated hydraulic properties have improved
dramatically since the Bingham Environmental data set was created.”
This statement is simply not true as there has been little change to the standard methods used to
determine saturated hydraulic conductivity, volumetric water content, and water retention
relations. The method commonly used to estimate the van Genuchten hydraulic function
parameters from water retention data is the RETC software published in 1991 by van Genuchten
et al. (1991). UDEQ states further “The quality and relevancy of the Bingham Environmental
data used by EnergySolutions is suspect.” And, with respect to van Genuchten parameters
derived from the testing, “This α is based in part on historic measurements made at Colorado
State University on core samples obtained at the Clive site by Bingham Environmental (1991),
which are known to be too small and too disturbed to adequately represent in-service conditions.
The relevancy of this historic data from Bingham Environmental is dubious, at best.”
There is no indication in the text of Bingham Environmental (1991) or in the test result sheets
from Colorado State University that the data were considered irrelevant, suspect, or dubious.
Colorado State University hosts an internationally recognized Civil and Environmental
Engineering Department. Their facilities include the Groundwater and Porous Media Laboratory,
where testing of the samples from the Clive Site was conducted. UDEQ has provided no basis
for disparaging the technical capabilities of this laboratory.
Following these statements, UDEQ reiterates their position that hydrogeologic models are not
site specific. They argue that statistical distributions for hydraulic property parameters used for
flow modeling of cover material described in Benson et al. (2011); SC&A (2015a, 2015b) should
be universally applied irrespective of climate, ecology, or geologic setting of the site. See the
responses to Interrogatory 05/2 and Interrogatory 153/2. UDEQ lists a number of waste disposal
sites that they consider to be analogs to the Clive Site: Monticello Uranium Mill Tailings
Repository, the Blue Water Uranium Mill Tailings Reclamation Site near Grants, New Mexico,
and the Cheney Disposal Facility near Grand Junction, Colorado. The response to Interrogatory
192/3 discusses why these sites are not suitable analogs for the Clive Site.
Homogeneous Cover
UDEQ raises objections in this interrogatory regarding how simulations were conducted using
in-service cover properties as requested by UDEQ. An alternative set of 50 HYDRUS
simulations was conducted using input parameters derived from the distributions and methods
described by Dr. Craig Benson in Volume 2, Appendix E, of the Safety Evaluation Report (SER)
prepared by SC&A (SC&A 2015a), consistent with the request of UDEQ to use this approach.
ET Cover Design Responses for the Clive DU PA Model
23 Feb 2018 53
These models represent modifications to previous models required in response to the SER issues.
These models are conservative and do not represent the likely evolution of the cover system.
UDEQ objects to use of an essentially hydrologically homogeneous cover for these alternative
simulations, stating that:
This was not the intent of our previous comments, and was misconstrued from the parameter
recommendations provided in Appendix E of the 2015 SER. The cover profile should retain
a layered structure representative of the materials planned for each layer, but with the
hydraulic properties of each layer adjusted to reflect pedogenesis. The parameters in the
2015 SER recommendations were presented as a guide for reasonable ranges consistent with
the recommendations in NUREG/CR-7028.
Given the UDEQ conceptual model of making no distinction between the properties of storage
and barrier layers, the cover can no longer be represented by a layered system. Please see the
response to Interrogatory 10/3 for a discussion of the consequence of using the Benson et al.
(2011) method for developing hydraulic input parameters.
The hydraulic property recommendations and cover material naturalization presented in Benson
et al. (2011) and in Appendix E (SC&A 2015a) are inappropriate for the Clive Site. See the
response to Interrogatory 05/2.
Correlation and Range of Hydraulic Properties
See the response to Interrogatory 05/2 regarding the lack of correlation of hydraulic properties of
Unit 4 soil.
In the application of the Benson et al. (2011) and Benson and Gurdal (2013) method in Appendix
E (SC&A 2015a), the low variance option was selected for the spreadsheet calculations specified
in Appendix E for creating sets of parameters for flow modeling. This choice was made because
the “typical” and “high” variance options produced parameter values outside the range of
recommended values. Though not clear in Appendix E (SC&A 2015a), from the references cited
the variances used in the spreadsheet appeared to be estimated from point data, not from mean
values. A distribution of means, associated with a lower variance than a distribution of point
values, is better aligned with the scale of the PA Model. Neither upscaling nor sub-sampling of
the data was used in developing the hydraulic parameter distributions for the alternative set of 50
HYDRUS simulations described in Appendix 21 (Neptune 2015f).
For Clive DU PA Model v1.4 (Neptune 2015c), estimates of the uncertainty in the hydraulic
properties of Unit 4 materials that compose the surface and evaporative zone layers of the ET
cover were needed. These were obtained by using the α and n values from the distributions
(mean and standard deviation) for each parameter from the Rosetta database of hydraulic
parameters for the textural class of silty clay (Schaap 2002). Details of the development of
distributions for α and n parameters for the surface and evaporative zone layers are given in the
response to Comment B.2.
ET Cover Design Responses for the Clive DU PA Model
23 Feb 2018 54
The α and n distributions in Appendix 5 of Clive DU PA Model v1.4 (Neptune 2015c) are based
on the standard error of the average from the Rosetta database, rather than the sample standard
deviation of the data. The reason for the use of the standard error of the average is to develop the
distribution at the spatial scale consistent with how the values will be used in the PA Model (see
Section 1.1 and Appendix 14 of the Clive DU PA Model v1.4 (Neptune 2015e) for more
discussion of scaling). The Rosetta database is comprised of measurements from point locations
or single cores, thus representing a smaller scale than the site-scale values needed for the PA
Model. Averages of the small-scale values are used to represent larger site-scales consistent with
the scale represented by GoldSim cells in the model. The standard error of the average represents
the expected variability among averages from different random datasets of 28 measurements
(i.e., the approximate sampling distribution of the average), or can be thought of as representing
uncertainty in the value of the true population mean, so that values from the distribution
represent possible site-scale averages (or means).
Unsaturated Flow Model Output
UDEQ compares the net infiltration rates simulated using the alternative distributions based on
the hydraulic property recommendations and cover material naturalization presented in Benson
et al. (2011) and in Appendix E (SC&A, 2015) with rates from a number of sites they regard as
analogs to the Clive Site. These hydraulic property recommendations are inappropriate for the
Clive Site. See the response to Interrogatory 189/3 for a discussion on the applicability of these
sites as analogs.
UDEQ requests plots of flow model water balance components on a daily basis. These
components are precipitation, runoff, infiltration, evaporation, transpiration, storage, and net
infiltration (deep drainage).
Steady-state annual averages of net infiltration and water content from the HYDRUS simulations
are the model results used to develop statistical distributions of these parameters for inputs to the
GoldSim model for the Clive DU PA. The GoldSim model samples inputs randomly at the
beginning of time for a single realization of the simulation, and those values that represent the
input distributions are used throughout the realization time. The model was deliberately run for a
long period of time (1,000 years) in order to reach a near-steady state net infiltration rate that is
not influenced by the initial conditions. Flow model results for use in the GoldSim DU PA model
consisted of the daily fluxes from the bottom of the Lower Radon Barrier and the water contents
for the Surface Layer, Evaporative Zone Layer, Frost Protection Layer, and Upper and Lower
Radon Barriers averaged over the years 900 to 1000 (36,525 days) of the simulation. Changes in
storage were zero when averaged over the last 100 years of the simulations. Runoff was also
negligible. The water balance plots showed the remaining components of water balance:
precipitation, evaporation, transpiration, and net infiltration as annual averages.
Water balance plots were provided to UDEQ for a range of net infiltration rates in Appendix 21
of the Clive DU PA v1.4 (Neptune 2015f). The results of five simulations were shown
representing the 10th, 30th, 50th, 70th, and 90th percentile net infiltration values from 50 HYDRUS
flow model simulations. Average annual values for each nonzero water balance component
(precipitation, evaporation, transpiration, and net infiltration) were presented. Using annual
ET Cover Design Responses for the Clive DU PA Model
23 Feb 2018 55
average parameter values and their associated uncertainty reflects the spatial and temporal scale
of the net infiltration and water content in the DU PA Model. The variability of data at points in
time and space is not the appropriate representation of variance in a PA model.
Regression Model
In this section of the interrogatory UDEQ raises four different issues regarding the model
abstraction that are intermingled in the interrogatory comments. This section of the response
begins by reiterating the goal of the model abstraction, and then addresses the issues separately
to avoid further conflating them.
Goal of model abstraction
As described in Section 1.1 and Appendix 14 (Neptune 2015e), the goal of “model abstraction”
in the context of PA modeling is to build a relatively simple statistical model to approximate the
relationship between the input parameters and an output of interest obtained from a sophisticated,
and usually computationally intensive, process-based model. For example, net infiltration may be
calculated using a process-based model like HYDRUS for a given set of input parameters.
The process model, HYDRUS in this case, is run at many combinations of input parameters,
chosen to explore the relevant region of the multi-dimensional parameter space of the inputs. The
goal is to explore the parameter space of the input parameters as much as is possible for a given
number of HYDRUS runs in order to gain as much information as possible to predict the
response for new combinations of input parameter values that may be used for a PA model run. It
is often the case that a relatively simple statistical model can be used to predict the output of a
very complicated process model, not by re-approximating the physical processes represented in
the process model, but just by capturing the overall relationship, perhaps as a function of few
input parameters and simple functional forms. The statistical model will predict a reasonable
value that would have been obtained from HYDRUS and is not expected to match the HYDRUS
result exactly. HYDRUS is an approximation of reality, and the model abstraction is a simple
approximation to HYDRUS. Uncertainty in predictions from the statistical model can also easily
be incorporated into distribution development if desired.
The success of the statistical model in predicting the output depends on the information
contained in the original runs of the HYDRUS model used to fit the model. There are three
settings to consider when evaluating the predictive ability of the statistical model for use in a PA
model: (1) predictive success at the same input parameter values used in the original HYDRUS
runs (in-sample prediction), (2) predictive success at new input parameter values within the
region explored in the original HYDRUS runs (out-of-sample prediction based on interpolation),
and (3) prediction success at input parameter values beyond the region explored in the original
HYDRUS runs (extrapolation).
Need for “significance of the regression” for assessing its efficacy
The goal of building a statistical model for prediction differs from the goal of developing a
model to understand and explain relationships (e.g., Shmueli (2010); Ramsey and Schafer
(2013); Rawlings et al. (1998)). The different goals lead to different strategies and statistical
ET Cover Design Responses for the Clive DU PA Model
23 Feb 2018 56
methods. Statistical models developed for the goal of prediction are often different from those
developed for the goal of explanation, and some of the tools developed for use when the goal is
explanation should not be used in the same way.
The linear regression models developed for predicting water content and net infiltration are
based on the results of modeling performed using the HYDRUS software. The model for net
infiltration is the focus of the interrogatory comments. Even when the goal is prediction,
statistical software typically provides p-values for individual regression coefficients; this does
not imply they should be used to assess the “efficacy of a regression” for prediction as suggested
in the interrogatory comments. In this case, the model for net infiltration contains α and n as
predictors, with p-values of 6.36×10-12 and 0.000108, respectively. The hypothesis tests
associated with these p-values for individual coefficients should not be used simultaneously to
assess predictive ability of the model (despite common misuses in practice). Instead, they simply
provide a summary of the degree of statistical evidence that the slope of the linear relationship
between the predictor (α or n) and the mean net infiltration differs from zero, while holding the
other predictor constant. For the goal of prediction, any variable thought to be useful for
prediction can be included in the model (Kutner et al. 2005; Ramsey and Schafer 2013). Use of a
large number of predictors, including some that contribute little to predictive performance, can
lead to larger standard errors of prediction, but in this case the standard errors of prediction are
not ultimately used and the number of predictors is so small (two or three) that there is no danger
in including too many predictors. Instead of relying on p-values, the usefulness of the regression
model should be primarily judged by the plot of the predicted values from the regression vs. the
original values obtained from HYDRUS, as well as by measures of out-of-sample predictive
performance, such as those obtained from hold-out datasets and/or cross-validation, both of
which are provided in this response.
A comparison of the predictions from the regression model for net infiltration rate (mm/yr) to the
original results from HYDRUS are shown in Figure 15. For simplicity, the linear regression
model was developed using the inputs and net infiltration results on the scales ultimately used as
inputs to the PA Model, rather than on a logarithmic scale. It is expected that the linear
regression will miss some curvature found in the Clive DU PA Model v1.4 HYDRUS results
(Neptune 2015c), but it is developed as a simple and easy-to-implement approximation that can
be easily incorporated in the PA Model using GoldSim. The R-squared value indicates the
proportion of variance in HYDRUS results explained by the linear regression predictions; 0.67 is
deemed reasonable given the observed relationship in Figure 15. The bulk of the net infiltration
predictions from the regression (between approximately .01 and 0.05 mm/yr) are slightly larger
than the values from HYDRUS, providing a conservative approximation within that range.
ET Cover Design Responses for the Clive DU PA Model
23 Feb 2018 57

Figure 15. The average annual net infiltration values obtained from HYDRUS compared to
the predictions from the linear regression model abstraction. The results from the 50
HYDRUS realizations were used to develop the regression model abstraction, and
therefore this plot depicts in-sample predictive performance. The one-to-one line is
shown for reference.
The existence of “typical standard deviations”
The standard deviation used for a distribution representing the current state of knowledge in an
input parameter for a PA model should be evaluated relative to the temporal and spatial scale it is
meant to represent in the PA model (see Section 1.1 and Appendix 14 (Neptune 2015e)) for more
discussion on scaling). Therefore, it is not productive to define a “typical standard deviation”
without reference to the scale it is meant to apply to. A “typical” standard deviation representing
large scales will not be the same as a “typical” standard deviation representing small scales. That
is, the variability among individual measurements taken at particular points in space and time is
not expected to be the same as that among quantities meant to describe large volumes of
materials over long periods of time. Hence, the adequacy of one distribution should not be
judged relative to a “typical” distribution developed for a different goal.
●●●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●●●●
●
●
●
●
●●
●
●
●
●
●●
●●
●
●●●
●
●●
r^2: 0.67
0.00
0.05
0.10
0.15
0.20
0.00 0.05 0.10 0.15 0.20
Predictions of Net Infiltration from Regression (mm/yr)
HY
D
R
U
S

N
e
t

I
n
f
i
l
t
r
a
t
i
o
n

(
m
m
/
y
r
)
Comparison of HYDRUS Results and Regression Predictions
ET Cover Design Responses for the Clive DU PA Model
23 Feb 2018 58
Evaluation of out-of-sample prediction outside the range needed for the PA Model
(extrapolation)
For the most efficient use of information and the best predictions within the context that the
model abstraction will be used, the abstraction should be developed based on the distributions
used in the PA model and not on different distributions that might be wider to reflect possible
smaller-scale values or different assumptions. Developing the model abstraction for wider
distributions is a waste of the limited resources available (mainly computational time running
HYDRUS in this case) to effectively explore the region of the parameter space needed for the PA
model. That is, it is more efficient to do a better job exploring the specific region of the input
parameter space that will actually be needed for the PA model than to spread the effort out over
values that will not be sampled in the PA model (thereby decreasing the predictive ability within
the region needed for the PA model). Statistical modeling for prediction should always be tied as
closely as possible to the region within which it will ultimately be used to obtain predictions.
Models cannot be assumed to perform well outside of the parameter space they were developed
for unless assumptions that the relationships should continue into other regions of the parameter
space can be justified (e.g., a linear relationship estimated for one region is expected to continue
into neighboring space). In the context of model abstraction for PA models, the original
HYDRUS runs should explore the parameter space covered by the distributions for inputs that
will ultimately be used for the PA model; this avoids the second issue as long as the number of
original runs of the HYDRUS model is large enough to adequately explore the multi-
dimensional space. The model abstraction is not designed to be used with different distributions
for input parameters, particularly if new distributions are wider than those used to inform the
original HYDRUS runs. As described above, it is actually a waste of resources to spread effort
out over a wider range than needed, at the expense of less information in the range of interest.
With this in mind, evaluation of out-of-sample prediction outside the range of inputs needed for
the PA model (i.e., extrapolation) is not considered a meaningful goal. The interrogatory
comments suggest this should be considered, but because the regression model is not designed to
be used with different input distributions, this is not undertaken. The values of the inputs for the
50 HYDRUS runs associated with v1.4 are based on the same distributions used in the PA
model.
Evaluation of out-of-sample prediction within range modeled (interpolation)
The model abstraction should be evaluated with respect to out-of-sample prediction within the
range covered by the original inputs to HYDRUS. The best-case performance of the regression
model will be its ability to predict the output for the values of the input parameters used to obtain
the original HYDRUS results and used to fit the regression model (termed “in-sample”
prediction) (see Figure 15). Therefore, it is important to check the ability to predict at new sets of
input values that fall within the region explored by the original HYDRUS runs; this is how the
model will be used to obtain predictions for the PA Model (out-of-sample prediction via
interpolation). The best method for evaluating “out-of-sample prediction” is to run the HYDRUS
model for an additional number of runs, use the model abstraction to predict the output, then
compare the predictions to the actual HYDRUS outputs (evaluation based on an independent
sample). This is often described as using a “hold-out” data set—the model is fit using a subset of
ET Cover Design Responses for the Clive DU PA Model
23 Feb 2018 59
the data available and then tested independently on the “hold-out” data set. Predictive
performance is not expected to be as good as for in-sample prediction, but if the model is
effective at predicting within the space explored in the original HYDRUS runs, it should be
similar. The downside of the “hold-out” method is that information potentially available to
inform the building of the regression model is not used because it is withheld only to be used for
model evaluation. Another method, k-fold cross-validation, allows use of information from all
the HYDRUS runs while still approximating out-of-sample prediction performance (described in
many statistics textbooks such as Hastie et al. (2009)). Cross-validation proceeds by repeatedly
splitting the available data into a “training set” and a “test set,” where the model is repeatedly fit
for each training set and predictive ability is tested for each test set. Then, the results from all the
predictions are combined to estimate out-of-sample predictive ability. The “k-fold” refers to the
size of the test sets relative to the training sets and also relates to the number of different splits,
and subsequent model fits, needed to carry out the procedure. For each split into a training and
test set, predictive accuracy can be summarized, and then combined across the splits. Each
observation will have a prediction, but the prediction will come from a model fit to a random
subset of 2/3 of the data instead of all the data. Then, typical summaries can be calculated, such
as root mean squared error (RMSE), and R-squared between the original values from HYDRUS
and the predictions.
For the net infiltration linear regression model, 3-fold cross-validation performed with the 50
v1.4 HYDRUS runs (Neptune 2015c) yields an RMSE for out-of-sample prediction of 0.018 and
an R-squared value of 0.57 (Figure 16). The regression model using all of the data for in-sample
prediction has an RMSE of 0.016 and an R-squared value of 0.67 (Figure 15). The in-sample
RMSE is expected to be less than that obtained via cross-validation to approximate out-of-
sample. In this case, the values are judged close enough to trust the out-of-sample prediction
implemented for new sets of input parameters used for the PA Model. A further check compares
the 1000 predictions of net infiltration actually used in the v1.4 PA Model with the 50 net
infiltration rates obtained from the original HYDRUS runs. Figure 17 displays these two sets of
values ordered from smallest to largest and spaced so that the 50 HYDRUS results cover the
same x-axis as the 1000 regression predictions from different inputs. The two lines are expected
to diverge near the minimum and maximum of the x-axis because the results based on 1000
different sets of inputs are expected to explore more extreme regions of the multivariate space
than are observed in the 50, even with the same input distributions. The goal of the model
abstraction is to produce a simple statistical model to provide reasonable predictions quickly and
easily within the PA Model, and the comparison in Figure 17 shows that the predictions are
consistent with those from the HYDRUS model, with a similar range and percentiles. This is
consistent with the cross-validation results. It is easy to think of HYDRUS results as the “truth,”
but it is important for the general context of the modeling to remember that they too are
approximations of reality and the goal of the model abstraction is reasonable predictions of net
infiltration.
ET Cover Design Responses for the Clive DU PA Model
23 Feb 2018 60

Figure 16. The average annual net infiltration values obtained from HYDRUS compared to
the predictions obtained from test datasets via 3-fold cross validation. The original
HYDRUS values predicted were not used to fit the regression equations used to
obtain the predictions. This approximates out-of-sample predictive performance of
the linear regression model abstraction. The one-to-one line is shown on each plot for
reference.
●●
●
●
●
●
●
●●●
●
●
●●●●●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●●
●●
●
●
●●
●
●
●●
●
●●
1 2 3
0.00 0.03 0.06 0.00 0.03 0.06 0.00 0.03 0.06
0.00
0.05
0.10
0.15
3−Fold Cross−Validation Predictions of Net Infiltration (mm/yr)
HY
D
R
U
S

N
e
t

I
n
f
i
l
t
r
a
t
i
o
n

(
m
m
/
y
r
)
Net Infiltration Regression Model Cross−Validation Results
ET Cover Design Responses for the Clive DU PA Model
23 Feb 2018 61

Figure 17. Comparison of 1,000 realizations of net infiltration using the linear model in
GoldSim with the results of the 50 HYDRUS simulations of infiltration.
2.5 Interrogatory CR R313-25-8(4)(a)-28/3: Bioturbation Effects and
Consequences
DEQ Critique of DU PA v1.4, Appendix 5: EnergySolutions/Neptune retain the same
assumptions with respect to biointrusion depths and potential impact on infiltration in v1.4 as
were provided in v1.2.
DEQ Critique of DU PA Appendix 21: EnergySolutions has conducted a series of analyses to
evaluate the impact of erosion on percolation rates from the cover. In one case, the simulation
included loss of 1.2 m of cover soil. EnergySolutions reports that percolation rates obtained for
the full thickness cover and a cover eroded by 1.2 m are essentially the same.
This is not logical, given that the soil in the cover is required to store the water during cooler and
wetter periods, and then to release the water during drier periods. The proposed cover is 1.52 m
thick. If the cover thickness is reduced by 1.2 m via erosion, then the soil water storage capacity
of the cover will be reduced by approximately 80%, and the percolation should change
accordingly. This result without supporting analysis makes all of the HYDRUS modeling suspect.
Additional quantitative and mechanistic evidence is needed to support the outcomes in this part
of Appendix 21. Water balance graphs, which depict the temporal variation in water balance
quantities (rather than a water balance quantity chart) could be used to illustrate whether the
outcomes are reasonable. Water balance graphs typically are created using daily output
ET Cover Design Responses for the Clive DU PA Model
23 Feb 2018 62
predicted from a water balance model and show the seasonal variation in each water balance
quantity.
Examples of water balance graphs are shown in Figure 20-1 (CR R317-6-2.1-20/2). These
graphs depict actual water balance data; water balance graphs from a model prediction would be
similar. The soil water storage record in the water balance graph would be compared to the soil
water storage capacity of the eroded profile.
Clive lies in an area having a semi-arid climate. Only certain types of plants grow readily at Clive.
Very little grass grows there. It’s difficult to see how the limited variety and density of plants will
provide adequate vegetative cover for erosion protection on an embankment. EnergySolutions
should find and document natural analogs in the area that support their predictions, particularly
since the predicted erosion rates appear too low to be realistic.
A related concern is the importance of the biological soil crust for sustaining plant growth and the
high uncertainty regarding its characteristics at the Clive site. EnergySolutions should provide
examples with quantitative data from sites in similar climate and with similar soils. These
examples should show how biological soil crust is preserved or re-established, the timeline for
re-establishment, and how presence (or not) of the biological soil crust affected erosion.
2.5.1 Interrogatory Response
This interrogatory begins by pointing to Section 4.4.3 of the SER (SC&A 2015b), stating that
EnergySolutions has not demonstrated that the ET cover design is adequate to protect against
intrusion by plants, animals, or ants. Field studies and analyses conducted by SWCA at the Clive
Site and nearby ecologically analogous sites demonstrate that disturbance of the cover by plants,
mammals, and ants will be negligible. See the response to Interrogatory 05/2.
UDEQ then raises concerns with the results of their requested simulations of the impact of
erosion on net infiltration using the conceptual model of in-service (naturalized) covers described
in Appendix E of the SER (SC&A 2015a). These simulations showed that a loss of 1.2 m would
have little effect on the net infiltration rate. UDEQ argues that “This is not logical given that the
soil in the cover is required to store the water during cooler and wetter periods, and then release
the water during drier periods.” This outcome is the result of applying UDEQ’s conceptual
model of soil formation processes in the cover materials. See the response to Interrogatory 20/2.
The next matter raised by UDEQ in this interrogatory is a request for daily water balance plots of
the flow model results. UDEQ has been provided with annual averages for water balance
components of precipitation, runoff, evapotranspiration, storage, and deep drainage. UDEQ
claims that these water balance plots are not adequate and have produced numerous examples of
daily water balance plots. Daily water balance is not the appropriate scale for evaluating a
performance assessment model. See the response to Interrogatory 21/2 for a discussion of the
evaluation of flow model water balance, and also see Section 1.1, Modeling for Probabilistic
Performance Assessment.
Next UDEQ argues that “It’s difficult to see how the limited variety and density of plants will
provide adequate vegetative cover for erosion protection on an embankment.” UDEQ requests
that natural analogs in the area be identified and documented to support predictions of adequate
plant cover. As described in the response to Interrogatory 71/1, SWCA Environmental
ET Cover Design Responses for the Clive DU PA Model
23 Feb 2018 63
Consultants (SWCA) assessed erosion under undisturbed conditions at Clive in June 2012
(SWCA 2012). Their observations were that the effects of wind and water erosion were limited.
There was minimal evidence found of water erosion even on the sloped study plots. This is
consistent with the Hansen, Allen, and Luce calculations of projected minimal sheet and rill
erosion loss described in the response to Interrogatory 191/3 and the evaluation of the potential
for gully erosion on the Federal Cell described in the response to Interrogatory 71/1.
The response to Interrogatory 71/1 describes the sequence of reclamation measures for quickly
re-establishing natural conditions on the cover developed by SWCA (2013). Ecological and
engineering measures described in the response to 171/1 will rapidly stabilize the cover in the
short term and continue to provide long-term stabilization. As SWCA (2013) states, “functioning
native ecosystems comprised of the borrow soils at the Clive site do not show erosion as the
DRC suggests.”
UDEQ expresses their concern about biological soil crust (BSC) at the Site. In particular, they
are concerned with “high uncertainty regarding its characteristics at the Clive site.” UDEQ
requests that EnergySolutions “provide examples with quantitative data from sites in similar
climate and with similar soils. These examples should show how biological soil crust is
preserved or re-established, the timeline for re-establishment, and how presence (or not) of the
biological soil crust affected erosion.”
BSC is common at the Clive Site. An average of nearly 80 percent of the soil surface in the
vicinity of the Clive Site is covered by BSC according to surveys done by SWCA (2013).
EnergySolutions has developed a comprehensive plan (SWCA 2013) for establishment of BSC
on the Federal Cell based on consideration of the literature (Belnap et al. 2001; Bowker 2007).
Plans for recovery of the BSC will not be passive. Active measures will be taken to accelerate
establishment, including topsoil inoculation with BSC organisms (Belnap et al. 2001; Bowker
2007) and stabilization of the soil surface by seeding with fast growing species. SWCA (2013)
describes the inoculation process used to spread BSC organisms collected from undisturbed
native soils on the topsoil (SWCA 2013), and the stabilization of the soil surface with fast
growing species shown to enhance BSC rehabilitation (Bowker 2007). Recovery time for BSC at
the Clive Site is projected by SWCA (2013) to be 3–5 years based on their Site analyses and the
use of active assistance measures.
Lastly, UDEQ requests information from similar sites regarding “how presence (or not) of the
biological soil crust affected erosion.” SWCA (2013) has provided discussion of the literature
describing the beneficial role of BSC in controlling soil erosion.
2.6 Interrogatory CR R313-25-7(2)-59/2: Bathtub Effect
DEQ Conclusion from April 2015 SER, Appendix C:
Until the issues are resolved regarding the design of the cover and infiltration rates (see the DU
PA SER, Section 4.1.1.1 and Appendix B) the potential for bathtubbing effects cannot be ruled
out. Therefore, this interrogatory remains open.
DEQ Critique of DU PA v1.4, Appendix 21: No further analysis has been performed since v1.2.
ET Cover Design Responses for the Clive DU PA Model
23 Feb 2018 64
2.6.1 Interrogatory Response
Net infiltration in the Clive DU PA Model v1.4 is calculated using stochastic inputs. The
implementation in the DU PA Model is described in detail in Appendix 5 of the Final Report for
the Clive DU PA Model, Clive DU PA Model v1.4 (Neptune 2015d). To evaluate the likelihood
of the bathtub effect occurring, a distribution of net infiltration rates for the ET cover was
developed from 1,000 realizations of the net infiltration model. The 99th percentile value of this
distribution was 0.106 mm/yr. The design value for the saturated hydraulic conductivity of the
clay liner below the waste is 1.0 × 10-6 cm/s (316 mm/yr) (Whetstone Associates 2011). At
steady state under unit gradient conditions this hydraulic conductivity corresponds to the flux of
water through the saturated clay liner.
Given the much greater capacity of the clay liner to allow water to flow through it in comparison
to the 99th percentile of net infiltration rates, the bathtub effect is not possible. Any increase in
saturated hydraulic conductivity of the clay liner below the waste due to naturalization will make
the bathtub effect even less likely. The largest net infiltration rate at the base of the cover
modeled using the Benson et al. (2011) approach for in-service cover properties was 1.31 mm/yr.
Even for this unlikely conceptual model of clay liner properties, the maximum net infiltration is
over 240 times smaller than the 316 mm/yr that can flow through the cover with its design value
saturated hydraulic conductivity.
2.7 Interrogatory CR R313-25-7(3)-60/2: Modeled Radon Barriers
DEQ Conclusion from April 2015 SER, Appendix C:
As described under Interrogatory 05, based on several unresolved issues related to the ET
cover, DEQ indicated in the DU PA SER Section 4.1.1.1 that the cover design was deficient and
that it cannot determine the adequacy of this portion of the Clive DU PA. (See the description
under Interrogatory 05 above for specific details.) Therefore, this interrogatory remains open.
DEQ Critique of DU PA v1.4, Appendix 5, Appendix 21: See Interrogatory 21 for discussion
regarding approach and concerns related to modeling the radon barriers.
2.7.1 Interrogatory Response
This interrogatory points to issues raised in Interrogatories 05/2, 21/1, and 90/2. See the
responses to Interrogatories 05/2, 21/1, and 90/2.
2.8 Interrogatory CR R313-25-7(1–2)-90/2: Calibration of Infiltration
Rates
DEQ Conclusion from April 2015 SER, Appendix C:
As noted in Sections 4.1.1.1, 4.1.1.3, and 4.4 of the DU PA SER, several issues (including
infiltration rates) regarding the ET cover remain unresolved. Therefore, this interrogatory remains
open.
ET Cover Design Responses for the Clive DU PA Model
23 Feb 2018 65
DEQ Critique of DU PA v1.4, Appendix 21: No further analysis has been performed on
calibration of infiltration rates since v1.2.
2.8.1 Interrogatory Response
Issues regarding infiltration rates raised in Sections 4.1.1.1, 4.1.1.3, and 4.4 of the DU PA SER
(SC&A 2015b) are:
Sensitivity analysis proposed by UDEQ
See the response to UDEQ Comment B.11.
Correlation between α and Ks parameters
See the response to Interrogatory 05/2.
Development of hydraulic parameter distributions
See the response to UDEQ Comment B.2.
Increased hydraulic conductivity of the clay liner over time is not accounted for in the
model.
See the discussion on the clay liner properties in the response to Interrogatory 05/2.
Only a single value of Ks was used for the Surface and Evaporative Zone layers and a
single value of alpha was used for the radon barriers.
See the response to UDEQ Comment B.2.
Disruption of cover due to erosion, frost, and biointrusion
See the discussion on biointrusion (Effect of Biologicals on Radionuclide Transport) in the
response to Interrogatory 05/2.
See the discussion on calculations of frost depth in the response to Interrogatory 05/2.
See the discussion on cover disruption due to erosion in the response to Interrogatory 05/2.
2.9 Interrogatory CR R313-25-7(2)-150/3: Plant Growth and Cover
Performance
DEQ Conclusion from April 2015 SER, Appendix C:
As discussed in the DU PA SER (Section 4.4.3), concerns remain regarding the potential
impacts of biointrusion on infiltration and this interrogatory is open.
DEQ Critique of DU PA v1.4 and Appendix 21: See responses to Interrogatories 10 and 28 for
further discussion.
ET Cover Design Responses for the Clive DU PA Model
23 Feb 2018 66
2.9.1 Interrogatory Response
In this interrogatory, UDEQ references issues in Section 4.4.3 of the SER (SC&A 2015b). This
section of the SER is titled “Effect of Biological Activity on Radionuclide Transport.” SC&A
(2015b) cite several examples of increased infiltration due to activity of ants and mammals.
These examples, however, are not specific to the Clive Site. SC&A (2015b) criticizes the number
of excavations and the mapping of roots at the Site conducted by SWCA (2013). Practical
considerations, however, dictate that sampling be limited.
Based on site-specific information collected by SWCA (2013), their opinion is that, while
greasewood will eventually become established at the Site, roots will follow available water and
will remain within the Surface, Evaporative Zone, and Frost Protection Layers growing laterally
along the clay barriers. For the flow modeling through the cover layers, a statistical distribution
for the saturated hydraulic conductivity based on the in-service cover properties described by
Benson et al. (2011) was used for the radon barriers. Use of these in-service hydraulic properties
accounts for changes in the hydraulic properties of the radon barriers due to biotic activity.
For additional discussion of these issues, see the responses to Interrogatories 05/2 and 10/3.
2.10 Interrogatory CR R313-25-8(4)(d)-153/2: Impact of Pedogenic
Processes on the Radon Barrier
DEQ Critique, v1.4 and Appendix 21: See responses to interrogatories 10 and 28 for further
discussion. In addition, alterations in the hydraulic properties of cover soils are due primarily to
changes in the size, shape, and connectivity of the pores in response to volume change.
Changes in hydrologic conditions within the cover profile (e.g., wetting or drying, freezing or
thawing) induce changes in pore water potential (aka pore water suction) that cause volume
change. Decreases in pore water due to drying or freezing cause the soil to shrink, resulting in
tensile stresses that form cracks and other macropores. Formation of macropores causes the
saturated hydraulic conductivity and the van Genuchten α parameter to increase. The
“macropores” formed by volume change are larger than the pores in the soil on completion of
construction, but generally are not large cracks that would be visible in a transect or test pit
excavated with a clay spade or similar tool.
Cover soils in more arid regions have a greater propensity for volume change and alterations in
hydraulic properties because very large changes in pore water potential occur seasonally. Plants
in arid regions have the ability to extract water to much higher potentials than plants in humid
regions (Gee et al. 1999), resulting greater volume change and more significant structural
changes. However, over time, cycling of pore water potential combined with the effects of biota
intrusion result in similar alterations in hydraulic properties regardless of climate (Benson et al.
2007, 2011).
2.10.1 Interrogatory Response
See the response to Interrogatory 05/2: Evapotranspiration Cover (ET Cover), which includes a
discussion of the conceptual model of soil formation at Clive based on field data analysis.
In this interrogatory, UDEQ attributes changes in hydraulic properties of a cover material to
volume changes of the material: “alterations in the hydraulic properties of cover soils are due
ET Cover Design Responses for the Clive DU PA Model
23 Feb 2018 67
primarily to changes in the size, shape, and connectivity of the pores in response to volume
change.”
Changes in volume due to expansion and shrinkage are characteristic of expansive clays. These
clays that develop a blocky structure due to vertical shrinkage cracks formed during dry periods
are described by Handy and Spangler (2007). Moisture cycles form a subangular blocky structure
characterized by aggregates approximately 1 cm in size. Expansive clays are composed of a
significant fraction of smectite group minerals. Volumetric shrinkage strain during drying was
measured by Albrecht and Benson (2001) on eight natural clay soils through cycles of wetting
and drying. These authors found that samples containing the larger percentages of smectite or
mixed illite/smectite had the highest volumetric shrinkage strains and showed the most extensive
cracking. The smallest shrinkage strains were measured for the soils with less smectite and more
illite, kaolinite, and quartz.
The mineralogy of Unit 4 soil clay composition was determined by x-ray diffraction through
analysis conducted by the University of Utah (Bingham Environmental 1996). The results shown
in Table 4 indicate a zero composition of minerals in the smectite group or, as it was formerly
named, montmorillonite. The absence of smectite minerals in the Unit 4 soil makes changes in
soil structure attributed to wetting and drying cycles as suggested by UDEQ unlikely.
Table 4. Minerals in Unit 4 soil clays.
Mineral Percent by Weight
Quartz 12
Plagioclase 2
K-Feldspar 3
Dolomite 4
Calcite 8
Aragonite 53
Kaolinite 2
Illite/Mica 1

2.11 Interrogatory CR R313-25-7(2)-175/1: Infiltration Rates for the
Federal Cell Versus the Class A West Cell
DEQ Conclusion from April 2015 SER, Appendix C:
As DEQ noted in the Round 3 Interrogatories:
ES notes that this interrogatory is no longer relevant since the Federal Cell will use an
ET cover. We agree with this position. However, a thorough discussion of the modeling
of infiltration rates, with soil hydraulic conductivity values as provided in NUREG/CR-
7028 (Benson et al., 2011), is expected in the report on the ET cover system.
ET Cover Design Responses for the Clive DU PA Model
23 Feb 2018 68
The role of hydraulic conductivity on infiltration rates is extensively discussed in the DU PA SER.
See Section 4.1.1.1 and Appendix B. As specifically noted in Section 4.1.1.1:
There are still a number of unresolved issues with respect to the selection of parameter
ranges, distributions, and correlations, as well as the modeling approach and predicted
sensitivities. These concerns are detailed in Appendix B. Further, because the model-
predicted infiltration rates may be sensitive to the hydraulic properties assigned to each
ET layer, the α and Ksat values assumed for modeling moisture in each soil layer within
the cover system must be correlated based on experimental data. Also, additional
justification is required for the soil property values used in the model by
EnergySolutions. Therefore, DEQ does not consider this portion of the performance
assessment resolved.
Therefore, this interrogatory remains open.
DEQ Critique of DU PA v1.4 and Appendix 21: See responses to Interrogatories 10, 21, 28, and
153 for further discussion.
2.11.1 Interrogatory Response
This interrogatory states that there are issues detailed in Appendix B. See the responses to SER
Appendix B, Comments B.1 through B.9 and B.11.
For discussion of the use of correlated α and Ks parameter values for flow modeling, see the
response to Interrogatory 05/2.
For discussion of the use of hydraulic parameter selection and distribution development, see the
response to UDEQ Comment B.2.
2.12 Interrogatory CR R313-25-8(5)(a)-176/1: Representative Hydraulic
Conductivity Rates
DEQ Conclusion from April 2015 SER, Appendix C:
At this time, DEQ does not accept the EnergySolutions position that infiltration results are
insensitive to radon barrier changes. As discussed under Interrogatory CR R313-25-7(2)-05/2:
Radon Barrier, an appropriate modeling analysis needs to be performed with DEQ agreement as
to values of in-service hydraulic conductivity and correlation between Ksat and α (see Appendix
E to the DU PA SER). Until that study is performed and the results analyzed, this interrogatory
remains open. (See also Appendix B to the DU PA SER.)
DEQ Critique of DU PA v1.4 and Appendix 21: See responses to Interrogatories 10, 21, 28, and
153 for further discussion.
2.12.1 Interrogatory Response
The saturated hydraulic conductivity, Ks, was not included in the regression equation for net
infiltration because it was found not to be a predictor (that is, not close to statistical significance).
Ks is, however, included in the regression models for volumetric water content of the radon
barriers.
ET Cover Design Responses for the Clive DU PA Model
23 Feb 2018 69
Saturated hydraulic conductivity is the ratio of water flux to hydraulic gradient under saturated
conditions. The hydraulic conductivity of an unsaturated soil is strongly dependent on the
volumetric water content. As air enters the pores of a saturated soil, the hydraulic conductivity
decreases rapidly. For example, a plot showing the relationship of average hydraulic conductivity
to water content for the Unit 4 silty clay material modeled for the Surface and Evaporative Zone
Layers is shown in Figure 18. This plot uses the mean value of Ks determined from site-specific
core data and hydraulic function properties from the United States Department of Agriculture
(USDA) Rosetta database (USDA 2017).

Figure 18. Relationship between hydraulic conductivity and water content used for Unit 4
material.
For this example, if the water content is slightly reduced from its saturated water content of
0.481 to a water content of 0.4, the hydraulic conductivity is greatly reduced from its saturated
value of 4.46 cm/day to a value of 0.04 cm/day. This reduction of hydraulic conductivity with
reduced water content is even more pronounced with coarser textured soils. In this example, a
small reduction in water content was shown to produce a 100-fold reduction in hydraulic
conductivity. Since most of the flow in the cover layers occurs under unsaturated conditions, it is
not surprising that net infiltration is not sensitive to the value of saturated hydraulic conductivity.
Another reason for the lack of sensitivity of the value of the saturated hydraulic conductivity to
net infiltration is likely the low moisture availability at the Site. Due to the climate there is not
enough available water for saturated conditions to occur extensively. This can be illustrated with
a simple comparison of precipitation volume to soil storage volume. If all of the water from the
entire mean annual precipitation at Clive was instantly infused into the cover, the porosity of the
layers could contain the saturation within the upper 19 inches of the cover, which is to the top
one inch of the Frost Protection Layer.
Changes in the hydraulic conductivity of the radon barrier were accounted for in the flow
modeling. A statistical distribution for the saturated hydraulic conductivity, Ks, of the radon
1.0E-15
1.0E-13
1.0E-11
1.0E-09
1.0E-07
1.0E-05
1.0E-03
1.0E-01
1.0E+01
0.1 0.2 0.3 0.4 0.5Hy
d
r
a
u
l
i
c

C
o
n
d
u
c
t
i
v
i
t
y

(
c
m
/
d
a
y
)
Volumetric Water Content (-)
Hydraulic Conductivity
silty clay
ET Cover Design Responses for the Clive DU PA Model
23 Feb 2018 70
barriers was developed using values from a range of in-service (“naturalized”) clay barrier Ks
values described by Benson et al. (2011, Section 6.4, p. 6-12). Details of the hydraulic property
distribution development of this distribution are described in the response to Interrogatory 21/2.
Correlation between the α and Ks parameters is discussed in the response to Interrogatory 05/2.
2.13 Interrogatory CR R313-25-7(2)-189/3: Modeling Impacts of
Changes in Federal Cell Cover-System Soil Hydraulic
Conductivity and Alpha Values
DEQ Critique of DU PA v1.4 and Appendix 21: See responses to Interrogatories 10, 21, 28, and
153 for further discussion. In addition, data from other facilities in the region near the Clive site
also confirm that changes in the hydraulic properties of cover soils occur, and the effectiveness
of a cover can change in response to changes in the hydraulic properties.
For example, Benson et al. (2008) report on an assessment of hydraulic properties in the fine-
textured layers in the cover over the uranium mill tailings facility in the Monticello, Utah. The
investigators found that the saturated hydraulic conductivity of the cover soils in the upper 1.5
meters increased by approximately 10x. Similarly, α increased by approximately 5x. Excavation
of caisson lysimeters at the site also showed roots and cracks present in the radon barrier, which
was 1.6–1.9 m bgs (Figure 10-1).
The radon barrier at the Grants, New Mexico, reclamation site was evaluated in the summer of
2016, 20 years after completion, by investigators sponsored by the U.S. Nuclear Regulatory
Commission (NRC) and DOE’s Office of Legacy Management (LM). At this site, the radon barrier
is closer to the surface, with 12 inches of riprap and a sand bedding layer placed directly over the
radon barrier. The capillary break provided by the riprap and the sand bedding layer were
believed to prevent drying and cracking of the radon barrier.
Large block samples were collected from the radon barrier at Grants, New Mexico, for
assessment of field-scale saturated hydraulic conductivity in the laboratory. Block samples were
also collected from an analog site representing conditions anticipated in the long term. A
summary of the hydraulic conductivities reported to date is included in Figure 189-1 below. All of
the saturated hydraulic conductivities are greater than 10-6 cm/s. Most are within or close to the
range described in NUREG/CR-7028 and are approaching the saturated hydraulic conductivity
measured at the analog site. None are less than 1 × 10-7 cm/s as assumed for the lower radon
barrier at Clive.
At the Cheney Disposal Facility near Grand Junction, Colorado, data from two large-scale
lysimeters indicate that the percolation rate from the cover profile has increased substantially
over time, most likely due to structural development within the frost protection layer and the
radon barrier at the site. A summary of the water balance data from these lysimeters is shown in
Table 189-1. This cover employs a rock armor layer, a sand bedding layer, and a frost protection
layer over the radon barrier. Herbicide is used to prevent plant intrusion and root development.
Thus, conditions at this site should minimize the possibility for pedogenesis and alterations in
hydraulic properties. Initially, percolation was on the order of 1 mm/y and less than about 1% of
precipitation. In less than a decade, however, the percolation rate has risen substantially and
was nearly 20% of precipitation in Water Year 2016.
As illustrated in NUREG/CR-7028, changes in hydraulic properties occur at sites more arid and
more humid than Clive. At the hyperarid Apple Valley site in the arid High Plains desert in
ET Cover Design Responses for the Clive DU PA Model
23 Feb 2018 71
southern California, the saturated hydraulic conductivity of a clay barrier similar to the radon
barrier at Clive increased from 1.5 × 10-8 to 1.2 × 10-5 cm/s, or 800x (Benson et al. 2011).
These examples illustrate that structural changes, alterations in hydraulic properties, and
alterations in the water balance occur at other sites in the region near Clive, Utah, and at more
arid locations. Accordingly, changes in the hydraulic properties should be anticipated in the cover
proposed for the Clive site.
2.13.1 Interrogatory Response
In this interrogatory, UDEQ again states their perception that a correlation between the van
Genuchten α parameter and the saturated hydraulic conductivity of cover materials should be
included in the flow model. The lack of correlation between the α and Ks parameters for the Unit
4 soil is discussed in the response to Interrogatory 05/2.
In addition, UDEQ maintains that statistical distributions of hydraulic properties developed by
Benson et al. (2011) should be used to represent the degradation of cover performance with time.
The basis for this assertion is saturated hydraulic conductivity data and water balance estimates
from other sites provided by UDEQ that in their words “confirm” the general application of a
conceptual model of soil structure formation described in Benson et al. (2011).
UDEQ argues that, since soil structural changes and resulting changes in hydraulic properties
and deep drainage are observed at these sites “in the region near Clive Utah,” that these same
changes should be expected at the Clive Site. The sites listed are the Monticello Mill Tailings
Repository south of the town of Monticello, Utah (290 miles from Clive), the Blue Water
disposal site near Grants, New Mexico (485 miles from Clive), the Cheney disposal site near
Grand Junction Colorado (270 miles from Clive), and the Apple Valley Alternative Cover
Assessment Program (ACAP) site referenced by Benson et al. (2011).
Of these sites, two do not have ET covers. At the Cheney site the upper layer is described as rock
armor treated with herbicide to eliminate vegetation. At the Blue Water site the tailings are
covered by a 1.7 to 2.6 ft radon barrier (DOE 2017). The radon barrier is overlain by a 4 to
12-inch thick layer of riprap. Neither of these sites features layers that would function as an ET
cover.
No information was found on the full facility name, identity of the owner/operator, setting of the
site, construction details, or operation of the site named Apple Valley in Benson et al. (2011).
Searches were unable to find more than a general location for this site near Apple Valley,
California, provided in Benson et al. (2011). The relevance of measurements of this site’s
performance to the Clive Site is uncertain without more information.
ET Cover Design Responses for the Clive DU PA Model
23 Feb 2018 72
SWCA (2013) describe the Monticello site as having similar seasonal precipitation, rainfall
patterns, and vegetation conditions (Waugh et al. 2008) to the Clive Site. However, SWCA
(2013) describe three important differences between the sites:
1. Monticello receives approximately 50% greater average annual precipitation
(15.4 in) than Clive.
2. The Monticello ET cover is comprised of clay-loam to sandy-loam soils that are
less alkaline and more fertile than the saline, alkaline silty-clay soils at Clive
(Waugh et al. 2008).
3. The native vegetation at Monticello is dominated by big sagebrush shrublands
and grasslands that are more diverse and of larger stature—with greater target
plant densities and cover for the ET cover—than those proposed at Clive.
Site-specific observations of soil formation at the Clive Site that differ significantly from those
described in Benson et al. (2011) are discussed in the response to Interrogatory 05/2.
2.14 Interrogatory CR R313-25-7(2)-192/3: Implications of Great Salt
Lake Freezing on Federal Cell Performance
DEQ Critique of DU PA v1.4, Appendix 2: In the Updated Site-Specific Performance Assessment
(EnergySolutions 2013), Appendix E, EnergySolutions presents a calculation of frost depth at the
Clive site based on the modified Berggren equation, which first presented by Berggren (1943),
refined by Aldrich and Paynter in 1953, and later adopted by the U.S. Army Corps of Engineers
and other agencies as their preferred method for frost depth determination (Departments of the
Army and Airforce, 1988).
In their July 8, 2014 (EnergySolutions 2014), response to this interrogatory, EnergySolutions
points to Appendix E to the Updated Site-Specific Performance Assessment (EnergySolutions
2013) for the calculation of the potential frost depth; however, that reference (nor any other
estimation of frost depth) is not provided in v1.4, Appendix 2.
Therefore, this interrogatory will remain open until an estimate of the potential frost depth has
been incorporated into DU PA Appendix 2, either by reference to or reproducing EnergySolutions
2013, Appendix E, or by providing a similar calculation of the potential frost depth. Additionally, if
EnergySolutions 2013, Appendix E, is referenced or reproduced, any open interrogatories
against Appendix E must be resolved before it is incorporated into DU PA Appendix 2.
DEQ Critique of DU PA v1.4, Appendix 21: An estimate of the potential frost depth has not been
provided in Appendix 21.
2.14.1 Interrogatory Response
See the response to Interrogatory 05/2 for a description of frost depth calculations for the Clive
Site.

ET Cover Design Responses for the Clive DU PA Model
23 Feb 2018 73
2.15 SER B.1 Supplemental Interrogatory Comment 1
1) Demonstrate why 20 HYDRUS runs are sufficient to capture the parameter uncertainty.

DEQ Critique from April 2105 SER, Appendix B:
EnergySolutions’ response provided to this comment did not address the comment satisfactorily.
DEQ understands that the regressions [Equations 39 and 40 of Appendix 5 to the depleted
uranium performance assessment (DU PA) (Neptune 2014b)] were created as simplified
surrogate models that relate percolation from the base of the cover and water content in each
layer of the cover profile to hydraulic properties of the cover soils. This regression model was
developed based on output from HYDRUS from 20 sets of input parameters.
Because only 20 cases were used for the simulations, the tails of the distributions describing the
hydraulic properties are poorly sampled, and more extreme cases may be inadequately
represented. Consequently, the regressions may represent average or mean conditions
sufficiently but may not adequately represent the more extreme cases. No information has been
provided to demonstrate that the extreme cases in the tails of the distributions are adequately
represented by the regression, or that 20 cases are sufficient to capture the effects of the tails of
the distributions. For heavy-tailed distributions such as those used for hydraulic properties, many
more simulations would be needed to adequately represent events driven by properties
associated with the tails of the distributions.
The predictions in EnergySolutions (2014) Figure 5 (see the discussion on Comment 7 below)
suggest that the process of developing the regression model has resulted in predictions that are
centered more around the mean behavior and that are insensitive to the tails. The percolation
predicted from the regression varies within a narrow range of around 0.3 millimeters per year
(mm/yr), whereas percolation predicted by HYDRUS predictions for all realizations ranges from
approximately 0.01 mm/yr to 10 mm/yr. The response suggests that this insensitive behavior is
due to the variance reduction in the hydraulic properties to account for spatial averaging, but
another plausible reason is that the regression is based on mostly mean behavior and is
relatively insensitive to extremes represented by the hydraulic properties in the tails of the
distributions.
A well-documented justification is needed that demonstrates that Equations 39 and 40, based on
predictions from 20 simulations using 20 sets of randomly sampled properties, adequately
predict the percolation rate and the water contents for cases near the mean and more extreme
cases in the tails of the distributions.
In addition, the analysis fails to adequately account for (1) correlations between parameters α
and Ksat in the same soil layer, and (2) correlations between the values of each parameter within
different soil layers. These deficiencies need to be resolved. DEQ also notes that the
EnergySolutions response contains no substantive discussion of how and why scaling was
conducted and how it impacts the results. This discussion must be provided.
DEQ Critique of DU PA v1.4, Appendix 21: See Interrogatory 21 for a description of the
HYDRUS model parameter uncertainty.
ET Cover Design Responses for the Clive DU PA Model
23 Feb 2018 74
2.15.1 Interrogatory Response
In response to supplemental interrogatories submitted to EnergySolutions on August 11, 2014
(EnergySolutions 2014), additional capability was developed to allow for more HYDRUS
simulations of the cover system to be completed within a practical time period. Fifty HYDRUS-
1D simulations were conducted to evaluate the uncertainty in infiltration flux into the waste zone
and water content within each ET cover layer as a function of hydraulic property uncertainty;
these simulations were documented in DU PA v1.4 Appendix 5 (Neptune 2015c). Equation
numbers 39 and 40 from Appendix 5 of DU PA v1.2 (Neptune 2015c) correspond to equation
numbers 41 and 42 from Appendix 5 of DU PA v1.4 (Neptune 2015c).
The development of hydraulic property input statistical distributions for this modeling is
described in the response to UDEQ Comment B.2. The values of α, n, and Ks used as inputs for
each of the 50 simulations are listed in Table 9 of DU PA v1.4 Appendix 5 (Neptune 2015c) and
are reproduced here in Table 6.
Simulations were run for 1,000 years. The mean of the fluxes into the top of the waste layer and
the mean water contents for the Surface Layer, Evaporative Zone Layer, Frost Protection Layer,
and Upper and Lower Radon Barriers over years 900 to 1000 were calculated.
The 50 HYDRUS-1D simulations resulted in a distribution of average annual infiltration into the
waste zone, and average volumetric water contents for each ET cover layer. Infiltration flux into
the waste zone ranged from 0.0067 to 0.18 mm/yr, with an average of 0.024 mm/yr, and a log
mean of 0.018 mm/yr for the 50 replicates.
Multiple linear regression models were fit to the HYDRUS infiltration results, and water
contents for each ET cover layer. The general form of the regression was:
𝑌=βV +β<∗𝐾)+β>∗α +βX ∗𝑛
Net infiltration is in units of mm/yr and volumetric water content is dimensionless. For the net
infiltration flux regressions, Ks was dropped as a predictor due to poor fit of the models. The
regressions were fit using the ‘lm()’ function in the software package R (R Core Team 2015),
which uses least squares optimization for estimating parameters. All values of β coefficients are
summarized in Table 5.
ET Cover Design Responses for the Clive DU PA Model
23 Feb 2018 75
Table 5. Coefficients calculated from multiple linear regression models.
Coefficient βo β1 β2 β3
SurfaceWC 0.48155 0.00000 0.54920 -0.20020
EvapWC 0.57947 0.00000 0.73997 -0.24790
FrostWC 0.04282 0.00000 0.43297 0.01617
Rn1WC 0.14737 -0.00076 1.70702 0.06353
Rn2WC 0.14740 -0.00076 1.70648 0.06351
Flux (mm/yr) -0.32921 N/A 5.56826 0.19538

Implementation in GoldSim
Average annual infiltration flux into the waste zone, and the volumetric water content of each ET
cover layer, were calculated using Equations 41 and 42, developed from HYDRUS-1D
simulation results (Neptune 2015c). GoldSim calculates values using Equations 41 and 42 from
Appendix 5 of DU PA v1.4 (Neptune 2015c) for each ET cover layer.
These equations for solving infiltration and water content in GoldSim are:
𝐼𝑛𝑓𝑖𝑙=βV +β>∗α +βX ∗𝑛 (41)
𝑊𝐶=𝛽_,V +𝛽_,<∗𝐾)+𝛽_,>∗𝛼+𝛽_,X ∗𝑛 (42)
where Infil is net infiltration in mm/yr, WC is average volumetric water content, and β values are
linear regression coefficients with the subscript i corresponding to Surface, Evaporative Zone,
Frost Protection, Upper Radon Barrier, and Lower Radon Barrier Layers. The necessary
distributions in GoldSim are VG_logAlpha, VG_logN, and RnBarrierKsat_Natdist.
α and n are calculated from values drawn from distributions using:
𝛼=10VG_logAlpha,𝑤ℎ𝑒𝑟𝑒 VG_logAlpha ~ 𝑁𝑜𝑟𝑚𝑎𝑙(𝑚𝑒𝑎𝑛: −1.79,𝑠𝑒: 0.121) and
𝑛=10VGlogN,𝑤ℎ𝑒𝑟𝑒 VGlogN~ 𝑁𝑜𝑟𝑚𝑎𝑙(𝑚𝑒𝑎𝑛: 0.121,𝑠𝑒: 0.019).
Ks is sampled using:
RnBarrierKsat_Natdist =𝐾7,~𝐿𝑜𝑔𝑛𝑜𝑟𝑚𝑎𝑙(𝑔𝑒𝑜𝑚.𝑚𝑒𝑎𝑛:3.37 𝑐𝑚/
𝑑𝑎𝑦,𝑔𝑒𝑜𝑚.𝑠𝑑: 3.23 𝑐𝑚/𝑑𝑎𝑦), with right shift of 0.00432 cm/day.
See the response to Interrogatory 05/2 regarding correlation of the van Genuchten α parameter
and saturated hydraulic conductivity.
ET Cover Design Responses for the Clive DU PA Model
23 Feb 2018 76
Figure 19 shows a comparison of the net infiltration results calculated using HYDRUS and using
the regression equation in the Clive DU PA Model v1.4. Clearly, the comparison shows an
excellent fit to the HYDRUS results, demonstrating that the use of a regression equation to
approximate the HYDRUS simulations resulted in a successful model abstraction in this case.

Figure 19. Comparison of 1,000 realizations of net infiltration using the linear model in
GoldSim with the results of the 50 HYDRUS simulations of infiltration.
As described in EnergySolutions (2014), the discrepancy between the net infiltration rates
obtained from the 20 HYDRUS simulations and the GoldSim estimated net infiltration using the
regression model was due to not scaling the α and n distributions used as inputs for the
HYDRUS simulations. This discrepancy has been resolved for the 50 HYDRUS simulations
described in Appendix 5 of the DU PA Model v1.4 (Neptune 2015c), and the comparison
between the HYDRUS results and the 1,000 realizations using the regression equation are shown
here in Figure 19.
The 50 cases were run at sets of values chosen to explore the part of the input parameter space
used for the DU PA Model, thus putting the effort and resources into the portion of the parameter
space most important to support the model. The values chosen are directly from the distributions
developed for Clive DU PA Model v1.4, and the use of values is adequate to expect to sample
from the tails of the distributions. The model abstraction and connections to input parameter
ET Cover Design Responses for the Clive DU PA Model
23 Feb 2018 77
distributions were described in detail earlier in this response, as well as the reasons for
developing distributions on a scale appropriate for the PA Model.
UDEQ concludes this interrogatory stating that the “EnergySolutions response contains no
substantive discussion of how and why scaling was conducted and how it impacts the results.”
An explanation regarding how and why scaling was conducted in multiple sections of the
responses. In addition, Section 1.1 of this document provides a high-level discussion, with
examples.
In summary, the reason for “scaling” is to take the available information about the input
parameters and use it to develop a distribution at a scale consistent with the scale represented by
the PA Model. In other words, the random draws from the distributions should be values that are
consistent with the scales they are supposed to represent (e.g., an extreme value possible from a
point location at one point in time should not be used to represent a large spatial volume over a
long period of time).
Available data often represent much smaller spatial volumes and time periods than are needed
for the PA Model; therefore, the data should not be used directly to develop a distribution, but
instead “scaling” should be employed to attempt to build a distribution representing the PA
Model scale. Averaging is one simple method of scaling information up by aggregating
information available at smaller scales.
Larger scales are typically associated with less variability, and thus narrower distributions, but
the statistical method used for scaling should depend on the properties of the available data and
on uncertainty in assumptions needed to justify the scaling. The end goal is distributions that
produce draws representing reasonably realistic values to apply to the spatial volumes
represented in the PA Model and long time periods over which they are held constant within a
run of the PA Model.
2.16 SER B.2 Supplemental Interrogatory Comment 2
2) The Table 9 HYDRUS parameters do not appear to “bound” the α, n, and Ksat
distributions. For example, in the distribution, Ksat ranges from 0.0043 to 52 cm/day,
but in the 20 HYDRUS runs Ksat only ranged from 0.16 to 10.2 cm/day.
DEQ Critique from April 2015 SER, Appendix B:
EnergySolutions’ response indicates that the input “values are considered sufficiently extreme to
evaluate the influence of Ks on the HYDRUS model outputs, and hence to determine the
influence of Ks on the water content and infiltration model outputs.” The basis for the conclusion
“considered sufficiently extreme” needs to be demonstrated rather than stipulated.
As cited in the response to Comment 1 (above), a well-documented justification is needed that
demonstrates that Equations 39 and 40, based on predictions from 20 simulations using 20 sets
of randomly sampled properties, adequately predict the percolation rate and the water contents
for cases near the mean and more extreme cases in the tails of the distributions. This
demonstration should also provide a physical basis for excluding some of the variability in key
ET Cover Design Responses for the Clive DU PA Model
23 Feb 2018 78
hydraulic properties normally considered to affect percolation strongly, such as Ksat in the
shallow cover-system layers (i.e., the Surface Layer and the Evaporative Zone Layer). Any
exclusion of this parameter or its full range of variability from other aspects of modeling,
correlation, or sensitivity analysis should also be justified. Although the Clive DU PA v1.2
appears superficially to have illustrated that the output was not sensitive to Ksat, this conclusion
may be the result of predictions from a cover hydrology model for which unrealistic parameters
were used as input (e.g., changing some parameter values but not others for a given soil layer).
A separate quantitative demonstration is needed showing that Equations 39 and 40, based on
the 20 sets of hydraulic properties used as input, are representative.
DEQ Critique of DU PA v1.4, Appendix 21: See Interrogatory 21 for a description of the
HYDRUS model input distribution, ranges and bounds.
2.16.1 Interrogatory Response
In response to supplemental interrogatories submitted to EnergySolutions on August 11, 2014,
additional capability was developed to allow for more HYDRUS simulations of the cover system
to be completed within a practical time period. Fifty HYDRUS-1D simulations were completed
to evaluate the uncertainty in infiltration flux into the waste zone and water content within each
ET cover layer as a function of hydraulic property uncertainty and were documented in DU PA
v1.4 Appendix 5 (Neptune 2015c). See the response to UDEQ Comment B.1.
HYDRUS Hydraulic Property Input Parameters
Parameter values and parameter statistical distributions for the hydraulic properties of the cover
layers shown in Figure 1 were developed based on site-specific data, engineering specifications,
widely used soil hydraulic property databases, and consideration of the function of the layer.
These are the parameters used for the hydraulic conductivity model described in Equations (1),
(2), and (3) in Section 1.3 needed for the HYDRUS flow modeling.
Surface Layer and Evaporative Zone Layer
These two layers are composed of the same Unit 4 silty clay. The source of site-specific material
properties for Unit 4 is Bingham Environmental (1991) (pp. B-20 and B-26). They report results
from measurement of water retention for two of nine cores sampled from Unit 4 at the Site. The
water retention relation is the correspondence between the tension of water held in the pores of
the material and the water content of the material. These data are used to estimate the parameters
of the van Genuchten-Mualem hydraulic conductivity model.
The functions of these layers, beyond supporting vegetation and providing erosion control, are to
store water from precipitation events within the layers and slow deeper drainage so that the water
is available for release to the atmosphere through evaporation and plant transpiration. The α and
n values of the van Genuchten water retention model (Section 1.3) influence the capacity of the
Unit 4 material used for the Surface and Evaporative Zone Layers to hold water. To provide a
better estimate of the uncertainty of these Unit 4 properties, statistical distributions were
developed for α and n to be used for inputs for the flow modeling.
ET Cover Design Responses for the Clive DU PA Model
23 Feb 2018 79
To develop the distributions for flow model input, α and n values were taken from the
distributions of the mean and the standard deviation for each parameter from the Rosetta model
database of hydraulic parameters for the textural class of silty clay (USDA 2017). The standard
deviations were converted to standard errors by dividing by √n, where n is the number of
samples (28 in this case).
The Rosetta model provided estimates of van Genuchten water retention parameters, saturated
hydraulic conductivity, and unsaturated hydraulic conductivity parameters. The Rosetta model is
widely used and has been successful in many applications. In addition to other predictive
features, the Rosetta model provides a database containing the class average values of soil
hydraulic function parameters for the 12 USDA soil textural classifications. These average
values are based on 2,134 soil samples for water retention and 1,306 soil samples for saturated
hydraulic conductivity (USDA 2017). These data were obtained from the RAWLS, AHUJA, and
UNSODA databases (USDA 2017).
To provide consistency with the conceptual model, probability distributions needed to be
specified that matched the spatio-temporal scale of the model. The fitted regression models were
used in the Clive DU PA GoldSim model, but the distributions of α and n were re-scaled to
match the structure of the GoldSim model. Scaling in this way is inherently an averaging
process, although some care needs to be taken to ensure that the immediate response reacts
linearly to the inputs (expectation is a linear operator). The Rosetta database indicates that 28
samples were used to develop the mean and standard deviation estimates. Consequently, scaling
was performed by dividing the standard deviation by the square root of 28, which represents
using the standard error of the Rosetta data for the parameter distributions implemented in
GoldSim. This provides an appropriate distribution for the Clive DU PA GoldSim model given
the structure and scale of that model.
The distributions for A and N given in terms of log transforms of α and n were described in the
response to UDEQ Comment B.1. Normal distributions of A and N were sampled 50 times, and
then transposed from log space by calculating 10A and 10N to provide 50 sampled values of α
and n. In addition, N was truncated such that it could not be less than or equal to 0.0. A
correlation between Ks and α was not used. (See the response to Interrogatory 05/2.)
To manage computational burden, the saturated hydraulic conductivity, Ks, for the Surface Layer
and Evaporative Zone Layer was assigned a single deterministic value of 4.46 cm/day based on
the mean value determined from analysis of the Unit 4 cores (Equation 29 of Neptune (2015c)).
See the response to Comment B.5.
Radon Barriers
An expanded assessment of the performance of the radon barriers was made possible by
developing a distribution for the saturated hydraulic conductivity (Ks) of the radon barriers to use
for the modeling. Development of this distribution is described in the response to Interrogatory
21/2.
ET Cover Design Responses for the Clive DU PA Model
23 Feb 2018 80
For all HYDRUS simulations, the same Ks value was applied to both the Upper and Lower
Radon Barriers. To manage computational burden, deterministic values for θr, θs, α, and n from
Table 17 of Whetstone Associates (2011) were used for the radon barriers and are listed in Table
8 of Neptune (2015c). A correlation between Ks and α was not used (see the response to
Interrogatory 05/2).
The equivalent of Table 9 of Appendix 5 of DU PA Model v1.2 (Neptune 2014) cited in the
interrogatory is reproduced here from Appendix 5 of DU PA Model v1.4 (Neptune 2015c). This
table, Table 6 below, contains the values of α, n, and Ks drawn from the distributions described
above and used in 50 HYDRUS simulations of net infiltration and volumetric water content.
Table 6. Parameter sets of van Genuchten α, n, and Ks used for HYDRUS modeling.
Realization α (1/cm) n Ks (cm/d)
1 0.013091 1.359766 3.285794
2 0.014317 1.371086 12.497148
3 0.010969 1.357776 3.736272
4 0.018089 1.342287 5.162964
5 0.019954 1.316356 2.325706
6 0.010797 1.279182 4.168751
7 0.016004 1.396199 2.595876
8 0.012816 1.308572 0.838501
9 0.014744 1.372326 2.055096
10 0.014791 1.360367 5.052781
11 0.020639 1.276159 3.234858
12 0.019501 1.327968 2.194697
13 0.015766 1.334194 1.307280
14 0.019048 1.373538 1.719640
15 0.018539 1.338996 1.635838
16 0.017045 1.267606 1.749758
17 0.019983 1.413655 5.126214
18 0.012494 1.326223 10.753272
19 0.019503 1.356646 1.845171
20 0.028186 1.378016 3.643845
21 0.010929 1.244500 6.738214
ET Cover Design Responses for the Clive DU PA Model
23 Feb 2018 81
Realization α (1/cm) n Ks (cm/d)
22 0.020973 1.282170 6.943533
23 0.017971 1.372107 1.099495
24 0.016549 1.467656 3.648668
25 0.012120 1.330512 6.338780
26 0.011984 1.382991 0.792890
27 0.012782 1.382761 7.005276
28 0.017094 1.275082 4.768674
29 0.013032 1.382671 9.861743
30 0.024165 1.349583 7.758327
31 0.016054 1.386282 1.478986
32 0.024889 1.310637 2.501489
33 0.017247 1.320670 2.459523
34 0.014338 1.265236 66.503659
35 0.016633 1.286526 31.683457
36 0.014343 1.383885 1.005712
37 0.022207 1.236303 3.733521
38 0.012511 1.317326 4.565641
39 0.018395 1.333180 6.167757
40 0.013735 1.294514 2.206236
41 0.015243 1.229113 4.106400
42 0.018063 1.282922 3.299065
43 0.017010 1.326811 32.484809
44 0.020072 1.323515 31.128008
45 0.015950 1.357247 2.326748
46 0.018944 1.252554 2.976567
47 0.015677 1.301147 1.241111
48 0.024293 1.287802 4.617869
49 0.018819 1.264178 0.737824
50 0.017781 1.263628 2.880623

ET Cover Design Responses for the Clive DU PA Model
23 Feb 2018 82
The 50 random draws from the distribution for Ks have a minimum of 0.73 cm/day, a maximum
of 66.5 cm/day, and a mean value of 6.76 cm/day. The maximum of 66.5 cm/day is greater than
the maximum of the range reported by Benson et al. (2011) of 52 cm/day (6.0 × 10-6 m/s). Also
note that the lowest Ks value drawn was 0.738 cm/day, a value more than 100 times larger than
the minimum of the distribution.
The sensitivity of the modeled net infiltration rate to the value of the saturated hydraulic
conductivity is discussed in the response to Interrogatory 176/1.
2.17 SER B.3 Supplemental Interrogatory Comment 3
3) NUREG/CR-7028 (Benson et al. 2011) gives the “in-service hydraulic conductivity” as
ranging from 7.5 × 10-8 to 6.0 × 10-6 m/s [0.7 to 52 cm/day], with a mean of
4.4 × 10-7 m/s [3.8 cm/day]. Instead of using the provided distribution (i.e., log-triangular
with a minimum, maximum, and most likely), ES/Neptune constructed a lognormal
distribution with a mean and standard deviation of 0.691 and 6.396 cm/day,
respectively. Provide the justification for this approach. For example, the selection of
0.0043 cm/day as the lower end of the Ksat distribution requires justification (Appendix
5, p.41). It is not clear why a design parameter value should be used when adequate
field data are available. The number chosen by the Licensee for the lower end of the
distribution range in the GoldSim implementation is 163 times lower than the lowest
value in the range specified within the NUREG guidance (see Section 13.0 of Appendix
5, Unsaturated Zone Modeling to the Clive DU PA). We believe that use of the design
parameter biases the Ksat distribution in a non-conservative manner.
DEQ Critique from April 2015 SER, Appendix B:
EnergySolutions’ response to Comment 3 has not demonstrated that the distribution of Ksat
used for the HYDRUS modeling adequately represents the range of conditions that might be
realized for a “naturalized” cover, i.e., one that has undergone pedogenesis as described in
NUREG/CR-7028 (Benson et al. 2011). To account for the higher Ksat in NUREG/CR-7028
(Benson et al. 2011), the lognormal distribution for Ksat was re-fit by the Licensee using an
abnormally large log(s) of 6.396. This provides an unrealistic distribution of Ksat that
substantially overweights Ksat in the lower range.
This, in turn, has the general effect of artificially increasing apparent capillary barrier effects in
the DU PA Model v1.2, i.e., at the interface between a relatively lower-permeability zone (the
combined Surface Layer and the Evaporative Zone Layer, having a mean Ksat value in the DU
PA Model v1.2 of 4.46 cm/day) and a relatively higher-permeability zone (the Frost Protection
Layer, having a mean Ksat value in the DU PA Model v1.2 of 106.1 cm/day). When
EnergySolutions assumes in HYDRUS that the Ksat value for the lower-permeability zone can be
as small as 0.0042 cm/day, the ratio in hydraulic conductivity between the higher-permeability
zone and the lower-permeability zone can thus be as large as 25,000. This creates in the model
an extremely potent artificial, non-realistic capillary barrier at the Evaporative Zone Layer/Frost
Protection Layer interface that, in an unrealistic way, reduces infiltration below that interface to
extremely small or even negligible values.
The primary model hydraulic conductivity value for the higher-permeability zone in the DU PA
Model v1.2, 106.1 cm/day, may already be unrealistic, since the assemblage of soil particles in
the Frost Protection Layer is proposed to be a random, poorly-sorted mixture of grain sizes, with
ET Cover Design Responses for the Clive DU PA Model
23 Feb 2018 83
smaller grains being as small as clay. The Frost Protection Layer is not characterized in terms of
actual grain size distribution in the DU PA Model v1.2, other than to say that particle sizes can
range from 16-inch diameter to clay size. The hydraulic conductivity assigned to it is arbitrary.
The assigned value is representative of a sandy loam, which is a very poor representation of the
proposed Frost Protection Layer. A mixture of poorly-sorted grain sizes, as found in the Frost
Protection Layer, tends to greatly diminish the hydraulic conductivity of a soil compared to a
relatively well-sorted mixture. Further exacerbating the problem in the DU PA Model v1.2 is that
the hydraulic conductivity values assumed in HYDRUS for the lower-permeability zone are
additionally allowed to be 163 times lower than the lowest specified value in the NUREG range
for in-service hydraulic conductivity (Benson et al. 2011).
The rationale for dramatically increasing log(s) to account for the higher Ksat associated with
pedogenesis or “naturalization” has not been provided and is counterintuitive. The log(s) should
at least be similar for as-built and naturalized covers and may, in fact, be lower for naturalized
covers because pedogenic processes ameliorate hydraulic anomalies inherent in the cover from
construction. NUREG/CR-7028 (Benson et al. 2011) indicates that pedogenesis tends to
transform in-service hydraulic conductivity values to as-built values found in a much higher, but a
more restricted, range. The mean should shift upward during naturalization as structure
develops, reflecting overall increase in Ksat and α rather than a broader range.
As noted previously, while the Clive DU PA Model v1.2 may have illustrated that the output was
not sensitive to Ksat, this conclusion may be the result of predictions from a cover hydrology
model for which unrealistic parameters were used as input. Insensitivity of infiltration to hydraulic
conductivity would be expected if inappropriate input parameter values are used so as to create
in the model an unjustified, artificial capillary barrier effect. Normally, in the absence of a capillary
barrier, infiltration is very sensitive to hydraulic conductivity. As stated by Alvarez-Acosta et al.
(2012):
A soil hydraulic property that is often a required input to simulation models is the
saturated hydraulic conductivity, Ks. It is one of the most important soil physical
properties for determining infiltration rate and other hydrological processes…. In
hydrologic models, this is a sensitive input parameter and is one of the most problematic
measurements at field-scale in regard to variability and uncertainty.
Thus, the insensitivity of deep infiltration to Ksat reported in the Clive DU PA is not sufficient to
dismiss the need for demonstrating the efficacy of the parameters used for the HYDRUS input in
Appendix 5 to the DU PA Model v1.2.
DEQ Critique of DU PA v1.4, Appendix 21: See Interrogatory 21 for a description of the
EnergySolutions assumptions regarding the in-service versus naturalized parameters.
2.17.1 Interrogatory Response
In this interrogatory UDEQ questions the validity of the lognormal distribution fit by Neptune to
the in-service barrier saturated hydraulic conductivity measurements from Benson et al. (2011).
See the response to Interrogatory 21/2 for a description of the methods used for fitting the
distribution for DU PA Model v1.4 (Neptune 2015c).
ET Cover Design Responses for the Clive DU PA Model
23 Feb 2018 84
UDEQ states “Instead of using the provided distribution (i.e., log-triangular with a minimum,
maximum, and most likely), ES/Neptune constructed a lognormal distribution with a mean and
standard deviation of 0.691 and 6.396 cm/day, respectively.”
Saturated hydraulic conductivity, Ks, is generally modeled with a lognormal distribution, as
demonstrated through the figures and the reporting of the geometric mean, rather than the
arithmetic mean, for this parameter in Benson et al. (2011). Therefore, developing a lognormal
distribution for this parameter is consistent with previous work. Benson et al. (2011) provide a
median and range of likely Ks values, but do not explicitly state that the distribution should be
log-triangular. The information provided has been used for the median and the likely range, but
fit to a lognormal distribution. This assumes that the ends of the “likely range” should be close to
the 1st and 99th percentiles of the distribution, rather than a strict minimum and maximum that
would be assumed using the suggested log-triangular distribution.
The lognormal distribution has many advantages over the log-triangular distribution. More mass
of the distribution is concentrated close to the median; for the log-triangular distribution, more
mass is located in the tails where values should be less likely. In practice, this means that the
lognormal distribution will lead to draws that are closer to the median, while draws from the log-
triangular distribution would lead to more draws that are at the extreme edges of the expected Ks
values. The lognormal distribution does not exclude plausible Ks values.
Although Benson et al. (2011) provide a range of likely Ks values, they never state that it would
be impossible for Ks to be smaller or greater than the ends of the reported range. In fact,
throughout the document, the authors provide examples of other Ks values that are sometimes
beyond the given range, i.e., the in-service Ks for Cedar Rapids is 0.06 cm/day. The lognormal
distribution does not completely exclude values outside the range deemed plausible, while a log-
triangular distribution completely excludes them. This is true for the low end of the distribution,
as well as the high end, where the lognormal distribution has a long right tail, allowing for
approximately 1% of Ks values to be greater than the upper end of the range specified by Benson
et al. (2011). Finally, as stated previously, Benson et al. (2011) suggest a value for the median of
the distribution. It is often assumed the mode of a triangular distribution is equal to the median,
but this is not the case. The log of the median can be used as the mode on the log scale, but then
the median on the original scale will be close to 6, instead of the goal of 3.8.
UDEQ also has concerns with the use of 0.00432 as a minimum Ks value. The lognormal
distribution was not fit with the value of 0.0043 but this value was used to truncate the
distribution after fitting so that lower values could not be drawn at random. The lower bound
excludes Ks values below the design specification, and it barely affects the distribution. Without
the lower bound and distribution shift, the probability of a draw being at or below the boundary
of 0.0043 cm/day is less than 10-8. Setting a lower bound does not mean that the distribution
approaches that value, but it is a precaution to ensure it never goes below it, which is the use of
the lower bound here.
UDEQ argues that “EnergySolutions’ response to Comment 3 has not demonstrated that the
distribution of Ksat used for the HYDRUS modeling adequately represents the range of
conditions that might be realized for a “naturalized” cover, i.e., one that has undergone
ET Cover Design Responses for the Clive DU PA Model
23 Feb 2018 85
pedogenesis as described in NUREG/CR-7028 (Benson et al. 2011).” In general, it appears that
the distribution developed for Ks is not entirely understood. UDEQ claims that the work
“provides an unrealistic distribution of Ks,” but the distribution stems directly from the Benson et
al. (2011) field data that they would like used. The lognormal distribution has a similar median
value, 3.37 cm/day, and a similar range of values for the 1st to 99th percentile, 0.22 cm/day to
51.5 cm/day, compared to their range of 0.65 to 52 cm/day.
UDEQ then states “To account for the higher Ksat in NUREG/CR-7028 (Benson et al. 2011), the
lognormal distribution for Ksat was re-fit by the Licensee using an abnormally large log(s) of
6.396. This provides an unrealistic distribution of Ksat that substantially overweights Ksat in the
lower range.” See the response to Interrogatory 21/2 for a description of the lognormal
distribution used for Ks in Appendix 5 of the Clive DU PA Model v1.4 (Neptune 2015c)
developed based on the information in NUREG/CR-7028 (Benson et al. 2011). The distribution
has a geometric mean of 3.37 and a geometric standard deviation of 3.23 (log(3.23) = 1.17).
Therefore, the lognormal distribution used for the HYDRUS modeling to obtain the 50
realizations used for the model abstraction associated with the DU PA Model v1.4 has a log(s)
of 1.17, not 6.396. The 6.396 was used for the 20 HYDRUS realizations used for the model
abstraction in Appendix 5 of the DU PA Model v1.2 (Neptune 2014). The distribution was
revised when capability was developed to run many more HYDRUS simulations. See the
response to UDEQ Comment B.2.
UDEQ then raises the issue of a capillary barrier at the interface between the bottom of the
Evaporative Zone Layer and the top of the Frost Protection Layer. For the Frost Protection
Layer, hydraulic properties for a sandy loam were used as taken from the HYDRUS-1D pull-
down menu, which includes properties from the database of Carsel and Parrish (1988). The
parameters for a sandy loam were chosen because this texture represents a mixture of particle
sizes consisting of sand, silt, and clay in fractions representing the range seen in the Unit 3
material at the Site. This soil component fills the voids between the cobbles in the Frost
Protection Layer and determines the water flow properties of this layer.
Single deterministic values were used for the hydraulic properties θr, θs, α, n, and Ks. The
saturated hydraulic conductivity for the Frost Protection Layer was assigned a single,
deterministic value of 106.1 cm/day based on the sandy loam textural class average from the
Carsel and Parrish database. Statistical distributions were not developed for the hydraulic
properties of this layer since the function of this layer is protection against freezing and
biointrusion, and modeling uncertainty in the hydraulic properties of this layer would add
significant computational burden to the modeling.
UDEQ states that the properties of this layer contribute to the layer behaving as a capillary
barrier. A capillary barrier is a two-layered system consisting of a fine-grained layer overlying a
coarse-grained layer. In their comments UDEQ is referring to the Surface and Evaporative Zone
Layers as the upper layer and the Frost Protection Layer as the lower layer. They argue that the
hydraulic properties of these layers lead to “an extremely potent artificial, non-realistic capillary
barrier” at the interface between the Evaporative Zone Layer and the Frost Protection Layer.
ET Cover Design Responses for the Clive DU PA Model
23 Feb 2018 86
The capillary barrier phenomenon occurs due to differences in the relationship between the
hydraulic conductivity and the water tension of the two materials. Water moving in the
unsaturated zone is at a negative pressure that can be referred to as tension. For water applied to
the top of the fine layer in an initially dry profile at an initially large tension the fine-grained
material will have a hydraulic conductivity larger than the coarse-grained material. Due to this
difference, water can move down through the fine-grained layer but cannot enter the coarse-
grained layer immediately. As more water moves into the fine-grained layer the water content in
this layer increases and the corresponding water tension decreases. Water cannot enter the air-
filled pores of the coarse-grained layer until the tension decreases to the water entry value of the
coarse-grained layer. When this tension is reached, water infiltrates into the coarse-grained layer.
This phenomenon is commonly used in cover design to temporarily hold water in the upper
layers by capillarity until it is released to the atmosphere through evaporation and transpiration
(Morris and Stormont 1999). This is the same concept as the “store-and-release” cover described
by Benson et al. (2011): “The storage layer resides below the topsoil (approx. 300 mm thick) and
above a capillary break or foundation layer (if present).”
The hydraulic property values and statistical distributions assigned to the Evaporative Zone and
the Frost Protection Layers are reasonable estimates based on site-specific information and
commonly used soils databases. Given these property assignments, the Frost Protection Layer
behaves hydraulically by enhancing storage in the Evaporative Zone in a realistic manner, not as
“an extremely potent artificial, non-realistic capillary barrier.”
UDEQ states further in Comment B.3 with respect to capillary barriers, “When EnergySolutions
assumes in HYDRUS that the Ksat value for the lower-permeability zone can be as small as
0.0042 cm/day, the ratio in hydraulic conductivity between the higher-permeability zone and the
lower-permeability zone can thus be as large as 25,000.” The basis of this comment is not clear.
The saturated hydraulic conductivity of the Frost Protection Layer in the flow model is
106.1 cm/day, while the saturated hydraulic conductivity of the Surface and Evaporative Zone
layers is 4.46 cm/day. This is a ratio of 24 to 1 not 25,000 to 1.
UDEQ comments on the analysis of flow modeling results that the magnitude of net infiltration
is not sensitive to the magnitude of the saturated hydraulic conductivity of the radon barriers.
There are a number of factors that contribute to this. The sensitivity of the modeled net
infiltration rate to the value of the saturated hydraulic conductivity is discussed in the response to
Interrogatory 176/1.
2.18 SER B.4 Supplemental Interrogatory Comment 4
4) Provide justification for using the Rosetta database, as appropriate for an
engineering earthen cover.
DEQ Critique from April 2015 SER, Appendix B:
This interrogatory asked for justification for using the Rosetta database for an engineered
earthen cover. The response goes to great length comparing the attributes of the Rosetta
database to other databases, none of which are populated with data for engineered earthen
ET Cover Design Responses for the Clive DU PA Model
23 Feb 2018 87
covers. Most of the databases are for agricultural soils, many of which have been tilled. Their
relevance to an engineered earthen cover has not been demonstrated. The response has
shown, however, that many of the mean values of hydraulic properties used as input are, to
some extent, in reasonable agreement with those associated with engineered earthen covers, as
described in NUREG/CR-7028 (Benson et al. 2011). On the other hand, as discussed in the
Supplemental Interrogatory Comment 3 (see Section B.3), the low-end value in the range of
hydraulic conductivity used in the GoldSim model is 163 times lower than the lowest specified
value in NUREG/CR-7028 for in-service hydraulic conductivity. The low-permeability tail of the
distribution is overweighted, and variability is not properly accounted for.
One response to the interrogatory, if it could be substantiated using data, would be that the
Rosetta database is not based on engineered earthen cover soils and should not be assumed to
be representative, but point-wise comparisons between hydraulic recommended properties in
Rosetta and those in NUREG/CR-7028 demonstrate that the mean hydraulic properties are
similar in both cases. However, as pointed out above, the variability assumed in the hydraulic
properties chosen to represent the soils in the DU PA Model v1.2 is not appropriately
characterized, and this limitation in the model biases the modeling results greatly.
While it is true that engineered soils undergo pedogenesis and become more like natural soils
over time, it is important to follow NUREG/CR-7028 guidelines. The fact that the GoldSim model
uses values for its Ksat distribution that, at the low end, are two orders of magnitude lower than
specified in NUREG/CR-7028, and that the low-permeability range of values is overweighted,
does not lead to confidence that the GoldSim model is set up appropriately.
Furthermore, in the GoldSim model as implemented, it is assumed for the input parameter values
that there is no correlation between log(α) and log(Ksat). When databases based on natural soils
are used, it is important to account for correlation between these two parameters. Strong
correlation between log(α) and log(Ksat) (with R2 = 0.9) has been established for the largest
database in North America, as well as for the largest database in Europe [see Sections 4.1.1.1
and 4.4.1 of the safety evaluation report (SER)]. The two correlation equations are quite similar.
Furthermore, a mathematical relationship similar to the correlation equations has been
developed from fundamental soil physics theory by Guarracino (2007).
Failure to account for this correlation, or other, significant correlations (e.g., correlation in
individual parameter values between different cover-system soil layers), leads to unrealistic
modeling. As stated in GoldSim’s User Manual, Appendix A: Introduction to Probabilistic
Simulation (GTG 2013):
Ignoring correlations, particularly if they are very strong (i.e., the absolute value of the
correlation coefficient is close to 1) can lead to physically unrealistic simulations. In the
above example, if the solubilities of the two contaminants were positively correlated
(e.g., due to a pH dependence), it would be physically inconsistent for one
contaminant’s solubility to be selected from the high end of its possible range while the
other’s was selected from the low end of its possible range. Hence, when defining
probability distributions, it is critical that the analyst determine whether correlations need
to be represented.
The response has also clarified that the Surface Layer and Evaporative Zone Layer were each
assigned a geometric mean hydraulic conductivity of 5 ´ 10-7 meters per second (m/s). This
hydraulic conductivity is considered unrealistically low for in-service near-surface layers (e.g.,
< 10 feet deep) that will be densely structured due to wet-dry cycling, freeze-thaw cycling, and
biota intrusion by roots, insects, etc. This unrealistically low Ksat at or near the surface may have
ET Cover Design Responses for the Clive DU PA Model
23 Feb 2018 88
choked off infiltration in the HYDRUS model and exacerbated runoff, thereby limiting deeper
ingress of meteoric water in the profile and under-predicting percolation. As discussed in Section
4.1.1.1 of the SER, the unrealistically low near-surface Ksat value, combined with the
unrealistically high Frost Protection Layer Ksat value, which is inputted into the model, would
tend to create in the model an unrealistic, artificial capillary barrier at the top of the higher
permeability layer that would inappropriately render modeled values of infiltration extremely low.
Soils at the surface develop significant structure and generally are much more permeable than
those much deeper in the profile. EnergySolutions will need to provide additional evidence that
this assumed hydraulic conductivity did not artificially bias the HYDRUS modeling.
The response to Comment 4 also indicates that NUREG/CR-7028 recommends using a single
measurement from a single site to define α. This is an incorrect interpretation of the design
recommendations in NUREG/CR-7028. The recommendation in NUREG/CR-7028 to use α = 0.2
1/kilopascal (kPa) applies when reliable site-specific information is not available and when a
single typical value (not a range of values) is desired. It is based on an interpretation of the
dataset presented in NUREG/CR-7028 as accounting for scale-dependent hydraulic properties.
The HYDRUS modeling in Appendix 5 used an α that is approximately one order of magnitude
lower than the recommendation in NUREG/CR-7028. This α is based in part on historic
measurements made at Colorado State University on core samples obtained at the Clive site by
Bingham Environmental (1991), which are known to be too small and too disturbed to adequately
represent in-service conditions. The relevancy of this historic data from Bingham Environmental
is dubious, at best.
DEQ Critique of DU PA v1.4, Appendix 21: See Interrogatory 21 for a description of the
EnergySolutions assumptions regarding the in-service versus naturalized parameters.
2.18.1 Interrogatory Response
The class average values of soil hydraulic function parameters for the 12 soil textural
classifications in Rosetta were based on 2,134 soil samples for water retention, and 1,306 soil
samples for saturated hydraulic conductivity (Schaap et al. 2001). These data were obtained from
the RAWLS, AHUJA, and UNSODA databases (Schaap et al. 2001). Given the stronger
economic incentive for characterizing agricultural land than for rangeland, the more extensive
soils databases are derived from data obtained from agricultural lands. Soil textural
classifications are determined by particle size distributions, not by land use, so these databases
have utility for non-agricultural application. Since the objective of the revegetation plan (SWCA
2013) is to develop a sustainable, steady-state condition that mirrors the natural system, few
inputs used in the hydrologic modeling would be related to engineered properties.
The available Unit 4 soil texture data indicate that the sample represents an extreme of the range
of particle sizes that compose the silty clay textural class. Distributions were developed for the
van Genuchten α and n parameters for the Surface and Evaporative Zone Layers that represented
the entire range of the silty clay class by using the mean and standard deviation values provided
by the Rosetta database.
The Benson et al. (2011) report published by the NRC (NUREG/CR-7028) provides
recommendations for ranges of hydraulic parameters that may be used to represent in-service
conditions of store-and-release and barrier layers in covers.
ET Cover Design Responses for the Clive DU PA Model
23 Feb 2018 89
The Surface and Evaporative Zone Layers in the Clive ET cover system correspond to store-and-
release layers. The α and n values of the van Genuchten water retention model strongly influence
the capacity of the Unit 4 material in the Surface and Evaporative Zone Layers to hold water. To
provide a better estimate of the uncertainty of these Unit 4 properties, statistical distributions
were developed for α and n to be used for inputs for the flow modeling. To develop the
distributions for flow model input, α and n values were taken from the distributions of the mean
and the standard deviation for each parameter from the Rosetta model database of hydraulic
parameters for the textural class of silty clay (USDA 2017). The standard deviations were
converted to standard errors by dividing by √n, where n is the number of samples (28 in this
case). Values of the van Genuchten parameter α for these two layers were drawn from a
statistical distribution with a mean of 0.016 1/cm. The value for α recommended for in-service
layers by Benson et al. (2011) (p. 10-4) is 0.2 1/kPa, which corresponds to a value of 0.02 1/cm,
similar to the mean used for the infiltration simulations. The distribution used for the van
Genuchten n parameter for the flow model simulations had a mean of 1.32. The value for n
recommended for in-service layers by Benson et al. (2011) (p. 10-4) is 1.3.
To manage computational burden, the saturated hydraulic conductivity, Ks, for the Surface Layer
and Evaporative Zone Layer was assigned a single deterministic value of 4.46 cm/day
(5.16 × 10-5 cm/s) based on the mean value determined from analysis of the Unit 4 cores
(Equation 29 of Neptune (2015c)). The hydraulic properties for Units 3 and 4 are based on
laboratory measurements by the Colorado State University (CSU) Porous Media Laboratory for
the moisture retention and hydraulic conductivity of core samples from Units 3 and 4 at the Clive
Site (Bingham Environmental 1991). The Unit 4 material used for the Surface and Evaporative
Zone Layers is classified as a silty clay. The average saturated hydraulic conductivity assigned to
a silty clay is 0.48 cm/day from the Carsel and Parrish (1988) database and 9.61 cm/day from the
Rosetta database (USDA 2017). The value used for Ks is clearly realistic as it was derived from
site-specific data and falls between estimates from two widely used soil hydraulic parameter
databases. Comparison with results obtained by Benson et al. (2011) can be made with mean
values of the Ks of store-and-release layers of in-service covers listed in Table 6.6 of Benson et
al. (2011). The geometric mean of these results is 8.7 × 10-7 m/s or 7.5 cm/day. This value is less
than twice the value used for the infiltration modeling.
A distribution for the saturated hydraulic conductivity (Ks) of the radon barriers was developed
to use for the modeling. The resulting distribution described in the response to Interrogatory 21/2
is:
𝐾𝑠 ~ 𝐿𝑜𝑔𝑛𝑜𝑟𝑚𝑎𝑙(𝑔𝑒𝑜𝑚.𝑚𝑒𝑎𝑛:3.37 𝑐𝑚/𝑑𝑎𝑦,𝑔𝑒𝑜𝑚.𝑠𝑑: 3.23 𝑐𝑚/𝑑𝑎𝑦), with a right
shift of 0.00432 cm/day.
The value for Ks recommended by Benson et al. (2011), p. 10-3, for modeling “typical
conditions” of in-service cover layers is 5 × 10-7 m/s (4.32 cm/day), which is well within the
distribution used for the Clive DU PA infiltration modeling.
The minimum value of 0.00432 cm/day, corresponding to the design specification for the radon
barriers (Whetstone Associates 2011), was used as a shift. Note that the minimum value was not
used to fit the distribution, but simply to constrain the distribution by not allowing Ks values
ET Cover Design Responses for the Clive DU PA Model
23 Feb 2018 90
smaller than that. UDEQ suggests that truncating the distribution at the design Ks value
“overweights” the low end of the distribution. This is not correct, as the shift only makes it
impossible to choose a value for Ks that is less than the design value.
Single values of α and n determined from site-specific measurements were used for the radon
barrier in the infiltration modeling to manage computational burden. A value of 0.003 1/cm was
used for α and a value of 1.17 was used for n. These values were used in previous modeling by
Whetstone Associates (2011). Benson et al. (2011) describe α as varying between 0.001 1/cm
and 0.032 1/cm. A range from 1.2 to 1.4 is recommended by Benson et al. (2011) for the n
parameter. The value used for the infiltration modeling is slightly below the low end of that
range.
Correlations between hydraulic parameters were not included in the distribution development.
See the response to Interrogatory 05/2 for a discussion of why hydraulic parameter correlations
are not included.
In this comment UDEQ has raised a number of objections to the choice of parameter values and
statistical distributions used for the Clive DU PA infiltration modeling. These objections appear
to be based on the belief that the recommendations contained in NUREG/CR-7028 (Benson et al.
2011) should be rigidly applied to the Clive Site. Hydrogeologic models, however, are site-
specific. Climate conditions and soil formation processes at the Clive Site contradict the
assumptions of rapid soil structure formation in the cover layers observed by Benson et al.
(2011) at other sites and demonstrate the inapplicability of the conceptual model to the Clive
Site. The one-size-fits-all generalization of hydraulic properties does not appropriately represent
the unique conditions at the Clive Site. See the response to Interrogatory 05/2:
Evapotranspiration Cover (ET Cover) and the response to Interrogatory 153/2.
For discussion of capillary barrier behavior in the cover system, see the response to Comment
B.3.
In this interrogatory UDEQ again questions the quality of the hydraulic property measurements
made by the Groundwater and Porous Material Laboratory at Colorado State University. See the
discussion under Naturalized Cover in the response to Interrogatory 21/2.
2.19 SER B.5 Supplemental Interrogatory Comment 5
5) a) Provide additional explanation/justification for the assumed surface boundary
condition and the sensitivity of the HYDRUS results to the boundary conditions.
b) Also, why is a linear regression the optimal surface response for the design?
DEQ Critique from April 2015 SER, Appendix B:
The interrogatory asked for additional justification for the assumed surface boundary condition.
EnergySolutions’ response explains how the boundary condition was created but does not
provide justification for the boundary condition. Two shortcomings need to be addressed
explicitly.
ET Cover Design Responses for the Clive DU PA Model
23 Feb 2018 91
First, the repetition of the same 100-year periods 10 times to represent the climatic conditions
over a 1000-year period of climatic input will need to be justified quantitatively. For all practical
purposes, this simulation strategy will provide essentially the same output for each 100-year
period in the record. This demonstration should show that the meteorological conditions over a
1000-year period, including extreme events expected over a 1000-year period, can be
represented adequately using a sequence of repeated 100-year records. Normally, longer
periods of time involve greater variability in the data. This requested demonstration should also
show that the impacts of these extremes on the hydrological response of the cover are
adequately represented.
Second, the justification should show that the hydrological behavior at the upper boundary (i.e.,
surface of the cover) is reasonable and within expected norms. This has not been demonstrated
in Appendix 5 (Neptune 2014b), and the unrealistically low Ksat assigned to the Surface Layer
(see Comment 4) in combination with likely capillary-barrier effect artifacts in the model may
have choked off infiltration into the cover profile. At a minimum, water-balance graphs should be
presented for typical and wet years showing the temporal behavior of each of the primary
cumulative water-balance variables for the cover (e.g., precipitation, runoff, soil water storage,
evapotranspiration, percolation). These graphs, and their associated discussion, should
demonstrate that the surface boundary is represented adequately and that predictions are within
expected norms.
The absence of climate change considerations should also be presented in the context of the
most recent climate science, which does show systematic shifts in climate throughout North
America within the next 10,000 years, if not sooner. An explanation should also be provided as to
why climate change is not relevant at the Clive site when it has been considered in performance
assessments for other disposal facilities in the region (e.g., the Monticello U mill tailings disposal
facility).
EnergySolutions’ response also provides an extensive discussion to justify the efficacy of
Equations 39 and 40 in Appendix 5. However, these outcomes may have been biased by the
unrealistically low Ksat assigned to the Surface Layer and Evaporative Zone Layer (see
Comment 4), which, in combination with likely capillary-barrier effect artifacts in the model, may
have choked off infiltration into the cover profile. The efficacy of Equations 39 and 40 should be
revisited once the impacts of the unrealistically low Ksat assigned to the Surface Layer and
Evaporative Zone Layer (see Comment 4) have been investigated.
As an alternative to the linear regression, DEQ/SC&A fit an exponential equation to the van
Genuchten α, n, and Ksat input data and the HYDRUS-calculated fluxes (Figure B-1). The
triangles shown in Figure B-1 are the fluxes calculated using the following exponential fit:
Flux = 45.465 ´ α1.4408 ´ n-1.332 ´ Ksat-0.445. For large fluxes, the exponential fit does not appear to
be much better than the linear fit, but for small fluxes (which tend to result when the van
Genuchten α is small), the exponential fit is much better than the linear fit.
DEQ Critique of DU PA v1.4, See Interrogatory 21 for a description of the EnergySolutions
assumptions regarding the linear regression of the GoldSim versus HYDRUS infiltration rates.
DEQ Critique of DU PA V1.4, Appendix 21: No changes have been made with respect to the
treatment of the surface boundary conditions.
ET Cover Design Responses for the Clive DU PA Model
23 Feb 2018 92
2.19.1 Interrogatory Response
a) Atmospheric Boundary Conditions
The methods used to develop atmospheric boundary conditions applied to the top of the Surface
Layer in the HYDRUS models were described in Unsaturated Zone Modeling for the Clive PA,
Clive DU PA Model v1.4 (Neptune 2015c). Precipitation, surface runoff, and evaporation under
time-varying climate conditions are boundary condition information required by HYDRUS. The
data required for the atmospheric boundary condition includes daily values of precipitation,
potential evaporation, and potential transpiration to represent the time-variable boundary
conditions on the upper surface of the cover. The long-term evaluation period for this analysis
makes it necessary to generate a representative climate record with a longer term than any
existing daily data record. The model is deliberately run for a long period of time (1,000 years)
in order to reach a near-steady state net infiltration rate that is not influenced by the initial
conditions.
The WGEN model (Richardson and Wright 1984) was used to generate a 100-year synthetic
precipitation record for the Site. The WGEN model is a component of the widely used HELP
model (Schroeder et al. 1994a; Schroeder et al. 1994b). A 100-year precipitation record was
generated using the monthly average values from measurements at the Site based on 17 years of
observations. Meteorological measurements at the Site have been shown to be consistent with
observations at National Oceanographic and Atmospheric Administration (NOAA) stations
(Whetstone Associates 2011). Use of the WGEN model is consistent with U.S. NRC guidance
(Meyer et al. 1996). These authors found a good comparison between observed and WGEN
simulated monthly average precipitation and histograms of daily precipitation for an arid site
example.
Daily potential evapotranspiration (PET) was calculated with values of daily maximum (Tmax),
minimum (Tmin), and mean (Tmean) temperatures and extraterrestrial radiation using the
Hargreaves method. This approach is used extensively in accepted modeling platforms such as
the Soil and Water Assessment Tool (Neitsch et al. 2011) and is also documented in the
HYDRUS manuals (Šimůnek et al. 2013). The use of appropriate meteorological data and
accepted methods for estimating daily boundary condition values supports the confidence in the
application of these boundary conditions. Repetition of a long-term daily record to provide a
boundary condition for an even longer-term simulation is often necessary for the large timescales
required by PA models. Examples can be found in EPA (2002) and Levitt (2011).
UDEQ argues that “an explanation should also be provided as to why climate change is not
relevant at the Clive site when it has been considered in performance assessments for other
disposal facilities in the region.” This response provides a brief discussion of potential climate
change impacts on unsaturated zone hydrology taken from a more detailed consideration of
issues involved with representing global climate change in performance assessment models by
Crowe et al. (2017).
In evaluating the influence of climate change on the unsaturated zone, the standard approach in
past performance assessments is to assume bounding or worse-case effects of future climate
ET Cover Design Responses for the Clive DU PA Model
23 Feb 2018 93
changes through model evaluations of the wettest and coolest glacial climate in order to
maximize future precipitation and infiltration. While potentially useful, these
bounding/conservative approaches do not consider the range of possible site-specific effects
from both long-term and abrupt anthropogenic driven climate change. Instead, climate scenarios
used in performance assessments evaluate natural variations in past climate associated with
glacial and interglacial conditions. Further, the timing of future climate change or climate-driven
events are not assessed; transition to a glacial climate is assumed during the isolation period of
disposed waste.
The earth is currently in an interglacial period and the timing of the inception of the next glacial
period is critically important for forecasting future climate states. Climate modeling studies
coupling changes in the earth’s orbital configurations with variable atmospheric CO2 content
demonstrate conclusively that the earth is very unlikely to return to a glacial state when CO2
concentrations remain above pre-industrial levels (~ 280 parts per million; current levels exceed
400 parts per million). A return to pre-industrial concentrations will likely require in excess
of 50,000 years and may require hundreds of thousands of years even if current emissions were
to drop dramatically.
The most likely scenario for the future global climate is continuation of the current interglacial
climate under conditions of variable but progressive global warming. Previously used
assumptions in performance assessments of bounding glacial climates (coldest/wettest
conditions) are no longer applicable and under some conditions may require design of overly
protective and costly closure covers to reduce infiltration.
UDEQ maintains that “unrealistically low Ks” values have been assigned to the surface layer
material in the flow model. This value comes from testing of core samples from the Clive Site.
See the response to UDEQ Comment B.4 for further discussion. UDEQ argues that “likely
capillary-barrier effect artifacts in the model, may have choked off infiltration into the cover
profile.” The hydraulic property values and statistical distributions assigned to the Evaporative
Zone and the Frost Protection Layers are reasonable estimates based on site-specific information
and commonly used soils databases. Given these property assignments, the Frost Protection
Layer behaves hydraulically by enhancing storage in the Evaporative Zone in a realistic manner.
See the response to UDEQ Comment B.3 for further discussion on capillary barriers.
UDEQ has been provided with annual averages for water balance components of precipitation,
runoff, evapotranspiration, storage, and deep drainage. UDEQ claims that these water balance
plots are not adequate and have produced numerous examples of daily water balance plots. Daily
water balance is not the appropriate scale for evaluating a performance assessment model. See
the response to Interrogatory 21/2 for a discussion of evaluation of flow model water balance.
b) Use of Linear Model for Abstraction
UDEQ then discusses comparisons between linear and quadratic regression models developed
previously from the results of the 20 HYDRUS realizations.
ET Cover Design Responses for the Clive DU PA Model
23 Feb 2018 94
In response to supplemental interrogatories submitted to EnergySolutions on August 11, 2014
(EnergySolutions 2014), additional capability was developed to allow for more HYDRUS
simulations of the cover system to be completed within a practical time period. Fifty
HYDRUS-1D simulations were conducted to evaluate the uncertainty in infiltration flux into the
waste zone and water content within each ET cover layer as a function of hydraulic property
uncertainty; these simulations were documented in DU PA v1.4 Appendix 5 (Neptune 2015c).
Therefore, modeling of net infiltration and cover layer water content in Clive DU PA v1.4 is
based on 50 realizations/cases, not the 20 referred to in the interrogatory.
Predictions of average annual net infiltration from regression models with only linear terms, or
with quadratic terms, are compared to the net infiltrations actually obtained from HYDRUS
(Figure 20). The predictions from the regression models use the same input values used for the
50 HYDRUS cases (i.e., the predictions are simply the fitted values from the fitting of the
regression model to the data from HYDRUS).

Figure 20. A comparison of predictions (fitted values) from the linear and quadratic
regression models, against the HYDRUS results for net infiltration used in the model
fitting.
As seen in Figure 20, there is curvature in the HYDRUS net infiltrations that is not fully captured
in the fitted regression models, even with quadratic terms. However, HYDRUS is a complicated
process model, and an abstraction of reality itself, and the goal of the abstraction is not to capture
●●●
●●
●
●
●●●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●●●●●
●
●
●
●
●●●
●
●
●
●●
●●
●
●●●
●
●●●●●
●●
●
●
●●●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●●●●●
●
●
●
●
●●●
●
●
●
●●
●●
●
●●●
●
●●
0.00
0.05
0.10
0.15
0.20
0.00 0.05 0.10 0.15 0.20
Net Infiltration Predictions(mm/yr)
HY
D
R
U
S

N
e
t

I
n
f
i
l
t
r
a
t
i
o
n

(
m
m
/
y
r
)
Model
●
●
Linear
Quadratic
Linear and Quadratic Regression Fits
ET Cover Design Responses for the Clive DU PA Model
23 Feb 2018 95
the HYDRUS net infiltration response surface perfectly, but to provide reasonable infiltration
rates into GoldSim.
The ranges of the predictions from the two regression models are very similar, with most values
between 0 and 0.05 (Figure 21). The predictions from the model with only linear terms tend to be
slightly larger than those from the quadratic model by about 0.02 to 0.05, but the differences are
so small that they are not practically meaningful. The small differences in predictions suggest
both models are reasonable; the linear regression model was chosen for implementation in DU
PA Model v1.4 because it is simpler and matches model form and is also used for water content
across layers. Additional discussion of the predictions from the regression model with only linear
terms is provided in the response to Interrogatory 21/2.

Figure 21. A comparison of predictions from the linear and quadratic regression models
based on the input values used for the 50 HYDRUS runs.
As described in Section 1.1 and in the responses to Interrogatory 21/2 and Comment B.1, the
goal of the model abstraction is to develop linear regression models to predict reasonable net
infiltration rates and water contents for use in the DU PA Model v1.4. This comment (B.5) refers
to an “exponential fit” based on the results from an original 20 HYDRUS runs associated with
DU PA Model v1.2; the specific fit referenced is not relevant to the DU PA Model v1.4 because
of changes to distributions and the use of 50 HYDRUS cases for the model abstractions
described previously in this response.
●●●●
●
●
●
●
●●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●●
●
●●
●
●
●
●
0.00
0.05
0.10
0.15
0.20
0.00 0.05 0.10 0.15 0.20
Model Predictions with Linear Terms (mm/yr)
Mo
d
e
l

P
r
e
d
i
c
t
i
o
n
s

w
i
t
h

Q
u
a
d
r
a
t
i
c

T
e
r
m
s

(
m
m
/
y
r
)
Linear and Quadratic Regression Comparison
ET Cover Design Responses for the Clive DU PA Model
23 Feb 2018 96
The model form suggested by UDEQ is one of many possible reasonable alternatives and, as
expected, does capture some of the curvature in the HYDRUS results not captured by the linear
regression model used for DU PA Model v1.4. However, as described in other responses, the
goal of the model abstraction is to produce reasonable values of net infiltration and water content
to use for GoldSim realizations, not to match the HYDRUS results as closely as possible;
HYDRUS is itself a model of reality with its own approximations and assumptions.
For comparison, model form suggested by UDEQ was refit using the 50 HYDRUS realizations
for v1.4. The predicted values for net infiltration are very similar between the two models
(Figure 22), with the linear regression model predictions covering a slightly larger range of
infiltrations and having slightly higher infiltrations in the middle of the range covered. The small
differences in predictions suggest both models are reasonable; the linear regression model was
chosen for implementation in DU PA Model v1.4 because it is simpler and matches the model
form also used for water content across layers. Additional discussion of the predictions from the
regression model with only linear terms is provided in the response to Interrogatory 21/2.

Figure 22. Comparison of predicted net infiltration rates at the inputs used with the 50
HYDRUS runs for the linear regression model and the “exponential” model
suggested by UDEQ.

●●
●
●●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●●●●●
●
●
●
●
●
●●
●
●
●
●●
●●
●
●●●
●
●●
0.00
0.05
0.10
0.15
0.20
0.00 0.05 0.10 0.15 0.20
Model Predictions with Linear Terms (mm/yr)
Mo
d
e
l

P
r
e
d
i
c
t
i
o
n
s

w
i
t
h

E
x
p
o
n
e
n
t
i
a
l

T
e
r
m
s

(
m
m
/
y
r
)
Linear and Exponential Fits
ET Cover Design Responses for the Clive DU PA Model
23 Feb 2018 97
2.20 SER B.6 Supplemental Interrogatory Comment 6
6) To summarize the 20 HYDRUS results, Appendix 5, Section 12.9 states: “Infiltration flux
into the waste zone ranged from 0.007 to 2.9 mm/yr, with an average of 0.42 mm/yr, and
a log mean of 0.076 mm/yr for the 20 replicates.” In addition to this statement, provide
the results for each HYDRUS run so that the results can be matched to the input data.
DEQ Critique from April 2015 SER, Appendix B:
This interrogatory requested that the results be provided for each HYDRUS run so that the
results can be matched to the input data. The response included a spreadsheet summarizing
percolation from the base of the cover and water contents from the HYDRUS analysis. However,
the output from HYDRUS was not provided.
The output from HYDRUS should be included in the report and presented in a manner consistent
with the practice associated with design and evaluation of water-balance covers (i.e., ET covers).
Water-balance graphs should be reported showing the key water-balance quantities, and
discussion should be provided that demonstrates that the predictions are within expected norms
for water-balance covers. This type of presentation and discussion has not been provided in
Appendix 5 or in subsequent responses to interrogatories.
EnergySolutions’ response also discusses graphs in an attached spreadsheet and indicates that
these graphs demonstrate that there is no relationship between percolation from the base of the
cover and Ksat of the radon barrier. This finding may have been biased by the unrealistically low
Ksat assigned to the Surface Layer and Evaporative Zone Layer (see Comment 4), which, in
combination with likely capillary-barrier effect artifacts in the model, may have choked off
infiltration into the cover profile. This issue needs to be reevaluated once the impact of the Ksat
assigned to the near-surface layers has been addressed.
DEQ Critique of DU PA v1.4, Appendix 21: See Interrogatory 21 for a description of the
adequacy of the HYDRUS model output.
2.20.1 Interrogatory Response
In response to supplemental interrogatories submitted to EnergySolutions on August 11, 2014,
additional capability was developed to allow for more HYDRUS simulations of the cover system
to be completed within a practical time period. Fifty HYDRUS-1D simulations were completed
to evaluate the uncertainty in infiltration flux into the waste zone, and water content within each
ET cover layer as a function of hydraulic property uncertainty, and were documented in DU PA
v1.4 Appendix 5 (Neptune 2015c). See the response to UDEQ Comment B.1 (Section 2.15.1).
The input parameters for the 50 realizations were provided in Appendix 5 of DU PA Model v1.4
(Neptune 2015c) and are reproduced in this document in Table 6. Flow model results for use in
the GoldSim model consisted of the average annual fluxes into the top of the waste layer and
average annual water contents for the Surface Layer, Evaporative Zone Layer, Frost Protection
Layer, and Upper and Lower Radon Barriers averaged over the years 900 to 1000 of the
simulation. The outputs from HYDRUS for each realization are provided in Table 7. The results
of these 50 realizations were used to fit regression equations for net infiltration and water content
for the DU PA Model v1.4.
ET Cover Design Responses for the Clive DU PA Model
23 Feb 2018 98
Table 7. Results of 50 flow realizations described in Appendix 5 of DU PA Model v1.4.
Realization
Net
Infiltration
(mm/yr)
Volumetric Water Content (-)
Surface Evaporative
Zone
Frost
Protection
Upper
Radon
Lower
Radon
1 0.0107 0.22 0.25 0.07 0.25 0.25
2 0.0123 0.21 0.25 0.07 0.24 0.24
3 0.0085 0.21 0.25 0.07 0.25 0.25
4 0.0236 0.22 0.26 0.07 0.25 0.25
5 0.0279 0.23 0.27 0.07 0.27 0.27
6 0.0079 0.23 0.27 0.07 0.24 0.24
7 0.0230 0.21 0.25 0.07 0.26 0.26
8 0.0098 0.23 0.26 0.07 0.26 0.26
9 0.0146 0.21 0.25 0.07 0.26 0.26
10 0.0131 0.22 0.25 0.07 0.25 0.25
11 0.0221 0.24 0.28 0.07 0.26 0.26
12 0.0281 0.23 0.26 0.07 0.27 0.27
13 0.0144 0.22 0.26 0.07 0.26 0.26
14 0.0380 0.22 0.25 0.07 0.27 0.27
15 0.0254 0.22 0.26 0.07 0.27 0.27
16 0.0122 0.24 0.28 0.07 0.26 0.26
17 0.0663 0.21 0.25 0.07 0.27 0.27
18 0.0084 0.22 0.26 0.07 0.24 0.24
19 0.0364 0.22 0.26 0.07 0.27 0.27
20 0.1832 0.22 0.26 0.08 0.28 0.28
21 0.0075 0.24 0.28 0.07 0.24 0.24
22 0.0248 0.24 0.28 0.07 0.25 0.25
23 0.0301 0.22 0.25 0.07 0.28 0.28
24 0.0458 0.20 0.24 0.07 0.27 0.27
25 0.0084 0.22 0.26 0.07 0.24 0.24
26 0.0109 0.21 0.24 0.07 0.27 0.27
ET Cover Design Responses for the Clive DU PA Model
23 Feb 2018 99
Realization
Net
Infiltration
(mm/yr)
Volumetric Water Content (-)
Surface Evaporative
Zone
Frost
Protection
Upper
Radon
Lower
Radon
27 0.0104 0.21 0.25 0.07 0.24 0.24
28 0.0117 0.24 0.28 0.07 0.25 0.25
29 0.0106 0.21 0.25 0.07 0.24 0.24
30 0.0821 0.22 0.26 0.08 0.26 0.26
31 0.0220 0.21 0.25 0.07 0.27 0.27
32 0.0649 0.23 0.27 0.07 0.27 0.27
33 0.0168 0.23 0.26 0.07 0.26 0.26
34 0.0067 0.24 0.28 0.07 0.22 0.22
35 0.0101 0.23 0.27 0.07 0.23 0.23
36 0.0149 0.21 0.25 0.07 0.27 0.27
37 0.0210 0.25 0.29 0.07 0.26 0.26
38 0.0086 0.22 0.26 0.07 0.24 0.24
39 0.0233 0.22 0.26 0.07 0.25 0.25
40 0.0096 0.23 0.27 0.07 0.25 0.25
41 0.0088 0.25 0.29 0.07 0.25 0.25
42 0.0151 0.23 0.27 0.07 0.25 0.25
43 0.0155 0.22 0.26 0.07 0.23 0.23
44 0.0301 0.23 0.27 0.07 0.24 0.24
45 0.0168 0.22 0.25 0.07 0.26 0.26
46 0.0142 0.24 0.28 0.07 0.25 0.25
47 0.0122 0.23 0.27 0.07 0.26 0.26
48 0.0471 0.24 0.28 0.07 0.26 0.26
49 0.0157 0.24 0.28 0.07 0.27 0.27
50 0.0130 0.24 0.28 0.07 0.25 0.25

UDEQ requests plots of flow model water balance components on a daily basis. See the
discussion of the applicability of daily water balance plots to PA models in the response to
Interrogatory 21/2.
ET Cover Design Responses for the Clive DU PA Model
23 Feb 2018 100
UDEQ questions the outcome from the fitting of the regression equation that net infiltration rates
are not sensitive to the magnitude of the saturated hydraulic conductivity of the radon barriers.
See the response to UDEQ Comment B2 (Section 2.16.1) for discussion of this issue.
2.21 SER B.7 Supplemental Interrogatory Comment 7
7) The HYDRUS and GoldSim calculated infiltration rates (and perhaps other intermediary
results) need to be provided in the report, so that the reviewers do not have to delve into
the code’s output files. For example, provide dot plots of the infiltration rates through the
surface layer and/or provide a statistical summary of the infiltration rates that were
sampled in GoldSim.
DEQ Critique from April 2015 SER, Appendix B:
This interrogatory requested that the percolation rates reported by HYDRUS be presented
directly in the report. The response includes Figure 4, which shows “infiltration” in mm/yr for
various layers in the cover and Figure 5, which shows “infiltration” (interpreted as percolation
from the base of the cover) from HYDRUS and predicted with the regression equation, i.e.,
Equation 39 in Appendix 5.
The quantities shown in Figure 4 need more explanation. Infiltration is defined as the flux of
water across the atmosphere-soil interface in response to precipitation. Water movement below
the surface is a volumetric flux, and the flux from the base of the cover and into the waste is the
percolation rate for the cover. Do these quantities represent the net flux from the base of each
layer in the cover? The “infiltration” for the surface layer report in Figure 4 also raises concern, as
the results indicate that the unrealistically low Ksat assigned to the Surface Layer and
Evaporative Zone Layer (see Comment 4), in combination with likely capillary-barrier effect
artifacts in the model, may have choked off infiltration into the cover profile and unrealistically
limited downward movement of water. A discussion of the HYDRUS predictions in the context of
cumulative water-balance quantities and expected norms for water-balance covers could
address this issue.
As indicated in the discussion associated with Comment 1, the predictions shown in Figure 5
illustrate that the percolation rate from the regression used in GoldSim is considerably different
from the predictions made with HYDRUS and is essentially insensitive to the hydraulic properties
used as input. The lack of sensitivity is attributed to the reduction in log-variance to address
spatial averaging, but another plausible explanation is that Equation 39 reflects central conditions
adequately but extreme conditions in the tailings inadequately. Yet another plausible explanation
is the likely capillary-barrier effect artifacts in the model, which would minimize or possibly even
exclude infiltration of water to greater depths, so long as evaporation could remove it from the
upper two soil layers. Furthermore, evapotranspiration rates in the model are likely too high,
since they do not account for accumulation of gravel at the surface over time, which would tend
to greatly diminish evaporation. A quantitative demonstration and explanation is needed to
address this issue.
The response should also indicate how and why temporal scaling was incorporated into the
hydraulic properties, as indicated by the term “spatio-temporal” used in the response to the
interrogatory. Temporal scaling should account explicitly for the temporal evolution of the
distribution of hydraulic properties due to pedogenic effects. No discussion has been provided
regarding a temporal evolution of hydraulic properties. If temporal scaling has not been
incorporated, then scale matching should be described as spatial rather than spatio-temporal.
ET Cover Design Responses for the Clive DU PA Model
23 Feb 2018 101
EnergySolutions’ response should also indicate why conventional spatial averaging procedures
for correlated hydraulic properties were not used in the spatial scaling process from point scale
measurements in the Rosetta database to grid scale in the model. Spatial scaling from a point
measurement to model grid scale will need to account for upscaling of the mean to address
measurement bias as well as downscaling of the log-variance in a manner consistent with the
spatial correlation structure of engineered but degraded-over-time in-service earthen cover soils.
The response should indicate how these factors are addressed by reducing the log-variance by
the square root of the sample size in the Rosetta database.
The discussion below illustrates DEQ’s mathematical (as opposed to hydrogeologic) concerns
with the way infiltration is being abstracted into GoldSim from the HYDRUS results.
1) The linear regression equation that has been programmed into GoldSim does not give
results that are consistent with what is calculated by HYDRUS (i.e., for a given pair of α and
n, the regression equation result in GoldSim does not approximate the HYDRUS result).
This is demonstrated by Figure B-1 (See DEQ Critique to Supplemental Interrogatory
Comment 5).
2) As acknowledged by EnergySolutions in its responses to Supplemental Interrogatories 1
and 2, due to scaling effects the ranges for α and n that have been programmed into
GoldSim are more narrow than those in HYDRUS (i.e., in HYDRUS, α ranges from
0.001883 to 0.3021, but in GoldSim, α only ranges from 0.005 to 0.0493; likewise, in
HYDRUS, n ranges from 1.029 to 1.883, but in GoldSim n only ranges from 1.060 to 1.540).
See Figure B-2 and Figure B-3 for complementary cumulative distribution (CCD)
comparisons that were prepared by SC&A utilizing EnergySolutions HYDRUS results and
the Neptune (2014b), Table 1 GoldSim α and n distributions.

The CCD comparison in Figure B-4 shows the effect of these two mathematical considerations
on the resulting GoldSim infiltration rate. This infiltration CCD is very similar to Figure 5 of the
EnergySolutions Response to Supplemental Interrogatories, except that it is rotated 90 degrees.
Note that GoldSim was not re-run for these analyses. Instead, the GoldSim equations were
programmed into an Excel Crystal Ball file, and 10,000 realizations were run. Also, the reason
the GoldSim CCDs are smoother than the HYDRUS CCDs is that the GoldSim CCDs have
10,000 points, whereas the HYDRUS CCDs have only 20.
DEQ Critique of DU PA v1.4, Appendix 21: See Interrogatory 21 for a description of the
adequacy of the GoldSim and HYDRUS model output.
2.21.1 Interrogatory Response
In this comment UDEQ first references the regression models for net infiltration developed from
the HYDRUS simulations with 20 realizations from Appendix 5 of v1.2 of the Clive DU PA
Model (Neptune 2014). As described previously, in response to supplemental interrogatories
submitted to EnergySolutions on August 11, 2014 (EnergySolutions 2014), additional capability
was developed to allow for more HYDRUS simulations of the cover system to be completed
within a practical time period. Fifty HYDRUS-1D simulations were conducted to evaluate the
uncertainty in infiltration flux into the waste zone and water content within each ET cover layer
as a function of hydraulic property uncertainty; these simulations were documented in DU PA
v1.4 Appendix 5 (Neptune 2015c). Therefore, net infiltration and cover water content values
ET Cover Design Responses for the Clive DU PA Model
23 Feb 2018 102
used in Clive DU PA v1.4 are based on 50 realizations/cases, not the 20 referred to in this
comment.
UDEQ requests that HYDRUS and GoldSim calculated infiltration rates be provided in a report
and that a statistical summary of the infiltration rates that were sampled in GoldSim be provided.
The input parameters for the 50 realizations were provided in Appendix 5 of DU PA Model v1.4
(Neptune 2015c). Infiltration rates used in GoldSim can be calculated using the regression model
in Appendix 5 (Neptune 2015c). The net infiltration rates calculated in the 50 HYDRUS
realizations are given in Table 7 of this document. A statistical summary of net infiltration
calculated in the 50 realizations is given in Section 12.9 of Appendix 5 (Neptune 2015c).
UDEQ shows comparisons between the van Genuchten α and n values generated using the
distributions for these parameters from the GoldSim model and the values used for the 20
HYDRUS simulations in their Figures B-2 and B-3. In addition, UDEQ shows a comparison in
their Figure B-4 between values of net infiltration generated using the regression model and the
values resulting from the 20 HYDRUS realizations. These comparisons for input parameters and
net infiltration clearly differ.
UDEQ has suggested a number of causes for these differences but, as described in
EnergySolutions (2014), these differences were due to not scaling the α and n distributions used
as inputs for the HYDRUS simulations. The proper scaling has been applied for the 50
HYDRUS simulations described in Appendix 5 of the DU PA Model v1.4 (Neptune 2015a).
The same comparisons using v1.4 distributions for α and n are shown in Figure 23 and Figure 24,
corresponding to the v1.2 comparisons in UDEQ’s Figures B-2 and B-3. Comparisons between
the complementary cumulative distribution functions for the 50 net infiltration values from the
HYDRUS realizations and 1000 values generated from the regression model used in GoldSim
are shown in Figure 25. These results demonstrate that proper scaling of the input parameter
distributions has resulted in consistency between HYDRUS inputs and outputs and the simplified
model used in GoldSim for net infiltration.
ET Cover Design Responses for the Clive DU PA Model
23 Feb 2018 103

Figure 23. Re-creation of information in UDEQ Figure B-2 showing complementary
cumulative distribution functions (CDFs) for the 50 α values used in the HYDRUS
realizations and 1000 values drawn from the distribution used in GoldSim for the
Clive DU PA Model v1.4.

Figure 24. Re-creation of information in UDEQ Figure B-3 showing complementary
cumulative distribution functions (CDFs) for the 50 n values used in the HYDRUS
realizations and 1000 values drawn from the distribution used in GoldSim for the
Clive DU PA Model v1.4.
0.00
0.25
0.50
0.75
1.00
0.00 0.01 0.02 0.03 0.04 0.05
α(1/cm)
Co
m
p
l
e
m
e
n
t
a
r
y

C
D
F
Model
HYDRUS
GoldSim
Complementary CDF for α: GoldSim and HYDRUS
0.00
0.25
0.50
0.75
1.00
1.2 1.3 1.4 1.5
n
Co
m
p
l
e
m
e
n
t
a
r
y

C
D
F
Model
HYDRUS
GoldSim
Complementary CDF for n: GoldSim and HYDRUS
ET Cover Design Responses for the Clive DU PA Model
23 Feb 2018 104

Figure 25. Re-creation of information in UDEQ’s Figure B-4 showing complementary
cumulative distribution functions (CDFs) for the 50 net infiltration values from the
HYDRUS realizations and 1000 values drawn from the distribution used in GoldSim
for the Clive DU PA Model v1.4.
UDEQ then argues “Furthermore, evapotranspiration rates in the model are likely too high, since
they do not account for accumulation of gravel at the surface over time, which would tend to
greatly diminish evaporation. A quantitative demonstration and explanation is needed to address
this issue.”
The accumulation of gravel at the surface at the Clive Site is highly unlikely. Recent field studies
(Neptune 2015b) provide evidence that the Site is within a region of significant eolian activity
(wind driven) evidenced by locally thick accumulation of gypsum dunes west and southwest of
the Site and a laterally continuous layer of suspension fallout silts preserved beneath the modern
surface throughout the Clive Site. The eolian deposits in the upper part of the stratigraphic
section represent a 10,000-year record of deposition and soil formation (Neptune 2015b).
Observations of Holocene eolian silt throughout the Clive Site support a conceptual model of
long-term eolian deposition on a stable surface that promotes and preserves concurrent eolian
deposits which are only slightly modified by slow processes of soil formation. See the response
to Interrogatory 05/2 for further discussion of eolian deposits and soil formation at the Site.
UDEQ requests clarification of the term “spatio-temporal scaling” used in reference to
distribution development for hydraulic properties. UDEQ suggests that “temporal scaling should
account explicitly for the temporal evolution of the distribution of hydraulic properties due to
pedogenic effects,” and “If temporal scaling has not been incorporated, then scale matching
should be described as spatial rather than spatio-temporal.”
0.00
0.25
0.50
0.75
1.00
0.00 0.05 0.10 0.15
Net Infiltration (mm/yr)
Co
m
p
l
e
m
e
n
t
a
r
y

C
D
F
Model
HYDRUS
GoldSim
CCDF for Net Infiltration: GoldSim and HYDRUS
ET Cover Design Responses for the Clive DU PA Model
23 Feb 2018 105
UDEQ’s conceptual model of significant soil structural changes occurring during the
performance period of the Clive DU PA Model is inconsistent with field observations at the Site.
For a discussion of soil formation at the Site, see the response to Interrogatory 05/2.
As described in Section 1.1 and Appendix 14 of the DU PA Model v1.4 (Neptune 2015e), the
temporal scale represented in a distribution should align with the scale represented by a value
from the distribution in the PA Model. The PA Model takes a single draw from each of the
relevant hydraulic property distributions at the beginning of a model realization and typically
uses that value for the entire realization of the model (over all time represented by the model).
From a temporal perspective, the values should be judged to be reasonable and to be held
constant over long time periods, and therefore should not reflect extremes that are only realistic
at smaller time scales.
Thinking about plausible long-term averages for the values representing the spatial volumes of
the PA Model is the easiest way to judge reasonableness of the distribution relative to time.
However, this is challenging because it is clearly impossible to collect data over 100s or 1000s of
years into the future to capture potential soil forming effects when developing the distribution of
a long-term average. Therefore, temporal scaling is more abstract than the methods discussed for
spatial scaling, and actually uses spatial variability as a surrogate for variability that might be
observed over time. For example, using data from different locations and regions captures
conditions in different materials that may be observed at one location over time through soil
forming effects. Therefore, the heterogeneity over space is used as a surrogate for heterogeneity
over time for the volumes represented by the PA Model; this is necessary if distributions are to
be based on empirical information because it is not possible to collect the data needed to capture
the heterogeneity over time into the future. Therefore, while the scaling does appear to be simply
“spatial scaling” on the surface, it is not a misnomer to call it spatio-temporal scaling because
both scales are always simultaneously considered even if most of the information available is
spatial. The resulting distributions should be evaluated relative to both dimensions by asking
“does the distribution provide values that are plausible when applied to the spatial volumes of
PA cells over long periods of time?”
The “conventional spatial averaging procedures” referred to by UDEQ are useful when extensive
spatially referenced point-scale data are collected from the area of interest (the Site) and the goal
is to aggregate or “upscale” the point-scale data to reflect information at many larger grid cells
representing aggregated material over the same spatial area as covered by the point-scale data.
There are two main reasons this approach is not realistic for distribution development for the DU
PA Model: (1) the Rosetta data are not site-specific (i.e. not from Clive) and therefore are not
spatially referenced to the volumes of soil represented by cells of the PA Model, and (2) the
volumes and time periods represented by cells of the PA Model are very large relative to the
areas over which data are collected.
The Rosetta database represents a diverse collection of measurements from samples at different
locations and different points in time, with no spatial reference to the Clive Site. As described in
Section 1.1 and Appendix 14 of the DU PA Model v1.4 (Neptune 2015e), and in other responses
related to “scaling,” the goal is to capture values in the distributions that are deemed plausible
considering they must represent large volumes of soil and long periods of time. Measurements
ET Cover Design Responses for the Clive DU PA Model
23 Feb 2018 106
from single locations included in the Rosetta database will capture extreme values that would not
be representative of aggregated measurements representing large volumes of material, and
therefore the standard deviation of the raw measurements should not be used as the standard
deviation of the final distribution for the PA Model. Instead, the distributional goals for the PA
Model require aggregation over potential values from single locations and times that could be
observed at the Site to develop a distribution reflecting plausible aggregated values given the
current state of knowledge.
Given the lack of site-specific data and spatially explicit information, a reasonable and simple
way to proceed is to base the distribution on the average over the values from a variety of
conditions found in the Rosetta database. The goal is then to develop a distribution capturing
plausible values of averages of the input variables to apply over the spatial areas and time
periods used in the PA Model. An elegant way to obtain such a distribution is use of fundamental
statistical results regarding the distribution of plausible averages that could have been obtained
under many different random samples of point-locations and times (the sampling distribution of
the average), and its estimated standard deviation (the standard error of the average). That is,
based on the available data, statistical theory can be used to develop a distribution describing
other averages that could have been obtained had sampling differed. The sampling distribution
can also be thought of as communicating the uncertainty in the estimated mean. Greater
variability in the raw data or less information in available data (e.g., smaller sample sizes) leads
to more uncertainty, and therefore a larger standard error of the average.
A reasonable distribution is a normal distribution with standard deviation equal to the standard
error of the average based on the Rosetta database, calculated as the sample standard deviation
divided by the square root of the sample size. This is not the only approach that could be used to
distribution development, but its simplicity is appropriate given the distributional goal and
available information. The approach essentially assumes that the available data represent a
collection of measurements from point locations randomly selected (in space and time) from the
Site over the next 10,000 years. The sampling locations and occasions are assumed to be far
enough apart that they can be assumed to be independent. Then, the sampling distribution of the
average reflects plausible spatio-temporal averages that are reasonable to apply over the Site and
over the length of time the model is run. For the Clive DU PA Model v1.4 distributions, the
distributions for α, n, and Ks are developed on the log scale where the underlying population
distribution of Rosetta values is more symmetric, further justifying the use of the normal
distribution to approximate the sampling distribution of the average.
2.22 SER B.8 Supplemental Interrogatory Comment 8
8) a) Demonstrate that the fitted equations for water content and infiltration (Appendix 5,
Equations 39 and 40, and Table 10) give “reasonable” results when compared to
HYDRUS.
b) For example, provide an explanation for why Ksat is insensitive to the infiltration rates.
ET Cover Design Responses for the Clive DU PA Model
23 Feb 2018 107
DEQ Critique from April 2015 SER, Appendix B:
This interrogatory asked for demonstration that Equations 39 and 40 provide realistic predictions
relative to the predictions from HYDRUS. EnergySolutions’ response provides a number of
graphs showing that the predictions in the Clive DU PA Model v1.2 using Equations 39 and 40
are similar to those from HYDRUS in the sense of the mean but exhibit less variability than the
predictions in HYDRUS. The reduced variability in the percolation predicted by Equation 39 is
attributed to the reduction in log-variance to address spatial averaging, but another plausible
explanation is that Equation 39 reflects central conditions adequately, but extreme conditions in
the tailings inadequately. A quantitative demonstration and explanation is needed to resolve this
issue.
This interrogatory also asked for an explanation of the lack of sensitivity of percolation rate to
Ksat. The response on pages 25 and 26 (un-numbered figures) shows that water is isolated in
the surface layer. However, using an unrealistically low Ksat for the Surface Layer and
Evaporative Zone Layer, in combination with likely capillary-barrier effect artifacts in the model
(see Comment 4), may have choked off infiltration into the cover profile and trapped water at the
surface, thereby limiting downward movement of water unrealistically and artificially impacting
the significance of Ksat of the radon barrier. A discussion of the HYDRUS predictions in the
context of cumulative water-balance quantities and expected norms for water-balance covers
could address this issue.
DEQ Critique of DU PA v1.4, Appendix 21: See Interrogatory 21 for a description of the
verification of the HYDRUS results.
2.22.1 Interrogatory Response
UDEQ requests a comparison between the regression models in the GoldSim DU PA Model for
net infiltration and water content with the HYDRUS simulation results for these parameters. The
request refers to “Appendix 5, Equations 39 and 40, and Table 10.” Equations 39 and 40 are
found in Appendix 5 of the DU PA Model v1.2 (Neptune 2014). As described in the response to
UDEQ Comment B. 1, improved regression models were provided based on 50 HYDRUS
simulations in Appendix 5 of the DU PA Model v1.4 (Neptune 2015c). These models correspond
to Equations 41 and 42 in v1.4. The regression equation coefficients are provided in Table 10 of
both documents. This comparison of net infiltration between the regression models and
HYDRUS results is discussed in the response to Comment B.1.
A comparison between measurements of the water content of the Evaporative Zone Layer Unit 4
soil and water contents obtained from the regression model is discussed in the response to UDEQ
Comment B. 9.
UDEQ raises a concern about the sensitivity of modeled net infiltration to the value of the
saturated hydraulic conductivity. See the response to Interrogatory 176/1 for a discussion of the
sensitivity of the net infiltration at the Clive Site to the values of the saturated hydraulic
conductivity of the cover layers.
In this interrogatory UDEQ argues that the values of the saturated hydraulic conductivity
assigned to the Surface and ET Zone Layers are unrealistically low and that hydraulic properties
ET Cover Design Responses for the Clive DU PA Model
23 Feb 2018 108
used create an artificial and unrealistic capillary barrier. See the response to UDEQ Comment
B.4 for a discussion of Surface and ET Zone Layer saturated hydraulic conductivities.
See the response to UDEQ Comment B.3 for a discussion of capillary barrier effects.
2.23 SER B.9 Supplemental Interrogatory Comment 9
9) Compare the moisture contents calculated using the fitted equations to the Bingham
(1991, Table 6 and/or Appendix B) Clive site measured Unit 4 moisture contents,
and rationalize any differences.
DEQ Critique from April 2015 SER, Appendix B:
The comparison with HYDRUS is remarkably good. However, the comparison with Equation 39
is not good. Equation 39 seems to predict q between 0.27 and 0.31 for nearly all cases, whereas
the data are over a much broader range.
DEQ Critique of DU PA v1.4, Appendix 21: See Interrogatory 21 for a description of the
adequacy of the goodness of fit against the Bingham (1991) data.
2.23.1 Interrogatory Response
UDEQ requested a comparison between the regression model for water content in Appendix 5 of
DU PA v1.2 (Neptune 2014) with the water content data for Unit 4 material acquired by
Bingham Environmental (1991). Comparisons were made between the regression model for the
Evaporative Zone Layer and Unit 4 water contents listed in Table 6 of Bingham Environmental
(1991). The Surface Layer was not selected for this comparison because it has a reduced porosity
due to the gravel admixture.
Gravimetric water contents for Unit 4 soils, at depths less than or equal to 2 feet (near the depth
of the Evaporative Zone Layer (0.5 to 1.5 ft)), were taken from Bingham Environmental (1991),
(Table 6, pdf p. 42-43). Six values matched the depth constraint and those data are presented in
Table 8. Volumetric water contents for these six samples were calculated by multiplying the
gravimetric values by the bulk density of 1.397 g/cm3 reported on the Adobe pdf page 174 of
Bingham Environmental (1991) for sample GW19A-B1 (Unit 4 sample).
ET Cover Design Responses for the Clive DU PA Model
23 Feb 2018 109
Table 8. Water Content Data from Table 6 of Bingham Environmental (1991).
Bingham 1991 data
DH ID Sample ID Depth (ft) Unit Grav. WC
% Vol. WC

GW-17A L-1 2 4 27.8 0.39
GW-19B L-1 2 4 17.5 0.24
SLC-203 NA 2 4 21.7 0.30
SLC-204 NA 2 4 15.3 0.21
SLC-205 NA 2 4 20.7 0.29
SLC-206 NA 2 4 19.6 0.27

Avg 20.43 0.285
Min 15.30 0.214
Max 27.80 0.388

Volumetric water contents calculated using the regression model were extracted from the
GoldSim DU PA Model v1.2 by adding a result element for the Expression
“WaterContentETCover_regr”. Then the model was run for 1,000 simulations to generate 1,000
values of water content for the Evaporative Zone Layer (Unit 4 soil).
Volumetric water contents from the regression model used in GoldSim (1000 replicates), from
HYDRUS (20 replicates), and from the six measured values from Table 8, are shown in Figure
26. For the x-axis, each of the 6 values in Table 8 were plotted at increments of ~ 167 in order to
show the data on the x-axis with 1,000 values (for the GoldSim results). Similarly, the
HYDRUS-1D values were plotted at increments of 50. As shown in Figure 26, the volumetric
water contents calculated with the fitted equation in GoldSim are well-bounded by the Bingham
data from Table 8. The mean volumetric water content value in Table 8 is 0.285 while the mean
from the GoldSim model 1,000 replicates is slightly higher at 0.294. The mean value of the 20
HYDRUS-1D replicates is 0.286, effectively identical to the Bingham Environmental (1991)
samples.
UDEQ states that, while the comparison of the sampled water contents with the HYDRUS
results is remarkably good, the comparison with Equation 39 (regression model) is not good.
They state that “Equation 39 seems to predict q (water content) between 0.27 and 0.31 for
nearly all cases, whereas the data are over a much broader range.” This observation on their
part is incorrect. Examination of the results from the regression model (labeled “GoldSim”) in
Figure 26 show water content values ranging from slightly greater than 0.1 to a maximum of
0.35. It is also important to keep in mind that the parameter input distributions to the regression
model have been upscaled to specify the model at the appropriate spatial scale of the Site and
the temporal scale of the model. Using proper upscaling the regression model output represents
ET Cover Design Responses for the Clive DU PA Model
23 Feb 2018 110
uncertainty in the system’s average response rather than the variability of data collected at
points in time and space.

Figure 26. Comparison of Bingham Environmental (1991) water content data with water
content calculated using the regression equation for the DU PA GoldSim model and
with the results of the 20 HYDRUS simulations.
2.24 SER B.11 Supplemental Interrogatory Comment 11
DWMRC provided EnergySolutions with an Excel file, “Clive Hydrus Sensitivity Recommend
REV2.xlsx,” which contains suggested or proposed combinations of input values for the
HYDRUS runs used to support the Clive DU PA.
DEQ Critique from April 2015 SER, Appendix B:
DEQ requested a sensitivity analysis for a reasonable range of parameters to evaluate whether
the model responds within expected norms for a water-balance cover. This request has been
made in part because Appendix 5 provides inadequate documentation to demonstrate the
efficacy of the HYDRUS model and its realism relative to expected norms for a water-balance
cover.
Moreover, Appendix 5 indicates that predictions made by the model are insensitive to hydraulic
parameters (notably Ksat) generally known to have a strong influence on predictions made by
HYDRUS and similar models. For example, the unrealistically low Ksat for the Surface Layer and
Evaporative Zone Layer (see Comment 4) may have choked off infiltration into the cover profile
and trapped water at the surface, thereby limiting downward movement of water unrealistically
and artificially impacting the significance of Ksat of the radon barrier. As explained throughout
ET Cover Design Responses for the Clive DU PA Model
23 Feb 2018 111
this document, there are significant concerns that the HYDRUS model may not be realistic and
may be biasing the analyses in the performance assessment. An assessment of the efficacy of
the HYDRUS model in the context of expected norms is essential to resolve this issue.
EnergySolutions’ response goes to great length to dismiss the requested sensitivity analysis as
not based on reasonable soil properties and as being inconsistent with a performance
assessment approach. The response justifies the criticism of the soil properties by citing
databases for soil properties unrelated to engineered earthen covers (e.g., the National
Resource Conservation Service database) or data reports known to contain measurements on
samples that are too small to represent in-service conditions and collected with antiquated
techniques that are known to cause disturbance of soil structure (e.g., the 1991 Bingham
Environmental report).
Despite these criticisms, the requested analyses apparently were conducted, but the output was
not included or presented comprehensively in the responses. The findings from these
simulations should be tabulated and reported, and water-balance graphs should be prepared and
discussed in the context of the mechanisms known to influence the hydrology of water-balance
covers. A thoughtful discussion would help justify the use of the HYDRUS model and build
confidence in the output.
DEQ Critique of DU PA v1.4, Appendix 21: See Interrogatory 21 for a description of the
adequacy of the range, distribution and bounds on the HYDRUS input data. The type of output
that should be provided is also presented.
2.24.1 Interrogatory Response
UDEQ requested responses to 10 supplemental interrogatories on August 11, 2014. Comment 11
in Appendix B of the SER (SC&A 2015a) contains Comment 10 of the 10 supplemental
interrogatories.
In this interrogatory, UDEQ provided a table with sets of flow model input parameters for the
cover layers. Single values were specified for the van Genuchten n parameter, the residual water
content, and the saturated water content. These same values were to be applied to all layers.
Values of the van Genuchten α parameter and the saturated hydraulic conductivity, Ks, were
given for each layer in categories labeled “low,” “typical,” and “high.” The values for α and Ks
for the radon barrier and the Frost Protection Layer were identical. UDEQ requested that
HYDRUS simulations be conducted using multiple combinations of parameter values for a total
of nine simulations.
For each simulation, UDEQ instructed the modelers to “run ‘warm up’ simulation 5 times back
to back beforehand using meteorological year having annual precipitation closest to long-term
average. Use heads from end of this 5 yr simulation as initial conditions for the performance
simulation.”
ET Cover Design Responses for the Clive DU PA Model
23 Feb 2018 112
EnergySolutions expressed their concern with this approach but did fulfill UDEQ’s request to
run the requested simulations. EnergySolutions provided the following response
(EnergySolutions 2014).
In general, EnergySolutions strongly disagrees with the request of running highly
speculative, unsupported, one-off cases suggested in the subject request. This is not
consistent with the intent of the Utah regulation nor the meaning or application of a
“sensitivity analysis.” In practice, an appropriate sensitivity analysis would consider only
combinations of input values that are plausibly visible at the site under study. Whereas
the concept of plausibility in this context is applied based on available data and
professional judgment, the values that are suggested in the subject document (and
repeated above) are not plausible for this site.
EnergySolutions also disagrees with the intent given that the site will return to natural
conditions. In fact, the Division-suggested input values do not match natural conditions,
whereas the probability distributions used in the Clive DU PA Model provide reasonable
bounds for site conditions projected into the future given the available information and
data.
There are significant limitations in assessing the effects of parameter and conceptual
uncertainty using deterministic modeling with specified (discrete) cover designs and
bounding transport parameters and assumptions. Any more comprehensive sensitivity
analysis for the infiltration modeling should not be based on selective, unrepresentative,
and non-systematic changes in physical properties of cover materials.
In accordance with well-documented NRC guidance, the probabilistic approach models
future conditions by projecting current knowledge/conditions as reasonably as possible
while capturing uncertainty in the parameters or assumptions of the model. This is
distinctly differentiated from “conservative” (i.e., supposedly biased towards safety)
modeling that is occasionally seen, typically using point values for parameters (implying
a great deal of confidence; i.e., no uncertainty, or conditioning). This type of conservative
modeling is often termed “deterministic” modeling, and has often been used to support
compliance decisions. However, supposed conservatism in parameter estimates (or
distributions) is often difficult to judge in fully coupled models in which all transport
processes are contained in the same overall PA model. More importantly perhaps, actual
conservative dose results from PA models do not support the full capability of a disposal
facility, which leads to sub-optimal decisions for disposal of legacy waste and for the
nuclear industries that need a disposal option. Conservative, deterministic models may
have utility at a “screening” level, and they are often useful during probabilistic model
building, but they do not provide the full range of information that is necessary for
important decisions such as compliance or rule-making (cf., Bogen (1994), Cullen
(1994)).
Analysis of non-representative, arbitrarily selected one-off cases that are based on
unrealistic conditions easily lead to misinterpretation of the performance of the disposal
system.
ET Cover Design Responses for the Clive DU PA Model
23 Feb 2018 113
What is proposed by the Division is not a sensitivity analysis. Rather, the Division
proposes an analysis of separate implausible combinations of input parameter values.
Some details are provided below:
Saturated Hydraulic Conductivity (Ks)—Surface Layer
The surface layer in the ET cover functions as a store and release layer. Proposed values
for this layer are 86.4 cm/day for a low value, 864 cm/day for a typical value, and
864 cm/day for a high value. The typical and high values proposed exceed the values for
the Ks of sand provided in both the Rosetta (USDA 2017) and the Carsel and Parrish
(1988) databases (712.8 cm/day and 643 cm/day respectively) and are not appropriate for
characterizing a silty clay.
These values are also inconsistent with the measurements provided by Benson et al.
(2011) for store and release covers. Table 6.6 of Benson et al. (2011) contains geometric
mean values of measurements of in-service Ks for store and release covers at 10 sites.
The geometric mean values of Ks ranged from 0.65 cm/day to 45.79 cm/day with a
geometric mean of all sites of 7.5 cm/day. The proposed low value is an order of
magnitude larger and the typical and high values are more than two orders of magnitude
larger than the mean of the measured values.
The National Resources Conservation Service (NRCS) Web Soil Survey (WSS)
(http://websoilsurvey.sc.egov.usda.gov/App/HomePage.htm) provides online access to
Soil Reports containing soil property data. The most extensive surface soil type at the
Clive Site corresponding to Unit 4 is classified as the Skumpah by NRCS. NRCS assigns
Ks values for the upper 5 feet of the Skumpah ranging from 3.6 cm/day to 35 cm/day.
These Ks values represent natural conditions.
Again, these values are orders of magnitude lower than the proposed values.
α Values—Surface Layer
The α values recommended for the low, typical, and high cases for the Surface Layer are
0.3 1/kPa (0.03 1/cm). These values are too large when compared to the values of
0.00295 1/cm and 0.0012 1/cm measured by Bingham Environmental (1991) on two
cores taken from Unit 4 at the site.
All Hydraulic Model Parameters —Frost Protection Layer
All hydraulic parameter values for the Frost Protection Layer are set to the identical
values recommended for the radon barriers. These two materials are quite different, and
treating them as identical is unrealistic; even after naturalization the Frost Protection
Layer will not reach the conditions of the current radon barrier. This would artificially
induce more flow through the frost layer, but would not represent the naturalized system.
ET Cover Design Responses for the Clive DU PA Model
23 Feb 2018 114
Added Gravel
For the HYDRUS simulations a mean value of 0.48 for the porosity of the Unit 4 silty
clay used for the Surface Layer was obtained from the Rosetta database (USDA 2017).
The effect of the addition of 15 percent gravel to the Surface Layer on porosity was
calculated using ideal packing equations (Koltermann and Gorelick 1995), giving a value
of 0.41. If adding gravel and naturalizing the layer have compensating effects, then the
saturated water content should have remained 0.48. Their recommended value is 0.4,
nearly identical to what we used.
The influence of change in soil structure on saturated hydraulic conductivity of the radon
barriers was included in the Clive DU PA Model by sampling from a distribution of
saturated hydraulic conductivity developed from measurements of barrier layers in
service covers (Benson et al. 2011).
Warm-up Simulations
UDEQ included requests for “warm up” simulations. Specifically, the request was “For
each case above, run ‘warm up’ simulation 5 times back to back beforehand using
meterological year having annual precipitation closest to long-term average. Use heads
from end of this 5 yr simulation as initial conditions for the performance simulation.”
The 20 HYDRUS-1D simulations were conducted with, essentially, a 900-year warm-up
period, which is a considerably longer time period than the 5 average years requested by
the Division. Neptune used a 100-year synthetic record that was repeated 10 times to
generate a 1,000-year record of atmospheric boundary conditions. All 20 simulations
were run for 1,000 years but only the time series of average water content and infiltration
for the last 100 years were used as results. This was done because the initial conditions
for all simulations were set to a water potential of -200 cm, which is wetter than steady-
state conditions. The long simulation time allowed for equilibration to steady-state. So,
essentially, there is a 900-year warm up period. Figure 27 shows the time series of
infiltration through the ET cover and into the waste zone. It is apparent that, even after
900 years, the line is not quite flat, indicating that our infiltration estimates are slightly
over-estimated.
ET Cover Design Responses for the Clive DU PA Model
23 Feb 2018 115

Figure 27. Time series of infiltration into the waste zone for one of the 20 HYDRUS-
1D simulations.
Nevertheless, the nine HYDRUS-1D simulations requested by the Division were run and
results showing the range from minimum to maximum infiltration (into the waste zone),
along with the results from the original 20 HYDRUS-1D simulations, are shown in
Figure 27. Despite the implementation of the high Ks values requested by the Division,
infiltration in the new nine simulations is generally lower than for the original 20
HYDRUS-1D simulations. This is largely due to setting residual water content to zero,
which effectively increases the water holding capacity of each soil layer.
Overall, the Clive DU PA Model provides a reasonable range for the input parameters for
the hydraulic properties given the currently available data and information, and the
HYDRUS runs for the nine additional combinations of single values for inputs adds no
further insight.

Additional issues raised in Supplemental Interrogatory Comment B.11 are: observed insensitivity
of net infiltration to the value of the saturated hydraulic conductivity, a capillary barrier formed
by the upper cover layers, criticism of the NRCS soil properties data base, and criticism of
sampling and testing methods used by the CSU Porous Media laboratory.
ET Cover Design Responses for the Clive DU PA Model
23 Feb 2018 116
Sensitivity of Net Infiltration to Saturated Hydraulic Conductivity
See the response to Interrogatory 176/1.
Capillary Barrier Effects
See the response to UDEQ Comment B.3.
Antiquated Techniques for Sampling and Analysis of Core
See the response to Interrogatory 21/2.
3.0 Conclusion
Many issues relating to ET cover design and modeling are interwoven throughout the UDEQ
interrogatories discussed in this document. The following summarizes critical aspects of these
responses:
Geologic and Depositional Setting
• Observations of eolian (wind carried) silt deposited in the last 10,000 years across the
Clive Site support a conceptual model of long-term eolian deposition on a stable surface
that promotes and preserves concurrent eolian deposits that are only slightly modified by
slow processes of soil formation. Based on these observations, modeling substantial
cover material hydraulic property changes within 5 to 10 years after construction as
presented in Benson et al. (2011) and in Appendix E of SC&A (2015a) is not appropriate
for the Clive Site.
• UDEQ attributes changes in hydraulic properties of a cover material to volume changes
of the material. Smectite minerals responsible for shrink/swell behavior in clays
attributed to wetting and drying cycles are absent from the Unit 4 clay soils at the Site.
• Sites listed by UDEQ as being comparable to Clive are the Monticello Mill Tailings
Repository south of the town of Monticello, Utah (290 miles from Clive), the Blue Water
disposal site near Grants, New Mexico (485 miles from Clive), the Cheney disposal site
near Grand Junction Colorado (270 miles from Clive), and the Apple Valley Alternative
Cover Assessment Program (ACAP) site referenced by Benson et al. (2011). These sites
are not comparable for a number of reasons including different cover types, different
soils, and different ecology.
Disruptive Processes
• Based on data collection and analysis and the proposed cover design, SWCA evaluated
the potential for disturbance by plant roots, mammals, and ants to result in increased
infiltration (SWCA 2013) and determined that soil disturbance and increased infiltration
due to biotic activity would be minute. Their evaluation does not support UDEQ’s
assertion that biointrusion of plants and animals will substantially increase infiltration.
ET Cover Design Responses for the Clive DU PA Model
23 Feb 2018 117
• SWCA (2013) conclude from their observations at the Clive Site and ecological analog
sites that the low fertility and alkalinity of the soils and the aridity of the climate limit the
growth of plants at the Site and would inhibit the development of large deep-rooted plants
in the cover. The Clive and Monticello Sites are not comparable with respect to rooting
depths.
• The frost depths calculated as part of this analysis show results that are consistent with
the depths of cover and frost protection proposed in the ET Cover design. Disruption of
the cover due to freeze/thaw is not likely.
Flow Model Input Parameter Distribution Development
• Probability distributions used in a PA model incorporate uncertainty in the mean rather
than variability among values from different point locations and times. The distribution
should represent plausible means for the entire volume of material over the time frame of
the model, rather than variability among measurements taken from individual locations
within a site. Practically, this often translates into using standard errors for the estimated
mean over a larger scale to define distributions, rather than standard deviation of
available small-scale measurements. The relevant question to ask is “Does the
distribution cover all plausible values of the input variable that could be representative of
the large volume of material represented by the PA Model cell over long periods of
time?”
• Parameter values and parameter statistical distributions for the hydraulic properties of the
cover layers were developed based on site-specific data, engineering specifications,
widely used soil hydraulic property databases, and consideration of the function of the
layer.
• Estimates of the uncertainty in the hydraulic function α and n values of Unit 4 materials
used for the surface and evaporative zone layers of the ET cover were obtained by using
α and n values from distributions (mean and standard deviation) for each parameter from
the Rosetta database of hydraulic parameters for the textural class of silty clay.
• The radon barrier saturated hydraulic conductivity distribution was based on 1st, 50th, and
99th target percentiles elicited from Benson et al. (2011) to capture plausible values for
in-service (degraded) barriers that could apply to the entire cover described by the
distribution. To manage computational burden, deterministic values for θr, θs, α, and n
from previous site-specific modeling were used for the radon barriers.
• Distributions were developed using the open-source statistical software package R (R
Core Team 2017), which functions to facilitate fitting of distributions to target quantiles,
or percentiles.
• The net infiltration through the clay liner in the flow model is identical to the value of the
net infiltration at the bottom of the radon barriers. This value is drawn from the net
infiltration distribution which is based on measurements of in-service barriers from
Benson et al (2011). Thus, the modeling approach does take into account clay liner
degradation.
ET Cover Design Responses for the Clive DU PA Model
23 Feb 2018 118
Model Abstraction—Development of Simplified Models for GoldSim
• The goal of “model abstraction” in the context of PA modeling is to build a relatively
simple statistical model to approximate the relationship between the input parameters and
an output of interest obtained from a sophisticated, and usually computationally
intensive, process-based model. Using the flow modeling as an example, the goal of the
statistical effort is to build a statistical model to capture the main structure in the
relationship between input parameters and results produced by HYDRUS (needed as an
input to the GoldSim model). The statistical model is then used to predict HYDRUS-like
outputs within GoldSim.
• Fifty HYDRUS-1D flow simulations were conducted to evaluate the uncertainty in net
infiltration into the waste zone and water content within each ET cover layer as a
function of hydraulic property uncertainty. Model input parameter distributions and
HYDRUS flow model results were fit with multiple linear regression models to develop
simplified models for the GoldSim PA model.
• Comparisons showed consistency between α and n inputs to the HYDRUS model and α
and n generated from the distributions for these parameters used in the GoldSim model.
Similarly, the net infiltration values from the HYDRUS realizations were consistent with
values generated using the regression equation in the GoldSim model.
• A linear regression model was chosen to represent net infiltration for implementation in
the DU PA Model because it was simpler and matched the model form also used for
water content across layers.
• The saturated hydraulic conductivity, Ks, was not included in the regression equation for
net infiltration because it was found not to be a predictor (that is not close to statistical
significance).
• Contrary to the opinion of UDEQ, there is not evidence of a strong linear relationship
between α and Ks that should be incorporated into PA distributions, particularly for the
Unit 4 soil within the silty clay soil textural class found at the Clive Site. These
parameters are demonstrated to be statistically independent, implying they are
theoretically uncorrelated and distributions can be developed separately for each
parameter.
Other Issues
• The hydraulic property values and statistical distributions assigned to the Evaporative
Zone and the Frost Protection Layers are reasonable estimates based on site-specific
information and commonly used soils databases. Given these property assignments, the
Frost Protection Layer behaves hydraulically by enhancing storage in the Evaporative
Zone in a realistic manner, not as described by UDEQ as “an extremely potent artificial,
non-realistic capillary barrier.”
• Steady-state annual averages of net infiltration and water content from the HYDRUS
simulations are the model results used to develop statistical distributions of these
ET Cover Design Responses for the Clive DU PA Model
23 Feb 2018 119
parameters for inputs to the GoldSim model for the Clive DU PA. Daily water balance is
not the appropriate scale to evaluate a performance assessment model.
• Common, accepted methods were used for developing atmospheric boundary conditions
for the HYDRUS flow model.
• Given the much greater capacity of the clay liner to allow water to flow through it in
comparison to the 99th percentile of net infiltration rates, the bathtub effect is not
possible.

4.0 References
Albrecht, B.A., and C.H. Benson, 2001. Effect of Desiccation on Compacted Natural Clays,
Journal of Geotechnical and Geoenvironmental Engineering 127 (1) 67–75
Anderson, M.D., 2004. Sarcobatus vermiculatus. In: Fire Effects Information System [Online],
United States Department of Agriculture (USDA), Forest Service, Rocky Mountain
Research Station, Fire Sciences Laboratory, accessed 2/11/2017, available from
https://www.feis-crs.org/feis/
Belnap, J., et al., 2001. Biological Soil Crusts: Ecology and Management, Technical Reference
1730-2, United States Department of the Interior, Bureau of Land Management, Denver
CO, 2001
Benson, C.H., et al., 2011. Engineered Covers for Waste Containment: Changes in Engineering
Properties and Implications for Long-Term Performance Assessment, NUREG/CR-7028,
United States Nuclear Regulatory Commission, Washington DC, December 2011
Benson, C.H., and T. Gurdal, 2013. Hydrologic Properties of Final Cover Soils. In Foundation
Engineering in the Face of Uncertainty, Honoring Fred H. Kulhawy, Geotechnical
Special Publication No. 229, edited by J.L. Withiam, et al., pp. 283–297, American
Society of Civil Engineers (ASCE), Reston VA
Bingham Environmental, 1991. Hydrogeologic Report, Envirocare Waste Disposal Facility,
South Clive, Utah, prepared for Envirocare of Utah, Bingham Environmental, Salt Lake
City UT, October 1991
Bingham Environmental, 1994. Hydrogeologic Report, Mixed Waste Disposal Area, Envirocare
Waste Disposal Facility, South Clive, Utah, prepared for Envirocare of Utah, Bingham
Environmental, Salt Lake City UT, November 1994
Bingham Environmental, 1996. Revised Hydrogeologic Report, prepared for Envirocare of Utah
Inc., Bingham Environmental Inc., Salt Lake City UT, February 1996
ET Cover Design Responses for the Clive DU PA Model
23 Feb 2018 120
Black, B.D., et al., 1999. Geology and Geologic Hazards of Tooele Valley and the West Desert
Hazardous Industry Area, Tooele County, Utah, Special Study 96, Utah Geological
Survey, Utah Department of Natural Resources, Salt Lake City UT, 1999
Bogen, K.T., 1994. A Note on Compounded Conservatism, Risk Analysis 14 (4) 379–381
Bowker, M.A., 2007. Biological Soil Crust Rehabilitation in Theory and Practice: An
Underexploited Opportunity, Restoration Ecology 15 (1) 13–23
Carsel, R.F., and R.S. Parrish, 1988. Developing Joint Probability Distributions of Soil Water
Retention Characteristics, Water Resources Research 24 (5) 755–769
Cressie, N., and C.K. Wikle, 2011. Statistics for Spatio-temporal Data, Wiley Series in
Probability and Statistics, John Wiley & Sons Inc., Hoboken NJ
Crowe, B., et al., 2017. Representation of Global Climate Change in Performance Assessment
Models for Disposal of Radioactive Waste—17183, proceedings of the WM2017
Conference, March 5–9, Phoenix AZ, 2017
Cullen, A.C., 1994. Measures of Compounding Conservatism in Probabilistic Risk Assessment,
Risk Analysis 14 (4) 389–393
Departments of the Army and the Air Force, 1988. Arctic and Subarctic Construction
Calculation Methods for Determination of Depths of Freeze and Thaw in Soils, Technical
Manual No. 5-852-6, Air Force Regulation AFR 88-19, Volume 6, Departments of the
Army and the Air Force, Washington DC, January 1988
Diodato, D.M., 2000. Computer Spotlights in Ground Water (38(1) 10–11), PC-Progress,
available from https://www.pc-progress.com/en/Default.aspx?h3d-reviews-2000
DOE, 2017. UMTRCA Title II, Bluewater, New Mexico, Disposal Site Fact Sheet, United States
Department of Energy, Office of Legacy Management, Grand Junction CO, 2017
DWR, 2014. Water Right Records and Well Log Database, Utah Division of Water Rights,
accessed 3/18/2014, available from http://waterrights.utah.gov/wrinfo/query.asp
EnergySolutions, 2012. Utah Low-Level Radioactive Material License (RML UT2300249)
Updated Site-Specific Performance Assessment, prepared for Utah Division of Radiation
Control, EnergySolutions LLC, Salt Lake City UT, October 2012
EnergySolutions, 2014. Responses to August 11, 2014—Supplemental Interrogatories, Utah
LLRW Disposal License RML UT 2300249 Condition 35 Compliance Report, prepared
ET Cover Design Responses for the Clive DU PA Model
23 Feb 2018 121
for Utah Division of Radiation Control, EnergySolutions LLC, Salt Lake City UT,
August 2014
EnergySolutions, 2015. Utah Radioactive Material License (RML UT2300249) Updated Site-
Specific Performance Assessment, Revision 2, prepared for Utah Division of Radiation
Control, EnergySolutions LLC, Salt Lake City UT, January 2015
Envirocare, 2000. Application for License Amendment for Class B & C Waste, Envirocare of
Utah Inc., North Salt Lake UT, December 2000
Envirocare, 2004. Revised Hydrogeologic Report for the Envirocare Waste Disposal Facility,
Clive, Utah, Version 2.0, Envirocare of Utah Inc., August 2004
EPA, 2002. Simulating Radionuclide Fate and Transport in the Unsaturated Zone: Evaluation
and Sensitivity Analyses of Select Computer Models, EPA/600/R-02/082, United States
Environmental Protection Agency, National Risk Management Research Laboratory,
Cincinnati OH, July 2002
Gates, J.S., and S.A. Kruer, 1981. Hydrologic Reconnaissance of the Southern Great Salt Lake
Desert and Summary of the Hydrology of West-Central Utah, Technical Publication No.
71, prepared by the United States Geological Survey, State of Utah Department of
Natural Resources, Salt Lake City UT, 1981
Groeneveld, D.P., 1989. Shrub Rooting and Water Acquisition on Threatened Shallow
Groundwater Habitats in the Owens Valley, California, Symposium on Cheatgrass
Invasion, Shrub Die-off, and Other Aspects of Shrub Biology and Management, 1989, Las
Vegas NV
GTG, 2014. GoldSim Contaminant Transport Module User's Guide, GoldSim Technology
Group, Issaquah WA, May 2014
Handy, R.L., and M.G. Spangler, 2007. Geotechnical Engineering: Soil and Foundation
Principles and Practice, Fifth Edition, McGraw-Hill Companies, New York NY
Hastie, T., et al., 2009. The Elements of Statistical Learning: Data Mining, Inference, and
Prediction, Second Edition, Springer-Verlag, New York NY
Higgs, M., et al., 2017. Statistical Methods for Effective Spatial and Temporal Scaling in
Support of Probabilistic Performance Assessments—17340, proceedings of the WM2017
Conference, March 5–9, Phoenix AZ, 2017
ET Cover Design Responses for the Clive DU PA Model
23 Feb 2018 122
ITRC, 2003. Technical and Regulatory Guidance for Design, Installation, and Monitoring of
Alternative Final Landfill Covers, The Interstate Technology & Regulatory Council,
Alternative Landfill Technologies Team, Washington DC, December 2003
Jury, W., and R. Horton, 2004. Soil Physics, Sixth Edition, John Wiley & Sons Inc., Hoboken NJ
Koltermann, C.E., and S.M. Gorelick, 1995. Fractional Packing Model for Hydraulic
Conductivity Derived from Sediment Mixtures, Water Resources Research 31 (12) 3283–
3297
Kutner, M.H., et al., 2005. Applied Linear Statistical Models, Fifth Edition, McGraw-Hill/Irwin,
New York NY
Levitt, D.G., 2011. Modeling the Movement of Transient Moisture through Disposal Units at Los
Alamos National Laboratory Technical Area 54, Area G, LA-UR-11-05424, prepared for
United States Department of Energy, Los Alamos National Laboratory, Los Alamos NM,
September 2011
McCray, J.E., 2007. Review of HYDRUS-2D/3D, Southwest Hydrology January/February
(2007) p. 41
Meyer, P.D., et al., 1996. Hydrologic Evaluation Methodology for Estimating Water Movement
Through the Unsaturated Zone at Commercial Low-Level Radioactive Waste Disposal
Sites, NUREG/CR-6346, PNL-10843, prepared for United States Nuclear Regulatory
Commission, Pacific Northwest Laboratory, Richland WA, January 1996
Morris, C.E., and J.C. Stormont, 1999. Parametric Study of Unsaturated Drainage Layers in a
Capillary Barrier, Journal of Geotechnical and Geoenvironmental Engineering 125 (12)
1057–1065
MSI, 2012. January 2011 through December 2011 and January 1993 through December 2011
Summary Report of Meteorological Data Collected at EnergySolutions' Clive, Utah
Facility, prepared for EnergySolutions Inc., Meteorological Solutions Inc., Salt Lake City
UT, February 2012
Mualem, Y., 1976. A New Model for Predicting the Hydraulic Conductivity of Unsaturated
Porous Media, Water Resources Research 12 (3) 513–522
Neitsch, S.L., et al., 2011. Soil and Water Assessment Tool Theoretical Documentation: Version
2009, Texas Water Resources Institute Technical Report No. 406, Texas A&M University
System, accessed 11/2/2017, available from http://swat.tamu.edu/media/99192/swat2009-
theory.pdf
ET Cover Design Responses for the Clive DU PA Model
23 Feb 2018 123
Neptune, 2014. Unsaturated Zone Modeling for the Clive PA, Clive DU PA Model v1.2, NAC-
0015_R1, prepared for EnergySolutions, Neptune and Company Inc., Los Alamos NM,
June 2014
Neptune, 2015a. Saturated Zone Modeling for the Clive DU PA, Clive DU PA Model v1.4, NAC-
0016_R4, Neptune and Company Inc., Los Alamos NM, October 2015
Neptune, 2015b. Neptune Field Studies, December, 2014, Eolian Depositional History Clive
Disposal Site, NAC-0044_R0, Neptune and Company Inc., Los Alamos NM, March 2015
Neptune, 2015c. Unsaturated Zone Modeling for the Clive PA, Clive DU PA Model v1.4, NAC-
0015_R4, prepared for EnergySolutions, Neptune and Company Inc., Los Alamos NM,
October 2015
Neptune, 2015d. Final Report for the Clive DU PA Model, Clive DU PA Model v1.4, NAC-
0024_R4, Neptune and Company Inc., Los Alamos NM, November 2015
Neptune, 2015e. Fitting Probability Distributions, Clive DU PA Model v1.4, NAC-0031_R1,
Neptune and Company Inc., Los Alamos NM, November 2015
Neptune, 2015f. Safety Evaluation Report Response, NAC-0053_R0, prepared for
EnergySolutions, Neptune and Company Inc., Los Alamos NM, November 2015
Neptune, 2015g. Embankment Modeling for the Clive DU PA, Clive DU PA Model v1.4, NAC-
0019_R4, Neptune and Company Inc., Los Alamos NM, October 2015
Neuman, S.P., et al., 2003. A Comprehensive Strategy of Hydrogeologic Modeling and
Uncertainty Analysis for Nuclear Facilities and Sites, NUREG/CR-6805, United States
Nuclear Regulatory Commission, Washington DC, July 2003
NRCS, 2016. Web Soil Survey (WSS), Natural Resources Conservation Service, accessed
2/29/16, available from http://websoilsurvey.sc.egov.usda.gov/App/WebSoilSurvey.aspx
Pavek, D.S., 1992. Halogeton glomeratus. In: Fire Effects Information System, [Online], United
States Department of Agriculture (USDA), Forest Service, Rocky Mountain Research
Station, Fire Sciences Laboratory, accessed 2/11/17, available from https://www.feis-
crs.org/feis/
Peace, J.L., et al., 2004. Vegetation Study in Support of the Design and Optimization of
Vegetative Soil Covers, Sandia National Laboratories, Albuquerque, New Mexico,
SAND2004-6144, Sandia National Laboratories, Albuquerque NM, November 2004
ET Cover Design Responses for the Clive DU PA Model
23 Feb 2018 124
R Core Team, 2015. R: A Language and Environment for Statistical Computing, R Foundation
for Statistical Consulting, available from https://www.r-project.org/
R Core Team, 2017. R: A Language and Environment for Statistical Computing, R Foundation
for Statistical Computing, Vienna, Austria
Ramsey, F.L., and D.W. Schafer, 2013. The Statistical Sleuth: A Course in Methods of Data
Analysis, Third Edition, Brooks/Cole, Cengage Learning, available from
https://www.cengage.com/c/the-statistical-sleuth-a-course-in-methods-of-data-analysis-
3e-ramsey
Rawlings, J.O., et al., 1998. Applied Regression Analysis: A Research Tool, Second Edition,
Springer-Verlag, New York NY
Richardson, C.W., and D.A. Wright, 1984. WGEN: A Model for Generating Daily Weather
Variables, United States Department of Agriculture, Agricultural Research Service,
Washington DC, August 1984
SC&A, 2015a. Utah Division of Radiation Control, EnergySolutions Clive LLRW Disposal
Facility, License No: UT2300249; RML #UT 2300249, Condition 35 Compliance Report;
Appendix A: Final Report for the Clive DU PA Model, Safety Evaluation Report, Volume
2, SC&A, Vienna VA, April 2015
SC&A, 2015b. Utah Division of Radiation Control, EnergySolutions Clive LLRW Disposal
Facility, License No: UT2300249; RML #UT 2300249, Condition 35 Compliance Report;
Appendix A: Final Report for the Clive DU PA Model, Safety Evaluation Report, Volume
1, SC&A Inc., Vienna VA, April 2015
Scanlon, B.R., et al., 2002. Intercode Comparisons for Simulating Water Balance of Surficial
Sediments in Semiarid Regions, Water Resources Research 38 (12) 59-1–59-16 doi:
10.1029/2001WR001233
Scanlon, B.R., et al., 2005. Evaluation of Evapotranspirative Covers for Waste Containment in
Arid and Semiarid Regions in the Southwestern USA, Vadose Zone Journal 4 (2005) 55–
71
Schaap, M.G., et al., 2001. ROSETTA: A Computer Program for Estimating Soil Hydraulic
Parameters with Hierarchical Pedotransfer Functions, Journal of Hydrology 251 (2001)
163–176
Schroeder, P.R., et al., 1994a. The Hydrologic Evaluation of Landfill Performance (HELP)
Model, Engineering Documentation for Version 3, EPA/600/R-94/168b, United States
ET Cover Design Responses for the Clive DU PA Model
23 Feb 2018 125
Environmental Protection Agency, Office of Research and Development, Washington
DC, September 1994
Schroeder, P.R., et al., 1994b. The Hydrologic Evaluation of Landfill Performance (HELP)
Model, User's Guide for Version 3, EPA/600/R-94/168a, United States Environmental
Protection Agency, Office of Research and Development, Washington DC, September
1994
Shabala, S., 2013. Learning from Halophytes: Physiological Basis and Strategies to Improve
Abiotic Stress Tolerance in Crops, Annals of Botany 112 (2013) 1209–1221 doi:
10.1093/aob/mct205
Shmueli, G., 2010. To Explain or to Predict?, Statistical Science 25 (3) 289–310 doi:
10.1214/10-STS330
Simonin, K.A., 2001. Atriplex confertifolia. In: Fire Effects Information System, [Online],
United States Department of Agriculture (USDA), Forest Service, Rocky Mountain
Research Station, Fire Sciences Laboratory, accessed 2/11/2017, available from
https://www.feis-crs.org/feis/
Šimůnek, J., et al., 2013. The HYDRUS-1D Software Package for Simulating the One-
Dimensional Movement of Water, Heat, and Multiple Solutes in Variably-Saturated
Media, Version 4.16, University of California Riverside, Riverside CA, March 2013
SWCA, 2011. Field Sampling of Biotic Turbation of Soils at the Clive Site, Tooele County, Utah,
prepared for EnergySolutions, SWCA Environmental Consultants, Salt Lake City UT,
January 2011
SWCA, 2012. Vegetated Cover System for the EnergySolutions Clive Site: Literature Review,
Evaluation of Existing Data, and Field Studies Summary Report, prepared for
EnergySolutions, SWCA Environmental Consultants, Salt Lake City UT, August 2012
SWCA, 2013. EnergySolutions Updated Performance Assessment—SWCA’s Response to First
Round DRC Interrogatories, SWCA Environmental Consultants, September 2013
USDA, 2017. ROSETTA Class Average Hydraulic Parameters, United States Department of
Agriculture, Agricultural Research Service, available from
https://www.ars.usda.gov/pacific-west-area/riverside-ca/us-salinity-
laboratory/docs/rosetta-class-average-hydraulic-parameters
Utah DEQ, 2017. Division of Waste Management and Radiation Control, EnergySolutions Clive
LLRW Disposal Facility License No: UT2300249; RML #UT 2300249, Amended and
ET Cover Design Responses for the Clive DU PA Model
23 Feb 2018 126
New Interrogatories Related to Clive DU PA Modeling Report Version 1.4 Dated
November 2015, Utah Department of Environmental Quality (DEQ), Salt Lake City UT,
May 2017
van Genuchten, M.T., 1980. A Closed-form Equation for Predicting the Hydraulic Conductivity
of Unsaturated Soils, Soil Sci. Soc. Am. J. 44 (1980) 892–898
van Genuchten, M.T., et al., 1991. The RETC Code for Quantifying the Hydraulic Functions of
Unsaturated Soils, EPA/600/2-91/065, United States Environmental Protection Agency,
Ada OK, December 1991
Wagner, C.H., 1982. Simpson’s Paradox in Real Life, The American Statistician 36 (1) 46–48
Waugh, W.J., et al., 2008. Monitoring the Performance of an Alternative Landfill Cover at the
Monticello, Utah, Uranium Mill Tailings Disposal Site, Waste Management 2008
Conference, Phoenix AZ
Whetstone Associates, 2011. EnergySolutions: Class A West Disposal Cell Infiltration and
Transport Modeling Report, prepared for EnergySolutions, Whetstone Associates Inc.,
Gunnison CO, November 2011

NAC-0108_R0

Erosion Responses for the Clive DU
PA Model
23 February 2018

Prepared by

NEPTUNE AND COMPANY, INC.
1435 Garrison St, Suite 201, Lakewood, CO 80215
Erosion Responses for the Clive DU PA Model
23 Feb 2018 ii
1. Title: Erosion Responses for the Clive DU PA Model
2. Filename: Erosion Responses for the Clive DU PA Model.docx
3. Description: Responses to UDEQ Interrogatories and Safety Evaluation Report Comments
received May 11, 2017

Name Date
4. Originator Mike Sully and Sean McCandless 12 Feb 2018
5. Reviewer Dan Levitt
6. Remarks

Erosion Responses for the Clive DU PA Model
23 Feb 2018 iii
CONTENTS
CONTENTS ............................................................................................................................. iii
FIGURES .................................................................................................................................. iv
TABLES ..................................................................................................................................... v
ACRONYMS AND ABBREVIATIONS ................................................................................... vi
1.0 Overview and Conceptual Model........................................................................................ 7
2.0 UDEQ Interrogatory Responses.......................................................................................... 8
2.1 Interrogatory CR R313-25-8(4)(a)-71/1: Biotic Processes in Gully Formation .............. 8
2.1.1 Interrogatory Response ............................................................................................ 8
2.2 Interrogatory CR R313-25-7(2)-191/3: Effect of Gully Erosion .................................. 13
2.2.1 Interrogatory Response .......................................................................................... 14
2.3 Interrogatory CR R313-25-25(4)-197/1: Properties of Embankment Side Slope
Materials..................................................................................................................... 20
2.3.1 Interrogatory Response .......................................................................................... 21
2.4 Interrogatory CR R313-25-25(4)-198/1: Gravel Content of Embankment Materials .... 21
2.4.1 Interrogatory Response .......................................................................................... 21
2.5 Interrogatory CR R313-25-25(4)-199/1: Uncertainties in Erosion Modeling ............... 25
2.5.1 Interrogatory Response .......................................................................................... 25
2.6 Interrogatory CR R313-25-25(4)-200/1: Use of RHEM to Develop Parameters for
SIBERIA .................................................................................................................... 25
2.6.1 Interrogatory Response .......................................................................................... 26
2.7 Interrogatory CR R313-25-25(4)-201/1: Estimating Rainfall Intensity ........................ 27
2.7.1 Interrogatory Response .......................................................................................... 27
2.8 Interrogatory CR R313-25-25(4)-202/1: Use of SIBERIA to Model Federal Cell
Erosion ....................................................................................................................... 30
2.8.1 Interrogatory Response .......................................................................................... 30
2.9 Interrogatory CR R313-25-25(4)-205/1: Erosion Analysis .......................................... 30
2.9.1 Interrogatory Response .......................................................................................... 31
3.0 Conclusion ....................................................................................................................... 34
4.0 References ........................................................................................................................ 35
Appendix A. HAL 2018 ............................................................................................................ 40
DRC Interrogatory CR R313-25-7(2)-191/3 ................................................................ 41
Appendix B. HAL 2017a........................................................................................................... 47
DRC Interrogatory CR R313-25-25(4)-198/1 .............................................................. 48
Appendix C. HAL 2017b .......................................................................................................... 53
DRC Interrogatory CR R313-25-25(4)-201/1 .............................................................. 54

Erosion Responses for the Clive DU PA Model
23 Feb 2018 iv
FIGURES
Figure 1. Utah-specific Iso-Erodent (R) mapping provided in the Utah Water Research
Laboratory report (Israelsen et al. 1984). .................................................................. 16
Figure 2. Wischmeier nomograph (Haan et al. 1994; Israelsen et al. 1984). ............................... 17
Figure 3. Table 9 and Table 10 of the US Department of Agriculture Handbook Number 537,
“Predicting Rainfall Erosion Losses, A Guide to Conservation Planning” (USDA
1978). ...................................................................................................................... 18
Figure 4. Empirical cumulative distribution functions (ecdf) of the values across realizations
found in the GoldSim look-up table for the seven deepest depth intervals. Note
that, in general, the deeper intervals have larger proportions of realizations with
smaller fractions. The two green curves associated with depth intervals (1.37m,
1.52m] and (1.52m, 2.03m] do not follow the order with depth, which can be
interpreted as consistent with the conclusion from Figure 205-1............................... 32

Erosion Responses for the Clive DU PA Model
23 Feb 2018 v
TABLES
Table 1. Gully erosion potential—permissible velocity analysis for Federal Cell. ...................... 12
Table 2. Effects of erosion—average soil loss analysis using USLE. Soil loss is calculated for
the top slope with Unit 4 clay with 15% gravel admixture and side slope with Unit
4 clay and 50% gravel admixture. ............................................................................ 18
Table 3. Calculated Time of Concentration Using TR-55 Method for Class A West Cell. .......... 28
Table 4. Calculated Time of Concentration Using Tr-55 Method for Federal Cell. .................... 28
Table 5. PMP Rainfall Intensity for the Class A West Cell. ....................................................... 29
Table 6. PMP Rainfall Intensity for the Federal Cell. ................................................................ 29
Table 7. Gully Erosion Potential—Velocity Analysis for Class A West Cell. ............................ 30

Erosion Responses for the Clive DU PA Model
23 Feb 2018 vi
ACRONYMS AND ABBREVIATIONS
CERCLA Comprehensive Environmental Response, Compensation, and Liability Act
CWCB Colorado Water Conservation Board
DEQ (Utah) Department of Environmental Quality
DOE (United States) Department of Energy
DU depleted uranium
DWMRC Division of Waste Management and Radiation Control
ecdf empirical cumulative distribution function
EPA (United States) Environmental Protection Agency
ET evapotranspiration
HAL Hansen, Allen, and Luce
HMR Hydrometeorological Report
LEM landscape evolution model
LLRW low-level radioactive waste
MPV maximum possible velocity
NOAA National Oceanographic and Atmospheric Administration
NRC (United States) Nuclear Regulatory Commission
NRCS Natural Resources Conservation Service
PA performance assessment
PDF probability density function
PMP probable maximum precipitation
PPA probabilistic performance assessment
PSB Prototype Surface Barrier
RCRA Resource Conservation and Recovery Act
RHEM Rangeland Hydrology and Erosion Model
RML Radioactive Material License
RUSLE Revised Universal Soil Loss Equation
sd standard deviation
SER Safety Evaluation Report
SWCA SWCA Environmental Consultants
SWCC soil water characteristic curve
UDEQ Utah Department of Environmental Quality
UMTRCA Uranium Mill Tailings Radiation Control Act
USDA United States Department of Agriculture
USLE Universal Soil Loss Equation

Erosion Responses for the Clive DU PA Model
23 Feb 2018 7
This document addresses Category 2, Erosion of the DU PA Version 1.4 Interrogatories (Utah
DEQ 2017). There are two open interrogatories and seven new interrogatories within this
category. Background information on the statistical approaches applied for the probabilistic PA
model is provided in Section 1.1 of NAC_0108, ET Cover Design. Section 1.2 of that document
also provides a summary of the cell design, climate, and vegetation.
1.0 Overview and Conceptual Model
There are three water erosion processes considered for the Site: sheet erosion, rill erosion, and
gully erosion (channel formation). Sheet erosion is the detachment of soil particles by water
flowing overland as a sheet. During rainfall events when rainfall exceeds infiltration, runoff can
occur, acting to erode cover materials. Sheet erosion is a process that extends over the entire area
of the cover. If flow concentrates on the surface, shallow channels called rills can also be
formed. Gully erosion is a process that occurs when flowing water forms larger narrow channels.
Gully erosion typically results in a gully that has an approximate “V” cross section that widens
and deepens with time until the gully stabilizes. The rate of erosion loss from all of these
processes depends on the steepness and length of the slope, soil texture, vegetation, and cover
surface characteristics, as well as rainfall intensity.
The stability of the cover with respect to erosion is evaluated for the Clive DU PA using several
modeling approaches. The Universal Soil Loss Equation (USLE) (USDA 1978) is used for
estimating losses due to sheet and rill erosion, and the permissible velocity method described in
NUREG-1623 (NRC 2002) is used for evaluating the potential for gully erosion. The USLE
provides estimates of long-term annual soil loss produced by rain drop impact and runoff. This
method is recommended by the EPA for evaluation of cover stability (EPA 1991) and is a
standard used across the United States. The NRC has recommended several methods for
evaluating the potential for gully erosion on soil covers for waste embankments (NRC 2002).
Modeling of sheet and rill erosion losses from the cover was done with USLE. The USLE uses a
conceptual model to evaluate stability that is described by Israelsen et al. (1984) as “if adequate
protection is provided to control sheet erosion, then rills and gullies will never form from
rainfall.” The USLE procedure is based on the Universal Soil Loss Equation—a combination of
parameters representing rainfall intensity, soil erodibility, slope length, slope steepness, cover
and management factors, and conservation practices. Values for the parameters are obtained
from tables and charts (or the computer program) applicable to the location, site-specific
materials, and cover design.
The annual erosion loss rates estimated by USLE for the cover were corroborated using the
Rangeland Hydrology Erosion Model (RHEM) (Nearing et al. 2011). RHEM is a process-based
runoff and erosion model developed specifically for rangeland applications.
The evaluation of design elements for long-term cover stability by the NRC (NRC 2002) is based
on the conceptual model that gully erosion differs from other design considerations in that the
development of gullies and the process of erosion are cumulative and can progress with
subsequent storm events. The method adopted by the NRC to provide long-term stability is to
design stable slopes that prevent the initiation of gullies by a single, very large storm event. By
Erosion Responses for the Clive DU PA Model
23 Feb 2018 8
designing to an unusually large single event, smaller more frequent events will have no
significant cumulative impact on stability.
NUREG-1623 (NRC 2002) details methods, guidelines, and procedures for designing erosion
protection for earthen covers at uranium mill tailings sites that apply to the Federal Cell at the
Clive Site. The stability of a design is evaluated by comparing the velocity of flow from an
extreme precipitation event on the embankment slopes to the maximum flow a channel can carry
without causing gullies. Site-specific information used to calculate the velocity on the
embankment includes the probable maximum precipitation intensity, catchment area, slope, and
a roughness coefficient.
Modeling of erosion in a borrow pit at the Site provided a line of evidence in addition to the
erosion modeling using USLE and the NRC gully projection method (NRC 2002) that
demonstrates the stability of the proposed ET cover design. Although the borrow pit is an
approximation to the embankment, this analysis using the landscape evolution model SIBERIA
provides insight into erosion projected out to 10,000 years.
2.0 UDEQ Interrogatory Responses
This section contains responses for Open Interrogatories CR R313-25-8(4)(a)-71/1 and CR
R313-25-7(2)-191/3, as well as New Interrogatories CR R313-25-25(4)-197/1, CR R313-25-
25(4)-198/1, CR R313-25-25(4)-199/1, CR R313-25-25(4)-200/1, CR R313-25-25(4)-201/1, CR
R313-25-25(4)-202/1, and CR R313-25-25(4)-205/1.
2.1 Interrogatory CR R313-25-8(4)(a)-71/1: Biotic Processes in Gully
Formation
DEQ Conclusion from April 2015 SER, Appendix C:
…the DU PA SER, Section 4.4.2 concluded the following:
Before the DU PA can be determined to be adequate, EnergySolutions needs to clarify
certain issues relating to Appendix 10 to the DU PA Model v1.2 (…)... DRC is currently
reviewing a license amendment request5 to use an ET cover of similar design to that
proposed for the Federal Cell in the DU PA. Any recommendations and conclusions
from that review must be applied to the proposed Federal Cell as well.
Therefore, this interrogatory remains open.
DEQ Critique of DU PA v1.4, Appendix 5, Appendix 21: No further analysis has been performed
pertaining to biotic processes in gully formation since v1.2.
2.1.1 Interrogatory Response
This interrogatory points to soil erosion issues raised in Section 4.4.2 of the Safety Evaluation
Report (SC&A 2015a). UDEQ expresses concerns that gullies will form and enhance radon
diffusion, deep infiltration, and contaminant transport. EnergySolutions plans both ecological
and engineering measures to address these concerns by minimizing gully formation.
Erosion Responses for the Clive DU PA Model
23 Feb 2018 9
Note that this interrogatory also attempts to link two distinct cell design reviews. Since issuance
of the 2015 SER, EnergySolutions has begun construction of rock armor cover on the CAW cell.
More importantly, the DU PA models the Federal Cell’s ET cover as designed. CAW cover
licensing is separate and distinct from the Federal Cell. Since their geometry differs, any attempt
to couple or join the reviews is unsupported and arbitrary. Therefore, this aspect of the question
is moot.
Ecological Measures
SWCA Environmental Consultants (SWCA) assessed erosion under undisturbed conditions at
Clive in June 2012 (SWCA 2012). Eight study plots and one soil borrow area were identified and
the location of wind and water erosion features were mapped. Each study plot was 0.25 ac
(1000 m2) with three plots located on hills. All plots were within the elevational range of the
Clive Site.
SWCA observed that the effects of wind and water erosion were limited. There was minimal
evidence found of water erosion even on the sloped study plots. This is consistent with the
Hansen, Allen, & Luce (HAL) calculations of projected minimal sheet and rill erosion loss
described in the response to UDEQ Interrogatory 191/3. The site-specific SWCA studies
indicated that a cover would remain stable with the established native vegetation and biological
soil crust mimicking undisturbed conditions.
SWCA has developed a sequence of reclamation measures for quickly re-establishing natural
conditions for the cover that will minimize erosion (SWCA 2013). During the first two years the
cover will be stabilized by seeding with fast-growing species. Establishment of these species will
also exclude invasive annuals. The planned seed mix includes native perennial shrubs, forbs, and
bunchgrasses that will become established during years three through five. The soil surface will
be treated with local biological soil crust to speed up crust recovery during these years.
Biological soil crusts are known to reduce wind and water erosion (Belnap and Gillette 1998;
Belnap et al. 2001). The upper Unit 4 soil layer of the cover will be mixed with 15 percent gravel
and the side slopes with 50 percent gravel as additional measures to prevent erosion.
Observations at a number of other sites have shown that a gravel admixture reduces erosion and
promotes evapotranspiration (SWCA 2013). SWCA (2013) concludes:
The proposed ET cover designs and associated installation procedures comply with UAC
R313-25-8(4)(a) and (b); UAC R313-25-18; UAC R313-25-19; UAC R313-25-20; UAC
R313-25-8(4)(d); and UACR313-25-22 by establishing a stable and functioning system
comprised of native vegetation and soil biota that minimizes any near-term episodic
erosional exposure of contaminated materials. A functioning native ecosystem also provides
long-term soil stabilization via soil development, plant roots in upper soil layers, and
biological soil crusts.
Engineering Measures
Calculations to evaluate the stability of the cover design with respect to rill and gully erosion for
the Class A West cell were provided in Appendix D of EnergySolutions (2015). Similar
calculations for gully erosion potential for the Federal Cell are presented here.
Erosion Responses for the Clive DU PA Model
23 Feb 2018 10
Methods for evaluating the design basis for erosion of soil covers for waste embankments are
given in NUREG-1623 (NRC 2002). This report details methods, guidelines, and procedures for
designing erosion protection for earthen covers at uranium mill tailings sites that apply to the
Federal Cell at the Clive Site.
The evaluation of design elements for long-term cover stability is based on the view taken by
NRC (2002) that gully erosion differs from other design considerations in that the development
of gullies and the process of erosion are cumulative and can progress with subsequent storm
events. The method adopted by the NRC to provide long-term stability is to design stable slopes
that prevent the initiation of gullies by a single, very large storm event. By designing to an
unusually large single event, smaller more frequent events will have no significant cumulative
impact on stability.
The potential for gully erosion on the Federal Cell was evaluated using the permissible velocity
method described in NUREG-1623 (NRC 2002).
Slope Description
The Federal Cell is designed as a covered embankment with relatively steeper sloping sides
nearer the edges. The upper part of the embankment referred to as the top slope has a slope of 2.4
percent, while the side slope is steeper with a slope of 20 percent. The length of the side slope is
175 ft. Top slopes of the embankment have different lengths. The longest top slope length is
521 ft.
Flow Concentration
The peak flow unit discharge, Q (cubic feet per second per foot [cfs/ft]), is calculated using the
Rational Formula (NRC 2002):
!=#× %× &× '
where
F is flow concentration factor,
c is dimensionless runoff coefficient,
i is rainfall intensity (inches/hour [in/hr]), and
A is catchment area (acres).
A default value of 3 is recommended in NRC (2002) for the flow concentration factor, F. A
value for the runoff coefficient of 0.5 is recommended for a graveled surface in Table 4.6 of
NUREG 4620 (NRC 1986). The rainfall intensity used for the projection is 18.4 inches for the
top slope and 19.8 inches for the side slope at the Clive Site. See the response to UDEQ
Interrogatory 201 for a description of the calculation of these values. The catchment area is the
area of a 1-ft wide strip along the length of the slope. Using these values, the peak flow unit
discharge, Q, for the top slope and side slope are found to be 0.328 and 0.122 cfs/ft respectively.
Erosion Responses for the Clive DU PA Model
23 Feb 2018 11
Flow Depth
The flow depth, y, is then calculated using the Manning equation for normal depth on a one-foot-
wide strip of the slope. This equation is given by NRC (2002) as
()*+=!,
-1.486 45 6+78
where
y is flow depth (ft),
n is Manning n, and
S is slope (ft/ft).
A value of 0.05 is used for the Manning’s n based on the calculation method of Bray for natural
channels described in Coon (1998).
Using the previously calculated values for Q and the Manning’s n, flow depths were calculated
to be 0.205 ft for the top slope and 0.060 ft for the side slope.
Maximum Permissible Velocity
A value of 5.0 ft/s was chosen as the maximum permissible velocity (MPV) based on the
characteristics of the channel. This is the value listed for gravel in Table CH13-T103 of Colorado
Water Conservation Board (CWCB 2006) and in Table 4.7 of Nelson et al. (1986). The NRC
(2002) method requires that the MPV be adjusted to account for the flow depth. Correction
factors developed by Chow are provided in Appendix A of NRC (2002). The correction factor
for flows less than 0.25 ft in depth is 0.5. The adjusted MPV values for both the top slope and the
side slope are adjusted to 2.5 ft/s.
Actual Flow Velocity
The actual flow velocity is determined by dividing the discharge by the flow depth:
9:=!(+
Using this equation, the top slope and side slope velocities are 1.60 ft/s and 2.03 ft/s.
These velocities for the top slope and side slope do not exceed the adjusted MPV, so the design
is acceptable (see Table 1). Both slope scenarios using the ET cover system provide acceptable
protection against gully erosion using these criteria by limiting the potential of gully formation
from high velocity channelization.
Erosion Responses for the Clive DU PA Model
23 Feb 2018 12
Table 1. Gully erosion potential—permissible velocity analysis for Federal Cell.
Slope
Description
Lengt
h (ft)
Slop
e
(ft/ft)
i
(in/hr
)
c Q
(cfs/ft
)
y
(ft)
Va
(ft/s
)
MP
V
(ft/s
)
Reductio
n Factor
Adjuste
d MPV
(ft/s)
Top Slope 521 0.02
4
18.3 0.
5
0.328 0.20
5
1.60 5.0 0.5 2.5
Side Slope 178.5 0.20 19.8 0.
5
0.122 0.06
0
2.03 5.0 0.5 2.5

Borrow Pit Model
Modeling of gully formation in a borrow pit at the Site provides another line of evidence in
addition to the sheet erosion modeling using USLE and the NRC gully projection modeling using
the permissible velocity method (NRC 2002) that demonstrate the stability of the proposed ET
cover design. (See the response to UDEQ Interrogatory 191/3.) Although the borrow pit is an
approximation to the embankment, the intent of this analysis is to provide insight into erosion
projected out to 10,000 years.
The SER (SC&A 2015a) states in Section 4.4.2 that results from the borrow pit modeling
described in Appendix 10 of Neptune (2014) indicate that at 10,000 years about 1 percent of the
surface will have gullies greater than or equal to 1 meter and suggest that, if this is the correct
interpretation of the results, the influence of gully formation on radon release and infiltration
should be minimal. Appendix 10 of the DU PA Model v1.2 presented a sampling of the results of
the 1,000 simulations, while the results from all 1,000 simulations were provided in Appendix 10
of the DU PA Model v1.4 (Neptune 2015a).
See the response to UDEQ Interrogatory 199 regarding issues raised in Section 4.4.2 of the SER
on the comparability of geometry and materials of the borrow pit and the embankment.
Section 4.4.2 of the SER noted that the description of how the results of the borrow pit erosion
modeling were used in the DU PA Goldsim Model needed clarification. A detailed description of
the analysis of erosion modeling results and their implementation in the GoldSim model was
included in Appendix 10 of the DU PA Model v1.4 (Neptune 2015b).
Lack of evidence of significant erosion activity at the Site, planned ecological remediation of the
soils used to construct the cover, and erosion models that demonstrate the stability of the cover
support the prediction that erosion of the embankment will be minimal and will not enhance
radon fluxes or deep infiltration.
Erosion Responses for the Clive DU PA Model
23 Feb 2018 13
2.2 Interrogatory CR R313-25-7(2)-191/3: Effect of Gully Erosion
DEQ Conclusion from April 2015 SER, Appendix C:
In conclusion, the analysis performed by HAL may or may not be correct, but before DEQ can
accept it, each value selected and used in the analysis needs to be justified.
EnergySolutions/HAL also needs to address how the embankment will be re-vegetated, how
much re-vegetation is necessary and how much is expected, and how long is it expected to take.
Therefore, this interrogatory remains open.
DEQ Critique of DU PA v1.4 and Appendix 21:
Gravel Surface
Gravel embedded in the upper layer may migrate upward over time due to environmental effects,
such as freeze/thaw or wet/dry cycling phenomena, bringing some particles to the surface. At the
same time, eolian erosion and deposition is likely to occur on the cover, potentially “silting in”
gravel particles that move to the surface. Formation of a “gravel mulch” layer (i.e., a clean coarse
layer of gravel at the surface) that would impede evaporation is unlikely. A more likely
phenomenon is formation of a desert pavement, with finer sands, silts, and clay particles
embedding around gravel particles. These finer materials provide a capillary conduit for
evaporation.
This phenomenon is observed at sites where riprap or cobbles are used as cover. Fines deposit
in the pores between the large particles, gradually accumulating and filling the pores. These fines
serve as a seed bed and as a capillary conduit, allowing water to flow upward. This was very
clear in the armored surfaces at both Uranium Mill Tailings Radiation Control Act (UMTRCA)
covers studied in 2016 for the NRC and DOE-LM. An example from the Uranium Mill Tailings
Reclamation site in Grants, New Mexico, is shown in Figure 191-1 below. Roots and structure in
the radon barrier are being mapped as shown in Figure 191-1(a) and brush growing in an
adjacent area of the riprap surface layer is shown in Figure 191-1(b).
The best approach to understand this issue, and to develop a suitable conceptual model for
Clive, is to seek out analogs in the area where undisturbed fluvial surficial soils exist with
appreciable gravel. Studying the surface of these soils will provide evidence regarding the long-
term surface characteristics that can be anticipated at Clive.
Gravel Fraction to Address Erosion
The appropriate gravel fraction necessary to prevent erosion has not been defined with
precision, nor has a validated methodology been developed to determine the appropriate gravel
fraction as a function of site-specific conditions. Models have been developed, but they have not
been validated in the field. For example, Smith and Benson (2016) used the model SIBERIA to
evaluate erosion from a top deck with a gravel amendment, but the model was not validated in
the field.
The gravel admixtures used at Hanford and Monticello have been effective in controlling erosion.
No major erosion issues have been encountered at either site on the shallow top decks. Riprap
is used on the steeper side slopes on both sites. There have been no quantitative field studies to
evaluate the reduction in erosion achieved with the gravel admixture on the top deck at either
site.
Erosion Responses for the Clive DU PA Model
23 Feb 2018 14
Gravel Fraction to Control or Prevent Biointrusion
There should be no expectation that 15% gravel, or even 50% gravel, will preclude biointrusion.
As noted previously, plants readily germinate and root in riprap layers when silt accumulates in
the pores (Figure 191-1). Vegetation is likely to be more robust in a gravel-amended surface
layer with smaller particles and more fine-textured particles.
Burrowing animals will readily penetrate a layer containing gravel particles, and plants will readily
grow in a fine-textured layer with as much as 50% gravel. Biointrusion design to prevent
burrowing requires particles larger than the breadth of the animal (precludes particles from being
moved through a burrow), and a gradation that results in pore sizes smaller than the breadth of
the animal (prevents burrowing between particles).
Homogenization
Pedogenic phenomena are known to create structure and alter the hydraulic properties of
earthen cover materials. There is no evidence in the literature that layering in covers diminishes
with time or that a homogeneous profile develops. For example, distinct layering has been
observed in recent excavations into UMTRCA covers that are 20 years old (Figure 191-2).
Structure has developed in these layers, and the hydraulic properties have changed, but the
profile is not homogeneous. A model for Clive should include a layered profile with appropriate
hydraulic properties assigned to each layer that reflect realistic development of structure.
2.2.1 Interrogatory Response
UDEQ raises a number of issues in this interrogatory, beginning with design for slope stability.
Erosion Analyses
The first request from UDEQ is to provide the rationale for selection of the Revised Universal
Soil Loss Equation (USLE) (USDA 1978) and the Rangeland Hydrology and Erosion Model
(RHEM) (Nearing et al. 2011) parameters used for erosion modeling. A detailed response to this
request was developed by Hansen, Allen, and Luce (HAL 2018). A summary of their final report
is provided here. The complete report is provided in Appendix A.
Universal Soil Loss Equation (USLE) Parameters
The USLE estimates average annual soil losses from erosion. While the USLE equations are the
same as the Revised Universal Soil Loss Equations (RUSLE), the use of the acronym RUSLE
could imply the use of the software program developed by the USDA for factor determination. In
order to clarify the methodology used in the analysis, all references are to USLE. The software
package was not used because factors from a Utah-specific publication were available for the
analysis as described below.
The USLE methodology is commonly used to determine the long-term stability of slopes and is
an industry-standard method for design of erosion control. Guidance given by the EPA states that
“The U.S. Department of Agriculture’s (USDA’s) Universal Soil Loss Equation is recommended
as the tool to evaluate erosion potential” (EPA 1991). The basis for this approach is that “if
adequate protection is provided to control sheet erosion, then rills and gullies will never form
from rainfall” (Israelsen et al. 1984). The USLE equation is defined as:
Erosion Responses for the Clive DU PA Model
23 Feb 2018 15
A = R*K*LS*C*P
where
A is the average soil loss per unit area, expressed in tons/acre/year,
R is the rainfall/runoff factor, which is the number of rainfall units for rainfall energy
and runoff and snowmelt,
K is the soil erodibility factor in tons per acre per year per unit of R,
LS is the topographic factor (length and steepness of the slope),
C is the cover and management factor, which is the ratio of soil loss from an area
with a given cover and management relative to that from an identical area in
continuous fallow, and
P is the supporting conservation practice factor, in this case assumed to be equal to 1.
This procedure and site-specific factors are described in “Erosion and Sedimentation in Utah—A
Guide for Control” (Israelsen et al. 1984) and “Design Hydrology and Sedimentation for Small
Catchments” (Haan et al. 1994).
R Factor
The R factor (6) is selected based on the Utah-specific Iso-Erodent (R) mapping provided in the
Utah Water Research Laboratory report (Israelsen et al. 1984) which is provided in Figure 1.
K Soil Erodibility Factor
The K values (0.18 and 0.07 for top slope and side slope, respectively) are based on the Unit
4-specific material characteristics with the top slope gravel admixture and characteristics for the
side slope gravel admixture, together with the Wischmeier nomograph as described in the
methodology presented in the Israelsen and Hann procedures and shown in Figure 2. A particle
size analysis using the hydrometer method was performed on the Unit 4 clay to determine the
percent silt and very fine sand (see Figure 2) for use with the Wischmeier nomograph.
Erosion Responses for the Clive DU PA Model
23 Feb 2018 16

Figure 1. Utah-specific Iso-Erodent (R) mapping provided in the Utah Water Research
Laboratory report (Israelsen et al. 1984).
Erosion Responses for the Clive DU PA Model
23 Feb 2018 17

Figure 2. Wischmeier nomograph (Haan et al. 1994; Israelsen et al. 1984).

C Factor (VM Factor)
The C parameters used in the equation for both the 4% and 20% slopes were based on Table 9
and Table 10 of the US Department of Agriculture Handbook Number 537, “Predicting Rainfall
Erosion Losses, A Guide to Conservation Planning” (USDA 1978), and are shown in Figure 3.
The C factor for the top slopes (0.2) is based on the sparse vegetative cover naturally found in
the areas immediately surrounding the Clive facility (Table 10 of Handbook Number 537, No
Appreciable Canopy, Type G with 20% ground cover) and the Unit 4 gravel admixture. The C
factor for the side slope is based on the higher percentage of gravel in the Unit 4 gravel
admixture (50% gravel). The 50% gravel admixture on the side slopes results in a pseudo-gravel
mulch once some of the fines have been removed. Therefore, a C factor of 0.02 was selected
(Table 9 Handbook Number 537, Crushed Stone ¼ to 1½ in, Mulch Rate of 240 tons/acre, Land
Slope of 21–33%, with a length limit of 200 feet).
Results
Soil loss calculated using these parameters in the USLE equation for the top slope and side slope
are shown in Table 2.
Erosion Responses for the Clive DU PA Model
23 Feb 2018 18

Figure 3. Table 9 and Table 10 of the US Department of Agriculture Handbook Number
537, “Predicting Rainfall Erosion Losses, A Guide to Conservation Planning” (USDA
1978).

Table 2. Effects of erosion—average soil loss analysis using USLE. Soil loss is calculated for
the top slope with Unit 4 clay with 15% gravel admixture and side slope with Unit 4
clay and 50% gravel admixture.
Slope
Segment
R
(ft tons/
ac/hr)
K
(tons/ac/EI)
L
(ft)
S
(%)
C A
(tons/ac/yr)
Total Soil
Loss
(mm/year)
Top Slope
(4%)
6 0.18 942 (4%) 4% 0.2 0.25 0.24
overall
0.026
Side Slope
(20%)
6 0.07 188 (20%) 20% 0.02 0.19

Erosion Responses for the Clive DU PA Model
23 Feb 2018 19
The Rangeland Hydrology and Erosion Model (RHEM) Model Parameters
The RHEM model was used in HAL’s interrogatory response entitled “DRC RFI Section 4.0
Erosion” completed in 2013 to corroborate the results of the USLE analysis (Appendix D of
EnergySolutions (2013)). This comparison showed a total loss from the embankment of 0.026
mm/yr from the USLE analysis and 0.016 mm/yr from the RHEM analysis. Since 2013, the
RHEM on-line model has been updated. The latest version of the model (Version 2.3) lacks
some of the functionality of the earlier model.
Specifically, slope lengths cannot be directly input as they were previously, so a direct
comparison with the calculation using USLE is not possible. This appears to be due to an error
within the program since the descriptions of the model available on the website include slope
length as an input parameter option. Attempts to reach the contacts listed for the RHEM model in
order to resolve the issue were unsuccessful. The model defaults to a slope length of 50 meters
(164 feet).
Additional parameter options for cover characteristics appear to have been added since 2013,
including options for a distinction between foliar cover and basal plant cover as well as rock
cover, litter cover, and biological soil crust cover. Since the tool is a web-based model and not
stand-alone software, the version of the online tool used in the 2013 analysis is no longer
available. The parameters described in the 2013 interrogatory for vegetative conditions, slope
grades, climate station, soil texture class, and slope lengths are still valid for the analysis
completed at the time but the results reported therein can no longer be validated.
Gravel Surface
Next UDEQ discusses the use of local analogs to study soils with gravel surfaces. There are no
questions or requests raised.
Gravel Fraction to Address Erosion
Then UDEQ goes on to argue that the appropriate gravel fraction necessary to prevent erosion
has not been defined with precision. Further, a validated methodology has not been developed to
determine this. Refer to the response to UDEQ Interrogatory 198/1 for further discussion.
Gravel Fraction to Control or Prevent Biointrusion
The next issue raised is a statement by UDEQ that gravel will not prevent biointrusion. This
observation is accurate but not relevant, as gravel used in the cover design is not assumed or
intended to inhibit biointrusion. In fact, as described in the revegetation plan developed by
SWCA (2013), one function of gravel mulch over the soil surface is to increase moisture
retention and seed establishment.
The proposed cover design uses the frost protection zone to provide the function of a
biointrusion barrier. This layer is described by SWCA (2013) as “18 inches (46 cm) of 10–16
inches (25–41 cm) gravel and cobble mixture in-filled with small gravel, sand, and other fines
(cobble and gravel to 16 inches [40.6 cm] diameter).” Please see the response to UDEQ
Interrogatory 05/2 for discussion of evaluation of the extent of expected biointrusion at the Site.
Erosion Responses for the Clive DU PA Model
23 Feb 2018 20
Homogenization
The last issue of this interrogatory raised by UDEQ is that the hydraulic parameter distributions
and correlations used for modeling of flow in the cover system based on the recommendations of
Benson et al. (2011) were not implemented in the way that was intended by UDEQ.
Contrary to UDEQ’s claim in this alternative modeling of net infiltration and water content of
the cover layers requested by UDEQ, EnergySolutions did in fact use the recommendations from
Section 10.2 “Parameters for Performance Assessments” in Benson et al. (2011), cited as the
source of recommendations in Appendix E of the SER (SC&A 2015b). These recommendations
for hydraulic parameters are:
• The saturated hydraulic conductivity of fine-textured earthen storage and barrier layers
can be assumed to range between 1 × 107 m/s and 5 × 106 m/s.
• The porosity of earthen storage and barrier layers will likely range between 0.35 and
0.45.
• The α-parameter in the van Genuchten equation, which is used to describe the soil water
characteristic curve (SWCC) for hydrologic simulations, varies between 0.01 and
0.33 kPa-1 for field-scale barrier and storage layers.
• The n-parameter in van Genuchten’s equation, which is used to describe the SWCC for
hydrologic simulations, varies over a very small range (typically between 1.2 to 1.4).
Except for the n-parameter, where UDEQ provides no reference as to which layer the
recommendation refers, parameter value recommendations for all other parameters are applied to
both barrier and storage layers without distinction. For the proposed Federal Cell cover design,
the layers of the evapotranspiration (ET) cover are all either storage or barrier layers; however,
UDEQ’s recommended method makes no differentiation between the layers with respect to
hydraulic parameters. The only distinguishing characteristics between depth zones of a variably
saturated flow model are the values of the hydraulic parameters. If a saturated hydraulic
conductivity value, for example, is drawn from a statistical distribution developed according to
the method of Benson et al. (2011), then that value will be assigned according to that method to
all cover layers in the model. Thus, the flow model of the cover will be homogeneous with
respect to hydraulic conductivity. UDEQ objects to this homogeneity in hydraulic properties in
this interrogatory, but that is the only viable outcome possible when applying the Benson et al.
(2011) method to the ET cover.
2.3 Interrogatory CR R313-25-25(4)-197/1: Properties of
Embankment Side Slope Materials
Interrogatory Statement:
Please explain and justify how, from an erosion perspective, the properties of Unit 4 material “are
sufficiently similar” to the Federal Cell side slope, which consists of a mixture of Unit 4 soil with
50% gravel, to support this “sufficiently similar” modeling assumption. Also, please explain how
the properties of Unit 4 material are sufficiently similar to Class A waste, which would be included
over the DU waste. See also Interrogatory 203/1 below.
Erosion Responses for the Clive DU PA Model
23 Feb 2018 21
2.3.1 Interrogatory Response
In this interrogatory UDEQ preferentially extracts two from a number of assumptions provided
in Appendix 10 of the Clive DU PA Model Final Report v1.4 (Neptune 2015a) as the basis for
applying the results from modeling of a borrow pit face at the Site to the projection of erosion on
the Federal Cell embankment. While there are differences in the characteristics of the borrow pit
and the embankment, this analysis provides another line of evidence in addition to the more
conventional modeling used to demonstrate the stability of the proposed ET cover design
described below. Notably, the multiple lines of evidence provide good agreement that the
proposed ET cover design will provide adequate erosion resistance.
Landscape evolution models such as SIBERIA are still in an early stage of scientific
development. Accordingly, the methods for building confidence in model results that are
commonly used in environmental modeling, such as rigorous calibration, sensitivity analyses,
and uncertainty analyses, are difficult to apply. These models, however can be used in an
exploratory context to gain insight into processes (Skinner et al. 2017) and to provide additional
evidence to support the results of cover stability analyses using methods recommended by EPA
and NRC. Calculations to evaluate the stability of the cover design with respect to gully erosion
for the Class A West cell were provided in Appendix D of EnergySolutions (2015). Similar
calculations for the Federal Cell are presented in the response to UDEQ Interrogatory 71/1.
2.4 Interrogatory CR R313-25-25(4)-198/1: Gravel Content of
Embankment Materials
Interrogatory Statement:
Please provide the design bases and justification for the amount and sizing of the gravel in the
top and side slopes of the Federal Cell. The proposal for the gravel admixture in the top slope
(15%) appears too small. Also, please provide evidence for existing semi-arid or arid sites where
only 15% gravel has been added to form a successful cover-system surface layer for a landfill.
Please describe actual analog sites where 50% gravel for side slopes has been demonstrated to
be effective against erosion.
2.4.1 Interrogatory Response
A detailed response to this interrogatory was developed by Hansen, Allen, and Luce (HAL
2017a). A summary of their final report is provided here. The complete report is provided in
Appendix B.
The design basis for the amount of gravel to add to the soil in the top and side slopes was first
determined based on the analysis of the long-term sustainability of the slopes exposed to sheet
erosion over time. The methods that were used are the Universal Soil Loss Equation (USLE)
(USDA 1978) and the Rangeland Hydrology and Erosion Model (RHEM) (Nearing et al. 2011).
These methods produce an estimate of the average annual soil loss in terms of tons per acre per
year. With the gravel admixed into the Unit 4 clay soil, average annual soil losses were
calculated to be about 0.24 tons/acre/year using USLE, which was also closely matched by the
RHEM. This amount of loss is almost 10 times less than what has been recommended in EPA
Erosion Responses for the Clive DU PA Model
23 Feb 2018 22
guidance for covers for hazardous waste facilities. A more detailed discussion of the
methodologies and results is provided in the response to UDEQ Interrogatory 191/3.
Once the slope was determined to be stable from an average soil loss perspective, checks on the
gully erosion potential were completed based on the calculation of predicted velocities by
comparing them to the maximum permissible velocity according to the method presented in
NUREG-1623, “Design of Erosion Protection for Long-Term Stabilization” (NRC 2002). This
second assessment based on an extreme intense rainfall event is recommended in NUREG-4620
(Nelson et al. 1986) because of the potential for significant damage to cover systems from such
large events. In fact, this extreme intense rainfall event is known as the Probable Maximum
Precipitation (PMP) and is so extreme that “the point precipitation data base, even if maximized
for PMP moisture potential, shows no observed values even close to the 10-inch PMP estimate
(the generalized, local storm PMP estimates in Utah)” (Jensen 1995). Additionally, the
methodology applies a flow concentration factor (F) of three (NRC 2002). Therefore, the
extreme event that is many times the highest ever recorded rainfall intensity in the area is also
compounded by an additional safety factor of a multiplier of three. The end result of this process
is an analysis that is meant to produce an extremely conservative value.
Flow velocities on the top and side slopes of the CAW embankment during the PMP event were
predicted to be 2.37 and 2.07 ft/sec, respectively, as discussed in more detail in the response to
UDEQ Interrogatory 191/3. The acceptable Maximum Permissible Velocity (MPV) was selected
from tables provided in NUREG/CR-4620 (Nelson et al. 1986). Under this method the slope is
stable if the calculated velocity (V, the velocity resulting from a PMP in this application) is less
than the MPV. By contrast, if velocities exceed the MPV, the slope will experience excessive
erosion that will lead to the formation of gullies. HAL found that the calculated velocities
resulting from the PMP did not exceed the prescribed permissible velocity.
Reference Review
The methods discussed above have been published and/or accepted by the EPA (1989) and the
US Nuclear Regulatory Commission (Nelson et al. 1986; NRC 2002). Other methods referred to
in the interrogatory, on the other hand, have not been adopted, published, or referred to in
guidance by federal agencies. While the works cited are part of productive ongoing academic
research and study, they are not conclusive in their findings and should instead be the focus of
additional research. A summary of each is provided below:
• “Gravel Admixtures for Erosion Protection in Semi-Arid Climates,” Erosion of Soils and
Scour of Foundations (Anderson and Stormont 2005). This paper was published as part
of the proceedings of sessions of the Geo-Frontiers 2005 Congress held in Austin, Texas.
Along with other conference proceedings articles, these early attempts to define a process
for admixture design are helpful in that they outline practical steps to determine gravel
size, gravel percentage, and admixture thickness. This approach, however, has limitations
that result from the lack of empirical evidence to back up the steps and applications of
many of the equations presented in the paper. The paper states that “the design method
for gravel admixtures presented here may serve as an outline for further erosion
investigations and provide guidance for future designs of gravel-soil admixture layers.”
Therefore, the paper was not meant to be used as a proven method for admixture design
but rather as a starting point for further research and investigation.
Erosion Responses for the Clive DU PA Model
23 Feb 2018 23
• “Design of Erosion Protection at Landfill Areas with Slopes Less than 10%” (Anderson
and Wall 2010a). Much of the background discussion of this paper is the same as
Anderson and Stormont (2005). The methodology presented is similar to the steps
provided in Anderson and Stormont (2005) with some changes. Again, the paper does not
claim this to be a proven method but indicates the authors are taking steps to test its
outcomes. Anderson and Wall (2010a) state “the procedure described here is being
applied to a landfill cover soon to be constructed in southern Nevada. The Nevada project
will provide the first large scale application of the procedure.” HAL was unable to find
information regarding empirical data on a smaller scale nor was final information on the
results of the large scale Nevada project able to be found. Without this information or
additional discussion in the paper, the limits of the equation and how the criteria for
success or failure are defined remain unknown. For example, assuming the Nevada
project site is shown to be successful with the 40% gravel admixture, there is nothing that
would indicate if a lower percentage of gravel admixture, such as 25%, would also be
adequate. Additionally, the suitability of the design and how success is defined is not
sufficiently described in the paper to know whether the absence of observed gully erosion
constitutes a successful design or if it is acceptable to have the formation of rills and
gullies as long as the depth does not exceed the admixture layer thickness calculated
using the method. This is an important distinction since the end result of the design
process presented in the paper is the calculation of the thickness of the admixture layer
which is presented as being dependent on the percentage of gravel in the admixture. This
specific paper does not provide limits or conclusive guidance to the reliability of the
method and therefore should serve as the starting point for further investigation and
research.
• “Erosion Protection at Landfill Slopes Greater than 10%” (Anderson and Wall 2010b).
This paper focuses on riprap on side slopes. The design proposed by EnergySolutions
does not include riprap side slopes; therefore, this paper does not apply.
• “Long-Term Cover Design for Low-Level Radioactive and Hazardous Waste Sites as
Applied to the Rocky Flats Environmental Technology Site Solar Evaporation Ponds”
(Stenseng and Nixon 1995). The paper describing the design used at the Rocky Flats
Environmental Technology Site was included as part of the proceedings from the 50th
Industrial Waste Conference held in May 1995. The paper includes a brief description of
the design basis for the 5% top slopes and 20% side slopes and the 40% gravel
admixture. The admixture gravel content appears to have been selected instead of
calculated as the result of a defined process. While instructive regarding the specific
cover design discussed in the paper, no additional guidance is provided to guide
application at other sites.
• “Ecology, Design and Long-Term Performance of Surface Barriers: Applications at a
Uranium Mill Tailings Site” (Waugh and Richardson 1997). This paper is similar to the
Stenseng and Nixon paper discussed previously in that it contains a summary of the
design parameters but does not provide a specific and applicable methodology for the
design of a soil gravel admixture with regard to erosion protection.
Erosion Responses for the Clive DU PA Model
23 Feb 2018 24
Sites with Top Slopes Comprised of 15% Gravel Admixture
There are some examples of sites that have employed slopes with 15% admixture and that give
some measure of proof that a 15% admixture can be effective at controlling erosion. A test site
was established at Hanford that is referred to as the Hanford Prototype Surface Barrier (PSB).
The purpose of the PSB test site was to evaluate surface barrier constructability, construction
costs, and physical and hydrologic performance at field scale (DOE 2016). This field-scale test
cover was installed in 1994 and is comprised of a top erosion protection layer made up of a silt
loam admixed with 15% pea gravel. The top slope of 2% in this case is slightly less than the top
slopes of 2.5% and 4% proposed for the Federal Cell, but is an example of a cover system that
relies on the same amount of gravel admixture. A review of the data collected at the site through
2015 (DOE 2016) concludes that:
The 19-year PHB record showed practically no evidence of wind or water erosion of the
ETC barrier, despite 3 years of triple the mean annual precipitation; three simulated 1000-
year-return, 24-hour precipitation events; and an intense, controlled fire that burned off all
vegetation across half the barrier surface….Even in the absence of vegetation (e.g., following
a fire), the pea gravel added to the silt loam protected the barrier surface from wind and
water erosion….Overall, the monitoring results have confirmed that the PHB design is
resistant to water and wind erosion and that resistance is expected throughout the barrier’s
1000-year design life.
Another test site using the same cover system was established at Hill Air Force Base in 1994,
though little information has been published about that site to date.
Sites with Side Slopes Comprised of 50% Gravel Admixture
No sites were found that have used a gravel admixture on side slopes at or above 20%. Similarly,
there were no methodologies found that specifically address the calculation of gravel admixtures
for slopes greater than 10%, other than the general methods found in NUREG-1623 and
NUREG/CR-4620 (Nelson et al. 1986; NRC 2002). For slopes over 9%, Simanton et al. (1984)
found that the rate of water erosion decreases exponentially with increasing rock fragment cover.
The effect of biological soil crust is also difficult to quantify. These crusts are expected to
become established along the slopes as has been observed along natural slopes in the Clive area.
The required lengthy post-closure care period provides the greatest opportunity for verification
of the design methodology described previously by Simanton et al. (1984).
The design process completed to determine the acceptability of the 50% gravel admixture on the
20% side slopes is in accordance with EPA and Nuclear Regulatory Commission guidance.
These guidance procedures contain no requirements that existing sites be used as evidence to
support proposed designs. While it is acknowledged that such evidence would be helpful, each
site is different and contains unique design constraints based on physical layout, climate,
material availability, and project goals that make determinations based on sound methodologies
necessary.
Erosion Responses for the Clive DU PA Model
23 Feb 2018 25
2.5 Interrogatory CR R313-25-25(4)-199/1: Uncertainties in Erosion
Modeling
Interrogatory Statement:
Please provide quantitative estimates of the uncertainties involved using the borrow pit model as
an analog for estimating erosion of the Federal Cell, including use of RHEM to develop input
parameters for SIBERIA, and modeling uncertainties inherent in the selection of SIBERIA.
2.5.1 Interrogatory Response
Landscape evolution models (LEMs) were developed and applied for a face of a borrow pit at the
Clive Site in order to predict the response of the pit face and upslope land surface to water
erosion during runoff events. The models provide a quantitative description of the evolution of
slopes and channels over time. While the embankment upper cover layers and the borrow pit
material were the same Unit 4 soil, there were differences between the embankment and the
borrow pit. The borrow pit was modeled as bare soil with no vegetation or gravel, with a much
smaller top slope but a much larger catchment area. The embankment will be vegetated, and will
have a surface layer with a gravel admixture but a steeper top slope. The catchment area of the
embankment is smaller and the side slope is much less steep than the borrow pit face.
Although the borrow pit is an approximation to the embankment, the intent of this analysis was
to provide insight into erosion projected out to 10,000 years. This analysis provides another line
of evidence in addition to the sheet erosion modeling using USLE and the gully projection
modeling using the permissible velocity method (NRC 2002) that demonstrate the stability of the
proposed ET cover design. (See the response to UDEQ Interrogatories 71/1, 191/3, and 197/1.)
This interrogatory requests that quantitative uncertainty estimates be provided for the use of the
borrow pit erosion model results as an analog to evaluate the influence of erosion on
embankment performance at 10,000 years. While potentially interesting, such estimates are not
necessary for demonstrating erosion resistance of the Federal Cell ET Cover. Furthermore,
LEMs have been used as exploratory models providing insight into landscape–climate processes
for many years, but they have not been developed to the level of other types of environmental
modeling (Skinner et al. 2017). Typical methods used for calibration, sensitivity analysis, and
uncertainty analysis are difficult to apply to LEMs since little data is available for calibration and
verification (Skinner et al. 2017); (Temme et al. 2009). LEMs, however, even at this early stage,
can be useful in providing insight into landscape-climate processes.
2.6 Interrogatory CR R313-25-25(4)-200/1: Use of RHEM to Develop
Parameters for SIBERIA
Interrogatory Statement:
Please remodel erosion of the Federal Cell cover using the newer version of the RHEM model
Al-Hamdan et al. 2015) applicable to disturbed soils and concentrated surface-water flow. The
SIBERIA model results in the DU PA v.1.4 should be compared with those of SIBERIA modeling
of erosion for the site based on the Grand Junction embankment modeling by Smith (2011).
Modeling of the latter embankment indicates that significant gullying can occur over time on side
Erosion Responses for the Clive DU PA Model
23 Feb 2018 26
slopes even with vegetated soil on the embankment having considerable (i.e., 40%) added
gravel (Smith 2011).
2.6.1 Interrogatory Response
UDEQ requests that modeling of erosion on the Federal Cell be redone using an updated version
of the RHEM model. As described by Al-Hamdan et al. (2014), “RHEM was initially developed
for functionally intact rangelands where concentrated flow erosion is minimal.” Al-Hamdan goes
on to describe the changes to RHEM as including accounting for changes in flow due to soil
disturbance: “Disturbance such as fire or woody plant encroachment can amplify overland flow
erosion by increasing the likelihood of concentrated flow formation.” UDEQ states that the
modeling for the Federal Cell should be redone because the previous version of RHEM used for
the modeling “has limited application to describing erosion by concentrated flow on disturbed
soils, as would be expected at Clive.”
The response to UDEQ Interrogatory 71/1 describes the sequence of reclamation measures for
quickly re-establishing natural conditions on the cover developed by SWCA (2013). Ecological
and engineering measures described in the response to 171/1 will rapidly stabilize the cover in
the short term and continue to provide long-stabilization. As SWCA (2013) states, “functioning
native ecosystems comprised of the borrow soils at the Clive Site do not show erosion as the
DRC suggests.” The cover surface will not have the characteristics of disturbed rangeland soils,
so the added features of RHEM for disturbed conditions are not relevant to modeling the Federal
Cell. Thus, redoing the SIBERIA erosion model calibration with RHEM will not add to the
information from the modeling results.
Erosion modeling of the Federal Cell and the Class A West Cell using methods in NRC- and
EPA-approved guidance has demonstrated the stability of the proposed covers. Sheet and rill
erosion modeling using USLE is described in the response to UDEQ Interrogatory 191/3.
Evaluation of the potential for gully erosion on the Federal Cell using the permissible velocity
method is performed as described in NUREG-1623 (NRC 2002) (response to UDEQ
Interrogatory 71/1).
In addition, changes to the RHEM model have made it currently inapplicable to modeling
scenarios at Clive. HAL (2018) note that slope length is no longer a functioning input variable to
RHEM; all simulations have a set slope length of 50 meters (164 ft). See the response to UDEQ
Interrogatory 191/3 for more information.
This interrogatory also requests a comparison of the borrow pit modeling results (DU PA v1.4
Appendix 10 (Neptune 2015c)) to the results of Smith and Benson (2016) for an embankment at
Grand Junction. Smith and Benson (2016) describe a modeling study comparing cover stability
for scenarios considering cover geometry, climate, rock fragment content of the surface layer,
and vegetation. The conditions considered by these authors were: semi-arid and humid climate;
surface layers of riprap, topsoil, and gravel admixture; and presence and absence of vegetation.
These scenarios were simulated to 1,000 years by Smith and Benson (2016).
There are a number of reasons why the borrow pit simulations (DU PA v1.4 Appendix 10
(Neptune 2015c)) and Smith and Benson (2016) are not comparable. The characteristics of the
topsoils used for the two covers are different. Smith and Benson described the properties of the
Erosion Responses for the Clive DU PA Model
23 Feb 2018 27
materials they used for the simulations in their Table 3.1. This table shows the particle size
distribution of the topsoil as 40 percent sand, 40 percent silt, and 20 percent clay. In contrast,
Unit 4 at the Clive Site is approximately 23 percent sand, 47 percent silt, and 30 percent clay (see
also the response to Interrogatory 191 and Appendix A).
The Grand Junction site is not ecologically analogous to the Clive Site (SWCA 2013).
Vegetation is an important factor in stabilizing covers. Smith and Benson (2016) assumed the
dominant vegetation for the semi-arid case to be Mountain Big Sagebrush. In contrast, SWCA
(2013) identifies the dominant vegetation community as Mountain Basins Mixed Salt Desert
Scrub on analogs to the Clive Site. The soils at Clive are saline; different chemistry and fertility
of soils at Grand Junction influence the characteristics of the native vegetation. SWCA (2013)
describe biological soil crusts as a “dominant feature of the vegetation communities in the Great
Salt Lake Basin.” Biological soil crusts, important in stabilizing soils, are not considered by
Smith and Benson (2016).
Clive borrow pit simulations projected to 10,000 years (DU PA v1.4 Appendix 10) considered
unvegetated, Unit 4 soils with no added gravel. Smith and Benson (2016) show erosion model
results for a cover in a semi-arid climate with a vegetated topsoil surface in Figure 4.1, and
results for a cover in a semi-arid climate with an unvegetated gravel admixure surface in Figure
4.6. No directly comparable results to the Clive borrow pit simulation conditions (semi-arid, no
vegetation, no gravel) were found in Smith and Benson (2016).
2.7 Interrogatory CR R313-25-25(4)-201/1: Estimating Rainfall
Intensity
Interrogatory Statement
It is not clear that the Probable Maximum Precipitation (PMP) was determined using the
procedures outlined in the National Oceanic and Atmospheric Administration and U.S. Army
Corps of Engineers publication Hydrometeorological Report No. 49 (HMR 49) (1977).
According to EnergySolutions, these procedures resulted in a “1-hour PMP rainfall intensity of
9.9 inches (Jones, 2012).” However, DWMRC finds that a value of 9.8 or 9.9 inches is not the
intensity, but rather the 1-hour PMP, or the maximum precipitation expected over 1 square mile
when averaged over an hour.
Please re-calculate the PMP using NUREG/CR-4620, as outlined below.
2.7.1 Interrogatory Response
A detailed response to this interrogatory was developed by Hansen, Allen, and Luce (HAL
2017b). A summary of their final report is provided here. The complete report is provided in
Appendix C.
Gully erosion potential was initially checked based on the calculation of permissible velocities
according to the method presented in NUREG-1623, “Design of Erosion Protection for Long-
Term Stabilization” (NRC 2002). As pointed out in the Interrogatory Statement, it is
acknowledged that the methodology utilized previously failed to incorporate the proper rainfall
intensity as outlined in NUREG/CR-4620 (Nelson et al. 1986).
Erosion Responses for the Clive DU PA Model
23 Feb 2018 28
In order to determine the probable maximum precipitation (PMP) intensity as outlined in the
guidance in NUREG/CR-4620 (Nelson et al. 1986), it is necessary to first calculate the time of
concentration for representative drainage areas for both the top and side slopes. The
methodology outlined in Technical Release 55, “Urban Hydrology for Small Watersheds”
(USDA 1986), was used to calculate the time of concentration for each representative slope.
TR-55 describes three types of drainage flow: sheet flow, shallow concentrated flow, and open
channel flow. Sheet flow is defined as flow over planar surfaces at very shallow depths for up to
300 feet. After a maximum of 300 feet, the flow transitions to shallow concentrated flow. Open
channels were not included in the calculations due to the absence of designed channels. The
same Manning’s roughness coefficient calculated to be 0.05 for use in the NUREG-1623 (NRC
2002) methodology using an empirical equation for channels with gravel beds with shallow flow
depths of Bray (Coon 1998) was also used to represent the sheet flow roughness. It was decided
to use the value calculated using the Bray method (0.05) instead of a higher value found in other
publications for overland flow using sparse vegetative cover in order to be more protective. The
above described methodology was applied to both the Class A West Cell and the Federal Cell to
calculate the time of concentrations shown in Table 3 and
Table 4.
Table 3. Calculated Time of Concentration Using TR-55 Method for Class A West Cell.
Slope Description Total
Length
(ft)
Slope
(ft/ft)
Sheet
Flow
Length
(ft)
Sheet Flow
Manning’s n
Shallow
Concentrated
Flow Length
(ft)
Total Time of
Concentration
(min)
Top Slope (4%) 942 0.04 300 0.05 642 17.4
Side Slope (20%) 188 0.20 188 0.05 0 4.6

Table 4. Calculated Time of Concentration Using Tr-55 Method for Federal Cell.
Slope Description Total
Length
(ft)
Slope
(ft/ft)
Sheet
Flow
Length
(ft)
Sheet Flow
Manning’s n
Shallow
Concentrated
Flow Length
(ft)
Total Time of
Concentration
(min)
Top Slope (2.5%) 521 0.025 300 0.05 221 16.9
Side Slope (20%) 178.5 0.20 178.5 0.05 0 4.4

The rainfall depth is 9.9 inches, determined using the methods outlined in Table 6.3a of the US
Army Corps of Engineers publication, Hydrometeorological Report No. 49 (HMR 49) (NOAA
1984), for determining an average 1-hour 1-square mile PMP. The steps were then followed from
NUREG/CR-4620 (Nelson et al. 1986) to determine rainfall intensity from the PMP. The
intensities for the Class A West and Federal Cells are shown in Table 5 and Table 6.
Erosion Responses for the Clive DU PA Model
23 Feb 2018 29
Table 5. PMP Rainfall Intensity for the Class A West Cell.
Slope Description Rainfall
Duration (Tc)
(minutes)
% PMP
(%)
Rainfall
Intensity (i)
(inches/hr)
Top Slope (4%) 17.4 53.8% 18.4
Side Slope (20%) 4.6 15.3% 19.8
Table 6. PMP Rainfall Intensity for the Federal Cell.
Slope Description Rainfall
Duration (Tc)
(minutes)
% PMP
(%)
Rainfall
Intensity (i)
(inches/hr)
Top Slope (2.5%) 16.9 52.1% 18.3
Side Slope (20%) 4.4 14.7% 19.8

Gully Erosion Potential
As opposed to the projection of the long-term effects of precipitation over time due to sheet
erosion, the effects of gully erosion are determined by the consideration of a large single rainfall
event. The procedure described in NUREG-1623 (NRC 2002) begins by calculating the peak
runoff rate considering the PMP rainfall intensity, the slope length, and the multiplication factor
of three recommended by NRC (2002). A flow depth is then calculated using the runoff rate,
slope, and a roughness coefficient for channels with gravel beds and shallow depths. The flow
rate and depth are then used to estimate a flow velocity.
The results for both the top slope and the side slope of the Class A West Cell using the vegetated
slope condition are summarized in Table 7. Flow velocities on the top and side slopes of the
Class A West Cell during the PMP event are predicted to be 2.37 and 2.07 ft/sec, respectively.
The permissible velocity method is commonly applied to determine channel stability. The slope
is assumed stable if the calculated velocity (V, the velocity resulting from a PMP in this
application) is less than the maximum permissible velocity (MPV). By contrast, if velocities
exceed the MPV, it is assumed that the slope will experience excessive erosion that will lead to
the formation of gullies. An MPV of 5.0 ft/s, appropriate for gravel channels, was selected from
tables provided in NUREG/CR-4620 (Nelson et al. 1986). This methodology then directs that the
estimates for the MPV be adjusted downward to account for the influences of flow depth. For the
flow depths calculated for this cell the adjustment factor is 0.5, reducing the MPV to 2.5 ft/s. The
side slope gully analysis was completed independently of the top slope.
Erosion Responses for the Clive DU PA Model
23 Feb 2018 30
Table 7. Gully Erosion Potential—Velocity Analysis for Class A West Cell.
Slope Description Length
(ft)
Slope
(ft/ft)
i
(in/hr)
c Q
(cfs/ft)
y
(ft)
V
(ft/sec)
Adjusted
MPV
(ft/sec)
Top Slope (4%) 942 0.04 18.4 0.5 0.60 0.25 2.37 2.5
Side Slope (20%) 188 0.20 19.8 0.5 0.13 0.06 2.07 2.5

Comparing calculated velocities to MPVs in Table 7 demonstrates that all slope scenarios using
the ET cover system provide acceptable protection against gully erosion using these criteria by
limiting the potential of gully formation from high velocity channelization.
2.8 Interrogatory CR R313-25-25(4)-202/1: Use of SIBERIA to Model
Federal Cell Erosion
Interrogatory Statement:
The Division is concerned that the SIBERIA model referenced in DU PA v.1.4 discussions
assumes a modeling-realm geometry inconsistent with that of the proposed Federal Cell. The
Federal Cell embankment is approximately 30 feet high (height of waste under top slope and
above grade) compared to the model analog height of 10 feet. Also, the SIBERIA model allows
for several hundred meters of ground surface upslope from the sloping pit face, but that ground
surface only has a 0.3% (0.003) grade in the model. By contrast, as described in Appendix 3 to
the DU PA v.1.4, the waste under the top slope above and upslope from the side slopes of the
embankment has a grade of up to 2.4%. This is about eight times greater. EnergySolutions
needs to explain how these differences affect the results and how the Federal Cell modeling
results can be reconciled against similar modeling studies conducted by Smith and Benson
(2016) for the Grand Junction Uranium Mill Tailings Disposal Site.
2.8.1 Interrogatory Response
SIBERIA is employed in the DU PA model as a supplementary line of evidence for embankment
stability. If SIBERIA results were the sole or primary basis for demonstrating embankment
stability, the distinctions noted in the interrogatory could be relevant; however, LEMs such as
SIBERIA are acknowledged to be subject to further development before their results should be
considered conclusive in licensing situations. See also the response to UDEQ Interrogatory
199/1.
2.9 Interrogatory CR R313-25-25(4)-205/1: Erosion Analysis
Interrogatory Statement:
As discussed below, there appears to be an issue with the FractionGully 1.52 m depth data.
Please explain why the 1.52 m depth percentages are smaller than the 1.97 m and 2.42 m depth
results.
Erosion Responses for the Clive DU PA Model
23 Feb 2018 31
2.9.1 Interrogatory Response
The interrogatory creates Figure 205-2 to summarize the percentage of realizations within
particular intervals/categories of “fraction of cover area” for six depth intervals. In general,
Figure 205-2 and the associated Table 205-1 are used to conclude that the percentage of
realizations with different fractions of cover area (below 0.03) are generally smaller for the depth
interval they label at “1.52” than for deeper depth intervals. It was expected that deeper depths
would consistently have a larger percentage of the realizations with smaller fractions. This
behavior can be attributed both to how Figure 205-2 was created (how the summaries were
calculated to create the curves), and also to the stochasticity incorporated in the creation of the
fractions at each depth interval within a single realization.
First, while it is interesting to summarize over realizations within a depth category, it is
important to keep in mind that, for one realization of the GoldSim model, a single realization is
selected from 1000 and the fractions of cover for each depth category are used for that one
realization. The fractions within each realization (one row of the GoldSim look up table) are
created as described in Appendix 10 of the Clive DU PA Model v1.4 (Neptune 2015c); they
come from the same distribution and sum to one. The fractions associated with the six or seven
deepest depth intervals are all small (generally less than ~ 0.006) and are close together due to
the relatively flat tail of the lognormal distribution used to generate them. Therefore, due to the
stochasticity in generating the fractions from the lognormal distribution, there are realizations
with a slightly higher fraction of cover in deeper intervals than in a shallower interval. For
example, for one realization the fraction is 0 for depth interval (1.52, 2.03m], but then 0.0018 for
the depth interval (2.03m, 2.53m], followed by 0 for the deepest two intervals. This stochasticity,
along with different distributions used for each realization, contributes unexpected relative
differences among depth intervals within the range of very small fractions and deep depth
intervals.
The implications of this on the PA Model are expected to be negligible given how the
realizations are used in the PA Model and the magnitude of the differences in fractions involved
in the behavior pointed out in the interrogatory. The interrogatory provides an extreme case of
erosion in Figure 205-4 and Table 205-2, created from summarizing over realizations within
depth intervals. The resulting curves are presented as the minimum, mean, and maximum, but
they do not represent particular realizations used in the GoldSim model and no longer meet the
sum-to-one constraint within individual realizations. The summaries presented in Figures 205-4
and 205-2 are interesting if the results are not over-sold as representing actual erosion behavior
of individual GoldSim realizations. For example, it would not be possible for a single realization
to have the maximum fraction observed over all realizations for every depth interval, and
statements regarding “mean erosion” using these summaries are only valid to the extent that
behavior of individual realizations used in the GoldSim model reflects the averages per depth
interval.
As described above, the general idea presented in Figure 205-2 and Table 205-1 may stem
mainly from a concern that fractions within each realization are not non-increasing with depth at
deeper depths. However, Figure 205-2 is not easy to replicate without additional information. In
particular, it is not clear how the labels for the depth intervals were chosen or how many
realizations were used to create it (just 250 to match the other plots, or all 1000 in the GoldSim
Erosion Responses for the Clive DU PA Model
23 Feb 2018 32
look up table?). The depth intervals (in mm) created for the look up table are (0,10], (10,152],
(152,305], (305,457], (457,610], (610,762], (762,914] , (914,1.07e+03], (1.07e+03,1.22e+03],
(1.22e+03,1.37e+03], (1.37e+03,1.52e+03], (1.52e+03,2.03e+03], (2.03e+03,2.53e+03],
(2.53e+03,3.04e+03], and (3.04e+03,3.54e+03]. Two of the labels used in Figure 205-2
correspond to endpoints of these intervals (1.07m and 1.52m), but the additional depth labels
(1.97m, 2.42m, 2.87m, and 3.32m) do not. They are also not midpoints. It is possible that
interpolation was used to get these from Figure 205-1, but it is not clear why this would be done
when the fractions provided specifically apply to the stated depth intervals. It is also not clear
over what intervals of fraction the percentages of realizations were calculated. Figure 4 has been
created to display the same information that Figure 205-2 is meant to represent, although using
cumulative proportions as the interrogatory did for Figure 205-3 because this is a more natural
and repeatable way to display the values in the look-up table. It uses all 1000 realizations and
seven deepest depth intervals. The “trends” in the plot over depth intervals do not match those
presented in Figure 205-1, but the general idea that curves are not “in order” by depth is the
same.

Figure 4. Empirical cumulative distribution functions (ecdf) of the values across
realizations found in the GoldSim look-up table for the seven deepest depth intervals.
Note that, in general, the deeper intervals have larger proportions of realizations with
smaller fractions. The two green curves associated with depth intervals (1.37m,
1.52m] and (1.52m, 2.03m] do not follow the order with depth, which can be
interpreted as consistent with the conclusion from Figure 205-1.
Erosion Responses for the Clive DU PA Model
23 Feb 2018 33
Using values of simulated cover area eroded by gullies from their Table 205-1, UDEQ states,
“An eroded area of 1,087 m2 means that the entire perimeter of the embankment has eroded back
0.6 m (1.9 feet), and a 5,433 m2 eroded area means the perimeter has eroded 2.9 m (9.5 feet).”
This statement is a gross misrepresentation of the modeled erosion processes. UDEQ is
presenting a false equivalence between areally distributed erosion losses due to sheet flow and
gullying on the embankment and losses due to cliff retreat, a mass wasting process.
Radon flux
UDEQ next presents comparisons between radon flux in the embankment predicted by the Clive
DU PA Model v1.2 and several analytical solutions.
Direct comparison of late-time results from the Clive DU PA Model v1.4 with steady-state
analytical solutions is made difficult by basic differences in the assumptions involved in the
approaches. Analytical solutions typically are restricted by simplifying assumptions that allow
for closed-form solutions. For example, a constant or uniform source concentration is a typical
assumption. The PA Model, conversely, allows for movement as well as decay of radium-226,
which amounts to a non-uniform source that analytical solutions do not account for.
The analytical solution procedure for multi-layered cover systems described in NRC’s
Regulatory Guide 3.64 (NRC 1989) and applied by UDEQ begins with calculating the flux from
the waste zone assuming no cover exists (Equation 1 in SER Section 4.2.1 (SC&A 2015a)), and
then repeating a similar calculation for each layer up the column. Reg. Guide 3.64 describes this
iterative solution procedure as an “approximate method” due to the assumed boundary conditions
at each step.
Typically, comparisons of analytical solutions to numerical solutions require that the numerical
model be carefully prepared to match the assumptions of the analytical solutions. Differences in
the assumed parameter values can also be important. For example, the SER states that since “the
diffusion coefficient for the Rogers 2002 curve is simply a fixed value, UDEQ believes that the
IAEA Equation radon flux is more representative of the conditions at the proposed Federal Cell.”
However, Table 205-3 of the interrogatory presents only values derived using the Rogers (2002)
method.
Nevertheless, it is acknowledged that cross-comparison with different modeling methods can be
valuable and instructive. The interrogatory states that radon flux, though at the ground surface,
compares well to UDEQ’s application of analytical solutions, but also mentions some
discrepancies at depth. It is noted that the results from the IAEA Equation are generally within
an order of magnitude of the DU PA Model v1.4 results at all depths despite possible differences
in assumptions. UDEQ correctly states that the DU PA Model v1.4 is calibrated to focus on
agreement with known analytical solutions of the flux at the ground surface rather than fluxes
deep in the waste zone, which are generally not important to dose calculations.
Lack of evidence of significant erosional activity at the Site, planned ecological remediation of
the soils used to construct the cover (response to Interrogatory 05/2), and erosion models that
demonstrate the stability of the cover (response to Interrogatory 71/1) support the prediction that
erosion of the embankment will be minimal and will not enhance radon fluxes. Given these site-
Erosion Responses for the Clive DU PA Model
23 Feb 2018 34
specific observations and modeling results, the differences in radon flux predictions between the
DU PA Model v1.4 and UDEQ’s application of analytical solutions at depths greater than 1 m
are not meaningfully different.
The April 2015 SER (SC&A 2015a) also notes that the embankment’s performance with respect
to radon flux would be adequate even with complete removal of the cover system, as predicted
doses are several orders of magnitude below regulatory limits. Thus, the comparison of the Clive
DU PA Model v1.4 with other radon flux predictions, which show only modest differences at
relevant depths, does not detract from the conclusion that radon fluxes are adequately attenuated
by the embankment due to the depth of burial of radon-generating wastes and the prevailing site
conditions.
3.0 Conclusion
UDEQ has raised many diverse issues in this set of interrogatories. The following points
summarize this response:
• Predicted erosion performance of the Federal Cell is reasonable. Lack of evidence of
significant erosional activity at the Site, planned ecological remediation of the soils used
to construct the cover, and erosion models that demonstrate the stability of the cover
support this position. Erosion of the embankment cover will be minimal and will not
enhance radon fluxes or deep infiltration.
• Using uniform hydraulic properties to model the entire ET cover represents the only
logical outcome of applying the Benson et al. (2011) method for assigning properties.
Given their conceptual model that makes no distinction between the hydraulic properties
of storage and barrier layers, the cover can no longer be represented by a layered system
in the flow model.
• The proper rainfall intensities for the embankment surfaces have been calculated
according to NUREG/CR-4620 (Nelson et al. 1986) and incorporated into the erosion
models.
• The design for the amount of gravel to add to the soil in the top and side slopes was
determined based on sheet erosion modeling using USLE and gully projection evaluation
using the permissible velocity method of the NRC.
• The results of the SIBERIA modeling of the borrow pit provide another line of evidence,
in addition to the more conventional modeling. This is useful in demonstrating the
stability of the proposed ET cover design despite differences in the characteristics of the
borrow pit and the embankment.
• The cover surface will not have the characteristics of disturbed rangeland soils, so the
added features of RHEM for disturbed conditions are not relevant to modeling the
Federal Cell. Thus, redoing the SIBERIA erosion model calibration with RHEM will not
add to the information from the modeling results.
Erosion Responses for the Clive DU PA Model
23 Feb 2018 35
• The SIBERIA models of the borrow pit (DU PA v1.4 Appendix 10 (Neptune 2015c)) and
a Grand Junction embankment (Smith and Benson 2016) are not comparable. No
conditions directly comparable to the Clive borrow pit simulations (semi-arid, no
vegetation, no gravel) were found in the Smith and Benson (2016) report.
• Radon fluxes are adequately attenuated by the embankment due to the depth of burial of
radon-generating wastes and the prevailing site conditions.
4.0 References
Al-Hamdan, O.Z., et al., 2014. Rangeland Hydrology and Erosion Model (RHEM)
Enhancements for Applications on Disturbed Rangelands, Hydrological Processes (2014)
1–13 doi: 10.1002/hyp.10167
Anderson, C., and J. Stormont, 2005. Gravel Admixtures for Erosion Protection in Semi-Arid
Climates, Erosion of Soils and Scour of Foundations, Proceedings of Sessions of the
Geo-Frontiers 2005 Congress, 2005, Austin TX
Anderson, C., and S. Wall, 2010a. Design of Erosion Protection at Landfill Areas with Slopes
Less than 10%. In Scour and Erosion, Geotechnical Special Publication (GSP) No. 210,
Proceedings of the 5th Annual Conference on Scour and Erosion (ICSE-5) 2010,
November 7–10, edited by S.E. Burns, et al., pp. 1054–1063, American Society of Civil
Engineers, San Francisco CA
Anderson, C., and S. Wall, 2010b. Erosion Protection at Landfill Slopes Greater than 10%. In
Scour and Erosion, Geotechnical Special Publication (GSP) No. 210, Proceedings of the
5th Annual Conference on Scour and Erosion (ICSE-5) 2010, November 7–10, edited by
S.E. Burns, et al., pp. 1064–1073, American Society of Civil Engineers, San Francisco
CA
Belnap, J., and D.A. Gillette, 1998. Vulnerability of Desert Biological Soil Crusts to Wind
Erosion: The Influences of Crust Development, Soil Texture, and Disturbance, Journal of
Arid Environments 39 (2) 133–142
Belnap, J., et al., 2001. Biological Soil Crusts: Ecology and Management, Technical Reference
1730-2, United States Department of the Interior, Bureau of Land Management, Denver
CO, 2001
Benson, C.H., et al., 2011. Engineered Covers for Waste Containment: Changes in Engineering
Properties and Implications for Long-Term Performance Assessment, NUREG/CR-7028,
United States Nuclear Regulatory Commission, Washington DC, December 2011
Coon, W.F., 1998. Estimation of Roughness Coefficients for Natural Stream Channels with
Vegetated Banks, U.S. Geological Survey Water-Supply Paper 2441, prepared in
Erosion Responses for the Clive DU PA Model
23 Feb 2018 36
cooperation with the New York State Department of Transportation, U.S. Geological
Survey, U.S. Department of the Interior, Denver CO, 1998
CWCB, 2006. Chapter 13, Hydraulic Analysis and Design, Section 1 Open Channels. In
Colorado Floodplain and Stormwater Criteria Manual, pp. CH13-100–CH13-F124,
Colorado Water Conservation Board, Denver CO
DOE, 2016. Prototype Hanford Barrier 1994 to 2015, DOE/RL-2016-37, Revision 0, United
States Department of Energy, Richland Operations Office, Richland WA, March 2016
EnergySolutions, 2013. Utah Radioactive Material License (RML UT2300249) Updated Site-
Specific Performance Assessment, Revision 1, prepared for Utah Division of Radiation
Control, EnergySolutions LLC, Salt Lake City UT, December 2013
EnergySolutions, 2015. Utah Radioactive Material License (RML UT2300249) Updated Site-
Specific Performance Assessment, Revision 2, prepared for Utah Division of Radiation
Control, EnergySolutions LLC, Salt Lake City UT, January 2015
EPA, 1989. Technical Guidance Document: Final Covers on Hazardous Waste Landfills and
Surface Impoundments, EPA 530-SW-89-047, United States Environmental Protection
Agency, Office of Solid Waste and Emergency Response, Washington DC, July 1989
EPA, 1991. Seminar Publication, Design and Construction of RCRA/CERCLA Final Covers,
EPA/625/4-91/025, United States Environmental Protection Agency, Washington DC,
May 1991
Haan, C.T., et al., 1994. Design Hydrology and Sedimentology for Small Catchments, 1st
Edition, Academic Press, San Diego CA
HAL, 2017a. Response to Interrogatory 198, Hansen, Allen, & Luce Inc., South Jordan UT,
December 2017
HAL, 2017b. Response to Interrogatory 201, Hansen, Allen, & Luce Inc, South Jordan UT,
December 2017
HAL, 2018. Response to Interrogatory 191, Hansen, Allen, & Luce Inc., South Jordan UT,
February 2018
Israelsen, C.E., et al., 1984. Erosion and Sedimentation in Utah: A Guide for Control, Reports.
Paper 372, Utah Water Research Laboratory, Utah State University, Logan UT, 1984
Erosion Responses for the Clive DU PA Model
23 Feb 2018 37
Jensen, D.T., 1995. Final Report, Probable Maximum Precipitation Estimates for Short-
Duration, Small-Area Storms in Utah, Utah Climate Center, Utah State University,
Logan UT, October 1995
Nearing, M.A., et al., 2011. A Rangeland Hydrology and Erosion Model, Transactions of the
ASABE 54 (3) 1–8
Nelson, J.D., et al., 1986. Methodologies for Evaluating Long-Term Stabilization Designs of
Uranium Mill Tailings Impoundments, NUREG/CR-4620, ORNL/TM-10067, United
States Nuclear Regulatory Commission (NRC), Washington DC, June 1986
Neptune, 2014. Final Report for the Clive DU PA Model, Clive DU PA Model v1.2, NAC-
0024_R2, Neptune and Company, Inc., Los Alamos NM, August 2014
Neptune, 2015a. Final Report for the Clive DU PA Model, Clive DU PA Model v1.4, NAC-
0024_R4, Neptune and Company Inc., Los Alamos NM, November 2015
Neptune, 2015b. FEPS Analysis for the Area G Radiological Risk Assessment, NAC-0038_R0,
Neptune and Company Inc., Los Alamos NM, March 2015
Neptune, 2015c. Erosion Modeling for the Clive DU PA, Clive DU PA Model v1.4, NAC-
0017_R4, Neptune and Company Inc., Los Alamos NM, October 2015
NOAA, 1984. Probable Maximum Precipitation Estimates, Colorado River and Great Basin
Drainages, Hydrometeorological Report No. 49, National Oceanic and Atmospheric
Administration, Silver Spring MD, 1984
NRC, 1986. Update of Part 61 Impacts Analysis Methodology, Methodology Report, Volume 1,
NUREG/CR-4370, United States Nuclear Regulatory Commission, Washington DC
NRC, 1989. Calculation of Radon Flux Attenuation by Earthen Uranium Mill Tailings Covers,
Regulatory Guide 3.64, United States Nuclear Regulatory Commission, Washington DC,
June 1989
NRC, 2002. Design of Erosion Protection for Long-Term Stabilization, NUREG-1623, United
States Nuclear Regulatory Commission, Washington DC, 2002
Rogers, T., 2002. A Change in Envirocare’s Disposal Cell Design, Waste Management 2002
Conference, February 2002, Tucson AZ
SC&A, 2015a. Utah Division of Radiation Control, EnergySolutions Clive LLRW Disposal
Facility, License No: UT2300249; RML #UT 2300249, Condition 35 Compliance Report;
Erosion Responses for the Clive DU PA Model
23 Feb 2018 38
Appendix A: Final Report for the Clive DU PA Model, Safety Evaluation Report, Volume
1, SC&A Inc., Vienna VA, April 2015
SC&A, 2015b. Utah Division of Radiation Control, EnergySolutions Clive LLRW Disposal
Facility, License No: UT2300249; RML #UT 2300249, Condition 35 Compliance Report;
Appendix A: Final Report for the Clive DU PA Model, Safety Evaluation Report, Volume
2, SC&A, Vienna VA, April 2015
Simanton, J.R., et al., 1984. Effects of Rock Fragments on Erosion of Semiarid Rangeland Soils,
SSSA Special Publication No. 13, Soils Science Society of America, Madison WI, 1984
Skinner, C., et al., 2017. LEMSI—The Landscape Evolution Model Sensitivity Investigation,
Geophysical Research Abstracts 19 (EGU2017-15699)
Smith, C.L., and C.H. Benson, 2016. Influence of Coupling Erosion and Hydrology on the Long-
Term Performance of Engineered Surface Barriers, NUREG/CR-7200, United States
Nuclear Regulatory Commission (NRC), Washington DC, May 2016
Stenseng, S.E., and P.A. Nixon, 1995. Long-Term Cover Design for Low-Level Radioactive and
Hazardous Waste Sites as Applied to the Rocky Flats Environmental Technology Site
Solar Evaporation Ponds, 50th Purdue Industrial Waste Conference Proceedings, 1995
SWCA, 2012. Vegetated Cover System for the EnergySolutions Clive Site: Literature Review,
Evaluation of Existing Data, and Field Studies Summary Report, prepared for
EnergySolutions, SWCA Environmental Consultants, Salt Lake City UT, August 2012
SWCA, 2013. EnergySolutions Updated Performance Assessment—SWCA’s Response to First
Round DRC Interrogatories, SWCA Environmental Consultants, September 2013
Temme, A.J.A.M., et al., 2009. Can Uncertain Landscape Evolution Models Discriminate
Between Landscape Responses to Stable and Changing Future Climate? A Millennial-
Scale Test, Global and Planetary Change 69 (2009) 48–58
USDA, 1978. Predicting Rainfall Erosion Losses, A Guide to Conservation Planning,
Agriculture Handbook Number 537, United States Department of Agriculture,
Washington DC, December 1978
USDA, 1986. Urban Hydrology for Small Watersheds, Technical Release 55 (TR-55), United
States Department of Agriculture, Washington DC, June 1986
Utah DEQ, 2017. Division of Waste Management and Radiation Control, EnergySolutions Clive
LLRW Disposal Facility License No: UT2300249; RML #UT 2300249, Amended and
New Interrogatories Related to Clive DU PA Modeling Report Version 1.4 Dated
Erosion Responses for the Clive DU PA Model
23 Feb 2018 39
November 2015, Utah Department of Environmental Quality (DEQ), Salt Lake City UT,
May 2017
Waugh, W.J., and G.N. Richardson, 1997. Ecology, Design and Long-Term Performance of
Surface Barriers: Applications at a Uranium Mill Tailings Site. In Barrier Technologies
for Environmental Management: Summary of a Workshop, edited by Committee on
Remediation of Buried and Tank Wastes and National Research Council, pp. D-54–D-67,
National Academy Press, Washington DC

Erosion Responses for the Clive DU PA Model
23 Feb 2018 40
Appendix A. HAL 2018

Erosion Responses for the Clive DU PA Model
23 Feb 2018 41

DRC Interrogatory CR R313-25-7(2)-191/3
Selected Statements from DEQ Conclusion from April 2015 SER, Appendix C:
Interrogatory 191 requested EnergySolutions to provide additional information about the
ability of steep side slopes to resist gully erosion. After reviewing…DEQ believes that the key
analysis is….Appendix D uses both RUSLE and REHM to calculate rill or sheet erosion, with
similar results. Both are well below the EPA’s criteria for ……cover systems. One problem
with the Appendix D analysis is that it does not describe how the values for the various
RUSLE and REHM parameters were selected. For example, the RUSLE has R, K, L, S, and C
parameters, but only L and S are functions of the embankment’s design, so the basis for
selecting the other parameters is not clear.
In conclusion, the analysis performed by HAL may or may not be correct, but before DEQ can
accept it, each value selected and used in the analysis needs to be justified.
HAL Response: February 12, 2018
Universal Soil Loss Equation (USLE)
The USLE estimates average annual soil losses from erosion. In order to clarify the
methodology used in the analysis, all references to RUSLE have been change to refer to USLE.
While the equations are the same, the use of the acronym RUSLE could imply the use of the
software program developed by the USDA for factor determination. The software package was
not used because factors from a Utah-specific publication were relied on for the analysis as
described below.
The review from DEQ indicated that the previous interrogatory response was lacking in the
description of the selection of the parameters. Therefore, the summary of the methodology has
been enhanced to provide additional background information. The USLE methodology is
commonly used to determine the long-term stability of slopes and is an industry-standard means
for design of erosion control. Guidance given by the EPA states that “The U.S. Department of
Agriculture’s (USDA’s) Universal Soil Loss Equation is recommended as the tool to evaluate
erosion potential” (US EPA Seminar Publication, 1991). The basis for this approach comes from
the theory that “if adequate protection is provided to control sheet erosion, then rills and gullies
will never form from rainfall” (Israelsen et al, 1984). Generally, the USLE equation is defined
as:
A = R*K*LS*C*P
Where:
A = the average soil loss per unit area, expressed in tons/acre/year
R = the rainfall/runoff factor, which is the number of rainfall units for rainfall energy and runoff
and snowmelt
K = soil erodibility factor in tons per acre per year per unit of R
LS = topographic factor (length and steepness of the slope)
C = the cover and management factor (equivalent to the VM factor), which is the ratio of soil
loss from an area with a given cover and management relative to that from an identical area in
continuous fallow
P = the supporting conservation practice factor, in this case assumed to be equal to 1
This procedure and site-specific factors are described in “Erosion and Sedimentation in Utah – A
Guide for Control” (Israelsen, 1984) and “Design Hydrology and Sedimentation for Small
Catchments” (C.T. Haan et al, 1994). The computed average sheet erosion soil loss is presented
in Table 1.
Erosion Responses for the Clive DU PA Model
23 Feb 2018 42
R Factor
The R factor (6) is selected based on the Utah-specific Iso-Erodent (R) mapping provided in the
Utah Water Research Laboratory report (Israelsen, 1984) which is also provided on the following
page.
K Soil Erodibility Factor
The K values (0.18 and 0.07 for top slope and side slope, respectively) are based on the Unit 4-
specific material characteristics with the top slope and side slope gravel admixtures together with
the Wischmeier nomograph as described in the methodology presented in the Israelsen and Hann
procedures and shown in the figure below. A particle size analysis using the hydrometer method
was performed on the Unit 4 clay to determine the percent silt and very fine sand (see figure
below) for use with the Wischmeier nomograph. The results of the hydrometer analysis are
attached to this response.
.
Erosion Responses for the Clive DU PA Model
23 Feb 2018 43

Erosion Responses for the Clive DU PA Model
23 Feb 2018 44
C Factor (VM Factor)
The C parameters used in the equation for both the 4% and 20% slopes were based on Table 9
and Table 10 of the US Department of Agriculture Handbook Number 537 “Predicting Rainfall
Erosion Losses, A Guide to Conservation Planning” and are shown in the figures below. The C
factor for the top slopes (0.2) is based on the sparse vegetative cover naturally found in the areas
immediately surrounding the Clive facility (Table 10 of Handbook Number 537, No Appreciable
Canopy, Type G with 20% ground cover) and the Unit 4 gravel admixture. The C factor for the
side slope is based on the higher percentage of gravel in the side slope gravel admixture (50%
gravel). The 50% gravel admixture on the side slopes results in a pseudo-gravel mulch once
some of the fines have been removed. Therefore, a C factor of 0.02 was selected (Table 9
Handbook Number 537, Crushed Stone ¼ to 1 ½ in, Mulch Rate of 240 tons/acre, Land Slope of
21-33%, with a length limit of 200 feet).

Erosion Responses for the Clive DU PA Model
23 Feb 2018 45
TABLE 1
EFFECTS OF EROSION - AVERAGE SOIL LOSS ANALYSIS USING USLE
(TOP SLOPE WITH UNIT 4 CLAY W/ 15% GRAVEL ADMIXTURE AND SIDE
SLOPE WITH UNIT 4 CLAY AND 50% GRAVEL ADMIXTURE)

Slope Segment
R
(ft tons/ac/hr)
K
(tons/ac/EI)
L
(ft)
S
(%)
C A
(tons/ac/yr)
Total Soil
Loss
(mm/year)
Top Slope (4%) 6 0.18 942 (4%) 4% 0.2 0.25 0.24
overall 0.026
Side Slope (20%) 6 0.07 188 (20%) 20% 0.02 0.19

The Rangeland Hydrology and Erosion Model (RHEM) Model Parameters
The RHEM model was used in HAL’s interrogatory response entitled “DRC RFI Section 4.0
Erosion” completed in 2013 in order to corroborate the results of the USLE analysis. Since the
completion of that effort, the RHEM on-line model has been updated. The latest version of the
model (Version 2.3) lacks some of the functionality of the earlier model. More specifically,
slope lengths are not able to be directly input as they were previously and therefore a direct
comparison with the calculation completed using USLE is not possible. This appears to be an
error within the program since the descriptions of the model available on the website include
slope length as an input parameter option. Attempts to reach the contacts listed for the RHEM
model in order to resolve the issue were unsuccessful. The model defaults to a slope length of 50
meters (164 feet). Additional parameter options for cover characteristics appear to have been
added since 2013, including options for a distinction between foliar cover and basal plant cover
as well as rock cover, litter cover and biological soil crust cover. Since the tool is a web-based
model and not stand alone software, the version of the online tool used in the 2013 analysis is no
longer available. The parameters described in the 2013 interrogatory for vegetative conditions,
slope grades, climate station, soil texture class, and slope lengths are still valid for the analysis
completed at the time but the results reported therein can no longer be validated.

Literature Cited

Bray, D. I., 1979. Estimating Average Velocity in Gravel-Bed Rivers. Proc. Am. Soc. Civ. Engrs, J.
Hydraul. Div., 105 (HY9), 1103-1122.

Haan, C.T., Barfield, B.J., Hayes, J.C., 1994. Design Hydrology and Sedimentology for Small
Catchments. Academic Press, Inc, San Diego, California.

Israelsen, C. Earl, Fletcher, Joel E., Haws, Frank W., Kisraelsen, Eugene K., 1984. Erosion and
Sedimentation in Utah: A Guide for Control. Utah Water Research Laboratory, Logan,
Utah.
Johnson, T. L. 2002. Design of Erosion Protection for Long-Term Stabilization, Final Report,
NUREG-1623, U.S. Nuclear Regulatory Commission, Office of Nuclear Material Safety
and Safeguards, Washington, D.C.
Erosion Responses for the Clive DU PA Model
23 Feb 2018 46

Jones, Gordon L., 2012. PMF and 100-year Storm Analysis for Clive, Utah. Hansen, Allen &
Luce, Midvale, Utah.

Mazor, G., G.J. Kidron, A. Vonshak, and A. Albeliovich, 1996. The role of cyanobacterial
exopolysaccharides in structuring desert microbial crusts. FEMS Microbiology Ecology.

Nearing, M.A., H. Wei, J.J. Stone, F.B. Pierson, K.E. Spaeth, M.A. Weltz, D.C. Flanagan and M.
Hernandez, 2011. A Rangeland Hydrology and Erosion Model. Soil & Water Division
of ASABE in March 2011 Edition, Vol. 54(3).

Nelson, J. D., R. L. Volpe, R. E. Wardwell, S. H. Schumm, and W. P. Staub. 1983. Design
Considerations for Long-Term Stabilization of Uranium Mill Tailings Impoundments.
NUREG/CR-3397 (ORNL-5979), U.S. Nuclear Regulatory Commission, Washington,
D.C.

U.S. Army Corps of Engineers, 1977. Probable Maximum Precipitation Estimates, Colorado
River and Great Basin Drainages, Hydrometeorological Report No. 49. US Departement
of Army Corps of Engineers, Silver Spring, Maryland.

U.S. Department of Agriculture, 1978. Predicting Rainfall Erosion Losses, A Guide to
Conservation Planning, Handbook 537. U.S. Department of Agriculture, Washington
D.C.

US EPA Seminar Publication, 1991. Design and Construction of RCRA/CERCLA Final Covers.

U.S. Geological Survey, 1973. Aragonite NW, Utah; Aragonite, Utah; Hastings Pass, Utah; and
Low, Utah Quadrangle Maps.

U.S. Geological Survey, 1987. Bonneville Salt Flats, Utah, 30 x 60 Minute Quadrangle Map.

U.S. Nuclear Regulatory Commission, 2002. NUREG-1623 - Design of Erosion Protection for
Long-Term Stabilization. Office of Nuclear Material Safety and Safeguards,
Washington, DC.

Erosion Responses for the Clive DU PA Model
23 Feb 2018 47
Appendix B. HAL 2017a
Erosion Responses for the Clive DU PA Model
23 Feb 2018 48
DRC Interrogatory CR R313-25-25(4)-198/1
DRC Interrogatory Statement:
Please provide the design bases and justification for the amount and sizing of the gravel in the
top and side slopes of the Federal Cell. The proposal for the gravel admixture in the top slope
(15%) appears too small. Also, please provide evidence for existing semi-arid or arid sites
where only 15% gravel has been added to form a successful cover-system surface layer for a
landfill. Please describe actual analog sites where 50% gravel for side slopes has been
demonstrated to be effective against erosion.
HAL Response: December 18, 2017
The design basis for the amount of gravel to add to the soil in the top and side slopes were first
determined based on the analysis of the long term sustainability of the slopes due to sheet erosion
over time. The methods that were used are the Revised Universal Soil Loss Equation (RUSLE)
and the Rangeland Hydrology and Erosion Model (RHEM). These methods produce an estimate
of the average annual soil loss in terms of tons per acre per year. With the gravel admixed into
the Unit 4 clay soil, average annual soil losses were calculated to be about 0.24 tons/acre/year
using RUSLE which was also closely matched by the RHEM. This amount of loss is almost 10
times less than what has been recommended in EPA guidance for covers for hazardous waste
facilities. A more detailed discussion of the methodologies and results is provided in the
response to DRC Interrogatory CR R313-25-7(2)-191-3.
Once the slope was determined to be stable from an average soil loss perspective, checks on the
gully erosion potential were completed based on the calculation of predicted velocities and
comparing them to permissible velocity according to the method presented in NUREG-1623
“Design of Erosion Protection for Long-Term Stabilization”. This second assessment based on
an extreme intense rainfall event is recommended in NUREG-1623 because of the potential for
significant damage to cover systems from such large events. In fact, this extreme intense rainfall
event is known as the Probable Maximum Precipitation (PMP) and is so extreme that “the point
precipitation data base, even if maximized for PMP moisture potential, shows no observed
values even close to the 10-inch PMP estimate (the generalized, local storm PMP estimates in
Utah)” (Jensen, 1995). Additionally, the methodology applies a flow concentration factor (F)
which NUREG-1623 recommends to be 3. Therefore, the extreme event that is multiples of the
highest ever recorded rainfall intensity in the area is compounded by an additional safety factor.
The end result of this process is an analysis that is meant to produce a highly conservative value.
Flow velocities on the top and side slopes of the CAW embankment during the PMP event were
predicted to be 2.37 and 2.07 ft/sec, respectively, as discussed in more detail in the response to
DRC Interrogatory CR R313-25-7(2)-191-3. The acceptable Maximum Permissible Velocity
(MPV) was selected from tables provided in NUREG/CR-4620. Under this method the slope is
stable if the calculated velocity (V, the velocity resulting from a PMP in this application) is less
than the MPV. By contrast, if velocities exceed the MPV, the slope will experience excessive
erosion that will lead to the formation of gullies. HAL found that the calculated velocities
resulting from the PMP did not exceed the prescribed permissible velocity.
Reference Review
Erosion Responses for the Clive DU PA Model
23 Feb 2018 49
The methods discussed above have been published and/or accepted by the US EPA and US
Nuclear Regulatory Commission. Other methods referred to in the interrogatory, on the other
hand, have not been adopted, published or referred to in guidance by federal agencies. While the
works cited are part of productive ongoing academic research and study, they are not conclusive
in their findings and should instead be the focus of additional research. A summary of each is
provided below:
“Gravel Admixture for Erosion Protection in Semi-Arid Climates, Erosion of Soils and Scour of
Foundations” (Anderson, C., and J. Stormont, 2005) – This paper was published as part of the
proceedings of sessions of the GeoFrontiers 2005 Congress held in Austin, Texas. Along with
some other conference proceedings articles, these early attempts to define a process for
admixture design is very helpful in that it outlines practical steps to determine gravel size, gravel
percentage and admixture thickness. This approach, however, has limitations that result from the
lack of empirical evidence to back up the steps and applications of many of the equations
presented in the paper. The paper states that “the design method for gravel admixtures presented
here may serve as an outline for further erosion investigations and provide guidance for future
designs of gravel-soil admixture layers”. Therefore, the paper as written was not meant to be
used as a proven method for admixture design but rather a starting point for further research and
investigation.
“Design of Erosion Protection at Landfill Areas with Slopes Less than 10%” (Anderson, C., and
Wall, S., 2010) – This paper is very similar to the aforementioned paper by Anderson and
Stormont where it comes from the proceedings of a conference, in this case the International
Conference on Scour and Erosion held in 2010. Much of the background language of this paper
is the same as the 2005 article by Anderson and Stormont. Similarly, the methodology presented
is very similar to the steps provided in 2005 with some changes. Again, the paper does not claim
this to be a proven method but indicates it is taking steps to test its outcomes. It states that “the
procedure described here is being applied to a landfill cover soon to be constructed in southern
Nevada. The Nevada project will provide the first large scale application of the procedure.”
HAL was unable to find information regarding empirical data on a smaller scale nor was final
information on the results of the large scale Nevada project able to be found. Without this
information or additional discussion in the paper, the limits of the equation and how the
parameters of success or failure are defined remain unknown. For example, assuming the
Nevada project site is shown to be successful with the 40% gravel admixture there is nothing that
would indicate if a lower percentage of gravel admixture, such as 25%, would also be adequate.
Additionally, the suitability of the design and how success is defined is not sufficiently described
in the paper to know whether the absence of observed gully erosion constitutes a successful
design or if it is acceptable to have the formation of rills and gullies as long as the depth does not
exceed the admixture layer thickness calculated using the method. This is an important
distinction since the end result of the design process presented in the paper is the calculation of
the thickness of the admixture layer which is presented as being dependent on the percentage of
gravel in the admixture. This specific paper does not provide limits or conclusive guidance to
the reliability of the method and therefore should serve as the starting point for further
investigation and research.
Erosion Responses for the Clive DU PA Model
23 Feb 2018 50
“Erosion Protection at Landfill Slopes Greater than 10%” (Anderson, C. and Wall, S.) – This
paper focuses on riprap on side slopes. The design proposed by EnergySolutions does not
include riprap side slopes; therefore, this paper does not apply.
“Long-Term Cover Design for Low-Level Radioactive and Hazardous Waste Sites as Applied to
the Rocky Flats Environmental Technology Site Solar Evaporation Ponds” (Stenseng, S.E. and
Nixon, P.A., 1997) – The paper describing the design used at the Rocky Flats Environmental
Technology Site was included as part of the proceedings from the 50th Industrial Waste
Conference held in May 1997. The paper includes a brief description of the design basis for the
5% top slopes and 20% side slopes and the 40% gravel admixture. The admixture gravel content
appears to have been selected instead of calculated as the result of a defined process. While
instructive regarding the specific cover design discussed in the paper, no additional guidance is
provided to guide application at other sites.
“Ecology, Design, and Long-Term Performance of Waste-Site Covers: Applications at a
Uranium Mill Tailings Site” (Waugh, W.J. and Richardson, G.N., 1997) – This paper is similar
to the Stenseng and Nixon paper discussed previously in that it contains a summary of the design
parameters but does not provide a specific and applicable methodology for the design of a soil
gravel admixture with regard to erosion protection.
Sites with Top Slopes Comprised of 15% Gravel Admixture
There are some examples of sites that have employed slopes with 15% admixture and give some
measure of proof that a 15% admixture can be effective at controlling erosion. A test site was
established at Hanford and is referred to as the Hanford Prototype Surface Barrier (PSB). The
purpose of the PSB test site was to evaluate surface barrier constructability, construction costs,
and physical and hydrologic performance at field scale (US Department of Energy, 2016). This
field-scale test cover was installed in 1994 and is comprised of a top erosion protection layer
made up of a silt loam admixed with 15% pea gravel. The top slope of 2% in this case is slightly
less than the top slopes of 2.5% and 4% proposed by EnergySolutions but is an example of a
cover system that relies on the same amount gravel admixture. A recent review completed by
the US Department of Energy in 2016 of the data collected at the site through 2015 concluded
that:
“The 19-year PHB record showed practically no evidence of wind or water erosion of the
ETC barrier, despite 3 years of triple the mean annual precipitation; three simulated
1000-year-return, 24-hour precipitation events; and an intense, controlled fire that
burned off all vegetation across half the barrier surface….Even in the absence of
vegetation (e.g., following a fire), the pea gravel added to the silt loam protected the
barrier surface from wind and water erosion….Overall, the monitoring results have
confirmed that the PHB design is resistant to water and wind erosion and that resistance
is expected throughout the barrier’s 1000-year design life”
Another test site using the same cover system was established at Hill Air Force Base in 1994
though little information has been published about that site to date.
Sites with Side Slopes Comprised of 50% Gravel Admixture
Erosion Responses for the Clive DU PA Model
23 Feb 2018 51
No sites were found that have used a gravel admixture on side slopes at or above 20%.
Similarly, there were no methodologies found that specifically address the calculation of gravel
admixtures for slopes greater than 10% other than the general methods found in NUREG-1623
and NUREG/CR-4620. For slopes over 9%, Simanton et al. found that the rate of water erosion
diminishes exponentially with increasing cover of rock fragments. The effect of biological soil
crust is also difficult to quantify but is expected to establish itself along the slopes as has been
observed along natural slopes in the Clive area. The observation of natural analogs in the Clive
vicinity could prove helpful in demonstrating the stability of steeper slopes although none would
be available with such a high gravel content (50%) as the proposed side slope design. The
required lengthy post-closure care period will perhaps provide the greatest opportunity for
verification of the design methodology described previously.
The design process that was completed to determine the acceptability of the 50% gravel
admixture on the 20% side slopes is in accordance with EPA and Nuclear Regulatory
Commission guidance. There was no requirement found in the guidance procedures that existing
sites be used as evidence to support proposed designs. While it is acknowledged that such
evidence would be helpful, each site is different and contains unique design constraints based on
physical layout, climate, material availability and project goals that make determinations based
on sound methodologies necessary.

Literature Cited

Anderson, C., and J. Stormont, 2005. Gravel Admixtures for Erosion Protection in Semi-Arid Climates,
Erosion of Soils and Scour of Foundations. Proceedings of Sessions of the Geo-Frontiers 2005
Congress, Austin, Texas, 1-12.

Anderson, C., and S. Wall, 2010. Design of Erosion Protection at Landfill Areas with Slopes Less than
10%. Scour and Erosion, Geotechnical Special Publication (GSP) No. 210, ed. S.E. Burns, S.K.
Bhatia, C.M.C. Avila, and B.E. Hunt, 1054-1063, American Society of Civil Engineers,
Proceedings of the 5th Annual Conference on Scour and Erosion (ICSE-5) 2010, November 7-10,
2010, San Francisco, California, United States.

Anderson, C. and S. Wall, 2010. Erosion Protection at Landfill Slopes Greater Than 10%. Scour and
Erosion, Geotechnical Special Publication (GSP) No. 210, ed. S.E. Burns, S.K. Bhatia, C.M.C.
Avila, and B.E. Hunt, 1064-1073, American Society of Civil Engineers, Proceedings of the 5th
Annual Conference on Scour and Erosion (ICSE-5) 2010, November 7-10, 2010, San Francisco,
California, United States.

Bray, D. I., 1979. Estimating Average Velocity in Gravel-Bed Rivers. Proc. Am. Soc. Civ. Engrs, J.
Hydraul. Div., 105 (HY9), 1103-1122.

Haan, C.T., Barfield, B.J., Hayes, J.C., 1994. Design Hydrology and Sedimentology for Small
Catchments. Academic Press, Inc, San Diego, California.

Jensen, Donald T., 1995. Probable Maximum Precipitation Estimates for Short-Duration, Small-
Area Storms in Utah. Utah Climate Center, Utah State University, Logan, Utah.

Erosion Responses for the Clive DU PA Model
23 Feb 2018 52
Johnson, T. L. 2002. Design of Erosion Protection for Long-Term Stabilization, Final Report,
NUREG-1623, U.S. Nuclear Regulatory Commission, Office of Nuclear Material Safety
and Safeguards, Washington, D.C.

Nelson, J. D., S. R. Abt, R. L. Voipe, D. van Zyl, N.E. Hikle, and W. P. Staub. 1986.
Methodologies for Evaluating Long-Term Stabilization Designs of Uranium Mill
Tailings Impoundments. NUREG/CR-4620, U.S. Nuclear Regulatory Commission,
Washington, D.C.

Simanton, J.R., E. Rawitz and E.D. Shirley, 1984. Effects of Rock Fragments on Erosion of
Semiarid Rangeland Soils. SSSA Special Publication No. 13 Soils Science Society of
America, Madison, Wisconsin.

Stenseng, S.E. and P.A. Nixon. Long-Term Cover Design for Low-Level Radioactive and
Hazardous Waste Sites as Applied to the Rocky Flats Environmental Technology Site
Solar Evaporation Ponds. Wukasch, R.F., 1995, Proceedings of the 50th Industrial Waste
Conference May 8-10, Purdue Research Foundation, Office of Technology Transfer,
West Lafayette, Indiana.

United States Environmental Protection Agency, 1989. Technical Guidance Document: Final
Covers on Hazardous Waste Landfills and Surface Impoundments. Office of Solid Waste
and Emergency Response, Washington, D.C.

Waugh, W.J. and G.N. Richardson, 1997. Ecology, Design and Long-Term Performance of
Surface Barriers: Applications at a Uranium Mill Tailings Site. Committee on
Remediation of Buried and Tank Wastes, National Research Council, eds. Barrier
Technologies for Environmental Management: Summary of a Workshop, The National
Academies Press, Washington, D.C.

Erosion Responses for the Clive DU PA Model
23 Feb 2018 53

Appendix C. HAL 2017b

Erosion Responses for the Clive DU PA Model
23 Feb 2018 54
DRC Interrogatory CR R313-25-25(4)-201/1
DRC Interrogatory Statement:
It is not clear that the Probable Maximum Precipitation (PMP) was determined using the
procedures outlined in the National Oceanic and Atmospheric Administration and U.S. Army
Corps of Engineers publication Hydrometeorological Report No. 49 (HMR 49) (1977).
According to EnergySolutions, these procedures resulted in a “1-hour PMP rainfall intensity
of 9.9 inches (Jones, 2012).” However, DWMRC finds that a value of 9.8 or 9.9 inches is not
the intensity, but rather the 1-hour PMP, or the maximum precipitation expected over 1
square mile when averaged over an hour.
Please re-calculate the PMP using NUREG/CR-4620 as outlined below.
HAL Response: December 19, 2017
Gully erosion potential was initially checked based on the calculation of permissible velocities
according to the method presented in NUREG-1623 “Design of Erosion Protection for Long-
Term Stabilization”. As pointed out in the Interrogatory Statement, it is acknowledged that the
methodology utilized previously failed to incorporate the proper rainfall intensity as outlined in
NUREG/CR-4620.
In order to determine the PMP intensity as outlined in the guidance in NUREG/CR-4620 it is
necessary to first calculate the time of concentration for representative drainage areas for both
the top and side slopes. The methodology outlined in Technical Release 55 “Urban Hydrology
for Small Watersheds” was used to calculate the time of concentration for each representative
slope.
TR-55 describes three types of drainage flow: sheet flow, shallow concentrated flow, and open
channel flow. Sheet flow is defined as flow over planar surfaces at very shallow depths for up to
300 feet. After a maximum of 300 feet, the flow transitions to shallow concentrated flow. Open
channels were not included in the calculations due to the absence of designed channels. The
same Manning’s roughness coefficient calculated to be 0.05 for use in the NUREG-1623
methodology using an empirical equation for channels with gravel beds with shallow flow depths
(Bray, 1979) was also used to represent the sheet flow roughness. It was decided to use the value
calculated using the Bray method (0.05) instead of a higher value found in other publications for
overland flow using sparse vegetative cover in order to be more protective. The above described
methodology was applied to both the Class A West Cell and the Federal Cell.
CALCULATED TIME OF CONCENTRATION USING TR-55 METHODOLOGY
FOR CLASS A WEST CELL

Slope Description Total
Length
(ft)
Slope
(ft/ft)
Sheet
Flow
Length
(ft)
Sheet Flow
Manning’s n
Shallow
Concentrated
Flow Length
(ft)
Total Time of
Concentration
(min)
Top Slope (4%) 942 0.04 300 0.05 642 17.4
Side Slope (20%) 188 0.20 188 0.05 0 4.6
CALCULATED TIME OF CONCENTRATION USING TR-55 METHODOLOGY
Erosion Responses for the Clive DU PA Model
23 Feb 2018 55
FOR FEDERAL CELL

Slope Description Total
Length
(ft)
Slope
(ft/ft)
Sheet
Flow
Length
(ft)
Sheet Flow
Manning’s n
Shallow
Concentrated
Flow Length
(ft)
Total Time of
Concentration
(min)
Top Slope (2.5%) 521 0.025 300 0.05 221 16.9
Side Slope (20%) 178.5 0.20 178.5 0.05 0 4.4

The rainfall depth determined using the methods outlined in Table 6.3a of the US Army Corps of
Engineers publication Hydrometeorological Report No. 49 (HMR 49) for determining an average
1-hour 1 square mile PMP is 9.9 inches. NUREG/CR-4620 gives the following steps for
determining rainfall intensity from the PMP:
PMP rainfall depth = (% PMP) x (PMP)

Where:
PMP rainfall depth = rainfall depth after adjusting the PMP for the duration, inches
% PMP = the region-specific % PMP based on the duration, %
PMP = Probable Maximum Precipitation, inches (9.9 inches using Table 6.3a from HMR 49)

The % PMP was obtained from a report commissioned by the State of Utah Division of Water
Resources and Division of Water Rights, State Engineer’s Office entitled “Probable Maximum
Precipitation Estimates for Short-Duration, Small-Area Storms in Utah” (Jensen, 1995). The
report states that “because the depth-duration values published in HMR 57 were very much like
those derived in this study, the published HMR 57 depth-duration values were accepted for
operational use in Utah.” Therefore, the % PMP was determined using the Utah-specific depth-
duration found in Figure 8 (Jensen, 1995) and HMR 57 instead of the general depth-duration
relationships described in HMR 49 and NUREG/CR-4620.
Rainfall intensity (i) for the purpose of use in the rational equation is then calculated as:
i = PMP rainfall depth x 60 / rainfall duration

Where:
i = rainfall intensity, inches/hour
rainfall duration = time of concentration calculated as described above, minutes
The results of the above calculation is summarized in the table below.

Erosion Responses for the Clive DU PA Model
23 Feb 2018 56
PMP RAINFALL INTENSITY
CLASS A WEST CELL

Slope Description Rainfall
Duration (Tc)
(minutes)
% PMP
(%)
Rainfall
Intensity (i)
(inches/hr)
Top Slope (4%) 17.4 53.8% 18.4
Side Slope (20%) 4.6 15.3% 19.8

PMP RAINFALL INTENSITY
FEDERAL CELL

Slope Description Rainfall
Duration (Tc)
(minutes)
% PMP
(%)
Rainfall
Intensity (i)
(inches/hr)
Top Slope (2.5%) 16.9 52.1% 18.3
Side Slope (20%) 4.4 14.7% 19.8

As opposed to the projection of the long-term effects of precipitation over time due to sheet
erosion, the effects of gully erosion are determined by the consideration of a large single rainfall
event. The equation given in NUREG-1623 for the procedure to determine the peak runoff flow
rate is:
Q = Fci A

Where:
Q = Runoff Rate, cfs/ft
F = Flow concentration factor (recommended to use a factor of 3 by NRC staff in NUREG-1623)
c = dimensionless runoff coefficient
i = rainfall intensity, in/hr
A = catchment area, acres (using a 1 foot wide strip along the length of the slope)
Using this flow rate, a flow depth is calculated by solving the Manning Equation for normal
depth on a one foot wide strip along the slope length. The Manning’s n value was calculated to
be 0.05 using an empirical equation for channels with gravel beds with shallow flow depths
(Bray, 1979). The derivation of the Manning Equation to solve for depth is given in NUREG-
1623 as:
y5/3 = Qn / (1.486 S1/2)
and V = Q/y where V is the flow velocity in ft/sec
The results for both the top slope and the side slope of the Class A West Cell using the vegetated
slope condition are summarized in the following table. Flow velocities on the top and side
slopes of the Class A West Cell during the PMP event are predicted to be 2.37 and 2.07 ft/sec,
Erosion Responses for the Clive DU PA Model
23 Feb 2018 57
respectively. The acceptable Maximum Permissible Velocity (MPV) was selected from tables
provided in NUREG/CR-4620. The permissible velocity method is a commonly applied method
to determine channel stability. Under this method the slope is assumed stable if the calculated
velocity (V, the velocity resulting from a PMP in this application) is less than the MPV. By
contrast, if velocities exceed the MPV, it is assumed that the slope will experience excessive
erosion that will lead to the formation of gullies. The methodology presented in NUREG-1623
then directs that the estimates for the MPV be adjusted downward to account for the influences
of flow depth. The side slope gully analysis was completed independently of the top slope.
GULLY EROSION POTENTIAL – VELOCITY ANALYSIS
CLASS A WEST CELL

Slope Description Length
(ft)
Slope
(ft/ft)
i
(in/hr)
c Q
(cfs/ft)
y
(ft)
V
(ft/sec)
Adjusted
MPV
(ft/sec)
Top Slope (4%) 942 0.04 18.4 0.5 0.60 0.25 2.37 2.5
Side Slope (20%) 188 0.20 19.8 0.5 0.13 0.06 2.07 2.5

Therefore, all slope scenarios using the ET cover system are assumed to provide acceptable
protection against gully erosion using these criteria by limiting the potential of gully formation
from high velocity channelization.

Literature Cited

Bray, D. I., 1979. Estimating Average Velocity in Gravel-Bed Rivers. Proc. Am. Soc. Civ. Engrs, J.
Hydraul. Div., 105 (HY9), 1103-1122.

Haan, C.T., Barfield, B.J., Hayes, J.C., 1994. Design Hydrology and Sedimentology for Small
Catchments. Academic Press, Inc, San Diego, California.

Jensen, Donald T., 1995. Probable Maximum Precipication Estimates for Short-Duration, Small-
Area Storms in Utah. Utah Climate Center, Utah State University, Logan, Utah.

Johnson, T. L. 2002. Design of Erosion Protection for Long-Term Stabilization, Final Report,
NUREG-1623, U.S. Nuclear Regulatory Commission, Office of Nuclear Material Safety
and Safeguards, Washington, D.C.

Jones, Gordon L., 2012. PMF and 100-year Storm Analysis for Clive, Utah. Hansen, Allen &
Luce, Midvale, Utah.

Nelson, J. D., S. R. Abt, R. L. Voipe, D. van Zyl, N.E. Hikle, and W. P. Staub. 1986.
Methodologies for Evaluating Long-Term Stabilization Designs of Uranium Mill
Tailings Impoundments. NUREG/CR-4620, U.S. Nuclear Regulatory Commission,
Washington, D.C.

Erosion Responses for the Clive DU PA Model
23 Feb 2018 58
U.S. Army Corps of Engineers, 1977. Probable Maximum Precipitation Estimates, Colorado
River and Great Basin Drainages, Hydrometeorological Report No. 49. US Departement
of Army Corps of Engineers, Silver Spring, Maryland.

U.S. Army Corps of Engineers, 1994. Probable Maximum Precipitation – Pacific Northwest
States Columbia River (including portions of Canada), Snake River and Pacific Coastal
Drainages, Hydrometeorological Report No. 57. US Department of Army Corps of
Engineers, Silver Spring, Maryland.

U.S. Department of Agriculture, 1986. Urban Hydrology for Small Watersheds, Technical
Release 55. Natural Resources Conservation Service, Conservation Engineering
Division, Washington D.C.

U.S. Nuclear Regulatory Commission, 2002. NUREG-1623 - Design of Erosion Protection for
Long-Term Stabilization. Office of Nuclear Material Safety and Safeguards,
Washington, DC.

NAC-0105_R0

Deep Time Supplemental Analysis
Responses for the Clive DU PA
Model
23 February 2018

Prepared by

NEPTUNE AND COMPANY, INC.
1435 Garrison St, Suite 201, Lakewood, CO 80215
Deep Time Supplemental Analysis Responses for the Clive DU PA Model
23 Feb 2018 ii
1. Title: Deep Time Supplemental Analysis Responses for the Clive DU PA Model
2. Filename: Deep Time Supplemental Analysis Responses for the Clive DU PA Model.docx
3. Description: Summary of how the Deep Time model represents future glacial cycles and
associated lake sedimentation in the Lake Bonneville basin, with a focus on the
representation of aeolian deposition and intermediate lake sedimentation.

Name Date
4. Originator Bruce Crowe 1 Feb 2018
5. Reviewer Dan Levitt
6. Remarks

Deep Time Supplemental Analysis Responses for the Clive DU PA Model
23 Feb 2018 iii
CONTENTS
CONTENTS ............................................................................................................................. iii
FIGURES .................................................................................................................................. iv
TABLES ..................................................................................................................................... v
ACRONYMS AND ABBREVIATIONS ................................................................................... vi
1.0 Overview and Conceptual Model........................................................................................ 1
1.1 Aeolian Deposition ....................................................................................................... 1
1.2 Intermediate Lake Sedimentation .................................................................................. 1
2.0 UDEQ Interrogatory Responses.......................................................................................... 3
2.1 Interrogatory CR R313-25-8(5)(A)-18/3: Sediment Accumulation Part One—Aeolian
Deposition .................................................................................................................... 3
2.1.1 Interrogatory Response ............................................................................................ 4
2.1.1.1 Factual Accuracy and Technical Applicability of the UDEQ Review
Comments ...................................................................................................... 4
2.1.1.2 The Effects of the Aeolian Deposition Rate on the Deep Time Model Results
and Uncertainty in the Results ........................................................................ 6
2.2 Interrogatory CR R313-25-8(5)(A)-18/3: Sediment Accumulation Part Two—
Intermediate Lake Sedimentation Rate .......................................................................... 6
2.2.1 Interrogatory Response ............................................................................................ 6
2.2.1.1 Intermediate Lake Sedimentation .................................................................... 6
2.2.1.2 Expanded Documentation of Lake Sedimentation Rates ................................. 8
2.2.1.3 Evidence of Clastic Sedimentation at the Clive Site ...................................... 10
2.2.1.4 Impact of Intermediate Lake Sedimentation on the Deep Time Model Results
..................................................................................................................... 10
3.0 Conclusion ....................................................................................................................... 11
4.0 References ........................................................................................................................ 13
Appendix A. Sedimentary Deposits of the Transgressive and Regressive Phases of Lake
Bonneville Near and Within the Clive Site (Deposits of Intermediate Lakes) .................... 15

Deep Time Supplemental Analysis Responses for the Clive DU PA Model
23 Feb 2018 iv
FIGURES
Figure 1. Deep Time ground surface flux of Rn-222 per NRC Reg. Guide 3.64. ........................ 11

Figure A-1. Roadcut exposures of the upper contact of the transgressive phase of Lake
Bonneville. .............................................................................................................. 16
Figure A-2. Annotated Google Earth image showing the Clive area and buried spits
developed on the southeast edges of the Grayback Hills. The spits prograded
toward and across what is now the Clive Site. .......................................................... 17
Figure A-3. (From Neptune (2015a).) View looking south from the north wall of Pit 29 at the
Clive Site. The dashed line marks the contact between the laminated marl of Lake
Bonneville (deep lake) and underlying transgressive sands with lenses of volcanic
conglomerate (deposits of an intermediate lake). ...................................................... 18
Figure A-4. (From Neptune (2015a).) Gully exposure of a lens of sandy conglomerate from
the outcrop shown in Figure A-3. The conglomerate contains distinctive clasts of
black volcanic rock derived from bedrock exposures in the Grayback Hills,
northwest of the Clive Site. The dashed line is the base of the laminated marl of
Lake Bonneville (deep lake). The gravel lens is here approximately 1 meter thick.
It is correlative with the Lake Bonneville Transgressive unit (0.76 m thick) listed
in Table 3 of Neptune (2015b). ................................................................................ 18
Figure A-5. (From Neptune (2015a).) Intermediate lake deposits of the regressive phase of
Lake Bonneville exposed in the north quarry wall of the Clive Site. The lake
deposits are exposed in the interval extending from the top of the hammer head to
the middle of the handle and are overlain by aeolian silt with local reddening from
soil alteration. .......................................................................................................... 19

Deep Time Supplemental Analysis Responses for the Clive DU PA Model
23 Feb 2018 v
TABLES
Table 1. Compilation of Lake Sedimentation Rates. .................................................................... 9

Deep Time Supplemental Analysis Responses for the Clive DU PA Model
23 Feb 2018 vi
ACRONYMS AND ABBREVIATIONS
DU depleted uranium
ka kilo-annum (1,000 years)
UDEQ Utah Department of Environmental Quality (also referred to in quoted
interrogatory text as “DEQ/SC&A,” reflecting authorship of DEQ’s consultant,
SC&A, Inc.)

Deep Time Supplemental Analysis Responses for the Clive DU PA Model
23 Feb 2018 1
1.0 Overview and Conceptual Model
1.1 Aeolian Deposition
During warm interglacial climates like the present, the predominant process of sedimentation at
the Clive Site is aeolian. (Note: the terms “aeolian” and “eolian” are used interchangeably in the
geological literature and refer to wind-generated sedimentary processes. The names are derived
from Aeolus, the Greek god of wind.) Aeolian source materials are transported and deposited by
prevailing winds sweeping across barren dry lake beds and the wetted fringes of playas where
gypsum crystals are supplied continuously by evaporation above sites of shallow groundwater.
Aeolian sedimentation occurs in two modes: saltation or bouncing of sand-sized particles
forming dune landforms, and suspension fallout that produces a near-uniform blanket of fine-
grained silt deposited at the ground surface.
The primary mode of aeolian deposition at the Clive Site is suspension fallout. The average
thickness of these deposits is about 72 cm (Neptune 2015a). The age of the aeolian silt deposits
at the Clive Site is estimated to be between 14,500 and 10,000 years before present, based
primarily on the estimates of the timing of final retreat of Lake Bonneville below the elevation of
the Clive Site (Neptune 2015a). Data for the thickness variations and ages of aeolian sediments
at the Clive Site are used to develop a distribution of the range of past aeolian sedimentation
rates (mean rate of 59 mm per 1,000 years with a standard error of the mean of 5 mm). This
distribution is used in the Deep Time model to project thicknesses of aeolian sediments over
time. The sediment thickness numbers are calculated in the model by multiplying the aeolian
deposition rates by the time when a future lake rises to the elevation of the Clive Site. Using
computer simulation, the Deep Time model samples a range of values of the depositional rate
and the time to the first lake arrival. A longer time to the first lake arrival and a higher aeolian
deposition rate result in a thicker blanket of aeolian deposits over the Clive Site, which in turn
reduces the Deep Time model projections of the peak ground surface radon flux. The peak radon
flux is the critical model output for assessing the Deep Time model results.
UDEQ agrees with the thickness estimates and time constraints for aeolian deposition in the
Clive area (see Interrogatory CR R313-25-8(5)(A)-18/3: Sediment Accumulation). They did not
comment on and presumably accept the model representation of the future time of the arrival of
the first lake. They disagree with the approach used to estimate the statistical variation in the
aeolian deposition rate (standard error of the mean; see Interrogatory CR R313-25-8(5)(A)-18/3:
Sediment Accumulation).
1.2 Intermediate Lake Sedimentation
The Lake Bonneville basin of Utah and Nevada has been occupied by fresh to saline lakes
ranging in size from the scale of Lake Bonneville during the Last Glacial Maximum to the
present day Great Salt Lake. Future glacial cycles will lead to the rise of the Great Salt Lake
above the elevation of the Clive Site. A glacial cycle refers to climate intervals associated with
long-term reduction in global temperature, reduction in atmospheric CO2, and growth of ice
sheets and glaciers with falling sea levels. An interglacial cycle refers to warmer periods between
glacial cycles with sea levels at or near current levels. The earth has been in an interglacial cycle
Deep Time Supplemental Analysis Responses for the Clive DU PA Model
23 Feb 2018 2
since the Holocene epoch, a subdivision of geologic time representing the last 12 ka years before
the present (ka is kilo-annum or 1,000 years).
The Deep Time model (Neptune 2015b) uses a combination of past patterns of glacial and
interglacial cycles over the last 2 million years, global ice core and oxygen isotopic data, and the
history of past lake cycles in the Lake Bonneville basin to develop model projections of future
lake cycles in the Clive region. These model projections are closely tied to a widely accepted
100 ka length of global glacial cycles during the last 1 million years (Crowe et al. 2017; Lisiecki
and Raymo 2005). Field observations and drill core studies of lake deposits from two locations in
the Lake Bonneville basin are used to correlate observed lake sediments to lake patterns during
glacial cycles.
Three processes of sedimentation will affect the Clive Site during future glacial and interglacial
cycles, and the integrated processes will result in progressive burial of the Site through time.
These processes are aeolian deposition, deep lake sedimentation, and intermediate lake
sedimentation.
Aeolian deposition: During warm interglacial cycles like present-day conditions, small lakes
like the modern Great Salt Lake are maintained (lake elevation about 1280 meters). Under these
conditions, the Clive Site (elevation about 1307 meters) is at the surface, and sedimentation at
the Site consists of deposition of aeolian silt as suspension fall-out from surface winds.
Deep lake sedimentation: During past glacial conditions, the Clive Site has been covered by
large/deep lakes with lake surface elevations ranging from about 1360 meters (Stansbury
Shoreline) to about 1620 meters (Bonneville Shoreline). These elevations are 53 to more than
300 meters above the elevation of the Clive Site. Sedimentation at the Site under these deep lake
conditions is primarily from slow precipitation of carbonate with secondary quantities of fine-
grained silt. The silt content of deep lake sediments varies depending on lake currents, distance
to shorelines or surface outcrops, and fluvial (stream) sources of sediment. The Deep Time
model uses a distribution of deep lake sedimentation rates that range from about 0.075 to
0.20 mm/yr (Neptune 2015b). The Clive Site is adjacent to areas of higher topography that are
likely to be local sources of clastic sediments during future deep lake cycles.
Intermediate lake sedimentation: Intermediate lakes are defined with respect to the elevation
of the Clive Site. They include lakes that rise to and slightly above the elevation of the Site.
Intermediate lakes transition to deep lakes when the water depth exceeds the depth of wave
agitation (arbitrarily defined as approximately 10 meters above the elevation of the Clive Site).
Thus defined, intermediate lakes form during periods of fluctuating lake levels and also during
transitory phases of transgressive and regressive deep lakes. Intermediate lake sedimentation is
affected by wave activity and sediment transport associated with wave activity (longshore drift).
Wave-driven currents transport clastic sediments. (Note: clastic sediments are defined as material
that is mechanically or chemically fragmented and transported/deposited by sedimentary
processes.) Intermediate lake sedimentation includes carbonate precipitation, but carbonate
deposits tend to be secondary in volume to clastic sediments in areas of significant clastic
sediment supply and/or wave activity.
Deep Time Supplemental Analysis Responses for the Clive DU PA Model
23 Feb 2018 3
Sedimentation associated with intermediate lakes in the Deep Time model is represented as a
sediment thickness per intermediate lake event. UDEQ argues that the basis for representation of
intermediate lake events is inadequately documented and that the depositional rates for the
intermediate lake events may be too high. Overestimation of depositional rates could lead to
Deep Time model projections that overestimate the depth of burial of the Clive Site over time.
2.0 UDEQ Interrogatory Responses
This section responds to Interrogatories CR R313-25-8(5)(A)-18/3 and CR R313-25-8(4)(D)-
132/2. As noted by UDEQ in Interrogatory 132/2, the issues raised therein “are similar to some
of the Interrogatory 18 responses. To reduce redundancy, consideration should be given to
closing this interrogatory.” Accordingly, this response is focused on the two main issues raised in
Interrogatory 18/3: aeolian deposition and intermediate lake sedimentation rate. The latter issue
covers the concerns raised in Interrogatory 132/2.
2.1 Interrogatory CR R313-25-8(5)(A)-18/3: Sediment Accumulation
Part One—Aeolian Deposition
After much exposition on the subject of aeolian deposition under the heading DEQ Critique of
DU PA v1.4, Appendix 13, this interrogatory states:
In conclusion, the above discussion presented four concerns that DEQ/SC&A has identified with
the EnergySolutions/Neptune aeolian deposition model. In order of perceived importance, these
are:
1) Deposition in the sub-areas of the embankment is likely correlated, rather than independent. A
correlated model would produce results that are more conservative than the current
EnergySolutions/Neptune model. However, the degree of correlation is presently unknown (and
perhaps unknowable).
2) The sample results do not represent a “point in time,” as EnergySolutions/Neptune indicated in
their previous response. Rather, the samples are an accumulation over 13,000 to 15,000 years
(Appendix 13, p. 38). Thus, the sample results can be thought of as being time averages.
3) Using the EnergySolutions/Neptune model results in a dose calculation means that the dose
receptor spends an equal amount of time in each embankment sub-area. The more conservative
assumption is that the dose receptor spends all of his time in the sub-area with the least amount
of deposition. Alternatively, the sub-area in which the dose receptor spends his time could be
randomly selected.
4) Dividing the embankment into 11 sub-areas based on the number of samples, as was done for
the EnergySolutions/Neptune model, appears reasonable. However, EnergySolutions/Neptune
should provide the rationale for selecting this approach.
DEQ/SC&A continues to believe that for nuclear licensing purposes the mean and standard
deviation aeolian deposition model should be used.
Deep Time Supplemental Analysis Responses for the Clive DU PA Model
23 Feb 2018 4
2.1.1 Interrogatory Response
2.1.1.1 Factual Accuracy and Technical Applicability of the UDEQ Review
Comments
Each of the four concerns outlined by UDEQ is addressed individually below.
Concern 1: UDEQ has questioned the approach used to select sub-areas for the embankment and
has suggested that the number of sub-areas should be established initially and used to identify
collection sites for each sub-area. They argue that sub-areas of the embankment are correlated
rather than independent, and that a correlated model would produce more conservative results
than the current Deep Time model. (Note: we assume “conservative” means thinner aeolian
deposits that would allow higher ground-surface radon flux in the Deep Time model results.)
Response to Concern 1
The Clive Site was not divided into sub-areas. UDEQ has incorrectly equated “sub-areas” with
field measurement sites. Field measurement sites for aeolian sediments at the Clive Site met two
selection criteria (Neptune (2015a), Section 2.1.4.1):
1. Contained exposures of the base of the aeolian sediments and the underlying sequence of
Lake Bonneville lake deposits.
2. Preserved the top surface of aeolian deposits to ensure the deposits were not changed by
post-deposition erosion or construction activities at the Site.
Additionally, the measurement data for the aeolian sediments were designed to be used in the
Deep Time model. The model requires thickness data for the aeolian sediments at the scale of the
DU disposal cell—the model does not calculate aeolian deposition rates for sub-areas. Field data
that met the two selection requirements were collected throughout the Clive Site (see Figure 8
and Table 2 in Neptune (2015a)).
Concern 2: UDEQ has argued that the field measurements of the thickness of aeolian sediments
across the Site do not represent a “point in time” as inferred in the Deep Time report. They
suggest instead that the sample sites are time integrated averages of aeolian deposition over
13,000 to 15,000 years.
Response to Concern 2
We agree in principle with the review comment. However, the Deep Time report did not claim
that the thickness of aeolian sediments was a “point in time.” UDEQ’s confusion may result from
an incorrect interpretation of the last paragraph of Section 4.0 of the Deep Time supplemental
model (Neptune (2015b), Section 7.1):
Therefore, the thickness of residual embankment material and sediment overlying the
disposed DU waste at the time when the first intermediate lake recedes will be effectively
equivalent to the thickness of aeolian sediments deposited up until that point in time,
represented by the rising elevation of the surrounding grade (emphasis added).
Deep Time Supplemental Analysis Responses for the Clive DU PA Model
23 Feb 2018 5
Here the “point in time” refers to the time of return of the first lake to the Clive elevation, not
processes of aeolian sedimentation. Aeolian sediments at the Clive Site represent an integrated
response of multiple processes (aeolian deposition and erosion, soil formation) occurring over
time at the spatial scale of the Clive Site. This is described in Section 5.1 of Neptune (2015c):
The preserved record of eolian deposition at the Clive site integrates variations in eolian
parameters and processes of deposition and burial over thousands of years. This integrated
record is assumed to provide the most consistent information for forecasting eolian
depositional rates over tens of thousands of years.
The Deep Time model uses a probability distribution of aeolian depositional rates projected
forward in time. The development of the probability distribution for the aeolian depositional rate
is described in Neptune (2015a), and its application is analogous to the spatial averaging of
aeolian sediment thickness across the Clive Site discussed in the response to Concern 1 above.
Concern 3: UDEQ argues that the aeolian deposition rate and Deep Time model results used
multiple sub-areas and should have included implicit assumptions about the location and
exposure time of a dose receptor in each sub-area.
Response to Concern 3
UDEQ’s objection is an improper interpretation of the radon flux and the Deep Time model
results. Sub-areas were not used in the Deep Time model. More importantly, the ground surface
radon flux has dimensions of area and time (pico-Curies per square meter per second). This
performance metric is not a dose and does not include or require a receptor or a receptor location.
Concern 4: UDEQ has requested that a rationale be provided for dividing the embankment into
11 sub-areas.
Response to Concern 4
The rationale for selection of measurement sites for aeolian sediments is described in Neptune
(2015a) and is repeated in the response to Concern 1 above (availability of suitable outcrops and
the scale of representation of the Clive Site in the Deep Time model). The probability
distribution used for the aeolian depositional rate in the Deep Time model is based on field data
from the Clive Site (11 field measurement sites and 21 clay resource test pits). These data
provide a statistically representative sample of the population of thickness data for aeolian
deposits across the Clive Site.
The Deep Time model requires upscaling of the field measurement data to the model scale, a
technically appropriate and established practice in environmental modeling studies (for example,
Blöschl and Sivapalan (1995); Neuman et al. (2003); Zhang et al. (2004)). The field
measurement scale of aeolian sediment thickness is applicable for areas of several square meters
(individual measurement site); the application of the aeolian rate calculation in the Deep Time
model is the DU disposal unit, a scale of thousands of square meters. The technically established
statistic for upscaling is the mean thickness of the field measurement and the standard error of
the mean to represent the variance of the averaged measurement data. Using the standard
deviation of the measurement data as suggested by UDEQ would incorrectly distort the variance
structure (representation of the variation of the measurement sites) in the Deep Time model
application.
Deep Time Supplemental Analysis Responses for the Clive DU PA Model
23 Feb 2018 6
2.1.1.2 The Effects of the Aeolian Deposition Rate on the Deep Time Model
Results and Uncertainty in the Results
UDEQ/SC&A have noted that differences between the standard error of the mean versus the
standard deviation of the measurements have a small impact on the average model results
(Interrogatory CR R313-25-8(5)(A)-18/3: Sediment Accumulation, p. 13). We agree. The
important Deep Time model result is the ground surface radon flux through time. This model
result is sensitive to the following: (a) the sub-model used to estimate the ground surface radon
flux, (b) the time to the return of the next lake at the elevation of the Clive Site, (c) the aeolian
deposition rate, and (d) lake sedimentation rates during glacial cycles. UDEQ comments critique
only the variance of the aeolian depositional rate, and we have demonstrated that their comments
are not applicable to the data used for aeolian sedimentation in the Deep Time model.
2.2 Interrogatory CR R313-25-8(5)(A)-18/3: Sediment Accumulation
Part Two—Intermediate Lake Sedimentation Rate
After discussing intermediate lake sedimentation rates under the heading DEQ Critique of DU PA
Appendix 21, this interrogatory concludes:
For all of these reasons, EnergySolutions/Neptune needs to either (1) provide independent
documentation that a sedimentation rate of 1.2 mm/yr is plausible or (2) define a plausible,
defensible intermediate lake sedimentation rate and redo the deep time analysis.
2.2.1 Interrogatory Response
This response addresses the following topics related to UDEQ concerns on the intermediate lake
sedimentation rate:
1. The technical rationale for the representation of intermediate lake sedimentation in the
Deep Time model.
2. New information is compiled from the scientific literature that expands documentation of
overall lake sedimentation rates and their application in the Deep Time model.
3. Field observations are summarized on the processes of intermediate lake sedimentation at
the Clive Site during the last transgressive (rising lake) and regressive (declining lake)
stages of Lake Bonneville (Appendix A).
4. The effects of intermediate lake sedimentation rates on the Deep Time model results, and
by extension the limited value of additional studies in this area.
2.2.1.1 Intermediate Lake Sedimentation
UDEQ has questioned the intermediate lake sedimentation “rate” used in the Deep Time model,
including both a reference to a sedimentation “rate” that is 10 times greater than the large lake
sedimentation rate (Version 1.4 of the Deep Time model; Neptune (2015b)) and a sedimentation
“rate” that they estimate to be as high as 5.64 mm/yr. They suggest the intermediate lake
sedimentation “rates” used in the model are unrealistic and cite literature from the Eastern Great
Basin of Utah supporting lake sedimentation rates that range from 0.12 to 0.83 mm/yr. (Note:
Deep Time Supplemental Analysis Responses for the Clive DU PA Model
23 Feb 2018 7
these sedimentation rates apply to lake sedimentation in general and are not specific to
intermediate lake sedimentation.) UDEQ also argues, citing input from Dr. Paul Jewell,
University of Utah, that lake sedimentation patterns can be complicated depending on proximity
to active faults and local sources of stream sediment. They have requested enhanced
documentation of sedimentation rates of 1.2 mm/yr or greater and development of a more
defensible “rate” of intermediate lake sedimentation. UDEQ also suggests that revision of the
intermediate lake sedimentation “rates” could require revision of the Deep Time model.
Establishing sedimentation rates for intermediate lakes is problematic for two reasons. First, the
lake sediments are reworked by wave activity that intermixes deep lake sediments (carbonate
marl), aeolian sediments, and intermediate lake clastic sediments. It is often difficult or
impossible to distinguish the different types of deposits in core studies. Second, intermediate
lakes are transitory features of limited duration. It is very difficult and impractical to attempt to
obtain the data necessary to constrain the ages of intermediate lake sediments and to estimate
depositional rates.
Recognizing these difficulties, the Deep Time model uses a pragmatic approach that represents
sedimentation by intermediate lakes as a depositional thickness per lake event. Estimates of
sediment thickness of intermediate lakes in the Deep Time model are derived primarily from
conceptual models of the processes of lake sedimentation, studies of sediment cores from drill
holes in the Lake Bonneville basin (Knolls and Burmester lake cores), and interpretations of lake
sediments in a Clive pit wall (Neptune 2015b). The latter interpretations are from a single
quarry-wall exposure that was studied in 1985. This combined information base is used to
establish a probability distribution of sediment thickness per intermediate lake event (see Figure
11; Neptune (2015b)).
UDEQ derives their cited intermediate lake sedimentation “rate” of 5.64 mm/yr by dividing the
geometric mean of the distribution of intermediate lake sediment thickness used in the Deep
Time model (2.82 m) by the estimated duration of an intermediate lake (500 years; Neptune
(2015b)). However, the Deep Time model does not use a sedimentation rate for intermediate
lakes and, in fact, the Deep Time model report specifically cautions (Neptune (2015b), Section
7.1):
There is virtually no information for the duration of intermediate lakes, due to the high
mixing rate of shallow lake sediments, which makes dating of times within a single
stratigraphic layer of a shallow lake sediment core extremely difficult. Thus, a distribution
was chosen to roughly calibrate with the heuristic model: lognormal with geometric mean of
500 y and geometric standard deviation of 1.5.
UDEQ, in their comment, incorrectly used information from the Deep Time probability
distribution of sediment thickness per lake event to calculate a sedimentation “rate” that is non-
applicable and was not used in the Deep Time model results. The technically correct approach
that UDEQ should have evaluated is a comparison of Deep Time model results for sedimentation
rates during 100-ka glacial cycles. Under this approach, the composite sediment thickness of
intermediate lake events is combined with aeolian and deep lake sedimentation. This composite
sediment thickness is constrained in the model by observed sediment thicknesses of 15 to 20
meters per glacial cycle based on core studies at the Knolls and Burmester sites (Neptune
(2015b), Section 7.4). This model interpretation of sediment thickness per glacial cycle is
Deep Time Supplemental Analysis Responses for the Clive DU PA Model
23 Feb 2018 8
equivalent to sedimentation rates of 0.15 to 0.20 mm/yr, and these rates are consistent with lake
sedimentation rates in new studies reported herein.
2.2.1.2 Expanded Documentation of Lake Sedimentation Rates
This section examines lake sedimentation rates in response to UDEQ’s request for
documentation of rates greater than 1.2 mm/yr. This new information does not apply directly to
the Deep Time model representation of intermediate lake sediment thickness per lake event. As
noted above, the data from Table 1 rates should be compared with sedimentation rates used in
the Deep Time model for100 ka glacial cycles.

Table 1 is a compilation of lake sedimentation rates from literature publications and calculations
derived from unpublished data and/or interpreted data. Sedimentation rates in Table 1 range from
as low as < 0.01 mm/yr to over 6.0 mm/yr. This range of variation is caused by a combination of
factors, including:

1. Rates of carbonate precipitation (fine-grained calcareous (CaCO3) mud).
2. Local availability of clastic sediment sources (gravel, sand, silt, and clay-sized particles),
primarily from proximity to areas of active faulting, higher topography near lake edges,
and/or near areas of river or stream that discharge into the lake.
3. Processes of wave activity that may produce strong longshore currents and may locally
transport and deposit clastic sediments.
4. Mass movement of lake bottom deposits.

Deep Time Supplemental Analysis Responses for the Clive DU PA Model
23 Feb 2018 9
Table 1. Compilation of Lake Sedimentation Rates.
Lake Location Min
Rate
(mm/yr)
Max
Rate
(mm/yr)
TimeSpan
(years)
Long/Short
(rate
duration)
Reference
Great Salt
Lake
U.S. 0.57 0.57 11,500 Long Thompson et al.
(2016)
Great Salt
Lake
U.S. 1.1 4.1 < 100 Short Naftz et al. (2000)
Great Salt
Lake
U.S. 0.1 6.7 < 10,000 Short/Long Oliver et al. (2009)
Lake Baikal Asia 0.05 0.05 > 100,000 Long Colman et al. (1995)
Lake
Bonneville
U.S. 0.8 0.8 12,000 Long Thompson et al.
(2016)
Lake
Bonneville
U.S. 0.04 0.94 12,000 Long Oviatt (2018 [in
press])1
Lake
Bonneville
U.S. 0.04 > 3.0 12,000 Long Oviatt (2018 [in
press])2
Lake
Michigan
U.S. < 0.01 0.84 < 10,000 Long Robbins and
Edgington (1975)
Lake
Michigan
U.S. 3.0 3.0 < 100 Short Robbins and
Callender (1975)
Lake
Tanganyika
Zaire 0.07 1.84 > 100,000 Long Cohen et al. (1993)
Owens Lake U.S. 0.4 0.4 > 100,000 Long Bischoff et al. (1997)
Red Lake Romania 1.17 3.2 < 100 Short Begy et al. (2009)
Taihu Lake China 0.1 0.42 < 100 yr Short Wang et al. (2001)
1 Open lake; only carbonate marl
2 Including deltas

High clastic sediment flux can cause very high local lake sedimentation rates. Very low
sedimentation rates generally occur in lake interiors that are distant from shorelines and sources
of clastic sediment (Oviatt 2018 [in press]). Significant variations in lake sedimentation rates can
occur over short horizontal distances depending on the local variations in processes of sediment
transport. During intervals of high wave activity, when intermediate lakes are present at the
elevation of the Clive Site, sedimentation rates can vary from negative (erosion) to very high
(progradation of gravel spits by longshore drift). Additionally, the dynamics of wave activity will
vary through time, with short-term variations in lake levels and changes in wind direction and
wind velocity. Maximum sediment deposition for intermediate lakes will most likely occur
during storm events (storm surge).
Deep Time Supplemental Analysis Responses for the Clive DU PA Model
23 Feb 2018 10
2.2.1.3 Evidence of Clastic Sedimentation at the Clive Site
Field studies of the lake sediments at the Clive Site provide important evidence of significant
past clastic sedimentation at the Clive Site associated with formation of transitory intermediate
lakes. These clastic sediments have been studied primarily for the transgressive (rising lake) and
regressive (declining lake) phases of Lake Bonneville (Neptune 2015a). Supporting
documentation for these observations, including photographs of the lake sediments and larger-
scale sedimentary features of longshore drift, were provided previously to UDEQ (Neptune
2015) and are repeated in Appendix A of this report.
The Clive Site has been and is expected to be an area of highly variable sedimentation during
future glacial cycles; this includes sedimentation associated with aeolian activity and formation
of intermediate and deep lakes. The optimal approach for representing the multiple dynamic
processes of lake sedimentation is through model simulations using a distribution of composite
sediment thickness for 100 ka glacial cycles. The sedimentation “rate” derived by UDEQ for
intermediate lakes is not applicable to the Deep Time model.
2.2.1.4 Impact of Intermediate Lake Sedimentation on the Deep Time Model
Results
The Deep Time model output of significance for the supplemental analysis is the ground surface
flux of radon (Figure 1 this paper; derived from Neptune (2015c)). Figure 1 shows that the peak
flux of radon occurs at about 60 ka, the model-projected time of the first occurrence of an
intermediate lake at the Clive Site.
The important Deep Time model assumptions and parameters that control the timing and
magnitude of the peak surface radon flux are:
1. The model representation of radon diffusion through the embankment cover and
overlying sediments.
2. The time of the first lake event at the Clive Site.
3. The aeolian deposition rate.
4. The intermediate lake sediment thickness for individual lake events.
Generally, the longer the time to the first event, the thicker the mantle of aeolian sediments on
top of the Clive Site. An intermediate lake event adds sediments above the top of the Site; the
added amount is determined by randomly sampling (computer simulation) the sediment
thickness probability distribution (see Figure 11; Neptune (2015b)). Overall, the Deep Time
model projections of combined processes of sedimentation (aeolian, intermediate, and deep lake
processes) affect the thickness of sediments covering the Clive Site over time. Increased depth
of burial will result in a systematic reduction in the radon flux (see the curve of the surface
radon flux in Figure 1).
An additional consideration is that the intermediate lake sediment thickness is not an
independent parameter in the Deep Time model results. It covaries with both the aeolian
sedimentation rate and the deep lake sedimentation rate. The three model parameters are
constrained in the Deep Time model by the composite thickness of sediments during the 100-ka
glacial cycles (15 to 20 meters thickness; Neptune (2015c)).
Deep Time Supplemental Analysis Responses for the Clive DU PA Model
23 Feb 2018 11

Figure 1. Deep Time ground surface flux of Rn-222 per NRC Reg. Guide 3.64.
3.0 Conclusion
UDEQ has raised several issues in the Deep Time interrogatories. The following points
summarize this response:
• The aeolian sediment data reported in Neptune (2015a) were used to establish statistical
parameters (sediment thickness and variations in sediment thickness) in the Deep Time
model. These measurements are a statistically representative sample of the population of
aeolian thickness data for the Clive Site and the DU disposal unit. UDEQ’s assertions
concerning sub-areas and sub-area correlations are not applicable to either the aeolian
field studies or the aeolian sedimentation parameters used in the Deep Time model.
• UDEQ has incorrectly assumed that the embankment is divided into 11 sub-areas in the
Deep Time model. The rationale for using the standard error of the mean in the
probability distribution for the aeolian depositional rate in the Deep Time model was
described in Neptune (2015a). The standard error of the mean is the correct statistical
Deep Time Supplemental Analysis Responses for the Clive DU PA Model
23 Feb 2018 12
parameter for representing the variance of the aeolian deposition rate for the scale of the
Deep Time model application. The recommendations by UDEQ do not follow established
modeling practices for environmental studies.
• The correct approach and statistical parameters are used for the aeolian depositional rate
in the Deep Time model and the impact of the variation in this rate on model results is
insignificant.
• Contrary to the interpretations in the review responses by UDEQ, the Deep Time model
does not use an intermediate lake sedimentation “rate.” Intermediate lakes are represented
by a distribution of sediment thickness per lake event.
• The distribution of sediment thickness for intermediate lake events is integrated with the
aeolian and deep lake sedimentation rates during 100-ka glacial cycles. The integrated
sedimentation rates during the glacial cycles (0.15 to 0.20 mm/yr) are consistent with
documented lake sedimentation rates, including rates cited in the review responses by the
UDEQ and the newly developed data shown on Table 1 in this report.
• We agree with the UDEQ comment that lake sedimentation patterns can be complicated
and are dependent on locations within lakes, interaction with wave-generated currents,
proximity to active faults, and local sources of sediment. These issues are already
addressed appropriately in the Deep Time model by integrating multiple sources of
information on lake sedimentation, including observations of lake and aeolian
sedimentary deposits at the Clive Site.
• Table 1 provides expanded documentation of lake sedimentation rates in response to the
UDEQ review comments, but these rates do not apply to the representation of
intermediate lakes in the Deep Time model. The data summarized in the table support
wide ranges in lake sedimentation rates and include rates greater than 1.2 mm/yr.
• The model distribution used to represent intermediate lake sediment thickness is based on
1985 data at the Clive Site for primarily the Lake Bonneville sedimentation history, with
limited local data for sedimentation associated with pre-Lake Bonneville glacial cycles.
The integrated net sedimentation rates used in the model for the 100-ka glacial cycles are
consistent with drill core data from the Clive region and published information on lake
sedimentation rates (see Table 1).
• The range of sediment thicknesses for intermediate lake events and integrated lake
sedimentation rates during glacial cycles used in the Deep Time model decreases the
peak ground surface radon flux and the radon flux over time. The data used for the lake
sedimentation rates are derived from and are consistent with core studies of deposits of
Lake Bonneville for past glacial cycles, published studies of lake sedimentation rates, and
field observations at the Clive Site.
On the basis of the concerns expressed by UDEQ as responded to herein, revision of the Deep
Time model is not justified.
Deep Time Supplemental Analysis Responses for the Clive DU PA Model
23 Feb 2018 13
4.0 References
Begy, R., et al., 2009. Recent Changes in Red Lake (Romania) Sedimentation Rate Determined
from Depth Profiles of 210Pb and 137Cs Radioisotopes, Journal of Environmental
Radioactivity 100 (2009) 644–648
Bischoff, J.L., et al., 1997. A Time-Depth Scale for Owens Lake Sediments of Core OL-92:
Radiocarbon Dates and Constant Mass-Accumulation Rate. In An 800,000-Year
Paleoclimatic Record from Core OL-92, Owens Lake, Southeast California, Geological
Society of America Special Paper 317, edited by G.I. Smith and J.L. Bischoff, pp. 91–98,
Geological Society of America, Boulder CO
Blöschl, G., and M. Sivapalan, 1995. Scale Issues in Hydrological Modelling: A Review,
Hydrological Processes 9 (1995) 251–290
Cohen, A.S., et al., 1993. Estimating the Age of Formation of Lakes: An Example from Lake
Tanganyika, East African Rift System, Geology 21 (6) 511–518
Colman, S.M., et al., 1995. Continental Climate Response to Orbital Forcing from Biogenic
Silica Records in Lake Baikal, Nature 378 (1995) 769–771
Crowe, B., et al., 2017. Representation of Global Climate Change in Performance Assessment
Models for Disposal of Radioactive Waste—17183, proceedings of the WM2017
Conference, March 5–9, Phoenix AZ, 2017
Lisiecki, L.E., and M.E. Raymo, 2005. A Pliocene-Pleistocene Stack of 57 Globally Distributed
Benthic δ18O Records, Paleoceanography 20 (2005) 1–17 doi: 10.1029/2004PA001071
Naftz, D.L., et al., 2000. Reconstructing Historical Changes in the Environmental Health of
Watersheds by Using Sediment Cores from Lakes and Reservoirs in Salt Lake Valley,
Utah, USGS Fact Sheet FS–164–00, United States Geological Survey (USGS),
December 2000
Neptune, 2015a. Neptune Field Studies, December, 2014, Eolian Depositional History Clive
Disposal Site, NAC-0044_R0, Neptune and Company Inc., Los Alamos NM, March 2015
Neptune, 2015b. Deep Time Assessment for the Clive DU PA, Deep Time Assessment for the
Clive DU PA Model v1.4, NAC-0032_R4, Neptune and Company, Inc., Los Alamos NM,
November 2015
Neptune, 2015c. Deep Time Supplemental Analysis for the Clive DU PA, Clive DU PA Model
vDTSA1, NAC-0043_R0, Neptune and Company, Inc., Los Alamos NM, March 2015
Deep Time Supplemental Analysis Responses for the Clive DU PA Model
23 Feb 2018 14
Neuman, S.P., et al., 2003. A Comprehensive Strategy of Hydrogeologic Modeling and
Uncertainty Analysis for Nuclear Facilities and Sites, NUREG/CR-6805, United States
Nuclear Regulatory Commission, Washington DC, July 2003
Oliver, W., et al., 2009. Estimating Selenium Removal by Sedimentation from the Great Salt
Lake, Utah, Applied Geochemistry 24 (2009) 936–949
Oviatt, C.G., 2018 [in press]. Geomorphic Controls on Sedimentation in Pleistocene Lake
Bonneville, Eastern Great Basin, Chapter in Geological Society of America Special
Paper, 2018
Robbins, J.A., and E. Callender, 1975. Diagenesis of Manganese in Lake Michigan Sediments,
American Journal of Science 275 (5) 512–533
Robbins, J.A., and D.N. Edgington, 1975. Determination of Recent Sedimentation Rates in Lake
Michigan Using Pb-210 and Cs-137, Geochimica et Cosmochimica Acta 39 (3) 285–304
Thompson, R.S., et al., 2016. Late Quaternary Changes in Lakes, Vegetation, and Climate in the
Bonneville Basin Reconstructed from Sediment Cores from Great Salt Lake. In Lake
Bonneville: A Scientific Update. Developments in Earth Surface Processes Vol. 20,
edited by C.G. Oviatt and J.F. Shroder Jr., pp. 221–291, Elsevier, Amsterdam,
Netherlands
Wang, J., et al., 2001. Taihu Lake, Lower Yangtze Drainage Basin: Evolution, Sedimentation
Rate and the Sea Level, Geomorphology 41 (2) 183–193
Zhang, X., et al., 2004. Scaling Issues in Environmental Modelling. In Environmental
Modelling: Finding Simplicity in Complexity, edited by J. Wainwright and M. Mulligan,
pp. 319–334, John Wiley & Sons, Chichester, England

Deep Time Supplemental Analysis Responses for the Clive DU PA Model
23 Feb 2018 15
Appendix A. Sedimentary Deposits of the Transgressive and
Regressive Phases of Lake Bonneville Near and Within the Clive Site
(Deposits of Intermediate Lakes)
Field studies of aeolian sediments and lake sediments in the Clive region (Neptune 2015a) have
described local exposures of gravel, sandy conglomerate, and sand in the Clive region. These
sediments were deposited primarily during the transgressive phase of Lake Bonneville, and
patterns of clastic sedimentation associated with past intermediate lake events were documented.
The cited studies were not focused on the lake events and thus detailed descriptions of the
sediments, including facies variations and thickness variations within and near the Clive Site,
have not been compiled.
The purpose of this brief summary is to document local characteristics of the deposits, including
evidence of energetic longshore drift with formation of local spits during the last transgressive
intermediate lake event near the Clive Site. The intermediate lake sediments contain distinctive
clasts of black volcanic rocks derived from local bedrock exposed in the Grayback Hills
northwest of the Clive Site. These relationships are consistent with southeast-directed wave
activity driving longshore drift that transported coarse sandy conglomerate and sand
southeastward from the Grayback Hills across the Clive Site.
Figure A-1 is a photograph of a roadcut exposure along Highway I-80 north of the Clive Site.
The outcrops expose the upper part of spit deposits of sandy conglomerate of the transgressive
phase of Lake Bonneville (deposited by longshore drift of an intermediate lake) overlain by Lake
Bonneville marl (deep lake deposits). The marl is overlain by Holocene aeolian silt.
Figure A-2 shows the locations of the Grayback Hills, the Clive Site, and the roadcut exposure of
Figure A-1. Buried ridges extend from the Grayback Hills toward the Clive Site and are upheld
by spit deposits of sandy conglomerate formed by longshore drift associated with an intermediate
lake. These ridges were interpreted originally as shorelines of the Gilbert phase of Lake
Bonneville. Recent field work has shown instead that the topography is upheld by spits buried
beneath the Lake Bonneville marl (Neptune 2015a).
Deep Time Supplemental Analysis Responses for the Clive DU PA Model
23 Feb 2018 16

Figure A-1. Roadcut exposures of the upper contact of the transgressive phase of Lake
Bonneville.
Deep Time Supplemental Analysis Responses for the Clive DU PA Model
23 Feb 2018 17

Figure A-2. Annotated Google Earth image showing the Clive area and buried spits
developed on the southeast edges of the Grayback Hills. The spits prograded toward
and across what is now the Clive Site.
Figure A-3 is from Neptune (2015a) and shows a quarry-wall exposure in the north part of the
Clive Site that contains exposures of the upper contact of the transgressive deposits of Lake
Bonneville.
Figure A-4 (also from Neptune (2015a) is a close-up view of the outcrop area shown in Figure
A-3. The sandy conglomerate was deposited by longshore drift that prograded sediment from the
northwest to southeast across the Clive Site. The thickness of this intermediate lake sediment is
expected to thin across the Clive Site (north to south) away from the sediment sources at the
Grayback Hills.
Deep Time Supplemental Analysis Responses for the Clive DU PA Model
23 Feb 2018 18

Figure A-3. (From Neptune (2015a).) View looking south from the north wall of Pit 29 at
the Clive Site. The dashed line marks the contact between the laminated marl of Lake
Bonneville (deep lake) and underlying transgressive sands with lenses of volcanic
conglomerate (deposits of an intermediate lake).

Figure A-4. (From Neptune (2015a).) Gully exposure of a lens of sandy conglomerate from
the outcrop shown in Figure A-3. The conglomerate contains distinctive clasts of
black volcanic rock derived from bedrock exposures in the Grayback Hills, northwest
of the Clive Site. The dashed line is the base of the laminated marl of Lake Bonneville
(deep lake). The gravel lens is here approximately 1 meter thick. It is correlative with
the Lake Bonneville Transgressive unit (0.76 m thick) listed in Table 3 of Neptune
(2015b).
Deep Time Supplemental Analysis Responses for the Clive DU PA Model
23 Feb 2018 19
Figure A-5 is a photograph of the deposits of the regressive phase of Lake Bonneville exposed in
the north wall of Pit 29 at the Clive Site. The sediments of this regressive intermediate lake are
preserved as a gradational zone between overlying aeolian silt (Holocene) and underlying
reworked marl of Lake Bonneville (light-colored sediments near the middle of the pick handle).
These regressive intermediate lake deposits are approximately 30 centimeters thick and illustrate
the variability of sedimentation patterns of intermediate lakes. The fine-grained silts of this
regressive intermediate lake are attributed to a combination of draping of the underlying
topography by fine-grained marl (the only available sediment source) and the probable rapid fall
of lake levels during the regressive phase of Lake Bonneville at the elevation of the Clive Site.

Figure A-5. (From Neptune (2015a).) Intermediate lake deposits of the regressive phase of
Lake Bonneville exposed in the north quarry wall of the Clive Site. The lake deposits
are exposed in the interval extending from the top of the hammer head to the middle
of the handle and are overlain by aeolian silt with local reddening from soil
alteration.

NAC-0102_R0

Other Wastes Responses for the
Clive DU PA Model
23 February 2018

Prepared by

NEPTUNE AND COMPANY, INC.
1435 Garrison St, Suite 201, Lakewood, CO 80215
Other Wastes Responses for the Clive DU PA Model
23 Feb 2018 ii
1. Title: Other Wastes Responses for the Clive DU PA Model
2. Filename: Other Wastes Responses for the Clive DU PA Model.docx
3. Description: Summary of how wastes other than DU are addressed in GoldSim model
v. 1.4 and supporting documentation.

Name Date
4. Originator Sean McCandless 12 Feb 2018
5. Reviewer Mike Sully 13 Feb 2018
6. Remarks

Other Wastes Responses for the Clive DU PA Model
23 Feb 2018 iii
CONTENTS
CONTENTS ............................................................................................................................. iii
ACRONYMS AND ABBREVIATIONS ................................................................................... iv
1.0 Overview and Conceptual Model........................................................................................ 1
2.0 UDEQ Interrogatory Responses.......................................................................................... 4
2.1 Interrogatory CR R313-25-9(5)(A)-196/1: .................................................................... 4
2.1.1 Interrogatory Response ............................................................................................ 4
2.2 Interrogatory CR R313-25-9(5)(a)-203/1: ..................................................................... 4
2.2.1 Interrogatory Response ............................................................................................ 4
3.0 References .......................................................................................................................... 5

Other Wastes Responses for the Clive DU PA Model
23 Feb 2018 iv
ACRONYMS AND ABBREVIATIONS
CAW Class A West embankment
CQA/QC Construction Quality Assurance/Quality Control
DEQ (Utah) Department of Environmental Quality
DOE (United States) Department of Energy
DU depleted uranium
DUO3 depleted uranium trioxide
ET evapotranspiration
GDP gaseous diffusion plant
LLRW low-level radioactive waste
NRC (United States) Nuclear Regulatory Commission
PA performance assessment
SER Safety Evaluation Report
SRS Savannah River Site
UDEQ Utah Department of Environmental Quality

Other Wastes Responses for the Clive DU PA Model
23 Feb 2018 1
1.0 Overview and Conceptual Model
The Clive DU PA is limited to the disposal of DU wastes of two general waste types: 1) depleted
uranium trioxide (DUO3) waste from the Savannah River Site (SRS); and 2) anticipated DU
waste as U3O8 from gaseous diffusion plants (GDPs) at Portsmouth, Ohio and Paducah,
Kentucky. The quantity and characteristics of DU waste from other sources that has already been
disposed of at the Clive Facility was not included. Similarly, other Class A LLRW anticipated
for disposal within the Federal Cell was not included.
This choice was made for several reasons:
1. Probabilistic PAs, while a mature approach within the Nuclear Regulatory Commission
(NRC) and the Department of Energy (DOE), are new to Utah LLRW regulators. This
reality is reflected by Utah’s need to utilize external contractor review support. A
probabilistic PA is initially a more complex and expensive modeling approach than
deterministic modeling; however, once the foundation is established it can easily be
adapted for additional wastes and new data. Due to shorter half-lives, other Class A
radionuclides have much shorter time periods for evaluation. Therefore, a relatively
simplified approach was selected, with the intent that, once the basic model structure and
approach was developed, additional Class A wastes could be more explicitly modeled as
a follow-on activity. Given the commitment to place DU in the bottom of the Federal
Cell, additional Class A wastes do not need to be modeled in order for embankment
construction and initial DU placement to begin.
2. Class A LLRW had already been modeled multiple times for the Clive Site and basic cell
design. These models demonstrated the Site’s ability to achieve the performance
objectives, albeit using deterministic rather than probabilistic performance assessments.
Note that deterministic performance assessments for the Clive Site are consistently
structured to demonstrate compliance even under extreme bounding conditions. These
models also explicitly included DU at its specific activity (i.e., elemental DU), and
existing disposals were placed in full compliance with a UDEQ-approved license. The
Class A West (CAW) embankment differs from the Federal Cell in physical geometry
and cover system design; however, the horizontal travel distances to compliance points
are the same, the vertical vadose zone thicknesses are the same, and the precipitation
falling on the cells is the same. Therefore, if the modeled infiltration is the same or less
for the Federal Cell than for the Class A West embankment, then Federal Cell disposal of
wastes already approved for disposal in the CAW can be expected to meet the
performance criteria.
3. By definition (via the regulatory basis for 10 CFR Part 61), Class A LLRW decays to
radiation levels that protect health and safety within the 100-year institutional control
period. Therefore, the majority of Class A contaminants of concern have decayed away
millennia before ingrowth concerns associated with DU would be exhibited.
In the final report for model version 1.4 (Neptune 2015a), Section 6.0 documents this scoping
constraint: “The disposal volume above the DU waste is assumed to be backfilled with clean
material for the purposes of this DU analysis.” This constraint was established and has been
Other Wastes Responses for the Clive DU PA Model
23 Feb 2018 2
carried forward unchanged from the initial submittal of version 1.0 of the DU PA Model
(Neptune 2011).
UDEQ acknowledged this approach in Section 1.0 of the 2015 SER (SC&A 2015):
It should be noted that this SER apples only to approval/disapproval of the DU PA as
required by UAC Rule R313-25-9(5)(a). That regulation does not consider “other wastes,”
which must be addressed in a separate performance assessment. If the DU PA performance
assessment is approved, it is expected that the next step in the regulatory process would be
submission of a proposed license amendment for review by DEQ.
Section 6.2.4 of the 2015 SER elaborates and provides Condition 4, which would apply to an
amended License that would permit DU disposal:
To meet the requirements of UAC R313-25-9(5)(a), EnergySolutions shall submit a revised
performance assessment that meets the requirements of that provision and addresses the total
quantities of concentrated DU and other radioactive wastes the facility now proposes to
dispose in the Federal Cell. This revised performance assessment shall be subject to notice
and comment and must be approved by the Director prior to the land disposal of other
radioactive waste [emphasis added].
EnergySolutions did not object to this aspect of the 2015 SER, as this accurately reflects their
intention and understanding of the regulatory process. Contrary to this agreed-upon regulatory
process, Utah DEQ (2017) introduced two new interrogatories relating to other Class A waste.
This is in spite of the basis for Interrogatory CR R313-25-9(5)(A)-196/1 correctly stating:
Therefore, to address the R313-25-9(5)(a) “other waste” requirement, a revised PA, or a
separate PA, must be prepared by EnergySolutions, and approved by DEQ, before any
“other waste” (understood to be DOE-generated Class A waste) is disposed of above the
DU [emphasis added].
EnergySolutions does not intend or request to dispose of other Class A LLRW above the DU
until such time as a PA accounting for the combined effects of DU and other Class A LLRW is
approved. This does not preclude approval to dispose of DU in the interim.
Current approved Clive deterministic PAs apply Unit 3 material properties to Class A waste. See
Whetstone’s Revised Western LARW Cell Infiltration and Transport Modeling Report (July 19,
2000) and their Class A West Disposal Cell Infiltration and Transport Modeling Report
(Whetstone Associates 2011), which both include discussion of model insensitivity to waste
thickness and other physical properties. This longstanding approach for the facility has been
acceptable to UDEQ, dating to the earliest deterministic PAs. In raising an objection, UDEQ
does not substantiate the change in approach from their prior approvals.
Modeling of waste layers as Unit 3 sandy soil has been accepted by the regulator precisely
because Class A unstable waste is heterogeneous and may have a variety of physical forms. It
would not be possible to model all possible waste form combinations; therefore, the model is
simplified to represent waste layers as more permeable than the overlying cover and the
underlying clay liner.
Other Wastes Responses for the Clive DU PA Model
23 Feb 2018 3
Furthermore, waste form can often contribute to delayed release of radionuclides for
environmental transport; however, credit is not taken for this mechanism in current approved
deterministic PAs nor in the DU PA v.1.4. For example, radionuclides present in waste as
activated metal or within a grouted monolith would not be available for transport as early after
disposal as radionuclides present as contamination on debris. By assuming that all activity is
available for environmental transport immediately, Clive PAs bias environmental release to
occur earlier than it may actually occur.
It is not necessary to model the potential for preferential pathways for radon or precipitation
within or between actual waste placement layers in order to evaluate performance of the Federal
Cell. For an evapotranspiration (ET) cover, infiltration is influenced by properties of the Unit 4
soil surface layer, while percolation is mainly controlled by Unit 4 Surface and Evaporative Zone
layers and the Frost Protection layer, with some influence possible from the radon barrier. In
other words, rates of infiltration, percolation, and exhalation are controlled by the cover system
and the clay liner. Model performance will be essentially the same regardless of the modeled
properties within the waste layers. See also ET Cover Design Responses for the Clive DU PA
Model (Neptune 2018a).
Sandy soil properties were conservatively selected for modeling the waste layers in order to
minimize potential effects of absorption/retardation that could occur if the waste layer were
assumed to be clay. Note that, in practice, clay is often used as fill material within the
embankment.
Appendix 2 to Neptune (2015a), Conceptual Site Model for Disposal of Depleted Uranium at the
Clive Facility, Clive DU PA Model v. 1.4 (Neptune 2015d), Section 8.1, remains consistent with
Clive’s prior approved PAs in this respect: “All wastes are assumed to have the characteristics of
local Unit 3 sandy soil.”
See also Appendix 16 to Neptune (2015a), Model Parameters for the Clive DU PA Model, Clive
DU PA Model v. 1.4 (Neptune 2015c), Section 4.16: “The current Clive DU PA Model has no
generic waste inventory, but this material is defined as a placeholder. Any layers to be filled with
generic LLW borrow material properties from Unit 3 (see Table 10).”
As noted in Interrogatory CR R313-25-9(5)(A)-196/1, Section 3.1 of Appendix 18 to Neptune
(2015a), Radon Diffusion Modeling for the Clive DU PA, Clive DU PA Model v. 1.4 (Neptune
2015b), states that other wastes are assigned Unit 4 properties. Appendix 18 has a typographical
error and should state that other wastes are assigned Unit 3 properties, consistent with the
Conceptual Site Model, Appendix 16, and the GoldSim model.
Note that the Federal Cell, as currently configured, isolates DU (and, ultimately, overlying Class
A LLRW) from the adjacent 11e.(2) embankment. The Federal Cell has no existing waste
placement of any type. Refer to Federal Cell Design Responses for the Clive DU PA Model
(Neptune 2018b) for discussion of the designs and related issues in response to interrogatories
under the category “Federal Cell Design.”
Other Wastes Responses for the Clive DU PA Model
23 Feb 2018 4
While not prohibiting placement of DU upon approval of the Clive DU PA Model v.1.4,
additional modeling is required by rule before Class A waste is disposed in the Federal Cell.
There are two reasons that this is the case:
1. The performance of the ET cover included in the DU PA should be verified for Class A
LLRW disposed in the same embankment as DU.
2. Incremental dose contributions from other Class A waste in the Federal Cell should be
quantified for a complete probabilistic PA.
Considering that current approved PAs for the Class A West, LARW, and Mixed Waste cells
include Class A LLRW in addition to DU, it is difficult to imagine a scenario where Class A
LLRW would require any limitation below the Class A limits when disposed with DU under a
probabilistic PA. Nonetheless, if demonstrated to be desirable under a combined probabilistic
PA, Federal Cell waste acceptance criteria can be set to limit sensitive Class A radionuclide
concentrations or placement location.
2.0 UDEQ Interrogatory Responses
This section contains responses for Interrogatories CR R313-25-9(5)(A)-196/1 and CR R313-25-
9(5)(a)-203/1.
2.1 Interrogatory CR R313-25-9(5)(A)-196/1:
Please provide an analysis to demonstrate that the DU PA v1.4 assumed homogeneous Unit 4
silty clay material used to model the layer above the DU is representative of the various types of
DOE-generated Class A waste EnergySolutions intends to dispose of in that layer. Density,
among other factors, should be considered.
2.1.1 Interrogatory Response
As discussed above, the reference to Unit 4 material is in error. The use of Unit 3 material
properties for Class A LLRW is reasonable and consistent with prior PAs for the facility.
2.2 Interrogatory CR R313-25-9(5)(a)-203/1:
Please describe how EnergySolutions proposes to address the requirements of R313-25-9(5)(a)
to demonstrate that PA requires consideration of the “total quantities of concentrated depleted
uranium and other wastes.”
2.2.1 Interrogatory Response
As provided in draft Condition 4 of the 2015 SER, prior to disposing of Class A LLRW other
than DU in the Federal Cell, EnergySolutions will obtain approval of additional modeling.
EnergySolutions does not intend or request to dispose of other Class A LLRW above the DU
until such time as a PA accounting for the combined effects of DU and other Class A LLRW is
approved. This does not preclude approval to dispose of DU in the interim.
Other Wastes Responses for the Clive DU PA Model
23 Feb 2018 5
3.0 References
Neptune, 2011. Final Report for the Clive DU PA Model version 1.0, Neptune and Company
Inc., Los Alamos NM, June 2011
Neptune, 2015a. Final Report for the Clive DU PA Model, Clive DU PA Model v1.4, NAC-
0024_R4, Neptune and Company Inc., Los Alamos NM, November 2015
Neptune, 2015b. Radon Diffusion Modeling for the Clive DU PA, Clive DU PA Model v1.4,
NAC-0033_R1, Neptune and Company Inc., Los Alamos NM, November 2015
Neptune, 2015c. Model Parameters for the Clive DU PA Model, Clive DU PA Model v1.4, NAC-
0026_R4, Neptune and Company Inc., Los Alamos NM, November 2015
Neptune, 2015d. Conceptual Site Model for Disposal of Depleted Uranium at the Clive Facility,
Clive DU PA Model v1.4, NAC-0018_R4, Neptune and Company Inc., Los Alamos NM,
November 2015
Neptune, 2018a. ET Cover Design Responses for the Clive DU PA Model, NAC-0106_R0,
Neptune and Company Inc., Lakewood CO, February 2018
Neptune, 2018b. Federal Cell Design Responses for the Clive DU PA Model, NAC-0101_R0,
Neptune and Company Inc., Lakewood CO, February 2018
SC&A, 2015. Utah Division of Radiation Control, EnergySolutions Clive LLRW Disposal
Facility, License No: UT2300249; RML #UT 2300249, Condition 35 Compliance Report;
Appendix A: Final Report for the Clive DU PA Model, Safety Evaluation Report, Volume
1, SC&A Inc., Vienna VA, April 2015
Utah DEQ, 2017. Division of Waste Management and Radiation Control, EnergySolutions Clive
LLRW Disposal Facility License No: UT2300249; RML #UT 2300249, Amended and
New Interrogatories Related to Clive DU PA Modeling Report Version 1.4 Dated
November 2015, Utah Department of Environmental Quality (DEQ), Salt Lake City UT,
May 2017
Whetstone Associates, 2011. EnergySolutions: Class A West Disposal Cell Infiltration and
Transport Modeling Report, prepared for EnergySolutions, Whetstone Associates Inc.,
Gunnison CO, November 2011

NAC-0104_R0

Groundwater Exposure Responses
for the Clive DU PA Model
23 February 2018

Prepared by

NEPTUNE AND COMPANY, INC.
1435 Garrison St, Suite 201, Lakewood, CO 80215
Groundwater Exposure Responses for the Clive DU PA Model
23 Feb 2018 ii
1. Title: Groundwater Exposure Responses for the Clive DU PA Model
2. Filename: Groundwater Exposure Responses for the Clive DU PA Model.docx
3. Description: Summary of how the deep and basal aquifers are addressed in GoldSim
model v. 1.4 and supporting documentation.

Name Date
4. Originator Gregg Occhiogrosso 12 Feb 2018
5. Reviewer Mike Sully and Dan Levitt 12 Feb 2018
6. Remarks

Groundwater Exposure Responses for the Clive DU PA Model
23 Feb 2018 iii
CONTENTS
CONTENTS ............................................................................................................................. iii
FIGURES .................................................................................................................................. iv
ACRONYMS AND ABBREVIATIONS .................................................................................... v
1.0 Overview and Conceptual Model........................................................................................ 1
1.1 Considerations for a Groundwater Dose Pathway.......................................................... 2
1.1.1 Water Quality .......................................................................................................... 2
1.2 Interaction of Shallow and Lower Aquifers ................................................................... 3
2.0 UDEQ Interrogatory Responses.......................................................................................... 3
2.1 Interrogatory CR R313-25-3 and R313-25-8-195/1: Aquifer Characterization and
Interrogatory CR R313-25-20-204/1: Exposure to Groundwater ................................... 3
2.1.1 Interrogatory Responses .......................................................................................... 4
2.1.2 Evaluation of Applicability of UDEQ Basal and Lower Aquifer Contamination
Scenarios ................................................................................................................. 4
2.1.3 Updated Groundwater Ingestion Tc-99 Concentration Estimate ............................... 9
3.0 Conclusion ....................................................................................................................... 10
4.0 References ........................................................................................................................ 10

Groundwater Exposure Responses for the Clive DU PA Model
23 Feb 2018 iv
FIGURES
Figure 1. Stratigraphic profile showing ET cover, waste zone, and stratigraphy below the
Federal Cell. The deep aquifer is subdivided into the lower confined aquifer and
the basal aquifer. ........................................................................................................ 1
Figure 2. Conceptual model of Johnson et al. (2011). .................................................................. 5
Figure 3. Wells near York, NE. Generated using Nebraska Department of Natural Resources
online mapping tools. The total area depicted is roughly 68 square miles. .................. 6
Figure 4. Production wells within a 5-mile radius (~78 square miles) of the Clive Site.
Generated using Utah Division of Water Rights online tools. Note that well 16-
816 was drilled on January 9, 1996 and abandoned on January 30, 1996. This well
was contemplated as a source of construction water for the EnergySolutions Clive
facility, but not used................................................................................................... 7

Groundwater Exposure Responses for the Clive DU PA Model
23 Feb 2018 v
ACRONYMS AND ABBREVIATIONS
DEQ (Utah) Department of Environmental Quality
DU depleted uranium
ECL Effluent Concentration Limits
EPA (United States) Environmental Protection Agency
ET evapotranspiration
GWPL groundwater protection limit
NRC (United States) Nuclear Regulatory Commission
PA performance assessment
PRA probabilistic risk assessment
PWS public water supply
RO reverse osmosis
SER Safety Evaluation Report
TDS total dissolved solids
UDEQ Utah Department of Environmental Quality

Groundwater Exposure Responses for the Clive DU PA Model
23 Feb 2018 1
1.0 Overview and Conceptual Model
This document provides a summary of the important context for a discussion of potential risks
associated with groundwater at the Clive Site. As discussed below, the hydrogeologic setting,
geochemical conditions, and known usage history of the Site largely preclude traditional
groundwater dose scenarios. The saline groundwater (discussed in Section 1.1.1) is of limited
utility without extensive treatment. Low well yield in the shallow subsurface due to the soil
texture necessitates that the withdrawal of any substantial quantity of water from beneath the Site
involves extraction from lower aquifers, providing dilution of exposure concentrations
(discussed in Section 1.2). Sections 2.0 considers the latest UDEQ interrogatories.
Figure 1 depicts the generalized stratigraphic profile of the Site subsurface; detailed discussion
of the hydrogeologic system beneath the Site can be found in ET Cover Design Responses for the
Clive DU PA Model (Neptune 2018). The water table aquifer which would be impacted by any
migration from the waste cells is referred to as the shallow unconfined aquifer. Groundwater
occurring at greater depths is referred to as the lower confined (~70–100 ft below ground
surface) and basal (450–750 ft below ground surface) aquifers.

Figure 1. Stratigraphic profile showing ET cover, waste zone, and stratigraphy below the
Federal Cell. The deep aquifer is subdivided into the lower confined aquifer and the
basal aquifer.

Groundwater Exposure Responses for the Clive DU PA Model
23 Feb 2018 2
Groundwater exposure as a dose pathway has been the subject of several past interrogatories.
Certain exposure scenarios have been evaluated in past rounds of interrogatories with UDEQ.
Round 2 Interrogatory Responses (EnergySolutions 2014) calculated doses for well drilling
scenarios (reflecting practices commonly seen in the Clive area drainage basin) based on results
from v1.2 of the Clive DU PA Model. The general characteristics of these aquifers that are
relevant to potential groundwater dose scenarios are described below.
1.1 Considerations for a Groundwater Dose Pathway
1.1.1 Water Quality
The State of Utah classifies groundwater resources into four classes according to total dissolved
solids (TDS) concentrations. Class I waters, referred to as “pristine,” “irreplaceable,” or
“ecologically important,” are most strongly protected and are subject to the most stringent
protections regarding TDS and contaminant levels. Class II and Class III waters are referred to as
“drinking water quality” and “limited use,” respectively, and are similarly subjected to specific
but less stringent standards. Class IV groundwater, defined by TDS concentrations greater than
10,000 mg/L, is referred to as “saline” and has no specific protection standards.
Typical TDS concentrations at the Clive Site range from about 10,000 mg/L to about
70,000 mg/L, and its water has been defined by UDEQ as Class IV (saline groundwater) (Utah
2014). For perspective, typical seawater TDS concentrations are about 35,000 mg/L, and the
Environmental Protection Agency (EPA) Secondary Standard for the TDS level in drinking
water is 500 mg/L. Poor water quality at all depths in the immediate vicinity of the Site limits the
utility of the groundwater without treatment. An EnergySolutions analysis (EnergySolutions
2013) concluded that consumption of untreated, native groundwater would result in 100%
mortality of the receptor due to salinity levels, making radiological risks irrelevant.
The lower confined aquifer exhibits lower, though still elevated, TDS levels and is also
considered Class IV. While treatment of the groundwater to potable TDS levels is possible in
principle by utilizing a desalination process such as reverse osmosis (RO), this practice is not
found in the Clive facility basin. It has been acknowledged that RO treatment would also reduce
contaminant concentrations, which is discussed in more detail below. Additionally, the
probability of such a practice is low due to the lack of population and the availability of less
costly and higher quality water sources throughout the west desert. Extraction wells in the area
around the Site do exist, but these are generally installed in more productive gravel zones in the
recharge areas of the surrounding foothills and they are generally used for industrial purposes.
As such, a groundwater ingestion scenario in the immediate vicinity of the Site is highly
unlikely, is outside of the known usage history of the Site, and has never been required in
EnergySolutions’ various other UDEQ-approved performance assessments (PA), nor is it
supported by the promulgation of any new regulatory requirement since prior PA approvals.

Groundwater Exposure Responses for the Clive DU PA Model
23 Feb 2018 3
1.2 Interaction of Shallow and Lower Aquifers
Groundwater level measurements and geochemical data suggest minimal flow from the shallow
aquifer to the basal and lower aquifers. This is important to any discussion of a groundwater dose
scenario because the shallow aquifer is not of sufficient yield to be a productive source and
would not produce a reliable water supply for an inadvertent intruder using treated groundwater
as a drinking water source. Therefore, a productive well in the vicinity of the Site would involve
pumping from the basal or lower confined aquifers. Significant contamination of the lower
aquifers would only occur with advective movement of radionuclides from the shallow aquifer
(i.e., via downward water flow), which would be possible via movement through the naturally
occurring material or some anthropogenic pathway like a multi-aquifer well. As discussed below,
an anthropogenic pathway is very unlikely based on the Site conditions and usage history.
Based on simultaneous water level measurements in well clusters GW-19A/B, I-1-30/50/100,
and I-3-30/50/100, Bingham Environmental (1996) reported an apparent upward hydraulic
gradient, with freshwater equivalent head differentials of about 1–2 ft between deep (~100 ft
depth) and shallower (~30–50 ft depth) well screens. Corresponding gradients (0.02 to 0.04) and
upward velocities were calculated (0.05 ft/yr to 0.10 ft/yr). Bingham Environmental (1996) also
noted that solute and isotopic data suggest “minimal or no vertical movement” from the shallow
unconfined aquifer to the deeper confined aquifer, stating that, “Lower TDS concentrations for
the few confined aquifer samples suggest the unconfined and confined aquifers may be
chemically as well as hydraulically distinct systems.” This evidence suggests that contamination
of the lower aquifer due to the natural flow from the shallow aquifer is unlikely. UDEQ has
acknowledged the upward gradient by including a performance standard in the Groundwater
Quality Discharge Permit (Utah 2014) for the Site which states that a neutral or upward
hydraulic gradient should be maintained in nested or paired monitoring wells.
2.0 UDEQ Interrogatory Responses
This section contains responses for Interrogatories CR R313-25-3 and R313-25-8-195/1 (a single
interrogatory referencing two parts of the Utah Administrative Code) and CR R313-25-20-204/1.
2.1 Interrogatory CR R313-25-3 and R313-25-8-195/1: Aquifer
Characterization and Interrogatory CR R313-25-20-204/1:
Exposure to Groundwater
• Interrogatory CR R313-25-3 and R313-25-8-195/1: Aquifer Characterization
- Statement: Please provide information assessing the aquifer hydraulic properties
and groundwater quality for the lower confined aquifer (e.g., at 70–100 ft) and
valley-fill or basal-aquifer-system aquifers (e.g., at 450–750 ft) at the Clive site.
Specific types of information include, for example, groundwater flow velocities,
aquifer transmissivities, water quality, sorption properties, and the degree of
hydraulic interconnection between the upper and basal aquifers. Calculations
should be shown for horizontal and vertical components of groundwater flow and
contaminant migration velocities.
Groundwater Exposure Responses for the Clive DU PA Model
23 Feb 2018 4
• Interrogatory CR R313-25-20-204/1: Exposure to Groundwater
- Statement: Please revise your June 8, 2014, partial response to Interrogatory 182
by extending it to 10,000 years and including the groundwater consumption
pathway, and include the results of the extended analysis in the next revision of the
DU PA, including the Appendix 19 sensitivity analyses.
- Excerpt of Basis: While preparing the April 2015 Safety Evaluation Report (SER),
the Department of Environmental Quality (DEQ)/SC&A extended the
EnergySolutions Interrogatory 182 partial response to 10,000 years and included
the groundwater consumption pathway, as well as several postulated scenarios
including a leaking well casing, a nearby failed or abandoned well that presents a
direct path between the upper and lower aquifer, and fresh water in the lower
aquifer. The results of this DEQ/SC&A analysis are given in the white paper,
“Groundwater Pathway Doses, Part 2,” Revision 2 (Marschke 2015).
2.1.1 Interrogatory Responses
These interrogatories share a common theme in their suggestion that various anthropogenic
scenarios might cause cross-contamination of the lower aquifer, and therefore a perceived need
for further development of the lower aquifer hydraulic properties. Because both stem from these
postulated scenarios, they are addressed together rather than as separate responses.
2.1.2 Evaluation of Applicability of UDEQ Basal and Lower Aquifer
Contamination Scenarios
Interrogatory R313-25-20-204/1 suggests the need to calculate groundwater consumption doses
based on the exposure mechanisms conjectured in Marschke (2015)1, which relied on two
literature sources not specific to Clive area practices to support the idea of cross-contamination
of a confined aquifer by the overlying unconfined aquifer (Johnson et al. 2011; Zinn and
Konikow 2007). The applicability of the scenarios presented in the literature sources is largely
unaddressed by Marschke (2015). Both sources, which consider mechanisms for contamination
from an unconfined aquifer moving to a confined aquifer via wells screened through both
aquifers, exclude justification of human groundwater consumption and are discussed below.
Johnson et al. (2011) conducted flow modeling for a synthetic domain meant to mimic the
hydrogeologic setting near York, Nebraska, an area where irrigation wells are very common (and
hydrogeologically different from the Clive drainage basin). The analysis attempted to quantify
the impacts of nearby irrigation wells to a public water supply (PWS) well placed in the same
vicinity. This analysis used hydraulic properties and hydrogeologic unit thicknesses specific to
the Nebraska site, and included multiple multi-aquifer wells near the pumping well that provided
hydraulic connection between the upper unconfined aquifer and the lower confined aquifer.
Figure 2 from Johnson et al. (2011) is reproduced in Figure 2 below to provide some context
regarding the conceptual model used in the study.

1 Note that this paper is prominently watermarked “WORKING DRAFT—DO NOT CITE OR QUOTE” yet was
cited by UDEQ.
Groundwater Exposure Responses for the Clive DU PA Model
23 Feb 2018 5
Figure 2. Conceptual model of Johnson et al. (2011).
The vicinity of the Clive Site does not have a similar hydrogeologic setting or a similar history of
closely spaced irrigation wells when compared to the study site in NebraskaError! Reference
source not found.. Figure 3, generated from well installation records of the Nebraska
Department of Natural Resources, shows hundreds of irrigation wells (purple). In the
hydrogeologic setting of York, NE, cross-contamination of aquifers via the dense network of
wells may indeed be a viable scenario warranting an analysis like that undertaken by Johnson et
al. (2011). The scenario is not readily applicable to the Clive Site, as production wells installed
in the Clive vicinity are comparatively rare and are primarily in upgradient gravel units within
the recharge zone of the surrounding foothills, as shown in Figure 4. The well nearest the Clive
Site, identified as 16-816 in Figure 4, was abandoned in 1996. The hydraulic gradient in the
shallow aquifer is generally toward the northeast.
Groundwater Exposure Responses for the Clive DU PA Model
23 Feb 2018 6

Figure 3. Wells near York, NE. Generated using Nebraska Department of Natural
Resources online mapping tools. The total area depicted is roughly 68 square miles.
Groundwater Exposure Responses for the Clive DU PA Model
23 Feb 2018 7

Figure 4. Production wells within a 5-mile radius (~78 square miles) of the Clive Site.
Generated using Utah Division of Water Rights online tools. Note that well 16-816
was drilled on January 9, 1996 and abandoned on January 30, 1996. This well was
contemplated as a source of construction water for the EnergySolutions Clive facility,
but not used.
The postulated mechanism for contamination of the lower aquifer with radionuclides from the
Clive Site would require not only a network of pre-existing multi-aquifer wells near a pumping
well, but also that the wells be in a specific geometry relative to one another and the Site. Multi-
aquifer wells might promote the flow of uncontaminated water to a pumping well for many
possible configurations. Additionally, given the low yields of the upper aquifer, it is unlikely that
Groundwater Exposure Responses for the Clive DU PA Model
23 Feb 2018 8
it could sustain prolonged elevated flows to the lower aquifer near the hypothesized multi-
aquifer wellbore. In short, the list of suppositions necessary for this mechanism of contamination
is long and specific. Johnson et al. (2011) also noted that:
Predicting the impact of contaminants on a PWS well due to leaking multi-aquifer wells is,
in general, not possible. However, the modeling steps described here can provide site-
specific insight into the potential for multi-aquifer wells to affect a confined-aquifer PWS
well.
Marschke (2015) made no attempt to apply the modeling steps laid out in Johnson et al. (2011) to
the Clive Site. Nevertheless, Marschke (2015) stated that “Johnson et al. (2011) determined that
a dilution factor of 10 would be appropriate” to calculate the relative flows from the upper and
lower aquifer to a well screened in the lower aquifer. Considering that no attempt was made to
apply the methodology on a site-specific basis, the dilution factor of 10 cited by Marschke
(2015) is arbitrary at best.
The second study cited, Zinn and Konikow (2007), is similarly a hypothetical model domain
meant to simulate the effects of existing, non-pumping wells on nearby pumping wells. The
authors noted the relevance of the study to regions with agricultural history like Nebraska and
Alberta, where numerous irrigation wells potentially affect local flow and transport. The model
domain size and shape are described as “arbitrary,” with the boundary conditions “conceptually
similar to a river valley”:
These conditions led to a predevelopment steady state regional flow field that was
conceptually similar to the side of a river valley or other simple groundwater basin—a
groundwater divide with primarily downward infiltration of water at the upstream end of the
system, predominantly horizontal flow near the middle part of the system, and predominantly
upward flow near the constant-head boundary.
The hydrologic conditions considered are not similar to the Clive Site. The model domain is
assigned homogenous properties, dissimilar to the Site (e.g., hydraulic conductivity of
76.2 m/day compared with Clive Site values on the order of 0.1 m/day). As noted above, the
combination of hydraulic properties and boundary conditions produces vertical hydraulic
gradients on either side of the domain. For certain configurations of pumping and non-pumping
wells, vertical flows in the non-pumping wells were altered. Zinn and Konikow (2007) noted that
“the effect of intraborehole flow is highly dependent on the location of the borehole within the
system, particularly in relation to the head field of the system.” This mechanism for
contamination of the lower aquifer is thus extremely improbable at the Clive Site, as it requires
an unlikely combination of pumping wells and non-pumping wells in a prescribed geometry.
Furthermore, the expense to install such a well pumping network far exceeds the cost of
transporting potable water from other sources outside of the Clive drainage basin. Moreover, as
described above in the discussion of Johnson et al. (2011), UDEQ made no attempt to apply this
scenario using site-specific hydrogeological unit structure or parameter values. As before, many
conditions which are inconsistent with the current use of the Clive Site are necessary for this
mechanism of contamination.
Groundwater Exposure Responses for the Clive DU PA Model
23 Feb 2018 9
In summary, the applicability of these contamination scenarios to the Clive Site is not well
supported upon closer examination of their bases. NRC’s Regulatory Guide 1.174 expresses the
need for site-specific information in probabilistic risk assessments (PRAs):
One overriding requirement is that the PRA should realistically reflect the actual design,
construction, operational practices, and operational experience of the plant and its owner.
Regulatory Guide 1.200 also states that a consensus PRA standard is that PRAs should use
“plant-specific information versus generic information to represent the as-designed, as-built and
as-operated plant.”
Interrogatories CR R313-25-3 and R313-25-8-195/1 express the perceived need for further
evaluation of aquifer properties and flow regimes for the aquifers below the shallow unconfined
aquifer, based on scenarios developed for markedly different hydrogeologic conditions rather
than on site-specific information. Given the improbability of any scenario at the Clive Site by
which deeper aquifers would become contaminated and thus impact the Federal Cell’s ability to
satisfy the required performance objectives, further evaluation is completely unnecessary.
2.1.3 Updated Groundwater Ingestion Tc-99 Concentration Estimate
A simple bounding calculation based on v1.4 of the Clive DU PA Model results shows that even
modest mixing with lower aquifer water and RO treatment process results in exposure
concentrations of Tc-99 associated with basal and deep groundwater below 3,760 pCi/L.
One thousand realizations of the Model were run with groundwater concentrations calculated
until 10,000 years, as requested in the interrogatory. The peak median concentration (i.e., the
median value of Tc-99 concentrations across the realizations at each time step) was about
4,000 pCi/L at the hypothesized shallow aquifer groundwater well location. The 95th percentile
values for all times were below 45,000 pCi/L. As such, a total dilution factor of ~12 would
reduce Tc-99 concentrations to below the limit of 3,760 pCi/L, even for the peak 95th percentile
concentration.
As discussed above, due to the high salinity of the groundwater, reverse osmosis or a similar
process would be required prior to consumption. Marschke (2015) assumed a conservative
dilution factor of 10 due to the RO treatment (i.e., 90% removal). RO Tc-99 removal rates as
high as 99% (dilution factor of 100) have been documented (Williamson 1992). As such,
removal of Tc-99 by RO is likely adequate to bring well water concentrations well below
3,760 pCi/L, even for the 95th percentile concentrations.
An additional dilution factor due to mixing with water from the lower aquifer was estimated in
the Round 2 Interrogatory Responses and was also utilized for calculations by Marschke (2015).
It was calculated using the Thiem-Dupuit equation for the most likely (though still highly
improbable) scenario of a well drilled near the Clive Site and screened in the lower aquifer. The
resulting dilution factor was about 300. Applying this factor along with the RO dilution factor
reduces Tc-99 concentrations to negligible levels, on the order of 10 pCi/L, even for the 95th
percentile case.
Groundwater Exposure Responses for the Clive DU PA Model
23 Feb 2018 10
In summary, though highly unlikely given the history and usage of the Clive Site, use of the
groundwater from the lower aquifer units would involve dilution of the contamination to an
extent that the risks to human health in a groundwater ingestion scenario would be negligible.
3.0 Conclusion
The poor water quality (Class IV) of water at the Clive Site limits its utility without treatment,
which would reduce radionuclide concentrations along with TDS concentrations. Substantial
withdrawals from the shallow aquifer are difficult due to low well yields, and cross-
contamination of the underlying aquifers is unlikely due to the natural hydrogeologic separation
and observed upward hydraulic gradients. Scenarios proposed by UDEQ involving
anthropogenic hydraulic connections are not applicable to the Clive Site, as these scenarios are
based on literature studies of areas with starkly contrasting usage histories and hydrogeologic
conditions.
4.0 References
Bingham Environmental, 1996. Revised Hydrogeologic Report, prepared for Envirocare of Utah
Inc., Bingham Environmental Inc., Salt Lake City UT, February 1996
EnergySolutions, 2013. Utah Radioactive Material License (RML UT2300249) Updated Site-
Specific Performance Assessment, Revision 1, prepared for Utah Division of Radiation
Control, EnergySolutions LLC, Salt Lake City UT, December 2013
EnergySolutions, 2014. Responses to August 11, 2014—Supplemental Interrogatories, Utah
LLRW Disposal License RML UT 2300249 Condition 35 Compliance Report, prepared
for Utah Division of Radiation Control, EnergySolutions LLC, Salt Lake City UT,
August 2014
Johnson, R.L., et al., 2011. Modeling the Potential Impact of Seasonal and Inactive Multi-
Aquifer Wells on Contaminant Movement to Public Water-Supply Wells, Journal of the
American Water Resources Association 47 (3) 588–596
Marschke, S., 2015. Groundwater Pathway Doses, Part 2, Revision 2, prepared for Utah
Department of Environmental Quality, SC&A Inc., Vienna VA, May 2015
Neptune, 2018. ET Cover Design Responses for the Clive DU PA Model, NAC-0106_R0,
Neptune and Company Inc., Lakewood CO, February 2018
Utah, 2014. Ground Water Quality Discharge Permit, Permit No. UGW450005, State of Utah,
Division of Water Quality, Salt Lake City UT, 2014
Williamson, D., 1992. Bench-Scale Testing of Reverse Osmosis to Remove Technetium 99 and
Trichloroethylene from Groundwater, Proceedings of National Research and
Groundwater Exposure Responses for the Clive DU PA Model
23 Feb 2018 11
Development Conference on the Control of Hazardous Materials, February 1992, San
Francisco CA
Zinn, B.A., and L.F. Konikow, 2007. Effects of Intraborehole Flow on Groundwater Age
Distribution, Hydrogeology Journal 2007 (15) 633–643 doi: 10.1007/s10040-006-0139-8

NAC-0103_R0

Recycled Uranium Responses for
the Clive DU PA Model
23 February 2018

Prepared by

NEPTUNE AND COMPANY, INC.
1435 Garrison St, Suite 201, Lakewood, CO 80215
Recycled Uranium Responses for the Clive DU PA Model
23 Feb 2018 ii
1. Title: Recycled Uranium Responses for the Clive DU PA Model
2. Filename: Recycled Uranium Responses for the Clive DU PA Model.docx
3. Description: Summary of how recycled uranium waste is addressed in GoldSim model
v. 1.4 and supporting documentation.

Name Date
4. Originator Gregg Occhiogrosso 13 Feb 2018
5. Reviewer Mike Sully and Dan Levitt
6. Remarks

Recycled Uranium Responses for the Clive DU PA Model
23 Feb 2018 iii
CONTENTS
CONTENTS ............................................................................................................................. iii
ACRONYMS AND ABBREVIATIONS ................................................................................... iv
1.0 Overview and Conceptual Model........................................................................................ 1
2.0 UDEQ Interrogatory Responses.......................................................................................... 2
2.1 Interrogatories CR R313-25-8(4)(a)-08/1 and CR R313-25-7(9)-51/3 ........................... 2
2.1.1 Interrogatory Response ............................................................................................ 2
3.0 Conclusion ......................................................................................................................... 4
4.0 References .......................................................................................................................... 5

Recycled Uranium Responses for the Clive DU PA Model
23 Feb 2018 iv
ACRONYMS AND ABBREVIATIONS
CAW Class A West embankment
DEQ (Utah) Department of Environmental Quality
DU depleted uranium
ET evapotranspiration
GDP gaseous diffusion plant
GWPL groundwater protection limit
PA performance assessment
PGDP Paducah Gaseous Diffusion Plant
PORTS Portsmouth Gaseous Diffusion Plant
SER Safety Evaluation Report
UDEQ Utah Department of Environmental Quality
WAC waste acceptance criteria

Recycled Uranium Responses for the Clive DU PA Model
23 Feb 2018 1
1.0 Overview and Conceptual Model
This document discusses issues related to the disposal of depleted uranium (DU) containing
recycled uranium in the Federal Cell. Recycled uranium waste contains fission products such as
Tc-99 which are not necessarily present in other DU wastes. UDEQ issued interrogatories related
to Tc-99’s source term concentration and the groundwater concentration at compliance
endpoints, but deemed the questions closed, stating that any license would be granted under the
condition that no DU containing recycled uranium would be disposed at the Site. Precedent
exists for disposal of Tc-99 at the Clive Site, as UDEQ has permitted Tc-99 disposal in the
adjacent Class A West (CAW) Disposal Cell. Issues surrounding disposal of DU containing
recycled uranium are discussed, including the mechanisms of exposure and responses to the
interrogatories (Section 2.0).
In the performance assessment (PA) model, all radionuclides, including Tc-99 and other fission
products, are transported via various transport processes which would cause contaminant
migration through the waste zone and into the underlying geologic units. Exposures are
evaluated at the hypothetical receptor locations (compliance points).
Contaminant transport includes transport media (water, air, soil), transport processes (advection-
dispersion, diffusion, plant uptake, soil translocation), and partitioning between phases.
Diffusion occurs in gas and water phases. Dilution occurs when mixing with less concentrated
water. Hydrodynamic dispersion is associated with water advection. Dissolution in water is
limited by aqueous solubility.
Infiltration of water through the cover, into wastes, and toward a hypothetical groundwater user
is the transport pathway for Tc-99 cited in UDEQ’s interrogatories related to recycled uranium.
Tc-99 is unique in that it is regarded as relatively long half-lived and mobile in the environment
because it does not generally sorb to the soil structure as readily as other radionuclides, meaning
it can be transported for longer distances.
The Clive DU PA Model calculates potential exposure concentrations in a probabilistic fashion
by incorporating uncertainty in each parameter value through Monte Carlo simulation, in which
the model is run thousands of times using different combinations of parameter values in each
realization. Details of the modeling processes can be found in the Conceptual Site Model for
Disposal of Depleted Uranium at the Clive Facility (Neptune 2015b).
Tc-99 has been accepted for disposal of waste with a limiting Tc-99 concentration in the adjacent
Class A West embankment, based on deterministic modeling using extreme bounding conditions.
The cited basis for the prohibition of DU containing recycled uranium is that the PA Model does
not adequately capture the range of possible waste concentrations. UDEQ goes on to claim that
the evapotranspiration (ET) cover performance may not adequately limit infiltration to assure
compliance with the groundwater protection limit (GWPL) at the hypothetical receptor location.
ET cover performance is addressed in ET Cover Design Responses for the Clive DU PA Model
(Neptune 2018a).
Recycled Uranium Responses for the Clive DU PA Model
23 Feb 2018 2
2.0 UDEQ Interrogatory Responses
This section contains responses for Interrogatories CR R313-25-8(4)(a)-08/1 and CR R313-25-
7(9)-51/3. UDEQ’s conclusions on the interrogatories regarding the disposal of wastes
containing recycled uranium are reproduced below. A single response is provided since the
interrogatories are essentially identical.
2.1 Interrogatories CR R313-25-8(4)(a)-08/1 and CR R313-25-7(9)-51/3
Interrogatory CR R313-25-8(4)(a)-08/1: Groundwater Concentration Endpoints
DEQ has stated that no depleted uranium (DU) waste containing recycled uranium will be
allowed to be disposed at Clive, so this interrogatory is closed.
Interrogatory CR R313-25-7(9)-51/3: Nature of Contamination
This interrogatory is closed because any license amendment will contain a license condition that
disposal of recycled uranium is not allowed in the DU waste. Furthermore, the license condition
will indicate that DU-waste containers shall contain neither heels of enriched uranium at average
concentrations greater than that allowed in the license nor heels of transuranic compounds at
average concentrations greater than 10 pCi/g (the Class A limit).
2.1.1 Interrogatory Response
Interrogatories 08/1 and 51/3 were deemed closed under the unsubstantiated condition that no
DU waste containing recycled uranium would be disposed at the Site, to be dictated by License
condition. These interrogatories are linked by their concern with groundwater exposures to
radionuclides, principally fission product radionuclides like Tc-99 which may be present in DU
wastes containing recycled uranium. UDEQ cited uncertainty in Tc-99 concentrations in
groundwater relative to the GWPL of 3790 pCi/L over the compliance period of 500 years to
justify the prohibition of waste containing recycled uranium (April 2015 SER, Section 6.1.2):
Because there is significant uncertainty regarding the Tc-99 concentration in the DU3O8
to be produced from the GDP tailings, and because Tc-99 and other mobile isotopes
may exceed the GWPL at 500 years, DEQ approves this portion of the DU PA with the
condition that no DU waste containing recycled uranium be accepted for disposal inside
the Federal Cell at Clive. Based on this restriction, GWPLs for the 500-year compliance
period can easily be met regardless of uncertainties in the infiltration rate through the ET
cover.
The concerns expressed are thus related to the waste concentrations upon emplacement (as
opposed to hypothetical ranges of concentrations in DU being stored at the Clive Facility) and
the performance of the cover system and the associated infiltration rates, which impact potential
exposure concentrations at hypothetical receptor locations.
From a risk perspective, this condition is arbitrary and unsupported, given that modeling results
demonstrate that DU containing Tc-99 and other fission products can be safely disposed at the
Federal Cell. Furthermore, Tc-99 in other Class A wastes is licensed for disposal in other
comparable embankments at the Clive Site. The relevant modeling concepts are briefly
summarized below, followed by responses to the concerns expressed in the SER which
precipitated the closure of these interrogatories.
Recycled Uranium Responses for the Clive DU PA Model
23 Feb 2018 3
Comparison of v1.4 Model Results to GWPL
Refinements to the PA Model incorporated in v1.4 have resulted in predicted Tc-99
concentrations in the upper aquifer well below the GWPL (3,790 pCi/L) over the 500-year
compliance period. As presented in the v1.4 Final Report for the Clive DU PA Model (Neptune
2015a), ten thousand realizations of the Model were run and the peak Tc-99 concentration at the
hypothetical groundwater compliance well was recorded for each run. The mean and median
peak values across the realizations were 26 pCi/L and 4.3E-2 pCi/L, respectively. The 95th
percentile peak value was 150 pCi/L. These results are a small fraction of the GWPL of
3790 pCi/L.
Due to these updated estimates, the prohibition of DU wastes containing recycled uranium is not
required to remain in compliance with the GWPL, as even the 95th percentile estimates of peak
Tc-99 concentrations are more than an order of magnitude below the GWPL for the 500-year
compliance period. Any refusal to remove a prohibition on disposal of recycled DU is arbitrary,
unsupported by the Model, and contrary to site precedence.
Responses to UDEQ Concerns
1) ET Cover Performance
Issues pertaining to the ET cover performance and net infiltration into the waste zone are
discussed in ET Cover Design Responses for the Clive DU PA Model (Neptune 2018a).
2) Tc-99 Waste Concentrations
Some uncertainty in the Tc-99 concentrations of wastes derived from gaseous diffusion plant
(GDP) operations is acknowledged, as direct measurements of fission product concentrations are
rare. However, a blanket prohibition of DU containing recycled uranium is not justifiable due to
uncertainties in Tc-99 concentrations. Tc-99 disposals at the Site in the CAW cell undergo
receipt and acceptance actions to confirm compliance with the waste acceptance criteria (WAC),
with WAC values established using a transport modeling approach not unlike the Federal Cell
PA Model. Federal Cell disposals would be subject to similar receipt and acceptance procedures,
and a WAC could be established in a similar fashion.
UDEQ’s contention that Tc-99 concentrations in the Model are not adequate is based on Section
4.1.2 of the April 2015 SER, which hypothesizes a possible Tc-99 concentration in cylinder
heels as high as 5,700,000 ppb. This value is based on a mass balance approach summarized in
Hightower et al. (2000); it should be interpreted as an extreme bounding value, and is described
as such by Hightower et al. (2000). As noted in NUREG/CR-2300 (NRC 1983):
The simplest quantitative measure of variability in a parameter or a measurable quantity is
given by an assessed range of the values the parameter or quantity can take. This measure
may be adequate for certain purposes (e.g., as input to a sensitivity analysis), but in general it
is not a complete representation of the analyst's knowledge or state of confidence and
generally will lead to an unrealistic range of results if such measures are propagated through
an analysis.
Recycled Uranium Responses for the Clive DU PA Model
23 Feb 2018 4
As such, a bounding, hypothetical maximum value is of limited utility and should be discounted
in favor of measured values from similar waste streams. A more realistic, measured value for an
upper Tc-99 cylinder heel concentration is presented in Appendix C of Hightower et al. (2000),
where measurements of heel concentrations were taken from a cylinder used multiple times at
both Portsmouth and Paducah:
Several empty feed cylinders have been identified that contain heels of feed and/or product
from the periods when reactor returns were being fed to the cascades. Cylinder 003174 is
such a cylinder with 13 lb of feed material heel remaining. This cylinder had been filled at
the PGDP [Paducah Gaseous Diffusion Plant] UF6 feed plant with UF6 prepared from
reactor returns material sometime after the cylinder’s purchase date of June 1954. After the
original charge in this cylinder had been fed to the cascades, the cylinder was filled with
enriched product from Paducah and shipped to the PORTS [Portsmouth Gaseous Diffusion
Plant], where the enriched material was fed to the Portsmouth cascade for further
enrichment. This cylinder was similarly refilled with enriched product and emptied two more
times without any washing, leaving its original heel of reactor returns feed material
combined with the heels left from emptying enriched product three times into the cascades.
(Hightower et al. (2000), Appendix C)
The highest, “first wash” Tc-99 concentration for this cylinder is reported as 270,000 ppb, a
factor of 20 lower than UDEQ’s hypothetical extreme bounding value derived from the mass
balance approach. The April 2015 SER claimed that the PA Model concentrations were as much
as 3.7 times too low when compared to the value of 5,700,000 ppb. When compared to the
measurements from this heavily used cylinder, however, the uncertainty currently incorporated
in the Model for Tc-99 concentrations is representative.
It is also worth noting that mobilization of the heel material for concentration measurement
required dissolving the heel with an acidic wash solution. The Model conservatively takes no
transportation rate credit for this solid waste form or waste packaging, and assumes all
radionuclides are immediately available for transport in the liquid phase at the first model time
step, a simple and conservative approach compared to more complicated release models that
incorporate mass-transfer limitations due to, for example, solubility and/or sorption
(NUREG/CR-5532, Kozak et al. (1990)). Additional discussion of the potential for doses via
groundwater is discussed in the response to Interrogatory CR R313-25-20-204/1: Exposure to
Groundwater in Groundwater Exposure Responses for the Clive DU PA Model (Neptune 2018b).
3.0 Conclusion
UDEQ’s stated prohibition of disposal of DU containing recycled uranium is based on
uncertainties in the performance of the cover system and in the concentration of fission products
in the wastes. As shown above, the latter concern is based on an extreme theoretical value.
Furthermore, the prohibition stands in contrast to site precedence, as disposals of Tc-99 have
been permitted in the CAW Cell, subject to WAC limitations. A similar framework could be
applicable to the Federal Cell, though DU PA Model v1.4 results suggest the risk associated due
to disposal of DU containing recycled uranium is minimal, and that no such WAC limitation is
warranted.

Recycled Uranium Responses for the Clive DU PA Model
23 Feb 2018 5
4.0 References
Hightower, J.R., et al., 2000. Strategy for Characterizing Transuranics and Technetium
Contamination in Depleted UF6 Cylinders, ORNL/TM-2000/242, Oak Ridge National
Laboratory, Oak Ridge TN, October 2000
Kozak, M.W., et al., 1990. A Performance Assessment Methodology for Low-Level Waste
Facilities, NUREG/CR-5532, SAND90-0375, prepared for United States Nuclear
Regulatory Commission, Sandia National Laboratories, Albuquerque NM, July 1990
Neptune, 2015a. Final Report for the Clive DU PA Model, Clive DU PA Model v1.4, NAC-
0024_R4, Neptune and Company Inc., Los Alamos NM, November 2015
Neptune, 2015b. Conceptual Site Model for Disposal of Depleted Uranium at the Clive Facility,
Clive DU PA Model v1.4, NAC-0018_R4, Neptune and Company Inc., Los Alamos NM,
November 2015
Neptune, 2018a. ET Cover Design Responses for the Clive DU PA Model, NAC-0106_R0,
Neptune and Company Inc., Lakewood CO, February 2018
Neptune, 2018b. Groundwater Exposure Responses for the Clive DU PA Model, NAC-0104_R0,
Neptune and Company Inc., Lakewood CO, February 2018
NRC, 1983. PRA Procedures Guide: A Guide to the Performance of Probabilistic Risk
Assessments for Nuclear Power Plants, Volume 2, Chapters 9–13 and Appendices A–G,
NUREG/CR-2300, Vol. 2, prepared by The American Nuclear Society and The Institute
of Electrical and Electronics Engineers, United States Nuclear Regulatory Commission,
Washington DC, January 1983

NAC-0101_R0

Federal Cell Design Responses for
the Clive DU PA Model
23 February 2018

Prepared by

NEPTUNE AND COMPANY, INC.
1435 Garrison St, Suite 201, Lakewood, CO 80215
Federal Cell Design Responses for the Clive DU PA Model
23 Feb 2018 ii
1. Title: Federal Cell Design Responses for the Clive DU PA Model
2. Filename: Federal Cell Design Responses for the Clive DU PA Model.docx
3. Description: Summary of Federal Cell design assumptions and considerations in GoldSim
model v. 1.4.

Name Date
4. Originator Sean McCandless 12 Feb 2018
5. Reviewer Mike Sully 12 Feb 2018
6. Remarks

Federal Cell Design Responses for the Clive DU PA Model
23 Feb 2018 iii
CONTENTS
CONTENTS ............................................................................................................................. iii
FIGURES .................................................................................................................................. iv
ACRONYMS AND ABBREVIATIONS .................................................................................... v
1.0 Overview and Conceptual Model........................................................................................ 1
2.0 UDEQ Interrogatory Responses.......................................................................................... 4
2.1 Interrogatory CR R313-25-7(2) and 7(6)-81/2: Comparison of Disposal Cell Designs .. 4
2.1.1 Interrogatory Response ............................................................................................ 4
2.2 Interrogatory CR R313-25-7(6)-84/3: Below-Grade Disposal of DU ............................. 4
2.2.1 Interrogatory Response ............................................................................................ 4
2.3 Interrogatory CR R313-25-7(2)-160/2: Comparison of Class A West and Federal Cell
Designs ......................................................................................................................... 5
2.3.1 Interrogatory Response ............................................................................................ 5
2.4 Interrogatory CR R313-25-22-162/2: Disposal Cell Stability ........................................ 6
2.4.1 Interrogatory Response ............................................................................................ 6
3.0 References .......................................................................................................................... 6

Federal Cell Design Responses for the Clive DU PA Model
23 Feb 2018 iv
FIGURES
Figure 1. Federal Cell and 11e.(2) Cell engineering drawing 14004 V1A (Neptune 2015b). ........ 2
Figure 2. Federal Cell and 11e.(2) Cell engineering drawing 14004 V3A (west-east cross
section) (Neptune 2015b). .......................................................................................... 3

Federal Cell Design Responses for the Clive DU PA Model
23 Feb 2018 v
ACRONYMS AND ABBREVIATIONS
CAW Class A West embankment
CQA/QC Construction Quality Assurance/Quality Control
DU depleted uranium
ET evapotranspiration
LLRW low-level radioactive waste
PA performance assessment
SER Safety Evaluation Report
UDEQ Utah Department of Environmental Quality

Federal Cell Design Responses for the Clive DU PA Model
23 Feb 2018 1
1.0 Overview and Conceptual Model
Version 1.4 of the Clive DU PA is based on a Federal Cell design that physically isolates DU
and Class A LLRW from the adjacent 11e.(2) embankment. This is a change from the approach
modeled in versions 1.0 and 1.2, which consisted of a single cell with a barrier wall intended to
isolate the different waste types.
The choice to revise the Federal Cell design to a standalone embankment was made for several
reasons:
1. The Federal Cell is proposed to have an evapotranspiration (ET) cover design, while the
existing 11e.(2) cell has a rock armor cover design. Final cover has been constructed over
a large portion of the currently open 11e.(2) cell footprint. Joining the two cover designs
would be difficult to engineer and model.
2. Remediating the existing completed 11e.(2) cover to an ET design would be costly in
terms of time, labor, and materials. The relatively small loss of overall site capacity,
compared with a single-cell design, was not judged to warrant the activity.
Appendix 3 to Neptune (2015a), Embankment Modeling for the Clive DU PA, Clive DU PA
Model v. 1.4 (Neptune 2015b), explains that version 1.4 of the DU PA considers a single,
standalone Federal Cell. This cell is located to the west of the existing 11e.(2) mill tailings cell
and to the south of the existing Class A West (CAW) cell. EnergySolutions engineering drawing
series 14004 was referenced and included in Appendix 3. Drawings 14004 V1A and 14004 V3A
are reproduced from this report as Figure 1 and Figure 2, respectively, below.
The Federal Cell isolates DU (and, ultimately, overlying Class A LLRW) from the adjacent
11e.(2) embankment. The Federal Cell has no existing waste placement of any type. It appears
that at least one of the interrogatories on this topic fails to recognize the design revision from
version 1.2 to version 1.4 of the DU PA Model.
Federal Cell Design Responses for the Clive DU PA Model
23 Feb 2018 2

Figure 1. Federal Cell and 11e.(2) Cell engineering drawing 14004 V1A (Neptune 2015b).
Federal Cell Design Responses for the Clive DU PA Model
23 Feb 2018 3

Figure 2. Federal Cell and 11e.(2) Cell engineering drawing 14004 V3A (west-east cross
section) (Neptune 2015b).
Federal Cell Design Responses for the Clive DU PA Model
23 Feb 2018 4
2.0 UDEQ Interrogatory Responses
This section contains responses for Interrogatories CR R313-25-7(2), CR R313-25-7(6)-81/2, CR
R313-25-7(6)-84/3, CR R313-25-7(2)-160/2, and CR R313-25-22-162/2.
2.1 Interrogatory CR R313-25-7(2) and 7(6)-81/2: Comparison of
Disposal Cell Designs
No further analysis has been performed on disposal cell designs since v1.2.
2.1.1 Interrogatory Response
This statement is incorrect. Version 1.2 of the DU PA proposed to dispose DU, Class A LLRW,
and 11e.(2) waste within a single embankment, with DU and Class A LLRW separated from
11e.(2) waste by a clay barrier wall. As discussed above, version 1.4 of the DU PA models a
distinct Federal Cell that is completely separate from the 11e.(2) cell.
The interrogatory basis furthermore states:
…EnergySolutions has advised that the proposed Federal Cell will be physically separated from
the 11e.(2) cell. EnergySolutions has provided only engineering drawings but no written
description of the new cell (i.e., Appendices 3 and 16 to the DU PA have not been revised).
This statement is also incorrect. Appendices 3 and 16 of the DU PA, v1.4, correctly incorporate
the Federal Cell design and dimensions as a single, standalone embankment separate from the
11e.(2) cell. There is no reason for this interrogatory to remain open.
2.2 Interrogatory CR R313-25-7(6)-84/3: Below-Grade Disposal of DU
No further analysis was performed in Appendix 21 on the below-grade disposal of DU.
2.2.1 Interrogatory Response
The interrogatory basis discusses a discrepancy between the reported potential number of DU
drums and cylinders in Appendices 3 and 4 to Neptune (2015a):
Please explain the difference between v1.4 Appendices 3 and 4 regarding the maximum number
of cylinders and drums, and demonstrate how the entire DU inventory can be disposed below
grade.
This discrepancy traces back to the use of different EnergySolutions drawings as points of
reference for the Neptune documents. Appendix 3 to Neptune (2015a) includes engineering
drawing 14004 L1A as Figure 7 and as the basis for the potential number of DU drums and
cylinders that could physically fit below grade in the Federal Cell. Appendix 4 to Neptune
(2015a) references drawing 14004 SK1 as its data source. Clearly, the different drawings are
intended to project different disposal scenarios.
While the difference between the scenarios is acknowledged, it is also irrelevant to the licensing
decision. Section 6.2.4 of the 2015 SER provides Condition 2 that would apply to an amended
Federal Cell Design Responses for the Clive DU PA Model
23 Feb 2018 5
License approving DU disposal: “All DU waste must be disposed of below the original-grade
level of the proposed Federal Cell (i.e., 4,272 ft-amsl).” This constraint will control waste
placement operations to be consistent with the important model assumption that DU waste is
placed below the original grade of the Federal Cell, regardless of the number or type of cylinders
and/or drums disposed.
There is no Utah or Federal Rule whereby EnergySolutions must demonstrate that all DU in
storage (or that will ever be created at any site within the United States) must completely fit
within the permitted capacity of the Federal Cell. However, this interrogatory demands, without
regulatory basis, that EnergySolutions “demonstrate how the entire DU inventory can be
disposed below grade.” By refusing to close interrogatories associated with DU capacity under
various disposal configurations, UDEQ’s reviewers imply a requirement that the Site be able to
dispose of the full national inventory of DU available for disposal. This is without regulatory
basis and is thus arbitrary and capricious on the part of UDEQ. Note also that, if approved, the
EnergySolutions Clive facility would not be the only U.S. site licensed for DU disposal.
EnergySolutions will only place DU below grade within the approved top slope footprint of the
Federal Cell, and in accordance with Construction Quality Assurance/Quality Control
(CQA/QC) requirements applicable for the waste form and container. These controls, together
with the type and timing of shipments for disposal, will drive actual cell utilization. Disposal
operations will be performed to optimize cell utilization within License constraints, but will often
result in under-utilized capacity due to material handling and spacing needs, use of clean fill for
shielding and contamination control, and other inefficiencies.
For example, the figures used to estimate potential total cylinders and drums that could be
disposed under various configurations assume regular, tight spacing for each container. It will
not always be operationally practical to maintain tight spacing, perhaps due to rigging or
equipment needs. Regardless of these realities and of the actual volume ultimately disposed,
further disposal of DU in the Federal Cell will cease once the below grade capacity is exhausted.
2.3 Interrogatory CR R313-25-7(2)-160/2: Comparison of Class A
West and Federal Cell Designs
See also Interrogatory CR R313-25-25(4)-202/1: Use of SIBERIA to Model Federal Cell Erosion.
2.3.1 Interrogatory Response
A response to Interrogatory 202/1 regarding use of SIBERIA to model erosion is provided in
Erosion Responses for the Clive DU PA Model (Neptune 2018).
The basis for Interrogatory 160/2 goes on to discuss a prior request (SC&A 2015) to compare
similar ET cover designs between the Class A West and Federal Cell:
DRC is currently reviewing a license amendment request to use an ET cover of similar design to
that proposed for the Federal Cell in the DU PA. Any recommendations and conclusions from
that review must be applied to the proposed Federal Cell as well.
Federal Cell Design Responses for the Clive DU PA Model
23 Feb 2018 6
Since submitting the request for approval of an ET cover design on the CAW cell,
EnergySolutions has begun construction of a rock armor cover on the CAW cell. More
importantly, the DU PA models the Federal Cell’s ET cover as designed. CAW cover licensing
is separate and distinct from the Federal Cell. Since their geometry differs, any attempt to couple
or join the reviews is unsupported and arbitrary. Therefore, this aspect of the question is moot.
2.4 Interrogatory CR R313-25-22-162/2: Disposal Cell Stability
This interrogatory can be closed because the same issues are raised in Interrogatory CR R313-
25-7(2)-160/2: Comparison of Class A West and Federal Cell Designs, which remains open.
2.4.1 Interrogatory Response
We concur that this interrogatory should be closed. See above for discussion of Interrogatory CR
R313-25-7(2)-160/2.
3.0 References
Neptune, 2015a. Final Report for the Clive DU PA Model, Clive DU PA Model v1.4, NAC-
0024_R4, Neptune and Company Inc., Los Alamos NM, November 2015
Neptune, 2015b. Embankment Modeling for the Clive DU PA, Clive DU PA Model v1.4, NAC-
0019_R4, Neptune and Company Inc., Los Alamos NM, October 2015
Neptune, 2018. Erosion Responses for the Clive DU PA Model, NAC-0108_R0, Neptune and
Company Inc., Lakewood CO, February 2018
SC&A, 2015. Utah Division of Radiation Control, EnergySolutions Clive LLRW Disposal
Facility, License No: UT2300249; RML #UT 2300249, Condition 35 Compliance Report;
Appendix A: Final Report for the Clive DU PA Model, Safety Evaluation Report, Volume
1, SC&A Inc., Vienna VA, April 2015