Cosler_Expert_Report_Marshall

Expert Report of
Douglas J. Cosler, Ph.D., P.E.
Chemical Hydrogeologist
Adaptive Groundwater Solutions LLC
Charlotte, North Carolina
Marshall Steam Station Ash Basin
Terrell, North Carolina
April 18, 2016

Introduction
Site Backqround
The Marshall Steam Station is a four -unit, coal-fired generating station owned by Duke Energy and
located on a 1,446 -acre site on the west bank of Lake Norman near the town of Terrell in Catawba
County, North Carolina. Coal combustion residuals ("coal ash") have historically been disposed in a
single -cell ash basin impoundment located north of the power plant. The ash basin was constructed in
1965 by building an earthen dike at the confluence of Holdsclaw Creek and the Catawba River (now Lake
Norman) and is generally located in historical depressions formed from Holdsclaw Creek and small
tributaries that discharged into the creek. Two unlined ash landfill units (Marshall dry ash landfill) are
located adjacent to the east (Phase 1) and northeast (Phase II) portions of the ash basin. A flue gas
desulfurization (FGD) landfill containing bottom ash and various other types of waste materials is located
to the west of the ash basin and is constructed with an engineered liner system. Industrial Landfill No. 1
(containing fly and bottom ash and various other types of waste materials) is located adjacent to the
northern portion of the ash basin and contains a liner and leachate collection system. The demolition
landfill (construction and demolition waste) is also located adjacent to the northern portion of the ash
basin (directly north of the dry ash landfill, Phase II). The photovoltaic farm structural fill (PV structural fill)
is constructed of fly ash and located adjacent to and partially on top of the northwest portion of the ash
basin.
Duke Energy performed voluntary groundwater monitoring at the site from November 2007 to October
2011 (nine sampling events) and NPDES permit -required compliance monitoring starting in February
2011 (sampling three times per year). Recent groundwater sampling results at Marshall indicate
exceedances of 15A NCAC 02L.0202 Groundwater Quality Standards (2L Standards). In response to
this, the North Carolina Department of Environmental Quality (NC DEQ) required Duke Energy to perform
a groundwater assessment at the site and prepare a Comprehensive Site Assessment (CSA) report. The
Coal Ash Management Act of 2014 (CAMA) also required owners of surface impoundments containing
coal combustion residuals (CCR) to conduct groundwater monitoring and assessment and prepare a CSA
report. The September 2015 CSA prepared by HDR Engineering, Inc. of the Carolinas (HDR) for the
Marshall site determined that ash handling and storage at the Marshall site have impacted soil and
groundwater beneath and downgradient from the ash basin. The CSA report identified Constituents of
Interest (COI) considered to be associated with potential impacts to soil and groundwater from the ash
basin and assessed COI concentration distributions in soil, groundwater, and seeps.
CAMA also requires the submittal of a Corrective Action Plan (CAP); the CAP for the Marshall site
consists of two parts. CAP Part 1 (submitted to DEQ in December 2015) provides a summary of CSA
findings, further evaluation and selection of COI, a site conceptual model (SCM), the development of
2

groundwater flow and chemical transport models of the site, presentation and analysis of the results of
the modeling, and a quantitative analysis of groundwater and surface water interactions. The CAP Part 2
contains proposed remedial methods for achieving groundwater quality restoration, conceptual plans for
recommended corrective action, proposed future monitoring plans, and a risk assessment.
Information Reviewed
My opinions are based upon an analysis and technical review of (i) hydrogeologic and chemical data
collected at the Marshall site; (ii) the analyses, interpretations, and conclusions presented in site -related
technical documents and reports; (iii) the groundwater flow and chemical transport models constructed
for the site (including model development, calibration, and simulations of remedial alternatives); (iv) the
effectiveness of proposed remedial alternatives to achieve groundwater quality restoration; and (v)
proposed future site monitoring. These opinions are subject to change as new information becomes
available.
As a basis for forming my opinions I reviewed the following documents and associated appendices:
(1) Comprehensive Site Assessment Report, Marshall Steam Station Ash Basin (September 8, 2015);
(2) Corrective Action Plan, Part 1, Marshall Steam Station Ash Basin (December 7, 2015);
(3) Corrective Action Plan, Part 2, Marshall Steam Station Ash Basin (March 3, 2016);
(4) Miscellaneous historical groundwater and soil concentration data for the Marshall site collected
prior to the CSA; and
(5) Specific references cited in and listed at the end of this report.
Professional Qualifications
I have advanced graduate degrees in Hydrogeology (Ph.D. Degree from The Ohio State University) and
Civil and Environmental Engineering (Civil Engineer Degree from the Massachusetts Institute of
Technology), and M.S. and B.S. degrees from Ohio State in Civil and Environmental Engineering. I have
36 years of experience as a chemical hydrogeologist and environmental engineer investigating and
performing data analyses and computer modeling for a wide variety of projects. These projects include:
investigation, remediation, and regulation of Superfund, RCRA, and other hazardous waste sites involving
overburden and bedrock aquifers; ground water flow and chemical transport model development; natural
attenuation/biodegradation assessments for chlorinated solvent and petroleum contamination sites;
volatile organic compound vapor migration and exposure assessment; exposure modeling for health risk
assessments; hydrologic impact assessment for minerals and coal mining; leachate collection system
modeling and design for mine tailings disposal impoundments; and expert witness testimony and
litigation support. I also develop commercial groundwater flow and chemical transport modeling software
for the environmental industry.
3

The types of sites I have investigated include: landfills, mining operations, manufactured gas plants,
wood -treating facilities, chemical plants, water supply well fields, gasoline and fuel oil storage/delivery
facilities, nuclear waste disposal sites, hazardous waste incinerators, and various industrial facilities. I
have investigated the following dissolved, nonaqueous-phase (LNAPL/DNAPL), and vapor -phase
contaminants: chlorinated solvents, various metals, gasoline and fuel oil constituents, wood -treating
products, coal tars, polychlorinated biphenyls, pesticides, dioxins and furans, phenolic compounds, flame
retardants (PBDE), phthalates, radionuclides, and biological constituents.
Summary of Opinions
The following is a brief summary of the opinions developed in my report:
• A total of 21 Compliance Boundary groundwater samples exceeded North Carolina groundwater
standards for these COI: antimony, boron, chromium, cobalt, iron, manganese, total dissolved
solids, and vanadium. Of these 21 exceedances, 19 were greater than the proposed provisional
background concentrations by HDR, which exaggerate background levels;
• The statistical analyses of historical, shallow -aquifer background groundwater concentrations at
the Marshall site (monitoring well MW -4) are invalid due to the characteristically slow rate of COI
migration in groundwater;
• There is a significant risk of chemical migration from the ash basin to neighboring private and
public water supply wells in fractured bedrock;
• Major limitations of the CAP Parts 1 and 2 groundwater flow and chemical transport models
prevent simulation and analysis of off-site migration;
• The CAP Closure Scenario simulations greatly underestimate (by factors of 10 or more) the time
frames required to achieve meaningful groundwater concentration reductions in response to
remedial actions;
• For either the Existing Condition or Cap -in -Place Scenario groundwater concentrations of many
coal -ash constituents much higher than background levels will continue to exceed North Carolina
groundwater standards at the Compliance Boundary because saturated coal -ash material and
secondary sources will remain in place;
• Source -area mass removal included in the Excavation Scenario results in COI groundwater
concentration reductions that are at least a factor of 10 greater compared to Cap -in -Place at
many locations. Additional excavation of secondary sources would further accelerate
concentration reductions;
• CSA data show 22 exceedances of groundwater standards in bedrock inside the Compliance
Boundary. However, the CAP Closure Scenarios do not address off-site chemical migration
control in the fractured bedrock aquifer;
4

• Due to an incorrect boundary -condition representation of the active ash basin, the CAP models
underestimate by a factor of three or more both the mass loading of COI into Lake Norman and
the corresponding Lake Norman water concentrations (attributable to coal ash ponds) estimated
by the groundwater/surface-water mixing model;
• The CAP Part 2 geochemical modeling and monitored natural attenuation (MNA) evaluations do
not provide the required quantitative analyses of COI attenuation rates necessary to support MNA
as a viable corrective action. The CAP 2 chemical transport modeling, which included attenuation
by sorption, demonstrated that MNA is not an effective remedial option for several COI (e.g.,
antimony, beryllium, boron, chromium, cobalt, hexavalent chromium, thallium, and vanadium);
and
• Future Compliance Monitoring at the Marshall site should include much more closely -spaced
Compliance Wells to provide more accurate detection, and groundwater sampling frequency
should be re-evaluated to allow valid statistical analyses of concentration variations.
Hydrogeology of the Marshall Site
Introduction
The groundwater system at the Marshall site is an unconfined, connected system consisting of three
basic flow layers: shallow, deep, and fractured bedrock. The shallow and deep layers consist of residual
soil, saprolite (clay and coarser granular material formed by chemical weathering of bedrock), and
weathered fractured rock (regolith). A transition zone at the base of the regolith is also present and
consists of partially-weathered/fractured bedrock and lesser amounts of saprolite. The single -cell ash
basin impoundment overlies native soil and was constructed in 1965 by building an earthen dike at the
confluence of Holdsclaw Creek and the Catawba River (now Lake Norman). Most of the coal ash lies
below the groundwater table and is saturated. Groundwater flow through saturated coal ash and
downward infiltration of rainwater through unsaturated coal ash leach COI into the subsurface beneath
the basin and via seeps through the embankments.
As described by HDR, groundwater flow in all three layers within the site boundary is generally from the
northwest and north to the southeast toward Lake Norman. Vertical groundwater flow between the three
layers also occurs, and surface water ponding in the ash basin effects flow directions locally. On the
downgradient eastern site boundary the CSA and CAP Parts 1 and 2 investigations assumed that all
groundwater (overburden and bedrock aquifers) discharges into Lake Norman. However, these studies
did not collect hydrogeologic data or perform data analyses or groundwater flow modeling to support this
assumption. The CSA and CAP Parts 1 and 2 also did not analyze potential changes to site groundwater
flow directions, or the risk of off-site migration of COI in the overburden or bedrock aquifers, caused by
5

groundwater extraction from numerous private and public water supply wells located close to the site
boundaries.
My report begins with a discussion of significant errors in CSA data analysis and conceptual model
development that contradict HDR's interpretation of three-dimensional groundwater flow patterns at the
Marshall site. This is followed by a presentation and discussion of measured exceedances of North
Carolina groundwater standards at multiple locations on the ash basin compliance boundary. I then
address several limitations of the CAP Parts 1 and 2 groundwater flow and chemical transport models
and identify various model input data errors. Finally, I present my evaluations of the CAP Closure
Scenario simulations and provide my opinions regarding the effectiveness of various remedial alternatives
for restoring groundwater quality to North Carolina standards.
Errors in Hydraulic Conductivity Test Analyses
Background
In the CSA and CAP reports HDR provides interpretations and conclusions regarding the horizontal and
vertical variations of groundwater flow directions and rates, and the fate and transport of COI dissolved in
groundwater. The most important site-specific parameter that controls these time -dependent flow and
transport mechanisms is the hydraulic conductivity (also referred to as "permeability") of the underlying
soils and fractured bedrock (Bear, 1979). Hydraulic conductivity (length/time) is a media -specific
measure of the rate at which water can flow through a porous (soil) or fractured (bedrock) porous
medium. Groundwater flow and chemical transport rates are directly proportional to the product of
hydraulic conductivity and the hydraulic gradient (hydraulic head difference between two points divided by
the separation distance; e.g., the water table elevation slope). Therefore, accurate measurement of
hydraulic conductivity is critical for understanding the current and future distributions of COI in soil and
groundwater and for evaluating the effectiveness (e.g., cleanup times) of alternative remedial measures.
In addition, the contrast in hydraulic conductivity between adjacent hydrogeologic units is the key factor in
determining three-dimensional groundwater flow directions and the ultimate fate of dissolved COI. For
example, at the Marshall site accurate measurement of hydraulic conductivity is critical in evaluating the
potential for: downward chemical migration into the fractured bedrock unit, off-site COI migration in the
overburden (soil) or fractured bedrock aquifers, and groundwater flow and COI transport into or beneath
Lake Norman.
A slug test is one of the standard field methods for measuring hydraulic conductivity (l) using a soil
boring or installed monitoring well. Slug tests were performed in most of the overburden and bedrock
wells at the Marshall site. In this test the static water level in the open hole (boring) or well casing is
N.

suddenly increased or decreased and the resulting transient change in water level is recorded. Two
commonly -used techniques for quickly changing the water level are the introduction (increases the water
level) and removal (decreases the water level) of a solid rod, or "slug" into the boring or well casing.
These tests are called "falling -head" and "rising -head" tests, respectively. Higher rates of water -level
recovery correspond to higher values of K. The measurements of water level versus time are analyzed
using mathematical models of the groundwater flow hydraulics and information regarding the well
installation (e.g., length of the slotted monitoring well screen and well casing diameter) to compute an
estimate of K.
As discussed below, HDR made significant errors in all of their analyses of field slug test data. Their
analysis errors caused the reported (CSA report) slug test hydraulic conductivity values to be as large as
a factor of 1.5 (almost 50 percent) smaller than the correct K values. I discuss the impacts of these
analysis errors on HDR's groundwater flow and chemical transport assessments and the CAP modeling
later in my report.
Overburden Slug Tests
HDR analyzed all of the CSA overburden slug tests in shallow and deep wells with the Bouwer-Rice
(1976) method using vertical anisotropy, Av = Kh,,n onrat 1 Kverticat , values that are as large as a factor of 20
lower than the values presented in the CSA report (e.g., compare geometric mean values in CSA Tables
11-9 and 11-10) and used in the CAP modeling (e.g., CAP 1 report Appendix C, Table 2), where K is
hydraulic conductivity. Comparing CSA Tables 11-9 and 11-10, the measured Av for overburden soil units
ranges from 2 to 50. In the calibrated CAP flow model Av = 10 for overburden soil. However, the
Bouwer-Rice slug test analyses assumed Av = 2 for every monitoring well (CSA Appendix H). Using the
measured M1 hydrostratigraphic vertical anisotropy, Av = 35, increases all of the M1 CSA overburden
hydraulic conductivity values (CSA Table 11-3) by almost 50 percent (factor of 1.5), depending on how
the slug -test radius of influence was computed. If the CAP 1 flow model results (Av = 10) are used in the
Bouwer-Rice analyses all of the measured overburden hydraulic conductivity values increase by almost
30 percent (factor of 1.3).
Since most of the reported overburden K values in the CSA report are up to 50 percent too low, the actual
average chemical transport rates in overburden soils are up to 50 percent greater than reported. This
site -wide data reduction error also affects the CAP flow and transport model calibrations. For example,
the transport model developers significantly reduced laboratory measurements of the soil -water partition
coefficient, Kd, for various COI during the transport model calibration based on comparisons of observed
and simulated chemical migration rates. However, if the correct (i.e., higher) overburden K values had
been used in the model calibration the Kd values would not have been reduced as much (compared to
laboratory values). The reason for this is, assuming linear equilibrium partitioning of COI with soil, the
7

chemical migration rate is proportional to K/Kd (except for Kd << 1). The CAP 1 and 2 transport model
history matching indicated that the simulated transport rate was too low, so HDR reduced the model Kd.
In other words, the reductions in calibrated Kd values would not have been as great if the correct (higher)
K values were used in the first place. As discussed below, the CAP Part 1 transport modeling used Kd
values that were typically factors of 10 - 100 smaller than the measured site-specific Kd's reported in CAP
Appendix D. In contrast, the CAP Part 2 transport modeling used Kd values that are generally a factor of
about 10 larger than the CAP 1 values; however, the CAP 2 Kd's remain on the order of 10 times smaller
than the measured site-specific Kd's for many COI. COI sorption to soil is important because, as
discussed later, aquifer cleanup times (i.e., chemical flushing rates) are generally proportional to the
chemical retardation factor, which is directly proportional to Kd , except when Kd << 1 (Zheng et al., 1991).
Groundwater Flow
In the CSA and CAP reports HDR made several critical assumptions, not supported by data, regarding
the horizontal and vertical groundwater flow directions near the boundaries of the Marshall site which
impacted their conclusions regarding the ultimate discharge locations for site groundwater and dissolved
COI. Two examples discussed in this section are (i) the relationship between site groundwater and Lake
Norman and (ii) groundwater flow directions and the potential for offsite migration of COI.
Lake Norman and the LeGrand Conceptual Model
All shallow, deep, and bedrock groundwater in the downgradient areas of the site was apparently
assumed to discharge into either Lake Norman or the unnamed stream located close to the northeastern
side of the ash basin before site-specific hydrogeologic data were analyzed. The rationale that HDR used
is a generalized, theoretical conceptual model (LeGrand, 2004) that relies on land -surface topographic
elevations rather than actual groundwater flow data (i.e., hydraulic head measurements); references to
this theory were made at numerous points in the CSA and CAP reports. However, HDR did not present
any site-specific data analyses or groundwater flow modeling that would support this assumption in either
report. In fact, as discussed below, the boundary conditions for the CAP Parts 1 and 2 flow models
effectively forced downgradient site groundwater in the shallow, deep, and bedrock aquifers to discharge
into Lake Norman and the unnamed stream.
HDR continued to state this assumption in the CAP 2 report (e.g., Section 3.3.2) even though strong
measured downward groundwater flow components from deep overburden to the fractured bedrock unit
exist in the southeastern part of the site next to Lake Norman. In the CAP 2 report Executive Summary
(page 2) HDR also claims that "Groundwater at the MSS site generally flows .... to the south and
southeast beneath the source areas toward Lake Norman, which serves as the hydrologic boundary
downgradient of the CCR source areas". I demonstrate below that this interpretation is incorrect because
F:3

CSA and CAP 2 hydraulic head data show that groundwater and dissolved COI in the fractured bedrock
aquifer flow off-site in the southeastern portion of the Marshall site toward neighboring water supply wells
The LeGrand (2004) guidance document presents a general discussion of groundwater flow patterns that
may occur near streams in the Piedmont and Mountain Region of North Carolina based on ground
surface elevations (i.e., site topography and surface watershed boundaries). However, surface water and
groundwater watersheds commonly do not coincide (Winter et al., 2003). Further, groundwater flow
patterns and rates in bedrock have been found to be poorly related to topographic characteristics (Yin
and Brook, 1992). LeGrand does not present or derive any mathematical equations or quantitative
relationships for groundwater flow near rivers or streams and does not address areally-extensive
waterbodies such as Lake Norman. The author emphasizes that site-specific data must be collected in
order to correctly evaluate river inflow or outflow. In strong contrast to the LeGrand generalizations,
numerous detailed and sophisticated mathematical (analytical and numerical) river -aquifer models and
highly -monitored field studies have been published in the scientific and engineering literature in the past
several decades. What these investigations and applied hydraulic models show is that the water flow rate
into or out of a river, stream, or lake and the depth of hydraulic influence within an underlying aquifer are
highly sensitive to several factors, including: the transient river/lake water surface elevation and slope;
river/lake bed topography; bed permeability and thickness; horizontal and vertical permeability (and
thickness) of the different hydrogeologic units underlying the river/lake; transient horizontal and vertical
hydraulic head variations in groundwater beneath and near the river/lake; and groundwater extraction
rates and screen elevations for neighboring pumping wells (e.g., Simon et al. 2015; McDonald and
Harbaugh, 1988; Bear, 1979; Hantush, 1964).
The CSA investigation did not: measure the bed permeability or thickness of Lake Norman or the
unnamed stream; characterize the lake/river bathymetry; monitor transient water surface elevation
variations at more than one location (one average value was used); collect lake/river bed hydraulic
gradient data; measure horizontal or vertical overburden or bedrock permeability beneath Lake Norman;
characterize the geology beneath Lake Norman; measure hydraulic heads in the overburden or bedrock
beneath Lake Norman or the stream; or consider the hydraulic effects of groundwater extraction from
nearby water supply wells [e.g., water supply wells NC0118676 (Duke Energy), NC0118736 (Old Country
Church), and NC0118622 (Midway Restaurant and Marina), located close to the southern section of the
ash basin compliance boundary; and the numerous private homes located south and west of the site;
see CSA Figures 4-1 and 4-5]. Well construction and well yield (gallons per minute) data for these wells
are listed in Table 1 of the 2014 receptor survey report (HDR, 2014).
Much of the CSA data contradict the LeGrand hypothesis at the Marshall site. For example, downward
flow components from deep to bedrock wells were measured at monitoring clusters AB -5 and AB -6
M

located in the southeastern portion of the site close to Lake Norman during the CSA and CAP 2 (compare
CSA Figures 6-6 and 6-7 and CAP 2 Figures 2-3 and 2-4). In addition, my corrected contouring of
bedrock hydraulic head data shows that groundwater flow in the bedrock aquifer is off-site and to the
south/southeast in this portion of the site (see next section). As discussed in other parts of my report,
HDR assumed in the CSA and the CAP Parts 1 and 2 modeling that groundwater flow in all three
hydrogeologic units is directly west to east near the southern part of the Compliance Boundary and used
a no -flow boundary in this model area, which prevents off-site flow. The CSA and CAP 1 and 2 also did
not evaluate the effect of the very large constant hydraulic head in the ash basin impoundment (30 feet
greater than the Lake Norman water level and the static bedrock hydraulic head in this area) as it relates
to vertical groundwater flow from overburden to bedrock and COI transport from ash basin water to
underlying groundwater. Maps of vertical hydraulic gradient variations (e.g., contour maps) were not
generated for the CSA or CAP 2, and HDR did not discuss the significance of downward hydraulic
gradients from overburden to bedrock, or from the impounded ash basin water to the aquifer system.
These downward groundwater flow measurements are consistent with the hydraulic conductivities of the
bedrock and overburden being of similar magnitude (CSA Table 11-9).
Groundwater Flow Directions
The CSA assumptions and analysis errors discussed above have had a strong effect on: the Conceptual
Model development; the site hydrogeologic and COI transport assessment; the construction/calibration
of the CAP flow and transport models; and the simulations of CAP Close Scenarios. The hydrogeologic
assumptions should have been carefully evaluated and tested during the performance of the CSA and as
part of the CAP groundwater flow model construction and calibration to determine whether they were
valid. Instead, the hypotheses appear to have effectively guided the model development and led to
inaccurate interpretations.
A good example of this is the bedrock aquifer hydraulic head and flow direction map developed by HDR
(CSA Figure 6-7). The hydraulic head contours and inferred bedrock flow directions in Figure 6-7 are
inaccurate because HDR did not account for the hydraulic impacts of three very important influences on
horizontal and vertical groundwater flow in the fractured bedrock aquifer: the ash basin impoundment,
Lake Norman, and groundwater extraction from off-site water supply wells (e.g., the wells mentioned
above). First, the 780 -foot head contour is incorrect because it cuts through the middle of the ash basin
where the constant hydraulic head is approximately 790 feet. Drawn correctly, the 780 -foot head contour
should mirror the northeastern boundary of the ash basin as shown in the deep aquifer map (CSA Figure
6-6 and CAP flow model Figure 11 of Appendix C). Second, the CSA bedrock flow map does not utilize
the hydraulic information (specifically, the groundwater/surface-water interaction) associated with the
Lake Norman shoreline area to the east and southeast of the Marshall site (e.g., CSA Figure 4-1). For
example, a U.S. Geological Survey study of overburden and bedrock groundwater flow near another part
10

of the Lake Norman shoreline (Pippin et al., 2008) determined that heads in bedrock were generally less
than a few feet greater than the Lake Norman water surface elevation, depending on how much rainfall
had occurred during the days and weeks preceding the measurement event (e.g., Figure 25 in the USGS
report). At the Marshall site, water -level measurements in monitoring well GWA-1 BR are consistent with
this pattern. Similarly, static water levels in the three bedrock pumping wells to the south (Duke Energy,
Old Country Church, and Midway Restaurant and Marina) are strongly influenced by hydraulic interaction
with Lake Norman. As shown in CSA Figure 4-1 these wells are located very close to the Lake Norman
shoreline. HDR (2014; Table 1) reports that the well yield for the Duke Well is 30 gallons per minute
(gpm), and Duke Energy has reported average annual water usage rates of about 10 gpm for this well in
annual water use reports submitted to the North Carolina Division of Water Resources (e.g., McGary,
2004). The reported well yields for Old Country Church and Midway Restaurant and Marina are 40 gpm
and 20 gpm, respectively.
To better understand the effects of these extraction wells on groundwater flow in the bedrock aquifer I
estimated hydraulic head reductions, relative to the Lake Norman static level, as a function of distance
from a pumping well using the exact mathematical model of Hantush (1964, Equation 73). This
analytical groundwater flow model predicts the steady-state, pumping -induced drawdown [i.e., reduction
in hydraulic head from static (non -pumping) conditions] in a confined aquifer (bedrock) overlain by a
"leaky layer", which I model as the overlying overburden aquifer (using the overburden aquifer vertical
hydraulic conductivity, as represented by data in CSA Table 11-9) and the Lake Norman water surface
elevation. I assumed a bedrock aquifer thickness of 100 feet, which is much larger than the
underestimated value used in the CAP models (i.e., 50 feet) but also is consistent with regional studies of
this bedrock aquifer system (Daniel et al., 1989). The cited depths of the Duke and Midway Marina
supply wells (HDR, 2014) are about 400 feet and 160 feet below the bedrock surface, respectively (540
and 245 feet below ground surface). For the overlying confining unit I used representative CSA values: a
50 -foot thick overburden aquifer with a 0.045 feet/day vertical permeability. The following bedrock
drawdown calculations depend on the product of bedrock permeability and thickness (i.e., transmissivity);
I varied permeability by a factor of seven, which has a similar hydraulic effect on drawdown as varying the
thickness by this amount. Therefore, my calculations also represent sensitivity analyses of both the
thickness and the permeability of the fractured bedrock unit. I used a 30-gpm pumping rate, which is
representative of the well yields for any of the three supply wells. The Hantush model is conservative
(i.e., underestimates drawdown) as applied to this analysis because the model assumes that Lake
Norman covers the entire area; in actuality the lake overlies only parts of the northeastern and
southeastern quadrants of the Duke Energy supply well zone of influence.
Figure 1 is a graph of the computed steady-state drawdown versus distance for three values of bedrock
hydraulic conductivity (Kb,): 0.68 (mean site value from CSA Table 11-9), 2.0, and 4.7 feet/day
11

(measured value for well GWA-1 BR, which is the Marshall -site bedrock well located closest to the Duke
Energy pumping well; CSA Table 11-3). The estimated drawdown ranges from 14 feet (higher Kb,) to 85
feet at the pumping well, and the radius of influence (e.g., drawdown = 0.01 feet) generally varies from
1800 feet (lower Kb,) to approximately 3,400 feet. Note that the distance from the Duke Energy pumping
well to monitoring well GWA-1 BR is about 2,000 feet. Therefore, these drawdown analyses indicate that
groundwater extraction from any of the supply wells NC0118676 (Duke Energy), NC0118736 (Old
Country Church), or NC0118622 (Midway Restaurant and Marina) will likely influence bedrock
groundwater flow directions in the southeastern portion of the Marshall site (as far north as the ash basin)
based on average groundwater withdrawal rates.
I used the CSA bedrock water -level data and all of the above information and analyses to re -construct the
bedrock hydraulic head map and generate groundwater pathlines, which are shown in Figure 2. 1
conservatively assumed a small 25 -foot drawdown in the Duke Energy supply well, five feet of drawdown
in the Old Church well, and no pumping in the Midway Marina well. Solid black circles represent bedrock
monitoring well data. Blue circles represent inferred bedrock heads beneath Lake Norman, which I
assumed were equal to 760 feet in areas located beyond the radius of influence of any of the three supply
wells. I experimented with lake elevations that were a few feet different than 760 feet and found that
these variations had little effect on my contour map. Within the radius of influence I subtracted the
drawdowns in Figure 1 (Kb, = 2.0 feet/day curve) from the 760 -foot lake level. I generated Figure 2 using
the kriging interpolation option in the Tecplot data visualization software package (Tecplot, Inc., Bellevue,
Washington). Kriging is an advanced geostatistical procedure that spatially interpolates between discrete
measurements (e.g., monitoring well water levels) by computing a weighted average of the known values
of the function (hydraulic head variation) in the vicinity of the measured value. Kriging is mathematically
closely related to regression analysis and generally gives the best unbiased prediction of interpolated
values (Deutsch and Journel, 1992).
The contours and pathlines in Figure 2 indicate that much of the bedrock groundwater flow in the
southeastern portion of the area enclosed by the ash basin compliance boundary is off-site and to the
south-southeast (e.g., areas west of GWA-1 BR and south of AB -613R). In the area generally northeast of
wells AB-5BR and GWA-1 BR groundwater in the bedrock aquifer flows to the east-northeast toward the
unnamed stream and Lake Norman. I further examined bedrock groundwater flow directions and the
hydraulic effects of off-site pumping by developing a second bedrock hydraulic head map (Figure 3) in
which the Duke Energy, Old Country Church, and Midway Restaurant and Marina water supply wells
were inactive. Without off-site groundwater extraction from these supply wells the bedrock groundwater
flow directions in the southeastern part of the Marshall site exhibit a small additional easterly component,
but overall are not significantly different than the south -southeasterly directions in this area of the site
shown in Figure 2.
12

Exceedances of Groundwater Standards
In this section I compare measured groundwater concentrations in shallow, deep, and bedrock
groundwater samples to North Carolina 2L and IMAC standards and show the following: (i) 15 measured
exceedances for several COI at multiple downgradient locations on the Compliance Boundary (CB); (ii)
an additional six (6) exceedances at CB locations based on chemical transport modeling I performed; (iii)
19 of the 21 downgradient Compliance Boundary exceedances were greater than the proposed
provisional background concentrations (PPBC) by HDR; (iv) 13 of the 21 CB exceedances were greater
than the maximum background concentration in the same hydrogeologic unit (e.g., shallow, deep, or
bedrock); (v) 22 additional exceedances were observed in wells screened in the highly -permeable
fractured bedrock unit underlying the ash basin and located inside the CB; and (v) the statistical analyses
of groundwater concentrations at wells MW -4 and MW -41D for purposes of defining background levels
were performed incorrectly.
Throughout this report I reference the ash basin compliance boundary and the Duke Energy property
boundary for the Marshall site as drawn on maps developed by HDR. My reference to the "compliance
boundary" is only for identification purposes and not an opinion that this boundary as drawn by HDR is
accurate or legally correct.
Summary of Exceedances
Table 1 summarizes exceedances of 2L or IMAC standards in shallow, deep, and bedrock groundwater
samples obtained from monitoring wells located: (i) on downgradient sections of the Ash Basin
Compliance Boundary (CB) as drawn by HDR; (ii) bedrock wells (BR) located inside the CB; and (iii)
modeled Compliance Boundary concentrations (CBM), using modeling techniques described below. The
proposed provisional background concentrations (PPBC) by HDR are also listed in Table 1.
A total of 15 Compliance Boundary groundwater samples exceeded North Carolina groundwater
standards for these COI: antimony, boron, chromium, cobalt, iron, manganese, total dissolved solids, and
vanadium. I estimated an additional six (6) exceedances at downgradient CB locations based on
chemical transport modeling and measured upgradient concentrations (CBM). In addition, 22
exceedances were observed in wells screened in the highly fractured bedrock unit located inside the CB.
A total of 19 of the 21 (measured plus modeled) Compliance Boundary exceedances were greater than
the proposed provisional background concentrations (PPBC) by HDR. A total of 13 of the 21 CB
exceedances were greater than background levels from the same hydrogeologic unit (e.g., shallow, deep,
13

or bedrock) for a particular constituent. Of the 22 bedrock exceedances, 14 were greater than PPBC
background levels.
Note that the iso -concentration contours in all of the CSA Section 10 figures and Figure ES -1 are not
consistent, and are in many cases misleading, with regard to chemical transport mechanisms in the
subsurface. For example, the iso -concentration contours in Section 10 generally closely encircle a
monitoring well and infer no subsequent transport downgradient from the well location. This contouring
problem is especially prevalent near downgradient sections of the Compliance Boundary (e.g., the
unnamed stream to the northeast and Lake Norman). Figure ES -1 is a good example of this practice.
These closed contours downgradient from the ash basin near the unnamed stream suggest that boron
transport beyond the farthest downgradient line of monitoring wells does not occur. This is not the case,
however, as demonstrated in the following section where boron exceedances at the Compliance
Boundary are demonstrated by modeling. Further, the CAP model simulated "existing conditions" plume
maps for boron and several other COI (CSA Appendix C) contain 'open contours" at these downgradient
parts of the CB, which confirm constituent transport to the boundary.
Modeled Compliance Boundary Exceedances
I computed Compliance Boundary concentrations labeled "CBM" with footnote "e" in Table 1 (wells MW-
14S,D; AL-1S,D; AB -1S; and MW -7S) using a one-dimensional, analytical chemical transport model
(van Genuchten and Alves, 1982; Equation C5) because the CB at these locations was 300-600 feet
downgradient from the wells and boron is highly mobile in the subsurface. I used site-specific
hydrogeologic data to determine input parameter values for groundwater pore velocity (Vp) and assumed
a boron retardation factor of unity (1.0), as HDR assumed in the CAP transport model. To calculate pore
velocity (Vp = K i/ne ), I used: mean hydraulic conductivity (K) values from CSA Table 11-9; average
effective porosity (ne) data in CSA Table 11-8; and measured horizontal hydraulic gradients (i) from the
hydraulic head maps in Section 6. 1 then calibrated the model to match observed 2015 boron
concentrations at the above wells. The concentrations labeled CBM in Table 1 are the modeled 2015
concentrations downgradient from each monitoring well at the Compliance Boundary.
Exceedances of Groundwater and Surface Water Standards in Seep Samples
As discussed in the introduction, the ash basin at the Marshall site was constructed above Holdsclaw
Creek and other tributaries that discharged to the Catawba River (currently Lake Norman). Seep
samples were collected during the CSA at two locations (S-2 and MSSW001 S001) on the downgradient
toe of the active ash basin dam during the CSA (CAP 1 Figure 2-2). After the CSA sample collection
HDR reported that location S-2 was not a separate seep from the ash basin, but rather pooled water from
the MSSW001 S001 seep. Nevertheless, HDR continued to show analytical results for both locations in
the CAP 1 report (Figure 2-2), and my following discussion refers to both locations.
14

Concentrations in the two seep water samples exceeded relevant NCAC 2B, 2L and/or IMAC standards
for various COI (e.g., CSA Table 7-8, CAP 1 Table 2-5, and CAP 1 Figure 2-2). Referring to my Table 1,
17 exceedances of North Carolina groundwater standards were detected in the seep samples for these
COI: arsenic, barium, beryllium, boron, chromium, cobalt, lead, manganese, selenium, thallium, total
dissolved solids, and vanadium. North Carolina surface water (2B) standards were exceeded for these
constituents: arsenic, beryllium, chromium, copper, lead, mercury, thallium, total dissolved solids, and
zinc.
Surface water sample SW -6, located in the unnamed tributary to Lake Norman immediately downgradient
from the active ash basin (CAP 1 Figure 2-2), was also considered by HDR to be representative of
groundwater quality at the site (CAP 2, Section 3.3.3). As shown in CAP 1 Table 2-6 North Carolina 2B
standards were exceeded in the July and October 2015 SW -6 samples for these COI: cobalt, iron,
manganese, sulfate, and total dissolved solids. Based on the significance of these exceedances, HDR
recommended in the CAP 2 report (Section 3.3.3) that further evaluation of the SW -6 surface-water/seep
fluid be conducted.
Statistical Analyses of Background Concentrations
Appendix G of the CSA report presents statistical analyses of historical concentrations from monitoring
wells MW -4 (shallow) and MW -41D (deep), which HDR described as following methods specified by the
U.S. Environmental Protection Agency (EPA, 2009), in an attempt to establish background groundwater
concentrations for the Marshall site. As outlined in Sections 3.2.1 and 5.5.2 of the EPA guidance
document these data must be checked to ensure that they are statistically independent and exhibit no
pairwise correlation. Groundwater sampling data can be non -independent (i.e., autocorrelated) if the
sampling frequency is too high (i.e., time interval between sampling events is too small) compared to the
chemical migration rate in the aquifer (groundwater pore velocity divided by chemical retardation factor).
Section 14 of the EPA guidance presents methods for ensuring that the data for background wells are not
autocorrelated, but the analyses in CSA Appendix G did not include evaluations for statistical
independence.
As an illustration, "slow-moving" groundwater combined with high chemical retardation (i.e., large soil -
water partition coefficients, Kd), which is the case for many COI in the shallow aquifer at the Marshall site,
can lead to the same general volume of the chemical plume being repeatedly sampled when the
monitoring events are closely spaced. Examining shallow background well MW -4, the measure hydraulic
conductivity at this location is 7.0E-5 to 3.7E-4 centimeters/sec (0.2 - 1.1 feet/day; CSA Tables 11-3 and
11-4). From CSA Table 11-8 the effective porosity at this location is about 0.25. Based on CSA Figure 6-
5 and Table 6-9 the horizontal hydraulic gradient in this area is about 0.02 feet/foot. Using these values
15

the estimated groundwater pore velocity (VP = K i/ne ) near MW -4 is on the order of 6 to 30 feet/year.
(Note that the groundwater velocities listed in CSA Table 11-12 are incorrect because this table is from
the Allen Steam Station CSA). Shallow pore velocities are generally factors of 40 to 200 greater
downgradient of the ash basin (between the ash basin and the Compliance Boundary) due to much
greater horizontal hydraulic gradients (-- 4x larger) and larger hydraulic conductivity (-- 10-50x greater) in
these areas.
The retardation factors, Rd, based on laboratory Kd measurements (Kd — 10 cm3/g, or greater) are on the
order of 100 (or greater) for many of the COI (except conservative parameters such as sulfate and
boron). Therefore, the average shallow chemical migration rate near monitoring well MW -4 (VP /Rd) is on
the order of 0.06 - 0.3 feet/year for many of the non -conservative COI, assuming linear equilibrium
sorption (refer to discussion below). With quarterly sampling, the chemical migration distance during the
time interval between sampling rounds is less than 0.1 feet (1 inch) for several COI, which is smaller than
the sand pack diameter for the monitoring wells. Therefore, based on either quarterly or annual
monitoring the shallow groundwater samples from monitoring well MW -4 are basically representative of
the same volume of the plume (i.e., the sandpack, depending on the well purge volume) for many COI,
and any measured sample concentration changes are not due to real chemical transport effects in the
aquifer. In this case, this means that the groundwater samples are non -independent and that the
statistical analyses of background concentrations do not satisfy the key requirements of the analysis
method.
CAP Groundwater Flow Model Underestimates Potential for Off -Site Chemical Migration
My discussions in this section focus on limitations of the CAP groundwater flow model. I focus specifically
on model boundary conditions representing Lake Norman; the overall size of the model grid and no -flow
boundary conditions on the western, southern, and northern grid boundaries; the misrepresentation of
groundwater flow in the fractured bedrock aquifer; and the potential for off-site groundwater flow in
relation to groundwater extraction from numerous private and public water supply wells located close to
the model boundaries, but not incorporated into the flow model.
Lake Norman and Southern No -Flow Boundary Conditions
The CAP Parts 1 and 2 groundwater flow models effectively force all groundwater (shallow, deep, and
bedrock aquifers) located beneath and downgradient from the ash basin to discharge into either Lake
Norman or the unnamed stream located close to the northeastern side of the ash basin, which discharges
as overland flow into the lake. The CAP 1 and 2 flow models also significantly underestimate the
potential for off-site flow and chemical migration in fractured bedrock. No -flow boundary conditions
16

defined along the entire western, northern, and southern model boundaries prevent any off-site
groundwater flow and chemical transport in these areas (e.g., refer to Figures 1 and 4 in Appendix C of
the CAP 1 Report). The CAP 1 and 2 models assign additional no -flow boundary conditions to all
downgradient bedrock cells located beneath Lake Norman, which prevents groundwater from flowing
beneath the lake in the bedrock aquifer. As discussed above (e.g., Figures 2 and 3), CSA hydrogeologic
data and available off-site hydraulic information clearly demonstrate that the CAP 1 and 2 models
boundary conditions are incorrect. The bottom surfaces (bedrock) of the flow models are also assumed
to be a no -flow boundary even though the hydraulic conductivity data and measured downward hydraulic
gradients at several monitoring well clusters do not support this assumption (e.g., downward flow from
deep to bedrock aquifers at well clusters AB-5D/BR and AB-6D/BR).
This hydraulic representation of Lake Norman in the flow models is inaccurate for many reasons. First,
the lake bottom is assumed to extend all the way through the overburden aquifer to the bedrock surface,
which is not the case. Second, groundwater flow beneath and adjacent to the lake is assumed to be
horizontal with zero vertical flow component. Because this boundary condition does not allow
groundwater to flow vertically in areas beneath and near the lake, the CAP models do not represent
actual site hydrologic conditions. Third, as represented in the CAP models, neither the lower -permeability
lake bed sediments nor the smaller vertically hydraulic conductivity of underlying soils restricts the
potential flow rate into or out of the lake (i.e., in the CAP models a perfect hydraulic connection exists
between the aquifer and Lake Norman). HDR did not evaluate the actual degree of aquifer -lake hydraulic
connection in the CSA or CAP 1 and 2. In summary, due to all of these factors the potential for site
groundwater and dissolved constituents to migrate off-site as underflow beneath Lake Norman and to the
south-southeast as groundwater flow cannot be evaluated with the model.
The CAP model should have represented Lake Norman using a "leaky -type" boundary condition in the top
model layer (McDonald and Harbaugh, 1988), and the model grid should have extended farther east and
south so that the above factors could have been evaluated during model calibration and sensitivity
analyses. The Electric Power Research Institute technical review committee, which included the
developer of the MT3D transport code used in the CAP modeling, made the same comment (December
02, 2015 memorandum entitled "Revised Marshall Model Review", which was submitted with the CAP
report). A leaky boundary condition incorporates the lake/river bed permeability and thickness, the
lake/river water surface elevation, and the simulated hydraulic head in the aquifer (at the base of the
lake/river bed) to dynamically specify a flux (flow rate per unit bed area) into or out of the groundwater
model depending on the head difference between the lake/river and aquifer. Typically, permeability and
vertical hydraulic gradient measurements for the lake bed (not collected in the CSA) and flow model
calibration (three-dimensional matching of simulated and measured hydraulic head measurements in the
17

aquifer) are used to determine a best -fit estimate of lake bed conductance (permeability divided by
thickness) in the model. HDR did not perform this routine analysis
Limitations of No -Flow Boundary Conditions and Small Model Domain Size
The limited areal extent and depth of the CAP Parts 1 and 2 flow and transport model grids prevent the
use of the models as unbiased computational tools that can be used to evaluate off-site migration of coal -
ash constituents. For example, the model grids should have extended farther west and incorporated
groundwater extraction from off-site private -home and public water -supply wells. The western no -flow
boundary in the current CAP models artificially prevents any off-site flow or transport to the west in either
the bedrock or overburden aquifers. The same is true for the entire northern and southern model
boundaries, as discussed above, and the artificial limitations created by the eastern Lake Norman
boundary condition. In addition, the downgradient boundaries of the CAP flow and transport models do
not extend to the Compliance Boundary along most of the Lake Norman boundary, which prevents their
use for estimating Compliance Boundary concentrations for various remedial alternatives.
The bottom boundaries of the CAP models should extend much deeper because the hydraulic
conductivity of the fractured bedrock zone is similar in magnitude to the overburden soils. In the present
configuration the lower boundary of the CAP 1 and 2 model grids is only about 50 feet below the bedrock
surface (Figure 2 in both the CAP 1 & 2 modeling appendices). Because several bedrock wells were
screened to this depth the bedrock hydraulic conductivity data collected for the CSA demonstrate that
imposing an impermeable model boundary at this depth is incorrect (compare similarities of mean
overburden and bedrock aquifer permeabilities in CSA Table 11-9).
Off -Site Groundwater Extraction Ignored
The CSA and CAP Parts 1 and 2 failed to examine the strong potential for coal -ash constituents from the
Marshall site to migrate with groundwater to private and public bedrock water supply wells located
immediately west, south, and north of the ash basin compliance boundary. CSA Figures 4-2 and 4-5
shows the locations of private and public water supply wells near the site. The basis of my opinion
includes the following: my hydraulic analyses of pumping -induced pressure reductions in the fractured
bedrock unit (Figure 1) and re-evaluation of bedrock hydraulic head variations and flow directions
(Figures 2 and 3), as discussed above; hydraulic conductivity measurements for the overburden and
bedrock formations; three-dimensional variations in measured hydraulic head in the bedrock and
overburden units; and groundwater concentration data. As discussed throughout my report, neither the
CSA nor CAP Parts 1 and 2 investigations addressed the potential for off-site migration.
COI's were detected in several water supply well samples (CSA Appendix B), but the CSA report did not
plot these detections on a map and did not discuss their possible relationship to the Marshall site.
18

Appendix B also did not present the well construction details (e.g., well diameter and elevation range of
the well screen or open bedrock interval) so that well dilution effects and potential chemical transport
pathways in the bedrock unit could be evaluated. In addition, the CSA investigations and CAP 1
modeling did not include these areas west, north, and south of the Marshall site. The CAP Part 2 flow
model did include a small number (four) of residential wells located inside the undersized model domain
(near the northern and western model domain boundaries), but the CAP 2 modeling report (CAP 2,
Appendix B) did not show simulated hydraulic head maps with these residential wells pumping and did
not provide any discussion or analyses of the potential for these wells to capture COI dissolved in
groundwater. The CAP Part 2 also did not increase the model grid size to incorporate the large number
of residential (89) and private (4) water supply wells located very close to the site boundary (CAP 2
Figure 3-3); fix the boundary condition problems; or correct the model input data errors I have outlined
so that the flow and transport models could be used to more accurately analyze the potential for off-site
chemical transport.
Another important model input data error is the bedrock hydraulic conductivity, which is assumed in the
CAP 1 flow model to be about a factor of three to five (3-5x) lower than the overburden aquifer in different
areas (Tables 2 in CAP 1 Appendix C and CAP 2 Appendix B). In the CAP 2 flow model the bedrock
permeability was further lowered to a value that is about a factor of 10 to 100 (10-100x) lower than the
overburden aquifer. However, the bedrock slug test results show that the mean bedrock permeability is
approximately the same as the overburden permeability (CSA Table 11-9). Therefore, the CAP 1 and 2
flow models significantly restrict (incorrectly) groundwater from flowing from the overburden aquifer into
the fractured bedrock unit, which causes the CAP transport models to underestimate the potential for off-
site chemical migration.
Ash Basin Hydraulic Boundary Condition and Groundwater Recharge Rates
Based on the model description in CAP 1 Appendix C [text and Figures 4 (boundary conditions) and 5
(groundwater recharge zones)] and CAP 2 Appendix B (Figure 5) the CAP 1 and 2 flow model
representations of the active ash basin are incorrect and significantly underestimate the recharge of ash
basin fluids into the underlying aquifer system (e.g., refer to my preceding discussion regarding
groundwater flow in bedrock and Figures 2 and 3). As noted in the CSA report (Section 2.6), the full
operating pond elevation (i.e., impounded fluid level) for the active ash basin is approximately 790 feet,
which is 30 feet greater than normal water level in Lake Norman. This very large hydraulic head increase
in the ash basin relative to the normal static conditions that existed in the underlying aquifer system prior
to construction creates very large downward groundwater flow components beneath areas of ponded
ash -basin fluid and causes the near -radial horizontal flow patterns observed in the bedrock (Figures 2
and 3) and overburden (CSA Figures 6-5 and 6-6) hydraulic head contour maps. The CAP 1 and 2 flow
19

models should have used constant -head or leaky -type boundary conditions in the top layer to represent
these areas of ponded fluid located inside the ash basin (e.g., McDonald and Harbaugh, 1988).
The CAP Part 2 flow model underestimates leachate discharge from the active ash basin by as much as
a factor of 34 in areas of ponded surface water (e.g., refer to CSA Figure 6-2). The CAP 1 model
underestimates active basin leakage by as much as a factor of 93. As shown in Figure 5 of CAP 1,
Appendix C, the CAP 1 flow model assumes a constant groundwater recharge rate (i.e., leakage rate)
equal to 4.5 inches/year in the active ash basin. In the CAP 2 flow model the active basin leakage rate is
assumed to be 12.3 inches/year (Figure 5 of CAP 2, Appendix B). However, CSA Figure 8-4.2 (cross-
section B -B) shows that the vertical hydraulic gradient through the coal ash in the downgradient portion of
the active ash basin is on the order of 0.5. Using Darcy's law and the mean vertical coal -ash permeability
of 6.7E-5 cm/sec in CSA Table 11-10, the approximate vertical leakage rate out of the active basin is
about 420 inches/year near Lake Norman (i.e., — 34 times greater than the specified CAP 2 recharge rate
of 12.3 inches/year; and — 93 times greater than the specified CAP 1 recharge rate of 4.5 inches/year).
Three related impacts of this incorrect active basin boundary condition are that the CAP models
significantly underestimate: (i) vertical groundwater flow rates (by up to a factor of 100) through coal -ash
source material in the vicinity of the downgradient portion of the active ash basin; (ii) horizontal
groundwater flow and chemical transport rates downgradient from the active ash basin (by at least a
factor of three; compare simulated hydraulic gradients near Lake Norman in Figure 13 of CAP 2
Appendix B with measured gradients in Figure 2-2 of the CAP 2 report); and (iii) vertical flow rates from
the overburden aquifer into the fractured bedrock unit beneath ponded areas. This incorrect boundary
condition representation of the active ash basin also causes the CAP models to significantly
underestimate (by at least a factor of three) both the mass loading of COI into the Lake Norman and the
corresponding Lake Norman surface water concentrations (attributable to coal ash ponds) that HDR
estimated with their mixing model (e.g., CAP 2 report Table 4-2 and Appendix D).
In addition, a site-specific distribution of groundwater recharge values should have been developed for
this and the other simulation scenarios to take into account site-specific topography and soil types (e.g.,
runoff estimation) and climate data (precipitation, evapotranspiration, etc.; e.g., using the U.S. Army
Corps of Engineers HELP Model; Schroeder et al., 1994). The predicted water table lowering due to
capping, discussed below, is very sensitive to the model recharge value; therefore, HDR should have
made more effort to develop a site-specific recharge -rate distribution.
20

CAP Chemical Transport Modeling
Due to model calibration, model construction, and boundary -condition and input -data errors the CAP
models significantly underestimate remediation time frames. As discussed in this section, reasons for this
include significant underestimation of the chemical mass sorbed to soil, failure to account for slow
chemical desorption rates, inaccurate analyses of water -table lowering due to capping, and flaws in the
transport model calibration.
Soil -Water Partition Coefficients and Model Calibration
The fraction of chemical mass sorbed to soil can be represented by the soil -water partition coefficient, Kd
(Lyman et al., 1982). Kd is an especially important parameter at the Marshall site because for most of the
COI the bulk of the chemical mass in the soil is associated with the solid phase (i.e., sorbed to soil grains
rather than dissolved in pore water). In effect, the solid fraction of the soil matrix acts as a large "storage
reservoir" for chemical mass when Kd is large [e.g., metals, many chlorinated solvents, and highly -
chlorinated polycyclic aromatic hydrocarbon (PAH) compounds associated with coal tars and wood -
treating fluids]. Kd is also a very important chemical transport parameter which is used to compute the
chemical retardation factor, Rd, assuming linear equilibrium partitioning of mass between the soil (solid)
and pore -water phases (Hemond and Fechner, 1994):
Rd = 1 + PbKd l ne
where Pb is the soil matrix bulk dry density and ne is the effective soil porosity. For example, the
chemical migration rate is directly proportional to hydraulic conductivity and inversely proportional to Rd .
The total contaminant mass in an aquifer is also directly proportional to Rd, as well as aquifer cleanup
times once the source is removed (e.g., Zheng et al., 1991).
Accordingly, it is very important to use accurate Kd values in the CAP Closure Scenario modeling.
Specifically, the CAP Part 1 transport modeling used Kd values that were typically factors of 10 - 100
(i.e., one to two orders of magnitude) smaller than the measured site-specific Kd's reported in CAP
Appendix D. In contrast, the CAP Part 2 transport modeling used Kd values that are generally a factor of
about 10 larger than the CAP 1 values (480x larger for arsenic; 70x larger for thallium; 25x larger for
vanadium); however, the CAP 2 Kd's remain on the order of 10 times smaller than the measured site-
specific Kd's reported in CAP 1 Appendix D and CAP 2 Appendix C for many COI. Further, soil -water
partition coefficients for the CAP Parts 1 and 2 models are much smaller than most values presented in
the literature for the COI (e.g., EPRI, 1984; Baes and Sharp, 1983). This means that, using the actual
measured Kd's for the Marshall site, the times required to reach North Carolina water quality standards at
21

the Compliance Boundary are more than a factor of 10 longer (see additional discussion below) than
cleanup times predicted by the CAP transport model.
The CAP Part 1 modeling report (CAP 1 Appendix C; Section 4.8) argues that the major Kd reductions
were needed due to the following:
"The conceptual transport model specifies that COls enter the model from the shallow saturated source
zone in the ash basin and beneath the CCR waste management units. When the measured Kd values
were applied in the numerical model to arsenic migrating from the source zones, this COI did not reach
the downgradient observation wells where it was observed in July 2015 at the end of the model simulation
period. The most appropriate method to calibrate the transport model in this case is to lower the Kd
values until an acceptable agreement between measured and modeled concentrations is achieved. Thus,
an effective Kd value results that likely represents the combined result of intermittent activities over the
service life of the ash basin, landfills, and structural fills. These may include pond dredging, dewatering
for dike construction, and ash grading and placement."
Considering the approach that HDR used to develop the chemical transport model (history matching), it is
not true that "the most appropriate method to calibrate the transport model is to lower the Kd values." The
CAP transport model used an incorrect value (2.12 g/cm3) for the bulk density of overburden materials.
The bulk density should have been computed using the total porosity (n) values in CSA Tables 11-1 and
11-7 using the following formula (e.g., Baes and Sharp, 1983):
Pb = 2.65 (1— n)
Based on the Table 11-7 mean total porosity values (--44%) Pb — 1.48 g/cm3 for overburden soil, which
means that the Rd values for the CAP 1 and 2 models were as much as a factor of 1.4 (2.12/1.48) too
high before HDR adjusted the Kd values during calibration. Also, as discussed earlier, the overburden
slug test values were about 50 percent too low due to HDR's data analysis errors. Both of these errors
(sorption rate and hydraulic conductivity) resulted in a modeled transport rate that was as much as a
factor of 2.1 too low before calibration simply due to data input errors.
At least two other important factors were not considered during the CAP 1 and 2 transport model
calibrations. First, the groundwater flow models are based on average hydraulic conductivity (K) values
within a material zone, but K distributions in aquifers are highly variable due to layering and other types of
heterogeneities (e.g., varying by factors of 3-10, or more, over distances as small as a few feet: Gelhar,
1984, 1986, 1987; Gelhar and Axness, 1983; Rehfeldt et al., 1992; Rehfeldt and Gelhar, 1992; Molz,
2015). The Marshall site hydrogeology certainly qualifies as "heterogeneous". This is very important to
consider for the CAP transport model calibrations because it is the high -permeability zones and/or layers
that control the time required (Tt,,,, l ) for a constituent to reach a downgradient observation point, and
22

HDR used differences in observed versus simulated Tt,,,,e, (i.e., time to travel from sources zones to
downgradient monitoring wells) as the justification for lowering measured Kd values.
Second, the history matching that HDR performed is very sensitive to the assumed time at which the
source (i.e., coal ash) is "turned on" and to the assumed distribution of source concentrations (fixed pore
water concentrations) in source area cells. Section 5.2 of CAP 1 Appendix C explains that different fixed
source concentration boundary conditions were specified for the Ash Basin (1965; model year 0), Dry
Ash Landfill Phases 1 & 2 (1984; model year 19), and the PV Structural Fill (2000; model year 35).
These different source concentrations were assumed to remain constant following their activation in the
model. For several reasons this source -area boundary condition approach is a major simplification and
generally inaccurate: coal ash was gradually and nonuniformly disposed of and distributed (spatially and
temporally) throughout the simulation period; it is very difficult (or not possible) to accurately extrapolate
geochemical or ash -water leaching conditions (i.e., predict COI pore -water concentrations) that existed
during the 2015 sampling round to conditions that may have existed in 1965 and thereafter; the actual
source -area concentration distributions are highly nonuniform, but it is not clear from the CAP modeling
reports how "... source concentrations were adjusted to match measured values... ", or if the source area
concentrations were nonuniform. All of these uncertainties are further magnified when using history
matching to calibrate a chemical transport model.
Based on the above model input errors and major uncertainties in hydraulic -conductivity variations and
source -term modeling, it is incorrect for HDR to simply reduce Kd values by factors of 10 to 100 below site
measurements (and the large database of literature Kd values) based only on the transport model "history
matching" exercises that HDR performed. My additional comments on the CAP Parts 1 and 2 transport
modeling of Closure Scenarios are listed in the following section.
Geochemical Modeling and Evaluation of Monitored Natural Attenuation
The CAP Part 2 geochemical modeling results do not include quantitative analyses of COI attenuation
rates at the Marshall site and are only qualitative in nature. In addition, HDR did not incorporate any
source/sink (e.g., precipitation/dissolution) terms representing geochemical reaction mechanisms in the
CAP 2 chemical transport model to evaluate whether such reactions are important compared to
groundwater concentration changes caused by advection, dispersion, and soil -water partitioning. In this
regard, HDR states in Section 2.10 of CAP 2 Appendix B: "A physical -type modeling approach was
used, as site-specific geochemical conditions are not understood or characterized at the scale and extent
required for inclusion in the model." Indeed, the Electric Power Research Institute (e.g., EPRI, 1984;
page S-8) has extensively reviewed subsurface chemical attenuation mechanisms applicable to the "utility
waste environment" and concluded: (i) precipitation/dissolution has not been adequately studied; and (ii)
"Quantitative predictions of chemical attenuation rates based upon mineralogy and groundwater
23

composition cannot be made because only descriptive and qualitative information are available for
adsorption/desorption mechanisms."
Nonetheless, HDR performed geochemical modeling to evaluate the technical basis for its monitored
natural attenuation (MNA) analysis; however, any quantitative MNA analysis must compare mass
transport rates and changes (e.g., grams/year per unit area normal to a groundwater pathline) in the
aquifer for the various active transport mechanisms in order to determine whether MNA is a viable
alternative (e.g., produces meaningful groundwater concentration reductions) at the Marshall site. In
Section 6.3.2 of the CAP 2 report HDR acknowledges that these quantitative evaluations were not
performed in CAP 2 and indicated that they would need to be completed as part of a Tier III MNA
assessment. Nevertheless, HDR suggested in the CAP 2 report that COI concentrations "will" or "may"
attenuate over time without completing the necessary evaluations to reach these conclusions. HDR also
states in CAP 2 Section 6.3.4 that ".... available assessment results indicate it is feasible that MNA can be
used partially or entirely to remediate groundwater at the MSS site .... ". HDR claims in CAP 2, Section
6.3.3 (page 47) that "The groundwater model did not allow for removal of COI via coprecipitation with iron
oxides, which likely resulted in a conservative prediction of COI transport." Finally, HDR concludes in
Section 7.2.2.1 of the CAP 2 report the following regarding the MNA assessment and Appendix H: "The
most significant finding was that the precipitation of iron and manganese serves to remove other COls
through co -precipitation and adsorption, thus confirming that attenuation is occurring." I saw no
quantitative analyses or evidence in the CAP 2 report or related appendices to support these claims. In
fact, the CAP 2 Appendix H emphasizes that much more geochemical data need to be collected and
chemical transport modeling with a source/sink term must be performed in a Tier III assessment to further
assess whether MNA is a viable remedial alternative.
Therefore, the CAP 2 report fails to provide any quantitative evidence supporting COI attenuation due to
co -precipitation with iron or manganese. The second component of COI attenuation evaluated in
Appendix H is chemical sorption to soil. It is important to note that, although the CAP models did not
incorporate a mechanism for co -precipitation with iron or manganese (or any COI sink term), the CAP
models did simulate attenuation due to sorption. Even with the sorption attenuation mechanism included,
CAP 2 Table 4-1 shows that for both the "existing conditions" and "cap -in-place" scenarios the following
COI will exceed North Carolina groundwater standards at the Compliance Boundary 100 years into the
future: antimony, beryllium, boron, chromium, cobalt, hexavalent chromium, thallium, and vanadium.
Further, my Table 1 shows that groundwater standards are currently exceeded at the Compliance
Boundary for iron, manganese, and total dissolved solids (i.e., iron, manganese, and TDS contaminant
plumes originating in the source areas have already reached the Compliance Boundary). The
conclusions of the MNA Tier I analyses (CAP 2 Appendix H, page 17) were that antimony, beryllium,
boron, cobalt, manganese, and selenium show limited evidence of attenuation and should not be
24

evaluated further for MNA. These Appendix H conclusions are consistent with the Table 4-1 results for
several of these COI (antimony, beryllium, boron, and cobalt). In addition, my Table 1 results support the
Appendix H conclusions that antimony, boron, cobalt, and manganese are not candidates for MNA
because these contaminant plumes have also already reached the Compliance Boundary. The Table 1
results contradict the Appendix H conclusion that chromium and vanadium should be further evaluated
because these COI currently exceed groundwater standards at the Compliance Boundary. All of these
data and CAP 2 modeling results strongly contradict the CAP 2 conclusion (e.g., Section 7.2.2.1) that "....
MNA is the recommended corrective action for the Marshall site."
Simulation of Closure Scenarios
As discussed below, CAP 1 Closure Scenario simulations greatly underestimate (by factors of 10 or
more) the time frames required to achieve meaningful groundwater concentration reductions in response
to remedial actions. Compared to the Cap -in -Place remedial alternative evaluated in the CAP Part 1, the
Excavation Scenario results in COI concentrations in groundwater that are in many cases at least a factor
of 10 smaller.
Although the CAP 1 modeling showed that Source Excavation outperforms CIP, the CAP 2 modeling did
not simulate an Excavation closure scenario. Nonetheless, the following comparisons between CIP and
Excavation impacts on groundwater concentrations are valid for both the CAP 1 and 2 model results.
This is because the main difference with the CAP 2 transport model (compared to CAP 1) is that
concentration changes resulting from either CIP or Excavation (if it was evaluated in CAP 2) occur much
more slowly (i.e., — 10x slower; 25-480x slower for arsenic, thallium, and vanadium) in the CAP 2 model
due to the much larger Kd (and Rd). The CAP 2 transport model also assumed uniform initial COI
concentrations equal to HDR's proposed provisional background concentrations (PPBC), even though the
PPBC exaggerate background levels (see above discussion) and there are no data to suggest that
background concentrations should be spatially uniform. Despite these changes in the CAP 1 and 2
models, the relative differences in groundwater concentrations between the two closure scenarios remain
about the same if the uniform starting (PPBC) COI concentrations are subtracted from the simulated
concentration versus time curves. For these reasons the following discussions focus on the CAP 1
modeling results.
Source Concentrations for Cap -in -Place Scenario
In this scenario the CAP 1 flow model predicts limited cap -induced water -table declines: approximately
14 feet (10 feet in the CAP 2 flow modeling) beneath the downgradient portion of the ash basin; and
about 2 feet, 3 feet, and 7 feet (about 1 foot in CAP 2) beneath the dry ash landfill (Phase 1), Phase II of
the dry ash landfill, and the PV structural fill, respectively. The geologic cross-sections presented in the
CSA show that the saturated coal ash thickness in the ash basin generally varies from 25 to 70 feet and
25

is on the order of 50 feet at many locations. Therefore, these cap -induced water table declines are about
a factor of four too small at most locations in the ash basin to dewater the source material. For example,
the CSA geologic cross-sections show saturated coal ash thicknesses in the active ash basin equal to
about 40-50 feet in cross-section A -A (northwest to southeast section generally through the middle of the
ash basin); 50-70 feet in cross-section B -B (southwest portion of the ash basin); and 25-45 feet in cross-
section C -C (north-northeast portion of the ash basin).
This means that under the simulated Cap -In -Place Scenario at the majority of locations in the ash basin
at least 75 percent of the coal ash, which is the source of dissolved COI, would remain saturated and
continue to leach constituents into groundwater. The CAP 1 simulations ignored this fact and set all
source concentrations equal to zero (i.e., assumed all coal ash was dewatered). Therefore, the simulated
Cap -in -Place concentrations should be much higher than the values presented in the CAP Part 1.
It is important to note, however, that the CAP Parts 1 and 2 groundwater flow model simulations
exaggerate the hydraulic effects of the cap (i.e., overstates water table lowering) due to the lateral no -flow
boundaries and the no -flow boundary at the base of the model, which is located only 50 feet below the
top of the highly -permeable, fractured bedrock aquifer (— 100 feet below ground surface). Note that the
depths of the neighboring Duke Energy and Midway Marina bedrock water supply wells, discussed
above, are 540 feet and 245 feet below ground surface (about 400 feet and 160 feet below the bedrock
surface), respectively, and provide high groundwater extraction yields of 30 and 20 gallons per minute,
respectively [Table 1 of the 2014 receptor survey report (HDR, 2014)]. The no -flow boundary conditions
along the entire western, southern, and northern grid boundaries prevent flow into the ash basin when
large, laterally inward hydraulic gradients are created by capping. In addition, the no -flow boundary
condition at the base of the flow model prevents upward flow from highly -permeable portions of the
bedrock aquifer that underlie the base of the CAP models, thus further exaggerating predicted water table
lowering.
Slow and Multirate Nonequilibrium Desorption of COI
Since the 1980's the groundwater industry has learned how difficult it is to achieve water quality
standards at remediation sites without using robust corrective actions such as source removal (Hadley
and Newell, 2012, 2014; Siegel, 2014). Two of the key reasons for this in aqueous -phase contaminated
soil are inherently low groundwater or remediation fluid flushing rates in low -permeability zones and slow,
nonequilibrium chemical desorption from the soil matrix (Culver et al., 1997, 2000; Zheng et al., 2010). A
good example of this is the "tailing effect" (i.e., very slow concentration reduction with time) that is
commonly observed with pump -and -treat, hydraulic containment systems. These factors are also related
to the "rebound effect" in which groundwater concentrations sometimes increase shortly after a
26

remediation system is turned off (Sudicky and Illman, 2011; Hadley and Newell, 2014; Culver et al.,
1997).
The CAP 1 and 2 flow models use different permeability (K) zones, but the scale of these zones is very
large and within each zone K is homogeneous even though large hydraulic conductivity variations (e.g.,
lognormal distribution) are known to exist at any field site over relatively small length scales (Molz, 2015).
Moreover, the CAP transport models assume linear, equilibrium soil -water partitioning which corresponds
to instantaneous COI release into flowing groundwater. The transport code (MT3D) has the capability of
simulating single -rate nonequilibrium sorption, but the Close Scenario simulations did not utilize this
modeling feature. Slow desorption of COI can also be expected at the Marshall site because sorption
rates are generally highly variable and multi -rate (Culver et al., 1997, 2000; Zheng et al., 2010), and Kd
values are nonuniform spatially (Baes and Sharp, 1983; EPRI, 1984; De Wit et al., 1995). The CAP flow
and transport models can be expected to significantly underestimate cleanup times required to meet
groundwater standards at the compliance boundary because they do not incorporate these important
physical mechanisms.
Adequacy of the Kd Model for Transport Simulation
The laboratory column experiment effluent data (e.g., CAP 1 Appendix D) generally gave very poor
matches with the analytical (one-dimensional) transport model used to compute Kd values. Since the
CAP transport models solve the same governing equations in three dimensions, the adequacy of the Kd
modeling approach for long-term remedial simulations should have been evaluated in much more detail in
the modeling appendix.
The transport modeling also did not evaluate alternative nonlinear sorption models such as the Freundlich
and Langmuir isotherms (Hemond and Fechner, 1994), which are input options in the MT31D transport
code. Several of the batch equilibrium sorption experiments (CAP 1 Appendix D) exhibited nonlinear
behavior, and such behavior is commonly observed in other studies (e.g., EPRI, 1984). However, HDR
only computed linear sorption coefficients (i.e., Kd) for the Marshall site in CAP Part 1. In CAP Part 2
HDR did fit Freundlich isotherms to the batch sorption data for selected COI (CAP 2 Appendix C, Tables
1-8) but did not use these Freundlich isotherm results in the CAP 2 transport modeling. De Wit et al.
(1995) showed that the nonlinear sorption mechanism is similar in importance to aquifer heterogeneities
in extending remediation time frames.
Closure Scenario Time Frames
Many of the simulated concentration versus time (CvT) curves presented in CAP 1 Appendix C are either
incorrect or highly misleading with respect to future groundwater concentration variations that may occur
under the Existing Conditions, Cap -In -Place (CIP), and Excavation scenarios. For example, several CvT
27

curves suggest that concentrations of several COI at various locations in the aquifer system would be as
much as factors of 10 to 100 greater if all of the ash source material is removed (Excavation) compared to
either doing nothing (Existing Conditions) or CIP for a period of 200 years into the future (e.g., arsenic,
antimony, boron, chloride, cobalt, chromium, hexavalent chromium, selenium, sulfate, thallium, and
vanadium at AB -1 BR; antimony at AB -SBR; arsenic, selenium, and vanadium at AB-6BR; barium at AL -
2D; beryllium at AB -1 S & AL -1 S; chromium at AB -3D; and selenium, sulfate, and vanadium at AL -4D).
Further, the model results for some of these CvT curves show groundwater concentrations increasing at a
dramatic rate 200 years into the future even if all source material is removed (e.g., selenium and
vanadium at AB-6BR; arsenic, cobalt, and hexavalent chromium at AB -1 BR; arsenic at AB -613R). The
CvT curves are also highly inconsistent at the same location (e.g., for the Excavation Scenario at AB -1 BR
arsenic concentrations increase significantly for 200 years, but boron concentrations decrease
significantly for the same time period) and between different locations (e.g., for the same COI the CvT
curves suggest that at some locations Excavation is much better than CIP, but at other locations CIP is
much better than Excavation).
In addition to these issues, the CAP 1 chemical transport model underestimates the time intervals
required to achieve groundwater concentration reductions (i.e., achieve groundwater quality restoration)
by a factor of 10 or more, as discussed in different parts of my report. In other words, the CAP 1 transport
model significantly overestimates the rate at which concentrations may reduce in response to remedial
actions such as capping or source removal. This is due to several factors, including major errors in model
input data, model calibration mistakes, field data analysis errors, and oversimplified model representation
of field conditions (e.g., hydraulic conductivity) and transport mechanisms (e.g., chemical
sorption/desorption). These limitations of transport models for realistically predicting cleanup times have
been recognized by the groundwater industry for the past few decades based on hands-on experience at
hundreds of extensively -monitored remediation sites.
Even if we ignore the above problems with the CvT curves in CAP 1 Appendix C and the factors of 10 or
more errors in cleanup time predictions with the CAP model, the remediation time frames for the
Excavation Scenarios are still more than two centuries for several constituents due to slow groundwater
flushing rates from secondary sources (surrounding residual soil) left in place after excavation and due to
high chemical retardation factors for most of the COI. However, excavation of secondary -source material
would further accelerate cleanup rates under this alternative. The simulated Cap -In -Place concentration
reduction rates are much slower, compared to excavation, but are also incorrect (i.e., overestimated)
because the cap -induced water -table lowering only dewaters a small percentage (— 25 percent) of the
source -area coal ash, as discussed above, and the CAP flow models overestimate cap -induced water -
table lowering due to boundary condition errors. Furthermore, these simulation time frames are well
beyond the prediction capabilities of any chemical transport model for a complex field site, especially one
28

that is as geochemically complex as the Marshall site. The historical model -calibration dataset (1965-
2015) is also significantly smaller than the predictive (remediation) time frames. In addition, the "history
matching" technique used to calibrate the transport model (e.g., major reduction in measured Kd values)
was not performed correctly by HDR.
Cap -In -Place versus Excavation Closure Scenarios
Even though the CAP 1 model significantly underestimates remediation time frames, the CAP 1 Closure
Scenario simulations (using CvT curves that do not appear to be suspect) demonstrate several significant
advantages of excavation for restoring site groundwater quality versus CIP. First, predicted COI
concentration reductions in groundwater are in many cases at least a factor of 10 greater with excavation
compared to CIP [e.g., antimony and boron (AB -31D, AB-5BR); arsenic, barium, cobalt, chromium,
hexavalent chromium, thallium (AB -3D); beryllium, cobalt, chromium, sulfate, thallium (AB-5BR); refer to
CAP Appendix C]. Further, if the cap -in-place simulations would have been performed correctly the
simulated cap -in-place concentrations would be much higher because predicted water -table lowering due
to the cap would be insufficient to dewater all of the coal ash.
Although the CAP 1 modeling demonstrated that the CIP closure alternative would be much less effective
than excavation, and that CIP would only dewater about one-fourth of the saturated coal -ash thickness in
many areas, HDR eliminated excavation from consideration in CAP 2. In Section 7.1 of the CAP 2 report
HDR assumes that "groundwater flow and geochemical modeling indicates that attenuation by a
combination of sorption, chemical precipitation, and dilution by surface water infiltration and fresh
groundwater effectively dissipates COls in groundwater beneath and downgradient of the source areas"
and that "... it is reasonable to assume that COls remaining in groundwater will decrease in concentration
overtime as upgradient non -impacted water moves through the aquifer." As discussed above, HDR
provided no quantitative analysis or evidence in the CAP 2 report or related appendices to support this
claim. Considering that up to 75 percent of the coal -ash source material would remain saturated with CIP
and that multiple exceedances of groundwater standards at the Compliance Boundary currently exist
(with no historical data to indicate that these Compliance Boundary concentrations are decreasing with
time), it is not reasonable to make sweeping assumptions about future concentration changes. Tier III
MNA analyses require rigorous quantitative evaluations using the CAP transport model with a source/sink
term that incorporates geochemical reactions to support MNA as a viable corrective action. CAP 2 did
not provide this information.
As discussed above, the CAP Part 2 flow model did include a very small number of residential wells (4 of
the 89 neighboring private wells and none of the four public water supply wells), but the CAP 2 modeling
report (CAP 2, Appendix B) did not show simulated hydraulic head maps with these residential wells
pumping and did not provide any discussion or analyses of the long-term potential for these wells to
29

capture COI dissolved in groundwater. Further, the private bedrock wells that HDR chose to include in
the CAP 2 model are located upgradient from the ash basins; HDR should have included all of the
private wells located immediately north, west, and south of the Duke Energy property boundary and the
four public water supply wells to the south (refer to my previous discussion revised maps of bedrock
groundwater flow directions). In CAP 2 section 4.1.5 HDR discusses that fact that the CAP 2 flow model
was used to compute 1 -year, reverse particle pathlines for these bedrock residential wells (Figure 15 in
CAP 2 Appendix B) to determine their short-term groundwater capture zones. However, the residential
well reverse pathline tracing should have been performed for a much longer time period (e.g., from the
beginning of coal ash disposal to the present) to evaluate whether COI may have migrated from source
areas to these wells. In addition, if HDR had extended the CAP 2 model grid much farther to the north,
west, and south the capture zones for the remaining 85 private and four public water supply wells could
have been determined, as I discuss earlier in my report.
The CAP Closure Scenarios do not include hydraulic containment remedial alternatives (e.g., gradient
reversal) for the bedrock aquifer that would address the risk of off-site COI transport. As discussed
above, the CSA data show many exceedances of groundwater standards in bedrock inside the
compliance boundary. In addition, downward groundwater flow components from the deep overburden to
bedrock aquifers were measured during the CSA in the southeastern portion of the site where
groundwater flow in the bedrock aquifer is off-site (Figures 2 and 3), and ponded water in the ash basin
impoundment has historically created a large downward hydraulic gradient beneath the ash basin. The
cap -in-place alternative does not address either concentration reduction or off-site chemical migration
control in the fractured bedrock aquifer.
Conclusions
Based on my technical review and analyses of the referenced information for the Marshall site I have
reached the following conclusions:
• A total of 21 Compliance Boundary groundwater samples exceeded North Carolina groundwater
standards for these COI: antimony, boron, chromium, cobalt, iron, manganese, total dissolved
solids, and vanadium. Of these 21 exceedances, 19 were greater than the proposed provisional
background concentrations by HDR, which exaggerate background levels;
• The statistical analyses of historical, shallow -aquifer background groundwater concentrations at
the Marshall site (monitoring well MW -4) are invalid. The time periods between groundwater
sample collection from this well are too small and the concentration data are not independent;
30

• There is a significant risk of chemical migration from the ash basin to neighboring private and
public water supply wells in fractured bedrock. The design of the CAP Parts 1 and 2 flow and
transport models prevents the potential for off-site migration from being evaluated;
• The limited CAP model domain size (Parts 1 and 2); the no -flow boundary conditions along the
western, southern, and northern boundaries; the impermeable boundary at the base of the model
at a depth of only about 50 feet below the top of the fractured bedrock aquifer; the incorrect
hydraulic boundary condition representations of Lake Norman and the active ash basin; and
HDR's failure to include bedrock -aquifer groundwater extraction from most neighboring private
and public water supply wells prevent simulation and analysis of off-site COI migration;
• The CAP Closure Scenario simulations greatly underestimate (by factors of 10 or more) the time
frames required to achieve meaningful groundwater concentration reductions in response to
remedial actions. This is due to oversimplification of field fate and transport mechanisms in the
CAP Parts 1 and 2 models and several model input errors;
• The simulated water table lowering for the Cap -in -Place Scenario is about a factor of four too
small at most locations in the ash basin in order to dewater all source material. In addition, the
actual cap -induced water table elevation reduction would be much less than predicted due to the
incorrect no -flow boundary conditions (3 of 4 lateral boundaries and the bottom bedrock surface)
in the flow model. Therefore, the remediation time frames for this scenario would be much
greater because most of the source zone would still be active with the cap installed;
• For either the Existing Condition or Cap -in -Place Model Scenario groundwater concentrations of
many coal -ash constituents much higher than background levels will continue to exceed North
Carolina groundwater standards at the Compliance Boundary because saturated coal -ash
material and secondary sources will remain in place;
• Source -area mass removal included in the Excavation Scenario results in COI groundwater
concentration reductions that are at least a factor of 10 greater compared to Cap -in -Place at
many locations. Additional excavation of secondary sources would further accelerate
concentration reductions;
• Due to an incorrect boundary -condition representation of the active ash basin, the CAP models
underestimate by a factor of three or more both the mass loading of COI into Lake Norman and
the corresponding Lake Norman water concentrations (attributable to coal ash ponds) estimated
by the groundwater/surface-water mixing model;
• The CAP Part 2 geochemical modeling and monitored natural attenuation (MNA) evaluations do
not provide the required quantitative analyses (e.g., numerical transport modeling) of COI
attenuation rates necessary to support MNA as a viable corrective action and are only qualitative
in nature. The CAP 2 chemical transport modeling, which included attenuation by sorption,
31

demonstrated that MNA is not an effective remedial option for several COI (e.g., antimony,
beryllium, boron, chromium, cobalt, hexavalent chromium, thallium, and vanadium);
• The CAP Closure Scenarios do not include hydraulic containment remedial alternatives for the
bedrock aquifer and do not address the risk of off-site COI transport. CSA data show 22
exceedances of groundwater standards in bedrock inside the Compliance Boundary. The cap -in-
place alternative does not address off-site chemical migration control in the fractured bedrock
aquifer; and
• Future Compliance Monitoring at the site should include much more closely -spaced Compliance
Wells to provide more accurate detection, and the time intervals between sample collection
should be large enough to ensure that the groundwater sample data are statistically independent
to allow accurate interpretation of concentration trends.
32

References
Baes, C.F., and R.D. Sharp. 1983. A Proposal for Estimation of Soil Leaching and Leaching Constants
for Use in Assessment Models. Journal of Environmental Quality, Vol. 12, No. 1. 17-28.
Barker, J.A., and J.H. Black. 1983. Slug Tests in Fissured Aquifers. Water Resources Research. Vol.
19, No. 6. 1558-1564.
Bear, J. 1979. Hydraulics of Groundwater. New York: McGraw-Hill.
Bouwer, H., and R.C. Rice. 1976. A Slug Test for Determining Hydraulic Conductivity of Unconfined
Aquifers with Completely or Partially Penetrating Wells. Water Resources Research. Vol. 12, No. 3.
423-428.
Culver, T.B., S.P. Hallisey, D. Sahoo, J.J. Deitsch, and J.A. Smith. 1997. Modeling the Desorption of
Organic Contaminants from Long -Term Contaminated Soil Using Distributed Mass Transfer Rates.
Environmental Science and Technology, 31(6), 1581-1588.
Culver, T.B., R.A. Brown, and J.A. Smith. 2000. Rate -Limited Sorption and Desorption of 1,2 -
Dichlorobenzene to a Natural Sand Soil Column. Environmental Science and Technology, 34(12),
2446-2452.
Daniel, C.C., D.G. Smith, and J.L. Eimers. 1989. Chapter C, Hydrogeology and Simulation of Ground -
Water Flow in the Thick Regolith -Fractured Crystalline Rock Aquifer System of Indian Creek Basin,
North Carolina. U.S. Geological Survey Water -Supply Paper 2341-C. Ground -Water Resources of
the Piedmont -Blue Ridge Provinces of North Carolina.
Deutsch, C.V., and A.G. Journel. 1992. GSLIB, Geostatistical Software Library and User's Guide.
Oxford University Press.
De Wit, J.C.M., J.P. Okx, and J. Boode. 1995. Effect of Nonlinear Sorption and Random Spatial
Variability of Sorption Parameters on Groundwater Remediation by Soil Flushing. Groundwater
Quality: Remediation and Protection, Proceedings of the Prague Conference, May 1995. IAHS
Publication No. 225.
EPA. 1985. Full -Scale Field Evaluation of Waste Disposal from Coal -Fired Electric Generating Plants.
Report EPA -600/7-85-028a, June 1985, Volume I, Section 5. Prepared by Arthur D. Little, Inc.
EPA. 2009. Statistical Analysis of Groundwater Monitoring Data at RCRA Facilities - Unified Guidance.
Office of Resource Conservation and Recovery, Program Implementation and Information Division,
U.S. Environmental Protection Agency. Report EPA 530/R-09-007. March 2009.
EPRI. 1984. Chemical Attenuation Rates, Coefficients, and Constants in Leachate Migration. Volume 1:
A Critical Review. Electric Power Research Institute Report EA -3356, Volume 1. Prepared by
Battelle, Pacific Northwest Laboratories, Richland, Washington. February 1984.
Gelhar, L.W. 1984. Stochastic analysis of flow in heterogeneous porous media. In Selected Topics in
Mechanics of Fluids in Porous Media, edited by J. Bear and M.Y. Corapcioglu, pp. 673-717, Martinus
Nijhoff, Dordrecht, Netherlands.
Gelhar, L.W. 1986. Stochastic subsurface hydrology from theory to applications. Water Resources
Research 22: 135S -145S.
33

Gelhar, L.W. 1987. Stochastic analysis of solute transport in saturated and unsaturated porous media.
In Advances in Transport Phenomena in Porous Media, NATO ASI Ser., edited by J. Bear and M.Y.
Corapcioglu, pp. 657-700, Martinus Nijhoff, Dordrecht, Netherlands.
Gelhar, L.W., and C.L. Axness. 1983. Three-dimensional stochastic analysis of macrodispersion in
aquifers. Water Resources Research 19, no. 1: 161-180.
Hadley, P.W., and C.J. Newell. 2012. Groundwater Remediation: The Next 30 Years. Groundwater,
Vol. 50, No. 5. 669-678.
Hadley, P.W., and C.J. Newell. 2014. The New Potential for Understanding Groundwater Contaminant
Transport. Groundwater, Vol. 52, No. 2. 174-186.
Hantush, M.S. 1964. Hydraulics of Wells. Advances in Hydroscience. Vol. 1. Academic Press. Ed.
V.T. Chow. 282-437.
Hemond, H.F., and E.J. Fechner. 1994. Chemical Fate and Transport in the Environment. Academic
Press.
HDR. 2014. Receptor Survey, Marshall Steam Station Ash Basin, NPDES Permit NC0004987.
Prepared for Duke Energy by HDR Engineering, Inc. of the Carolinas, October 10, 2014.
LeGrand, H.E. 2004. A Master Conceptual Model for Hydrogeological Site Characterization in the
Piedmont and Mountain Region of North Carolina. Prepared for North Carolina Department of
Environment and Natural Resources, Division of Water Quality, Groundwater Section.
Lyman, W.J., W.F. Reehl, and D.H. Rosenblatt. 1982. Handbook of Chemical Property Estimation
Methods. McGraw-Hill Book Company.
McDonald, M.G., and A.W. Harbaugh. 1988. MODFLOW, A Modular Three -Dimensional Finite -
Difference Ground -Water Flow Model. Techniques of Water -Resources Investigations of the United
States Geological Survey, Department of the Interior.
McGary, M. 2004. Annual Water Use Report, Marshall Steam Station, 2004. North Carolina Division of
Water Resources, Water Withdrawal and Transfer Registration.
Molz, F. 2015. Advection, Dispersion, and Confusion. Groundwater, Vol. 53, No. 3. 348-353.
Pippin, C.G., M.J. Chapman, B.A. Huffman, M.J. Heller, and M.E. Schelgel. 2008. Hydrogeologic Setting,
Ground -Water Flow, and Ground -Water Quality at the Langtree Peninsula Research Station, Iredell
County, North Carolina, 2000-2005. Scientific Investigations Report 2008-5055, U.S. Department of
the Interior, U.S. Geological Survey.
Rehfeldt, K.R., and L.W. Gelhar. 1992. Stochastic analysis of dispersion in unsteady flow in
heterogeneous aquifers. Water Resources Research 28, no. 8: 2085-2099.
Rehfeldt, K.R., J.M. Boggs, and L.W. Gelhar. 1992. Field study of dispersion in a heterogeneous aquifer,
3, geostatistical analysis of hydraulic conductivity. Water Resources Research 28, no. 12: 3309-
3324.
Schroeder, P.R., T.S. Dozier, P.A. Zappi, B.M. McEnroe, J.W. Sjostrom, and R.L. Peyton. 1994. The
Hydrologic Evaluation of Landfill Performance (HELP) Model: Engineering Documentation for
Version 3. Report EPA/600/R-94/168b, September 1994, U.S. Environmental Protection Agency,
Office of Research and Development, Washington, D.C.
34

Siegel, D.I. 2014. On the Effectiveness of Remediating Groundwater Contamination: Waiting for the
Black Swan. Groundwater, Vol. 52, No. 4. 488-490.
Simon, R.B., S. Bernard, C. Meurville, and V. Rebour. 2105. Flow -Through Stream Modeling with
MODFLOW and MT3D: Certainties and Limitations. Groundwater, Vol. 53, No. 6. 967-971.
Sudicky, E.A., and W.A. Illman. 2011. Lessons Learned from a Suite of CFB Borden Experiments.
Groundwater, Vol. 49, No. 5. 630-648.
van Genuchten, M.Th., and W.J. Alves. 1982. Analytical Solutions of the One -Dimensional Convective -
Dispersive Solute Transport Equations. U.S. Department of Agriculture, Agricultural Research
Service, Technical Bulletin No. 1661.
Winter, T.C., D.O. Rosenberry, and J.W. La Baugh. 2003. Where Does the Ground Water in Small
Watersheds Come From?. Groundwater, Vol. 41, No. 7. 989-1000.
Yin, Z. -Y., and G.A. Brook. 1992. The Topographic Approach to Locating High -Yield Wells in Crystalline
Rocks: Does It Work?. Groundwater, Vol. 30, No. 1. 96-102.
Zheng, C., G.D. Bennett, and C.B. Andrews. 1991. Analysis of Ground -Water Remedial Alternatives at a
Superfund Site. Groundwater, Vol. 29, No. 6. 838-848.
Zheng, C., M. Bianchi, S.M. Gorelick. 2010. Lessons Learned from 25 Years of Research at the MADE
Site. Groundwater, Vol. 49, No. 5. 649-662.
35

Figures
36

10'
10'
00 10,
C
0
i 10-1
0
10"
10'
1000 2000 3000
Distance from Pumping Well (feet)
Figure 1
Drawdown in Bedrock Versus Distance from Pumping Well
37
4000
KBR = 0.68 fWay
���• KBR=4.7ft/day
Q
K,=2.0ftlday
i
i4
i
• i
i
14
• i
■ r
14 ` • r
"O
r " r ■
•
i
r
i
1000 2000 3000
Distance from Pumping Well (feet)
Figure 1
Drawdown in Bedrock Versus Distance from Pumping Well
37
4000

BG-2BR
v
i:i Sfl4 1
GWA-9BR = 1
1 --a' AB•15BR
-..
Ate' i AL-2BR
�Jb„ ,. J ` �,. Y-* � _^��. 111 ���• �\ ".,_•,l
r
P tAB-GBRJ _` i 4 p
qpABABR.
` ... Lake Norman j 1
Ash Basin Waste Boundary - i, AB -CBR e`�'. t ' r I • /' j \
t �
e Lake Norman
~GWA-1BR % •j• •
tr
�$ DId Country Church V/
Duke Supply Well
•1 I TI f \ 1 `v
Figure 2
Bedrock Hydraulic Head Map with Off -Site Pumping
38

GlR.I B' R� I
w AB -68R
Ash Basin Waste Boundary
BG -26R
726
MW-14BR
Ip
SAB IBR
Lake Norman
Lake Norman
BR
*I Old CouELty Church]
F0
Figure 3
Bedrock Hydraulic Head Map without Off -Site Pumping
39

Table
40

Table 1. Exceedances of NC Groundwater Standards at Compliance Boundary and in Bedrock
for 2015 Monitoring Well and Seep Samples
Constituent
Antimony
Background
Sample Sample Location Concentration Concentrationsc,d
Name Depth Measured PPBC
GWA-2D Deep CB 4.1
41
Standard
2L IMAC
2_5 1 -
ND (BG -1D, BG -3D), 0.33 (MW -4D)
ND (BG-2BR)
5
AB-6BR
Bedrock
BR
1.3
Arsenic
S-2
Seep
Ash Basin
87.1
Barium
S-2
Seep
Ash Basin
990
Beryllium
S-2
Seep
Ash Basin
15.2
Boron
MW -14S
Shallow
CB Me
2,500e
AL -1 S
CB Me
4,300e
AB -IIS
CB Me
4,500e
MW -7S
CB Me
4,400e
MW -14D
Deep
CB Me
2,500e
AL -1 D
CB Me
1,100e
AL-2BR
Bedrock
BR
2,100
S-2
Seep
Ash Basin
4,000
MSSW001
S001
6,000
41
Standard
2L IMAC
2_5 1 -
ND (BG -1D, BG -3D), 0.33 (MW -4D)
ND (BG-2BR)
5
10 -
157.3
700 -
1
4 -
100
700 -
ND (BG -1S, -2S, -3S; MW -4)
ND (MW -4D, BG -1 D),
26 (BG -3D)
ND (BG-2BR)

Table 1. Continued
Constituent
Sample
Name
Sample
Depth
Location
Concentration'
Background
Concentrations,,d Standard'
Measured PPBC 2L IMAC
Chromium
GWA-7S
Shallow
CB
22.1
11.3 10 -
5.7 (BG -1S), 9 (BG -2S),
2.5 (MW -4), 73.7 (BG -3S)
GWA-2D
Deep
CB
182
3.1 (BG -1 D), 6.6 (BG -3D),
1.2 (MW -4D)
AL-2BR
BR
17.5
80.4 (BG-2BR)
S-2
Seep
Ash Basin
85.7
Cobalt
GWA-2S
Shallow
CB
2.6
2.5 - 1
1.2 (BG -1S), 0.38 (BG -2S),
ND (MW -4),5 (BG -3S)
AB-5BR
Bedrock
BR
7.9
GWA-1 BR
BR
1.7
11.9 (BG-2BR)
S-2
Seep
Ash Basin
333
MSSW001 S001
92
Iron
GWA-7S
Shallow
CB
12,300
467.1 300 -
480 (BG -1 S), 140 (BG -2S),
510 (MW -4), 3,400 (BG -3S)
GWA-7D
Deep
CB
2,200
310 (BG -1 D), 250 (BG -3D),
77 (MW -4D)
42

Table 1. Continued
Constituent
Sample
Name
Sample
Depth'
Location
Concentration`
Background
Concentrationsc,d
Measured PPBC
Standard
2L IMAC
AB -1 BR
Bedrock
BR
2,800
467.1
300 -
Iron
GWA-1 BR
BR
780
(continued)
MW-14BR
BR
320
18,200 (BG-2BR)
AB-15BR
BR
1,300
Lead
S-2
Seep
Ash Basin
227
35.9
15 -
Manganese
GWA-2S
Shallow
CB
92
48
50 -
GWA-7S
CB
510
160 (BG -1S), 20 (BG -2S),
19 (MW -4),180 (BG -3S)
GWA-7D
Deep
CB
76
54 (BG -1 D), 24 (BG -3D),
4.1 (MW -4D)
AB -1 BR
Bedrock
BR
54
GWA-1 BR
BR
780
380 (BG-2BR)
AB-5BR
BR
650
AB-15BR
BR
270
S-2
Seep
Ash Basin
11,600
MSSW001 S001
8,500
Selenium
AL-2BR
Bedrock
BR
24
1.6 (BG-2BR) 10
20 -
S-2
Seep
Ash Basin
25.1
43

Table 1. Continued
Constituent
Sample
Name
Sample
Depth'
Location
Concentration`
Background
Concentrationsc,d Standard
Measured PPBC 2L IMAC
Thallium
S-2
Seep
Ash Basin
8.6
0.5 - 0.2
MSSW001 S001
0.6
Total
GWA-2D
Deep
CB
650,000
85,400 500,000 -
Dissolved
142,000 (BG -1 D), 183,000 (BG -3D),
Solids
82,000 (MW -4D)
S-2
Seep
Ash Basin
892,000
MSSW001 S001
989,000
Vanadium
GWA-2S
Shallow
CB
0.33
3.9 - 0.3
GWA-7S
CB
23.7
3.8 (BG -1S), 44 (BG -2S),
2.2 (MW -4),14.4 (BG -3S)
GWA-7D
Deep
CB
6.1
3.5 (BG -1 D), 21.9 (BG -3D),
MW -10D
CB
3.2
2.8 (MW -4D)
GWA-2D
CB
12.2
GWA-1 BR
Bedrock
BR
3.5
100 (BG-2BR)
MW-14BR
BR
4.3
AB-6BR
BR
7.4
GWA-9BR
BR
3.9
AB-5BR
BR
0.76
AB -1 BR
BR
0.43
AB-9BR
BR
2.4
AL-2BR
BR
4.5
S-2
Seep
Ash Basin
566
44

a Refer to the CSA Report for Monitoring Well locations and screen intervals
b CB = Monitoring Well located on the Ash Basin Compliance Boundary; BR = Monitoring Well screened in Bedrock
`All values are total measured concentration in the water samples in units of micrograms per liter
d Refer to the CSA Report for information regarding placement of Background Wells.
PPBC = Proposed Provisional Background Concentration (CAP Report Table 2-2)
e CBM = Projected downgradient concentration at Compliance Boundary based on calibrated one-dimensional, analytical
chemical transport model discussed in this Report
45