HomeMy WebLinkAboutNC0038377_App B GW Modeling Rpts_20160229GROUNDWATER FLOW AND TRANSPORT MODELING
REPORT FOR MAYO STEAM ELECTRIC PLANT,
November 8, 2015
Prepared for
SynTerra
148 River Street
Greenville, SC 29601
Investigators
Lawrence C. Murdoch, Ph.D.
Scott E. Brame, M.S.
Ronald W. Falta, Ph.D.
Regina Graziano, M.S.
Contents
1.0
Introduction............................................................................................................. 5
1.1
General Setting and Background........................................................................... 5
1.2
Study Objectives................................................................................................... 6
2.0
Conceptual Model................................................................................................... 7
2.1
Aquifer System Framework.................................................................................. 7
2.2
Groundwater Flow System.................................................................................... 8
2.3
Hydrologic Boundaries...................................................................................... 9
2.4
Hydraulic Boundaries............................................................................................ 9
2.5
Sources and Sinks................................................................................................
10
2.6
Water Budget.......................................................................................................
10
2.7
Modeled Constituents of Interest........................................................................
10
2.8
Constituent Transport ..........................................................................................
11
3.0
Computer Model...................................................................................................
12
3.1
Model Selection...................................................................................................
12
3.2
Model Description...............................................................................................
12
4.0
Groundwater Flow and Transport Model Construction ..................................
13
4.1
Model Domain and Grid.....................................................................................
13
4.2
Hydraulic Parameters..........................................................................................
15
4.3
Flow Model Boundary Conditions......................................................................
15
4.4
Flow Model Sources and Sinks...........................................................................
16
4.5
Flow Model Calibration Targets.........................................................................
18
4.6
Transport Model Parameters...............................................................................
18
4.7
Transport Model Boundary Conditions...............................................................
19
4.8
Transport Model Sources and Sinks...................................................................
20
4.9
Transport Model Calibration Targets..................................................................
20
5.0
Model Calibration to Current Conditions..........................................................
21
5.1
Flow Model Residual Analysis...........................................................................
21
5.2
Flow Model Sensitivity Analysis........................................................................
24
5.3
Transport Model Calibration and Sensitivity......................................................
24
5.4
Result of the Transport Simulation..................................................................
25
6.0
Predictive Simulations of Corrective Action Scenarios .....................................
27
6.1
CAP 1 -No Action...............................................................................................
27
6.2
CAP2 - Capping Ash in Place.............................................................................
28
6.3
CAP3 - Complete Ash Removal.........................................................................
30
7.0
References..............................................................................................................
31
Tables................................................................................................................................
32
Figures...............................................................................................................................
41
LIST OF TABLES
Table 1. Observed and computed heads for the calibrated flow model.
Table 2. Calibrated hydraulic parameters.
Table 3. Recharge in different zones in the model.
Table 4. Flow parameter sensitivity analysis
Table 5. Comparison of observed and simulated boron concentrations (ug/L) in monitoring
wells.
Table 6. Comparison of observed and simulated arsenic concentrations (ug/L) in monitoring
wells.
LIST OF FIGURES
Figure 1. Map of the Mayo Steam Electric Plant showing ash basin, hydrologic features and
boundary of model region.
Figure 2. Air photo of the Mayo Steam Electric Plant showing ash basins and hydrologic
features and boundary of model region (yellow).
Figure 3. Fence diagram of the 3D hydrostratigraphic model (Solids) used to generate
hydrostratigraphy.
Figure 4a. Perspective of site looking SW along Crutchfield Branch showing computational grid
in cross section.
Figure 5. Distribution of recharge flux used in the Mayo model. Scale in ft/d.
Figure 6. Surface water features and wells in the vicinity of the Mayo Plant.
Figure 7. Hydraulic conductivity (ft/d) distributions in grid layers in the Mayo model.
Figure 8. Zones of different hydraulic conductivity enlarged and condensed from Figure 7.
Figure 9. Hydraulic head computed at steady state as a function of the observed head in
monitoring wells in the vicinity of the Mayo Plant.
Figure 10. Steady state hydraulic head distribution predicted in top of bedrock (layer 8 of the
model) in the Mayo model.
Figure 11. Steady state hydraulic head distribution in predicted in top of bedrock (layer 8 of the
model) in the Mayo model.
Figure 12. Steady state hydraulic head distribution in the vicinity of the ash basin (layer 4 of the
Mayo model).
Figure 13. Concentrations of boron in ash used to specify concentrations in the Mayo model.
Figure 14. Concentration of boron predicted to occur in Oct 2015 in different layers of the Mayo
model.
Figure 15. Concentrations of boron predicted to occur in Oct 2015 in different layers.
Figure 16. Map and cross -sections describing design of the Cap -in -Place design used to develop
the model of this scenario.
Figure 17. Distribution of recharge in simulations of corrective actions.
Figure 17a. Hydraulic head calculated in the vicinity of the ash basin for the different corrective
actions.
Figure 18-73. Distributions of concentrations of boron and arsenic at 2015, 2020, 2030, 2045,
layers in the model, and for different corrective actions.
1.0 INTRODUCTION
Duke Energy Progress, LLC. (Duke Energy) owns and operates the Mayo Steam Electric
Plant (Mayo Plant) located near Roxboro, in Person County, North Carolina. The Mayo Plant
began operations in the 1983. In 2013, the Mayo Plant converted from a wet to dry ash system.
Consequently, 90 percent of currently generated CCR is dry. Mayo Plant is a single unit facility.
Coal combustion residuals (CCRs) have been stored in one ash basin located northwest of the
Plant. The ash basin is approximately 144 acres in size, is constructed with an earthen dam, and
contains approximately 6,900,000 tons of CCR (Duke Energy, October 31, 2014).
1.1 General Setting and Background
The Mayo Plant ash basin is located in the northern portion of the Plant property,
northwest of the railroad line that effectively bisects the plant. The ash basin is impounded by an
earthen embankment system approximately 2,300 feet long, with a height of 110 feet. The entire
basin area is approximately 144 acres and contains approximately 6,900,000 tons of CCR
material (Duke Energy, October 31, 2014). The ash basin is the only ash storage area at the site
with the exception of the recently operational CCR monofill. The ash basin is approximately half
open water and half ash. Heavily wooded land surrounds the ash basin with the exception of the
wastewater treatment facility on the southeast side of the ash basin.
The Mayo Plant ash basin was constructed with two engineered toe drains located at the
base of the dam. In addition, ash basin water is diverted into a holding lagoon or forebay for
water quality treatment. Water flows from the ash basin into the forebay through a 48-inch pipe,
riser, and decant pipe through the forebay embankment. Discharge from the forebay is controlled
by a concrete overflow weir. Discharge from the forebay passes over the weir and flows through
NPDES Outfall 002, through a discharge canal and into Mayo Lake.
The Mayo Plant is in the eastern Piedmont Region, which is underlain by metamorphic
and igneous rocks (Trapp and Horn, 1997). The Piedmont region of North Carolina is
characterized by gently rolling hills and ridges, however the topography around the Mayo Plant
is steeper with incised streams and deeply cut ridges.
The ash basin was created by impounding Crutchfield Branch, a small stream that flows
to the north. Mayo Creek, the drainage to the east, was impounded to create Mayo Lake.
In general, the topography slopes away from the entrance road and the power plant area
towards Mayo Lake. The power plant area is situated at an approximate elevation of 520 feet
MSL and Mayo Lake is at an elevation of about 435 feet MSL, a vertical difference (relief) of 85
feet. The area is underlain by metamorphic rock, which is fractured and weathered at shallow
depths. Saprolite is up to 25 feet thick, but is absent locally. It is unsaturated in the uplands, but
is saturated at lower elevations streams. A transition zone of partially weathered rock underlies
the saprolite and is generally continuous throughout the Mayo Plant area. The transition zone at
the Mayo Plant is partially weathered rock that is gradational between saprolite and competent
bedrock. It is as thick as 22 ft in the vicinity of the Mayo Plant. The degree of weathering
decreases with depth (SynTerra, 2015). The water table in the upland areas is typically in the
fractured rock. Alluvium occurs along Crutchfield Branch and Mayo Creek.
1.2 Study Objectives
The purpose of this study is to predict the groundwater flow and constituent transport that
will occur as a result of different possible corrective actions at the site. The study consists of
three main activities: develop a calibrated steady-state flow model of current conditions, develop
a historical transient model of constituent transport that is calibrated to current conditions, and
create predictive simulations of the different corrective action options.
2.0 CONCEPTUAL MODEL
The site conceptual model for the Mayo Plant is primarily based on the Comprehensive
Site Assessment Report (CSA Report) for the Mayo Site (SynTerra, 2015). The CSA report
contains extensive detail and data related to most aspects of the site conceptual model.
2.1 Aquifer System Framework
The aquifer system at the site is unconfined and includes three main hydrostratigraphic
units, a saprolite/transition zone, upper bedrock, and lower bedrock (Legrand, 1988). The
saprolite/transition zone consists of partially to thoroughly weathered (SynTerra, 2015). It is
underlain by fractured metamorphic rock. The degree of fracturing is spatially variable and
generally decreases downward. Vertical and horizontal fractures zones can cause localized
zones of high permeability within the rock (Legrand, 1988; Miller, 1990). The permeability is
moderate in many of the bedrock wells, and it is inferred that the fracture density and hydraulic
conductivity decrease downward (Legrand, 1988).
The saprolite/transition zone is saturated in the vicinity of streams and lakes where
groundwater is discharging, but it is unsaturated in most upland areas. The water table occurs in
the fractured bedrock in most upland areas. Saprolite thickness ranges up to 66 feet, but it is less
than 25 ft thick in most locations. Alluvium was observed at two locations along the Crutchfield
Branch with a maximum thickness of 7.5 feet.
Ash is saturated in the ash basins. The water level in the ash basin is at approximately
480 ft and the ash surface rises up to approximately 490 ft at the southern end of the basin. The
elevation of Crutchfield Branch stream is approximately 390 ft at the ash basin dam.
Hydraulic conductivity was determined in the field using slug tests. The hydraulic
conductivity of the saprolite was 0.02 to 0.45 ft/day with a geometric mean of 0.2 ft/d. The
transition zone ranged from 0.02 to 3 ft/d with a geometric mean of 0.015 ft/d. These
measurements reflect the variability of the transition zone, where hydrologic properties
influenced by the localization of clays and fragments of unweathered rock.
The hydraulic conductivity of the bedrock spanned a broad range, from 0.003 ft/d to 5.6
ft/d, according to the results from 29 slug tests on 15 wells. The high values were probably
measured at wells that intersect fractures zones in the rock.
The hydraulic conductivity of the ash was measured by conducting a total of 5 slug tests
at 4 wells completed in the ash. Hydraulic conductivity spanned 2 orders of magnitude, from
0.05 to 4 ft/d, and the geometric mean was 0.4 ft/d.
2.2 Groundwater Flow System
The groundwater system is recharged by rain water and from water that infiltrates
through the ash basin. The average value of recharge was estimated from the map of recharge in
North Carolina by Haven (2003) and from analyzing stream hydrographs. A shapefile of the
recharge map by Haven (2003) was enlarged and features of the site were superimposed on it.
Colors on the map were compared to colors on the legend because quantitative data were
unavailable from the file. This indicated that recharge was in the range of 4-12 inch/yr (10-30
cm/yr) in the watershed draining into the ash basin (Figure 1).
The flow in Hyco Creek was obtained from measurements made at the gauging station
USGS 02077200 near Leasburg, NC. The gauging station is in Caswell County approximately
20 mi SW of the site and it measures flow from a watershed approximately 45.9 mil to the south.
Hyco Creek flows into Hyco Lake. The analysis was conducted on I I years of data starting in
January, 2002. The hydrograph was analyzed by separating stormflow and baseflow from the
hydrograph using the method described by the Institute of Hydrology (1980). This method of
hydrograph separation is widely used by the USGS and others. The separated hydrograph was
analyzed using methods described by Mau and Winter (1997), and Rutledge and Mesko (1996)
to estimate the recharge required to produce the observed baseflow. Recharge was estimated on
a monthly basis and then averaged over the time period of the dataset. This resulted in an
estimate of recharge that ranges from 3 to 7 inch/yr, depending on how the recharge is assumed
to occur between baseflow turning points.
Recharge estimated using the hydrograph from Hyco Creek was generally less than that
shown on from the map by Haven (2003), although the ranges from the two methods overlap.
Both of these methods of estimating recharge have advantages and disadvantages, so it was
assumed that the recharge the upland areas was 0.0018 ft/d (approximately 6.5 in/yr). Further, it
was assumed that recharge was negligible in the vicinity of the plant, the lined treatment lagoon,
surface water, and the low permeability dams. Specific values assumed for the recharge are
given later.
The Mayo Plant ash basin occupies the former stream valley of Crutchfield Branch, a
headwater stream flowing northeast. The basin was created by building a dam across the stream
valley. A small stream flows into the ash basin, and groundwater is inferred to discharge into
the basin from uplands to the west, south and east of the basin. The surface watershed that drains
into the basin is bounded on the east and south by a divide with drainage into Mayo Lake. It is
bounded on the west by a divide along US Highway 501 (SynTerra, 2015). The ash basin is
drained by Crutchfield Branch stream, which flows north off the Mayo Plant property into
Virginia (Figures 1 and 2).
Water levels in the wells completed in the ash basin were similar to the level of surface
water behind the dam. The water levels ranged from 480 ft to 485 ft in the ash basin wells, with
heads that increased toward the south where the headwaters of Crutchfield Branch enter the basin
(Figure 1).
The distribution of hydraulic head suggests that ground water is likely flowing into the
upstream end of the ash basin from the uplands to the west, south and east. The water level near
the dam is several 1 Os of ft higher than the water level in wells along Crutchfield Branch. It is
inferred that this large head difference drives groundwater flow north from the ash basin.
Approximately 21 water supply wells have been identified within one half mile of the
Mayo Plant (SynTerra, 2014). Most of the private wells are either northwest or south of the
Mayo Plant. The wells were sampled by NCDNER for chemical analysis, but measurement data
on the discharge rate from the wells was unavailable. The average daily water use is 80 to 100
gals per person, so a well providing water for a family of four people would be pumped at
approximately 350 gals per day.
2.3 Hydrologic Boundaries
The major discharging locations for the shallow water system serve as hydrologic
boundaries to the shallow groundwater system. These include lakes and streams.
2.4 Hydraulic Boundaries
The shallow groundwater system does not appear to contain impermeable barriers or
boundaries in the study area. The degree of fracturing, and thus the hydraulic conductivity, is
expected to decrease with depth in metamorphic rock. This will result in blocks of unfractured
rock where the hydraulic conductivity is quite low. However, isolated fractures may occur that
result in large local hydraulic conductivities, and the locations of these fractures is difficult to
predict. It was assumed that the rock was impermeable below mean sea level, and this elevation
was used as the bottom of the model.
2.5 Sources and Sinks
Recharge is the major source of water in the uplands and ash basins. Most of the
groundwater discharges to streams and lakes, as outlined above. Groundwater flows through the
ash basin, so it acts as both a source of, and sink for groundwater.
Domestic water supply wells act as sinks to groundwater to the south and north of the
plant. (SynTerra, 2015). Screen elevations and pumping rates from most of these wells are
unknown.
2.6 Water Budget
Over the long term, the rate of water inflow to the study area is equal to the rate of water
outflow from the study area. Water enters the groundwater system through recharge, and leaves
through discharge to surface water and wells.
2.7 Modeled Constituents of Interest
Antimony, arsenic, barium, cobalt, iron, manganese, pH, thallium, total dissolved solids
(TDS), and vanadium were detected in the ash basin (pore) water, near the bottom of the ash, at
concentrations greater than 2L or IMAC. Of these constituents, all but arsenic, barium, and TDS
were also detected in Site background wells upgradient of the ash basin at concentrations greater
than 2L or IMAC. Neither arsenic nor barium was detected greater than 2L or IMAC outside of
the ash basin.
Cobalt, iron, manganese, and vanadium are commonly detected in shallow groundwater
in the Piedmont of North Carolina. Site background concentration ranges for some of these
constituents are available from routine monitoring of the upgradient compliance boundary
monitoring wells and newly installed background wells.
Boron, cobalt, and TDS are the only COIs in groundwater detected downgradient of the
ash basin (SynTerra, 2015). Cobalt is common in background wells, so the occurrence
downgradient is difficult to distinguish from background. TDS is also difficult to distinguish
from background. Statistical analyses of the occurrence of cobalt and TDS would help identify
whether these compounds could be distinguished from background.
The COIs selected for modeling at the Mayo site are boron and arsenic. Boron is the
primary COI, but arsenic was also included because it is present in the ash at concentrations
several times greater than the 2L standard and it has a relatively low 2L standard. Other
constituents were not used in the modeling exercise for one or more of the following reasons: 1)
concentrations in the ash pore water do not greatly exceed likely background levels; and 2) there
is no discernable plume of the constituent extending downgradient from the ash basin.
2.8 Constituent Transport
The COIs that are present in the coal ash dissolve into the ash pore water. As water
infiltrates through the basins, water containing COIs can enter the groundwater system through
the bottom of the ash basins. Once in the groundwater system, the COIs are transported by
advection and dispersion, subject to retardation due to adsorption to solids. If the COIs reach a
hydrologic boundary or water sink, they are removed from the groundwater system, and they
enter the surface water system, where in general, they are greatly diluted.
At this site, boron is the primary constituent that is migrating from the ash basin.
3.0 COMPUTER MODEL
3.1 Model Selection
The numerical groundwater flow model was developed using MODFLOW (McDonald
and Harbaugh, 1988), a three-dimensional (3D) finite difference groundwater model created by
the United States Geological Survey (USGS). The chemical transport model is the Modular 3-D
Transport Multi -Species (MT3DMS) model (Zheng and Wang, 1999). MODFLOW and
MT3DMS are widely used in industry and government, and are considered to be industry
standards. The models were assembled using the Aquaveo GMS 10.0 graphical user interface
(http://www.aquaveo.com/).
3.2 Model Description
MODFLOW uses Darcy's law and the conservation of mass to derive water balance
equations for each finite difference cell. MODFLOW considers 3D transient groundwater flow
in confined and unconfined heterogeneous systems, and it can include dynamic interaction with
pumping wells, recharge, evapotranspiration, rivers, streams, springs, lakes, and swamps.
Several versions of MODFLOW have been developed over the years. This study uses the
MODFLOW-NWT version (Niswonger, et al., 2011). The NWT version of MODFLOW
provides improved numerical stability and accuracy for modeling problems with variable water
tables. That improved capability is helpful in the present work where the position of the water
table in the ash basin can fluctuate depending on the conditions under which the basin is operated
and on the corrective action activities.
MT3DMS uses the groundwater flow field from MODFLOW to simulate 3D advection
and dispersion of the dissolved COIs including the effects of retardation due to COI adsorption
to the soil matrix.
4.0 GROUNDWATER FLOW AND TRANSPORT MODEL
CONSTRUCTION
The flow and transport model for this site was built through a series of steps. The first
step was to build a 3D model of the site hydrostratigraphy based on field data. The next step was
determination of the model domain and construction of the numerical grid. The numerical grid
was then populated with flow parameters, which were adjusted during the steady-state flow
model calibration process. Once the flow model was calibrated, the flow parameters were used
to develop a transient model of the historical flow patterns at the site. The historical flow model
was then used to provide the flow field for the transient constituent transport simulations.
Calibration of the transport model required some adjustments in order to reproduce the
observed boron plume. This resulted in a second iteration of flow model calibration, so that the
calibrated flow model matches the observed heads, and the transient flow and transport model
reproduce the observed boron plume.
4.1 Model Domain and Grid
The first steps in the model grid generation process were the determination of the model
domain, and the construction of a 3D hydrostratigraphic model. The model has dimensions of
approximately 3 miles by 3 miles. The model domain was rotated 31 ° clockwise so boundaries
of the model were parallel with the ash basin dam. The shortest distance between the ash basin
and a model boundary was approximately 1 mile.
The ground surface of the model was interpolated from USGS NED n37w079 1/3 arc -sec
2013 1 degree IMG dataset obtained from http://viewer.nationalmgp.gov/viewer/. The elevations for
the top of the ash basin were modified using more recent surveying data.
The hydrostratigraphic model consists of six units: Ash, Saprolite, Transition Zone,
Upper Bedrock, Middle Bedrock, and Lower Bedrock.
The hydrostratigraphic model was developed using "Solids" in GMS (Figure 3). Five
solids were created and then subdivided after the computational mesh was developed. The solids
include Ash, Saprolite, Transition Zone, Fractured Rock, and Rock. The lower contact between
the ash basin and the underlying saprolite was assumed to be the ground surface prior to
construction of the ash basins. An electronic file describing this surface was created by
digitizing a preconstruction topographic map. The digitized points were interpolated to create a
continuous surface representing the preconstruction ground surface, and this was used as the
contact between the ash and the underlying saprolite. This created a surface that was consistent
with the borehole observations except at A13MW-4 where the contact was 20 ft lower than
expected. This may have resulted from construction -related excavation that was not recorded on
the pre -construction topo map. The bottom contact between the ash and underlying weathered
rock was lowered in the vicinity of ABMW-4 as a result. The lateral extent of the ash was
determined from aerial photographs and from maps in the CSA report (SynTerra, 2015).
The contacts between the saprolite, transition and underlying bedrock were determined
by interpolating data measured in borings described in the CSA report and historical data. This
produced two isopach maps, one showing the thickness of the saprolite and the other showing the
thickness of the partially weathered rock in the transition zone. The interpolated isopach surface
for the saprolite was subtracted from the ground surface to create a surface marking the contact
between the saprolite and transition zone. The isopach map for the transition zone was
subtracted from the surface describing the saprolite-transition zone contact. This created a
surface that was used as the contact between the transition zone and the underlying fractured
rock.
The methodology outlined above for creating a geologic model was done so the
interpolated contacts would follow the ground surface between boreholes, which is consistent
with the expectations based on the hydrogeology of the Piedmont region (e.g. LeGrand, 1988;
Miller, 1990). A contact between the fractured and relatively unfractured bedrock was assumed
to occur 100 ft below the bottom of the transition zone.
The numerical model grid consists of 15 layers representing the hydrostratigraphic units.
The model grid was set up to conform to the contacts from the solids. The model grid layers
correspond to the solids as follows:
H drostrati ra hic layer
Grid layer
Ash
1-4
Saprolite
5
Transition zone
6-7
Upper fractured rock
8-10
Middle fractured rock
11-12
Rock
13-15
Grid layers 1-4 were set as inactive outside of the region of the ash basin as determined
from aerial photos and the CSA report. Grid layers 1-15 were set as inactive in the eastern,
southern, and western corners of the model domain (Figure 1).
The numerical grid consists of rectangular blocks arranged in columns, rows and layers.
There are 171 columns, 232 rows, and 15 layers (Figure 4). The maximum width of the columns
and rows is 100 ft. The size of the grid blocks is approximately 50 ft in the vicinity of the ash
basins. The horizontal dimension of some of the grid blocks is as small as 25 ft in the vicinity of
the dams.
4.2 Hydraulic Parameters
The horizontal hydraulic conductivity and the horizontal to vertical hydraulic
conductivity anisotropy ratio (anisotropy) are the main hydraulic parameters in the model. The
distribution of these parameters is based primarily on the model hydrostratigraphy, with some
local variations. Many of the hydraulic parameter distributions in the model were uniform
throughout a model layer. Initial estimates of parameters were based on literature values, results
of slug and core tests, and simulations performed using a preliminary flow model. The hydraulic
parameter values were adjusted during the flow model calibration process described in Section
5.0 to provide a best fit to observed water levels in observation wells.
4.3 Flow Model Boundary Conditions
The outer lateral boundary conditions for the saprolite is almost entirely constant head,
with small areas of no -flow locally. Boundaries on the east side the model include parts of Mayo
Lake, which were held at specified head.
The boundaries on the south and east and north sides of the model are independent of
definitive hydrologic features. A specified head boundary condition with the head set in the
middle of the transition zone was used along these boundaries. This boundary condition forces
the water table to be in the transition zone along these boundaries, which is a reasonable
approximation of the expected conditions. The specified head boundary condition extends
along the upland areas, but it is terminated within a few hundred ft the locations of streams or
lakes. This is because streams or lakes that intersect the external boundary are defined by their
own boundaries conditions (as either constant head or drain -type boundaries). This creates short
intervals of no -flow conditions between streams or lakes and the uplands. The constant head
boundary condition along the outside of the model was assigned to Layer 7 in the grid, which is
the lowest layer in the transition zone. No -flow conditions were assumed along the boundary
beneath Layer 7.
4.4 Flow Model Sources and Sinks
The flow model sources and sinks on the interior of the model consist of recharge, lakes,
streams, and groundwater pumping.
Recharge is a key hydrologic parameter in the model (Figure 5). As described in Section
2.2, the recharge rate for upland areas in the vicinity of the Mayo Plant was assumed to be
0.0018 ft/d (6.5 inches/year). The recharge on exposed ash was assumed to be 0.0018 ft/d, the
same as in upland areas. This is because the shallow water table would have increased
evaporation, while the lack of vegetation would have decreased ET on the ash basin compared to
the upland area. As a result, without field data it was difficult to assess how the recharge on the
ash basin would have differed from the recharge on the uplands. The recharge was assumed to
be zero at the ash basin lake and on Mayo Lake. The recharge rate on the Mayo Plant was set to
0.0001 ft/d, due to the large areas of roof and pavement. The recharge beneath a lined treatment
pond north of the plant site was assumed to be 0.0001 ft/d (Figure 5). Recharge on the dams
themselves was set to 0.0001 ft/d, which is consistent with the low hydraulic conductivity
assumed for the dams.
Recharge was not adjusted much during the model calibration process, but it is included
in the sensitivity analysis. The reason for not including recharge as a calibration parameter is
that for steady-state flow, the hydraulic heads are determined primarily by the ratio of recharge
to hydraulic conductivity, so the two parameters are not independent. In situations where the
groundwater discharges to a flow measuring point (for example a gauged stream in a watershed),
the flow measurement can be used to calibrate the recharge value allowing both the recharge rate
and the hydraulic conductivity to be simultaneously calibrated. However, no streams were
gauged at the Mayo site, so this was impossible for this model and the recharge was fixed.
Lakes were represented as specified head boundaries with the head set to their stage
(Figure 6). This includes Mayo Lake and the ash basin lake. The stage of the ash basin was set
to 480 ft based on Lidar data and a surveying point. The stage of Mayo Lake was 432 ft.
Streams were represented as type 3 boundary conditions, called "drains" in MODFLOW
(Figure 6). The elevation of the streams was set to the ground surface elevation determined from
the Lidar data. The drain conductance was set to 100 ft2/day, a relatively large value that will
cause negligible head loss, and was not adjusted during calibration (Figure 6).
The ash basins were represented by simulating the observed surface water as specified
head and applying recharge on the exposed ash. This approach treats the ash basins in the same
way as other hydrogeologic components in the model, and it was selected as the best approach to
characterize current conditions. The hydrologic conditions when the ash was below the level of
the water early in the life of the basin differed from the current conditions. However, the
hydraulic heads in the ash today are only slightly above the level of the water. It is likely that the
ground water flow system created by the ash basin filled only with water would have been
essentially the same as the basin containing ash as it does today. As a result, the hydrologic
conditions in the ash basin will be assumed to be constant through the life of the basin.
The outflow channel was represented as specified head, with the stage set to elevations
measured by surveying. This was done because this engineered channel could exchange water
with the groundwater system, and the head was measured in the field.
Relatively little information was available about the wells in the model area. The
location of water supply wells in the model area are shown as yellow squares in Figure 6
(SynTerra, 2015). Most of the wells are probably open holes in the upper few 100 ft of bedrock.
However, it is common for drillers in the Piedmont to extend wells to depths of several 100 ft in
an effort to intersect permeable fractures and create more productive wells. As a result, the depth
of the wells probably ranges from 150 ft to 600 ft. The wells are assumed to be screened in grid
layers 10-12 in the model.
The pumping rates from the wells were unknown, but it was assumed that the wells were
pumped at about 350 gallons/day, which is an average water use for a family of four.
4.5 Flow Model Calibration Targets
The steady state flow model calibration targets were 47 water level measurements made
in observations wells in June, 2015. The flow model calibration target wells are listed in Table
1. In general, wells with a D designation at the end of the name are screened in the transition
zone, those with a BR designation are screened in the upper bedrock, and those with an S
designation are in saprolite. Wells with ABMW are screened in ash.
4.6 Transport Model Parameters
The transport model uses a MODFLOW simulation to provide a groundwater velocity
field. The transport simulation was started April, 1983, and it continued through Sept, 2015.
The Mayo Plant began operations in 1983, and it was assumed that the basin was filled with
water at this time. The flow model assumes that the ash basin filled with water quickly and the
heads were maintained at the same level as they are today. As a result, a steady state flow field
calculated by calibration to the current conditions was used to simulate water flow during
transport.
The key transport model parameters (besides the flow field) are the constituent source
concentration in the ash basin, and the constituent soil -water distribution coefficients (Kd).
Secondary parameters are the longitudinal, transverse, and vertical dispersivity, and the effective
porosity. The constituent source concentrations were estimated from recently measured ash pore
water concentrations in monitoring wells (SynTerra, 2015).
Linear adsorption Kd values for Mayo Plant COIs were measured in the laboratory using
core materials from the coal ash and native aquifer materials (Langley, et al., 2015). In general,
the measured Kd values for the constituents were highly variable, and the variability within a
given material type was larger than the variability between different materials.
In light of the variability of the measured Kd values, it was decided to use Kd as a
calibration parameter. The initial value used in calibration was 5 mL/g, which is the low end of
the range measured by Langley et al. (2015). It was found during the transport model
calibration that Kd = 0.12 mL/g for boron.
The Kd value used for arsenic in the simulations was 50 mL/gm. This is on the low end
of the range of Kd values determined in the laboratory by Langley et al. (2015), which ranged
from 45 to 1,800 mL/g in batch, and from 6 to 3,349 mL/g in column experiments. Arsenic was only
detected in significant concentrations below the ash basins where the uncertainty in the distance
between the bottom of the ash and the screen on the monitoring well was relatively large. As a result
of this uncertainty and the limited number of data points, it was decided to use a laboratory value for
Kd. A value that was on the low end of the range that included both column and batch data was
selected so that the transport calculations for arsenic would predict the most rapid transport feasible
within the range of the data. This was done to bound the maximum transport distances for arsenic.
The longitudinal dispersivity was assigned a value of 20 ft, the transverse dispersivity
was set to 2 ft, and the vertical dispersivity was set to 0.2 ft. The soil dry bulk density was set to
1.6 g/mL. The effective porosity was assumed to decrease with depth from 0.3 in the ash, to 0.2
in the saprolite and transition zone to 0.001 in the deep rock. The effective porosity was
assumed to decrease with depth based on the hydrogeologic conceptual model. It was assumed
the effective porosity was uniform within a grid layer and was distributed according to the
following table:
Layer
Effective porosity
1-4
0.3
5-7
0.2
8
0.05
9-10
0.01
11-15
0.001
4.7 Transport Model Boundary Conditions
The transport model boundary conditions are no flow on the exterior edges of the model.
The infiltrating rainwater is assumed to be clean, and enters with zero concentration from the top
of the model. Contaminants are assumed to leave the model when they arrive at a drain, or are
removed by flow that enters a constant head boundary.
The initial condition for the current conditions transport model is one of zero
concentration of COIs in groundwater. No background concentrations are considered. The
concentration in the ash basin is assumed to rapidly increase to the observed concentrations at
the start of the simulation.
4.8 Transport Model Sources and Sinks
The ash basins are the source of COIs in the model. These sources are simulated by
holding the COI concentration constant in cells located inside the ash basins (Figure 13). This
allows infiltrating water to carry dissolved constituents from the ash into the groundwater
system.
Chemical analyses from four wells were used to characterize the distribution of COI
concentration within the ash basins. The concentration observed in the wells was assumed to
represent the concentration in the vicinity of the well throughout the simulation. This resulted in
a patch -like distribution of concentration within the ash basins (Figure 12).
Layer 5 in the vicinity of ABMW-4 was assumed to be ash, as described above. Layer 5
is saprolite elsewhere in the model, but the contact between saprolite and ash was approximately
20 ft lower than estimated from the pre -ash -basin topography. It seems likely that this
discrepancy occurred because of excavation during construction. The concentration in the
vicinity of ABMW-4 in Layer 5 was held constant to resemble the behavior of ash. This created
a zone of relatively high concentration in Layer 5.
The outflow channel flowing south from the ash basin is represented as constant head and
it is possible that water could flow from this engineered feature to the ground water. It was
assumed that the concentration in the surface water at this location was zero. This assumption
had no bearing on the results, however, because the outflow channel gained ground water.
The transport model sinks are potentially lakes and streams. As groundwater enters these
features, it is removed along with any dissolved constituent mass. Similarly, if water containing
a constituent were to encounter an extraction well, the constituent would be removed with the
water.
4.9 Transport Model Calibration Targets
The transport model calibration targets are COI concentrations measured in monitoring
wells in June, 2015 (SynTerra, 2015).
5.0 MODEL CALIBRATION TO CURRENT CONDITIONS
5.1 Flow Model Residual Analysis
The flow model was calibrated in stages starting with a model that assumed
homogeneous conditions in most formations. In general, calibration was done by seeking the
simplest configuration of parameters that matched the observed hydrogeologic conditions and
the assumed or observed geologic conditions. Many of model layer properties were
homogeneous. Several heterogeneities were assumed to improve the fit between the simulated
and observed heads and concentrations.
The calibration was initiated using the geologic model to define the geometry of
hydrogeologic units and assigning hydraulic conductivities typical of the region. PEST was then
use to minimize the residual between predicted and observed heads. This resulted in reasonably
close matches, but there were several wells where the simulation significantly over- or under -
predicted the heads.
The next step was to infer heterogeneities that could reduce the residuals. The model
over -predicted the heads at several wells in upland areas. It was inferred that these wells
intersected, or were near zones of relatively high permeability that were broad enough to extend
beneath a nearby stream or lake. This configuration reduced the head in the well, and the
hydraulic conductivity of the zone was increased until either the head was reduced sufficiently,
or an upper limit of hydraulic conductivity was reached. This was done by including zones of
high hydraulic conductivity in the upper bedrock in the vicinity of wells MW-08BR and CW-4
(Figure 7).
The heads at MW-06BR and CW-4 were lower than predicted by the model, so a similar
approach was used. A high permeability zone was created that extended under the ash basin
lake. Slug tests at MW-06BR indicated the hydraulic conductivity of enveloping rock was quite
low, so it was assumed the high K zone extended below the depth of the well. This is the
motivation for including the high K zone in Layer 12 (Figures 7 and 8). The maximum hydraulic
conductivity determined from slug tests in fractured rock was 6 ft/d and this was the upper limit
of K assumed to represent these zones in the model.
The occurrence of the flat -lying zones of high hydraulic conductivity as shown in Figure
7 and 8 was inferred at the Mayo site based on head observations, but this inference is based on a
geologic style that is known from other locations. Flat -lying zones of interconnected fractures
several hundred ft or more across were described in crystalline rock at the USGS Mirror Lake
research site (e.g. Tiedeman et al. 2001), and similar fracture zones have been recognized at
other fractured rock sites that have been studied in detail. It was assumed that the fractures zones
in the model were shaped like flat -lying layers several hundred ft in maximum dimension,
similar to those described by Tiedeman et al. (2001).
The observed hydraulic heads in MW-4 were significantly above the heads predicted by
the model calibrated using PEST and adjusted using the fractures zones described above. The
head in this well was more than 15 ft above the water level of the nearby ash basin. To account
for this difference, a low permeability zone was assumed between MW-4 and the ash basin. This
zone was assumed to extend from the saprolite to the sparsely fractured rock at depth, and it was
assigned a hydraulic conductivity of 0.01 ft/d.
The zone of low permeability used in the model corresponds to a ridge where large
boulders of recrystallized quartz were observed in the field. Similar quartz -forming ridges are
known elsewhere in the Piedmont where they can form zones of low hydraulic conductivity (e.g.
Snipes et al. 1984), presumably because the quartz fills pore space and resists weathering. This
zone is intersected by MW-4 and MW-6BR. The water level in both wells changed slowly after
it was perturbed, an indication that the wells were completed in tight rocks. The water level
change was too slow to obtain meaningful estimates of hydraulic conductivity using slug tests.
The flow calibration was done iteratively with transient transport simulations. This was
necessary in order to match both the heads and the boron concentration distributions. The
fracture zone in Layer 12 plays an important role in the transport. Boron would not reach CW-2
in the absence of a fracture zone, according to the model. As a result, the position of the fracture
zone in Layer 12 was adjusted to extend the length of the boron plume to better match the
observed data.
The final calibrated flow model has the following volume balance:
Volume balance in steady state model in ft
Feature
Input
Output
Constant Head
59979
113144
Recharge
308190
0
Wells
3150
Drains (streams)
251875
Total
368170
368170
The difference between the input and output is 0.094 ft3, which is a volume balance error
of less than 10-6. The major input to the model is from recharge with a lesser amount from
constant head boundaries. The constant head boundaries creating input to the model are where
ground water is flowing into the model from the boundaries around the periphery. The ash basin
lake is also a constant head input. The output is split between groundwater discharging to
constant head boundaries and drains. The major constant head sink is Mayo Lake and it accounts
for about half the flow going to streams in the model. Less than 1 percent of the water input is
removed through wells, according to the model.
The final calibrated flow model has a mean head residual of -0.56 ft., and a root mean
squared head residual of 5.42 ft. The total span of head measurements ranged over 191 ft, from
365 ft to 556 ft. Using this range to normalize the residual gives a normalized root mean square
error of 2.84%. A comparison of the observed and simulated water levels is listed in Table 1,
and the observed and simulated levels are cross -plotted in Figure 9. Table 2 lists the best -fit
hydraulic parameters from the calibration effort.
The calibrated conductivity of the ash is 1 ft/d (Table 2). The calibrated conductivity of
the saprolite and transition zones are 3.0 and 1.3 ft/d, respectively. The hydraulic conductivity of
the upper fracture rock is 0.05ft/d and it decreases to 0.02 ft/d and 0.007 ft/d with depth.
The calibrated values of hydraulic conductivity are consistent with values from the slug
tests conducted in the ash, transition zone, and upper fractured rock. No tests were conducted
below the upper fractured rock hydrostratigraphic unit, and testing in saprolite was limited to 3
wells.
The calibrated model predicts the highest hydraulic head occurs south of the ash basin
and the lowest heads occur on Crutchfield Branch and Mayo Creek along the northern edge of
the model. The heads in the ash basin are in the midrange between these extremes and this is
consistent with the basin resembling a flow -through lake, with higher heads on the west, south
and east sides, and lower heads to the north (Figures 10, 11, and 12).
5.2 Flow Model Sensitivity Analysis
A parameter sensitivity analysis was performed on the calibrated model by systematically
increasing and decreasing the main parameters by factors of either 2 or 0.5 from their calibrated
value. Table 4 shows the results of the analysis, expressed in terms of the normalized root mean
square error (NRMSE) for each simulation. The baseline NRMSE is 2.84%.
The flow model showed the highest degree of sensitivity to the upland recharge and to
the hydraulic conductivities of the transition zone and saprolite stratigraphic unit. The saprolite
and transition zone were saturated beneath the ash basins and in the vicinity of Crutchfield
Branch and this accounts for the calibration sensitivity. The model was only weakly sensitive to
the hydraulic conductivities of the ash and the deep rock.
There is a weak sensitivity to the hydraulic conductivity of the dams. The heads of wells
downgradient of the dams are quite sensitive to the hydraulic conductivity of the dams.
However, the hydraulic conductivity of the dams is quite low and changing this value by a factor
of 2 has little effect on the overall residual.
The sensitivity of the model to the pumping rate of the domestic wells was small, but
detectable. This is because the heads in several monitoring wells (e.g. MW-12S, MW-14BR,
MW-5BR) can be influenced by pumping from the domestic wells.
5.3 Transport Model Calibration and Sensitivity
The transport simulations used a steady state flow model based on current conditions.
This assumes that the water level in the ash basin was maintained at approximately 480 ft since
the plant opened in 1983.
Initial simulations predicted that boron concentrations were absent from MW-3 and CW-
2 and CW-2D. Field observations indicated that boron concentrations were elevated in these
wells, so the model was adjusted to match these observations. The adjustments included
reducing the Kd to Kd=0.12 mL/gm. The position of a fracture zone inferred to occur in layer 12
was also adjusted by moving it closer to Crutchfield Branch. This affected the concentrations in
those wells, although it had little effect on the heads in MW-6BR.
5.4 Result of the Transport Simulation
The simulated concentrations reasonably match most of the observed concentrations
(Table 5). The concentrations that are an exact match in Table 5 are from wells in the ash basin
where the concentrations were set as boundary conditions in the model. Some of the wells where
boron was detected are in areas where the predicted concentrations gradients are steep, so small
changes in location result in significant changes in concentration. This is one factor that explains
the differences between predicted and observed concentrations. The simulations over -estimate
the observed arsenic concentrations (Table 6) because of the conservative value of Kd that was
used.
The primary area of boron groundwater contamination occurs beneath the ash basin dam,
and contamination beneath the ash basin on the upgradient end of the basin is also predicted
(Figures 14 and 15). The largest plume of boron is predicted to occur beneath the eastern side of
the dam. The plume extends under the dam and reaches wells CW-2 and CW-2D, approximately
750 ft along Crutchfield Branch from the edge of the water in the ash basin lake (Figures 14 and
15).
The model predicts that the overall footprint of the region above the 2L standard shrinks
with depth, but the length of the boron plume along Crutchfield Branch is approximately the
same from layers 5 through 12. Observations of boron concentration at CW-2 (804 µg/L) and
CW-21) (804 µg/L) suggest that boron concentrations decrease with depth, whereas the model
predicts that concentrations should increase with depth. This prediction is a consequence of the
geometry of hydraulic conductivity structure inferred during calibration. The high conductivity
fracture zone inferred in layer 12, and the low conductivity band have an important influence on
transport out of the ash basin. Evidence for this influence is the co -location of the zone of
highest concentration, the fracture zone, and the low conductivity band, as shown in Figure 15.
The leading edge of the plume as defined by the 2L contour lies between the CW-2 wells
and Mayo Lake Rd, according to the simulations. This is important because the property line
and compliance boundary are along Mayo Lake Rd.
Arsenic concentrations occurred above the 2L level in the saprolite beneath the ash basin.
Arsenic concentrations decreased sharply with depth, however, because in the lower half of the
transition zone (Layer 7) none of arsenic concentrations exceeded 2L standard.
6.0 PREDICTIVE SIMULATIONS OF CORRECTIVE ACTION
Once the flow model was calibrated with regard to water levels, and the simulated
concentrations in wells around the ash basin closely matched observed concentrations that
exceeded the 2L standards, the model was used to predict contaminant distributions for the next
5, 15, and 30 years. The dates for those simulations are referred to in this report as 2020, 2030,
and 2045 respectively. Given the limited information provided to the modeling team about the
timing of any actual on -site events, the results provided can be thought of as occurring 5, 15, and
30 years after any activities take place and the actual dates can be modified as needed.
Three future conditions were evaluated using the flow and transport model:
41' Corrective Action Plan #1 (CAP 1): No Action
41' Corrective Action Plan #2 (CAP 2): Capping Ash in Place
'61' Corrective Action Plan #3 (CAP 3): Complete Ash Removal
The distribution of recharge, locations of drains, and distribution of material were
modified to represent the different corrective actions. For example, the recharge was modified as
shown in Figure 17. The hydraulic head distribution was recalculated and then the transport was
simulated for each case. The corrective actions changed the hydraulic head in the vicinity of the
ash basin (Figure 18) as the engineered designs interacted with the hydrogeologic conditions.
This interaction altered the groundwater flow and the transport of dissolved compounds, as
shown in the results of the simulations.
6.1 CAP1 - No Action
This method relies on natural attenuation processes to reduce the contaminant
concentrations over time. In this scenario, the ash basin is left in place without modification and
the assumption is made that current recharge and contaminant loading rates from the ash to the
underlying formations are held constant.
The model of this scenario includes a distribution of recharge and hydraulic properties as
described above. The flow system was assumed to be at steady state with respect to the
conditions in 2015. Concentrations in the ash were held constant at the measured concentrations.
Boron concentrations delineated by the 2L contour are larger in the simulations from
2045 than they are in 2015, but the differences are relatively small. The most apparent
difference is beneath the southwest end of the basin where the 3500 mg/l contour has noticeably
expanded from 2015 and 2045 in Layer 7 (e.g. Figure 18b to Figure 33).
Another important feature is the leading edge of the boron plume along Crutchfield
Branch. The leading edge is on the south side of Mayo Lake Rd in 2015 and 30 years later the
leading edge has moved from the south to the north side of the road, according to the
simulations. For example, compare the leading edge of the plume in Layer 10 shown in Figure
18c to the same location in Figure 34. The location of the leading edge of the plume in these
Figures is important because the property line occurs along Mayo Lake Rd, and the compliance
boundary is on or near Mayo Lake Rd.
Simulated arsenic concentrations in saprolite beneath the ash increased in 2045 compared
to the 2015 values. The distribution of arsenic in the saprolite in 2015 is patchy, and by 2045 the
space between many of the patches has filled in (compare Figure 17 to Figure 35). The bottom
of the transition zone and underlying rock are less than the 2L standard in 2045, according to the
simulations. It is worth pointing out that the Kd value for arsenic was on the low side of the
range of lab values and the simulations over -estimated several of the observed arsenic
concentrations during calibration.
6.2 CAP2 - Capping Ash in Place
The ash basin capping in place method involves placing a low permeable liner over the
ash basin to contain the ash and to prevent rainwater infiltration. This scenario assumes that there
is no recharge within the ash basin. Initial conditions were applied using the existing
concentrations in the formations underneath and within the ash basin as of October 2015. The
grid was modified by removing grid blocks to simulate the grading operations described in a
design drawing provided by AECOM in October 2015 (Figure 16a).
The computational grid was modified to resemble the distribution of ash shown in the
design. This included removing grid blocks in the vicinity of the dam to account for grading.
Grid blocks above those in the baseline grid were not included in this simulation because it was
assumed that material would be unsaturated and would therefore not be active in this simulation.
Boundary conditions in the vicinity of the ash basin were modified to account for
drainage of the basin, removal of the dam, and grading. The constant head boundary conditions
representing the basin were removed. Drain boundary conditions were included to represent
drains extending across the graded basin. The drain boundary conditions were set up to provide
a flow path for water to exit the system, and to limit the development of impounded conditions
upstream from the ash.
It was assumed that the dam was removed and the vicinity of the dam was restored to the
original hydraulic conductivity, which depends on the hydrostratigraphic unit. Calibration to the
ground water heads indicated that the low permeability part of the dam extended at least through
layer 7 in the model (through the transition zone). The low permeability zone inferred from the
calibration was removed in this simulation.
The initial concentrations in this simulation were assumed to be the concentrations
simulated to occur in 2015 in the baseline model. Concentration in the ash were allowed to vary.
This relaxes the assumption made in the No -Action scenario that the concentrations in the ash
basin were held constant.
The boron concentrations decreased and the contours changed location from 2015 to
2045. The change is particularly noticeable in the saprolite (Layer 5). Compare Figure 18a to
Figure 50, for example. The leading edge of the boron plume recedes in this scenario. By 2045,
the leading edge has receded by 100 to 200 ft in the vicinity of Mayo Lake Rd (Figures 50, 51
and 52).
Arsenic concentrations increase in the saprolite, but the increase is smaller than for the
case of No Action. The arsenic concentration in Layer 7 remains below 2L in this scenario.
It is important to keep in mind that the simulation of the No Action scenario outlined
above assumed that the concentration in the ash basin remained constant at the current values,
whereas the Cap -In -Place scenario described in this section assumed the concentration in the ash
started at the observed concentration but could vary with time. This change allows the
concentration in the ash and at shallow depths to decrease more rapidly than in the No Action
scenario.
6.3 CAP3 -3 Complete Ash Removal
Ash basin removal involves removing all ash off -site. The location of the off -site landfill
was not considered in this modeling effort.
This scenario assumes that the initial concentrations in the formation were the
concentrations in the simulations as of October 2015. The upper four layers in the original
model were rendered inactive to simulate removal of the ash. "Drains" were added to the model
along the axes of drainages exposed when the ash was removed. The other major change is that
the recharge within the ash basin boundary was set to the ambient recharge value of 0. 00 18 ft/d
(6.5 in/yr).
Removing the ash significantly reduces the size and concentration of the boron plume in
saprolite and transition zone. For example, the size of the plume in the saprolite in Figure 68
appears to be less than 10 percent of the size of the plume in the No Action case in Figure 32.
However, the difference in the size of the plume diminishes with depth and the plume in the
bedrock 30 years after the ash was removed is approximately the same size as a comparable time
in the No Action case. The concentrations in the plume for the case when the ash has been
removed are less than those for the No Action case in the bedrock (Figure 18c and Figure 70).
Removing the ash has no effect on the arsenic concentration that can be detected from the
contour plots. For example, compare Figure 71 to Figure 17.
7.0 REFERENCES
Haven, W. T. 2003. Introduction to the North Carolina Groundwater Recharge Map.
Groundwater Circular Number 19. North Carolina Department of Environment and
Natural Resources. Division of Water Quality, 8 p.
Langley, W.G., J. Daniels, and S. Oza, 2015, Sorption Evaluation of the. Roxboro Steam Electric
Plant. Charlotte Department of Civil and Environmental Engineering, report prepared for
SynTerra,
Legrand, H. 1988. Piedmont and Blue Ridge. Back, W., J. Rosenshein, and P. Seaber, eds.
1988. Hydrogeology: The Geology of North America 0-2: The Decade of North
American Geology. Boulder, Colorado: Geological Society of America.. Geological
Society of America. P. 201-208.
McDonald, M.G. and A.W. Harbaugh, 1988, A Modular Three -Dimensional Finite -Difference
Ground -Water Flow Model, U.S. Geological Survey Techniques of Water Resources
Investigations, book 6, 586 p.
Miller, J.A. 1990. Ground Water Atlas of the U.S. South Carolina and vicinity (HA 730-G).
USGS. http://pubs.usgs.gov/ha/ha730/ch g/index.html.
Niswonger, R.G.,S. Panday, and I. Motomu, 2011, MODFLOW-NWT, A Newton formulation
for MODFLOW-2005, U.S. Geological Survey Techniques and Methods 6-A37, 44-.
Snipes, D.S., Burnett, L.L., Wylie, J.A., Sacks, L.A., Heaton, S.B., Dalton, G.A., and Israel,
B.A., 1984, Indicators of ground -water quality and yield for a public water supply in rock
fracture zones of the piedmont: Clemson, S.C., Water Resources Research Institute
Report 115, 80 p
SynTerra, 2014, L.V. Roxboro Steam Electric Plant, Semora, NC, Water Supply Well Survey
Report of Findings, September 30, 2014.
SynTerra, 2015, Comprehensive Site Assessment Report, Roxboro Steam Electric Plant, Semora,
NC. September 2, 2015.
Tiedeman, C.R. and P.A. Hseih. 2001. Assessing and open hole aquifer test in fractured
crystalline rock. Ground Water, v. 39, n.l, p 68-78.
Trapp, H. and M.A. Horn. 1997. Ground Water Atlas of the U.S. North Carolina and vicinity
(HA 730-L). USGS. http://pubs.usgs.gov/ha/ha730/ch_l/L-text4.html.
US EPA, 2015, http://www.epa.gov/watersense/pubs/indoor.html accessed 8/26/15.
Watermark Numerical Computing, 2004, PEST Model -Independent Parameter Estimation User
Manual: 5th Edition.
Zheng, C. and P.P. Wang, 1999, MT3DMS: A Modular Three -Dimensional Multi -Species
Model for Simulation of Advection, Dispersion and Chemical Reactions of Contaminants
in Groundwater Systems: Documentation and User's Guide, SERDP-99-1, U.S. Army
Engineer Research and Development Center, Vicksburg, MS.
Tables
Table 1. Hydraulic head observed at monitoring
wells in June 2015 in the vicinity of the Mayo
Plant, heads predicted using the steady state
model, and their difference shown as a residual.
Head (ft)
Well Id
Observed
Predicted
Residual
ABMW-01
477.93
482.08
-4.15
ABMW-02
481.56
485.94
-4.38
ABMW-02BR
482.75
485.89
-3.14
ABMW-03
483.02
487.88
-4.86
ABMW-03S
483.04
487.86
-4.82
ABMW-04
485.11
485.05
0.06
ABMW-04D
485.06
484.98
0.08
ABMW-04BR
483.42
485.04
-1.62
MW-03BR
421.83
416.21
5.62
MW-05BR
501.94
501.42
0.52
MW-06BR
450.48
443.00
7.48
MW-07D
444.46
447.45
-2.99
MW-07BR
445.94
447.29
-1.35
MW-08S
437.02
439.47
-2.45
MW-08D
433.53
439.53
-6.00
MW-08BR
432.54
439.56
-7.02
MW-09BR
470.42
463.61
6.81
MW-10BR
499.57
497.07
2.50
MW-11BR
488.05
501.27
-13.22
MW-12S
556.6
548.43
8.17
MW-12D
556.9
548.47
8.43
MW-13BR
495.47
495.33
0.14
MW-14BR
504.61
506.55
-1.94
MW-15BR
402.68
408.64
-5.96
MW-16S
366.92
368.23
-1.31
MW-16D
365.52
368.31
-2.79
MW-16BR
365.41
368.62
-3.21
BG-1
510.26
501.66
8.60
BG-2
511.16
519.08
-7.92
CW-1
472.2
473.55
-1.35
CW-1D
471.93
473.76
-1.83
CW-2
375.29
378.08
-2.79
CW-2D
375.29
384.68
-9.39
CW-3
421.88
415.02
6.86
CW-4
427.7
437.32
-9.62
CW-5
501.53
501.34
0.19
CW-6
450.81
443.55
7.26
MW-2
435.11
431.44
3.67
MW-3
377.32
375.59
1.74
MW-4
496.61
499.21
-2.60
P1
443.77
440.15
3.62
P1A
439.81
440.14
-0.33
P2
413.69
414.28
-0.58
P3
394.41
399.04
-4.63
P3A
394.86
399.05
-4.19
P4
417.02
408.49
8.53
MA
418.84
409.13
9.71
Table 2. Calibrated hydraulic parameters.
Unit
Kh
(ft/d)
Kh/Kv
Ash, West Basin
1
5
Saprolite
3
1
Transition zone
1.3
1
Upper Rock
0.05
1
fracture zone
4-6
1
Mid Rock
0.02
1
fracture zone
2
1
Rock
0.007
1
Dam
0.05
1
Table 3. Recharge indifferent
zones in the model.
Recharge zone
Flux (ft/d)
Regional
0.0018
Mayo Lake
0.0
Ash Basin Lake
0.0001
Ash Basin Dam
0.0001
Mayo Lake Dam
0.0001
Lined Pond
0.0001
Table 4. Flow parameter sensitivity analysis. Results are expressed as model normalized root
mean square error (NRMSE) of the simulated and observed heads.
NRMSE
Unit
Kh (ft/d)
0.5 x calibrated
Calibrated
2 x calibrated
Ash
1
2.90%
2.84%
2.84%
Saprolite
3
3.22%
2.84%
3.41%
Transition zone
1.3
2.94%
2.84%
3.40%
Upper Rock
0.05
2.99%
2.84%
2.91%
Mid Rock
0.02
2.84%
2.84%
2.90%
Rock
0.007
2.86%
2.84%
2.98%
Dams
0.05
3.01%
2.84%
2.85%
Upland Recharge
0.0018
5.02%
2.84%
6.29%
Domestic wells
50 ft3/d
2.98%
2.84%
3.10%
Table 5. Boron concentrations observed in 2015 and simulated
at observation wells at Mayo Steam Electric Plant.
Boron (mg/L)
Well Id
Observe
Simulated
ABMW-01
4
4560
ABMW-02
9
9200
ABMW-02BR
0
1
ABMW-03
1
1940
ABMW-03S
1
395
ABMW-04
4
4930
ABMW-04D
5
1970
ABMW-04BR
0
0
MW-03BR
0
0
MW-05BR
0
0
MW-06BR
0
0
MW-07D
0
0
MW-07BR
0
0
MW-08S
0
0
MW-08D
0
0
MW-08BR
0
0
MW-09BR
0
0
MW-IOBR
0
0
MW-11BR
0
0
MW-12S
0
0
MW-12D
0
0
MW-13BR
0
0
MW-14BR
0
0
MW-15BR
0
0
MW-16S
0
0
MW-16D
0
0
MW-16BR
0
0
BG-1
0
0
BG-2
0
0
CW-1
0
0
CW-ID
0
0
CW-2
8
717
CW-213
3
1315
CW-3
0
0
CW-4
0
0
CW-5
0
0
CW-6
0
0
MW-2
0
7
MW-3
8
1372
MW-4
0
0
Table 6. Arsenic concentrations observed in 2015 and
simulated at observation wells at Mayo Steam Electric
Plant.
Arsenic (µg/I)
Observed
Simulated
ABM W-01
3290
3290
ABM W-02
2491
2491
ABM W-02BR
<1
3516
ABM W-03
4232
4232
ABMW-03S
1
3706
ABM W-04
4281
4281
ABM W-04D
140
792
ABMW-04BR
1
518
MW-03BR
<1
1020
MW-05BR
<1
0
MW-06BR
<1
193
MW-07D
<1
0
MW-07BR
<1
0
MW-08S
<1
0
MW-08D
<1
0
MW-08BR
3
0
MW-09BR
<1
0
MW-10BR
<1
0
MW-11BR
<1
0
MW-12S
<1
0
MW-12D
<1
0
MW-13BR
<1
0
MW-14BR
<1
0
MW-15BR
<1
0
MW-16S
<1
0
MW-16D
<1
0
MW-16BR
2
0
BG-1
<1
0
BG-2
<1
0
CW-1
<1
0
CW-1D
<1
0
CW-2
<1
0
CW-2D
<1
0
CW-3
<1
0
CW-4
<1
0
CW-5
<1
0
CW-6
<1
0
M W-2
<1
0
M W-3
<1
0
M W-4
<1
0
Figures
Figure 1. Map of the Mayo Steam Electric Plant showing ash basin, hydrologic features and
boundary of model region.
Figure 2. Air photo of the Mayo Steam Electric Plant showing ash basins and hydrologic
features and boundary of model region (yellow).
Figure 3. Fence diagram of the 3D hydrostratigraphic model (Solids) used to generate
hydrostratigraphy.
I
.. i� � ... a;sllplien11213
• pi1671�
yur ql� � Illil 31 �f�1�'� YY
I lli' i,tL III
ly Inll
7 Ip IIIII illllll llllll � �`YII�1 I �Eu
egp
or
4[p milt' "I"
� �111��161U lj I'}j III I{Iy111Y I
Figure 5. Distribution of recharge flux used in the Mayo model. Scale in ft/d.
Figure 6. Surface water features and wells in the vicinity of the Mayo Plant. Streams are
represented as "drains" and shown in green. Lakes and canals are represented as specified head
cells (purple) with a stage equal to the ground surface. The head is specified to be in the middle
of the transition zone along the external boundary in upland areas (purple bands around the edge
of the model). Water wells are shown as yellow squares.
HK
'..
�� y
5.4
4.5n::.
3.6Js
2.7-
1.8
0.9
*`
0.0
ft/d
7'
d
0
5 r 4 T
I
ia
a�
L a y e r 5 L
Layer E-7
0,
Figure 7. Hydraulic conductivity (ft/d) distributions in grid layers in the Mayo model. Low
permeability bands are either dams or tight rock. High permeability zones are inferred flat -lying
fracture zones.
Figure 8. Zones of different hydraulic conductivity in the vicinity of the ash basin. High
hydraulic conductivity in individual layers representing fracture zones (white with orange
border), and a zone of low permeability from layers 5 through 12 (blue with white border) in the
Mayo model.
Mayo
560
520
500
U)
480
-0 460
a�
440
E
U 42€3
e1#1#1
;M
all#]
350 400 45U 500 55U
COZ Observed Heads (ft)
Figure 9. Hydraulic head computed at steady state as a function of the observed head in
monitoring wells in the vicinity of the Mayo Plant. The line has a slope of 1:1.
Head
580
540
500
460
420
380
340
300
�vTITr
IVO
04
/ O
! 413
C M _-15BR
-1 «1 VR
420 MW-13BR -4
W
� -5 Q
MW
6WR
09BR
Figure 10. Steady state hydraulic head distribution predicted in top of bedrock (layer 8 of the
model) in the Mayo model. Colored bars indicate difference between predicted and observed
hydraulic head at monitoring wells. Green = less than 4.2 ft (2.5% NRMSE); yellow = 4.2 to 9.5
ft (2.5% to 5% NRMSE); red = greater than 9.5 ft (5% NRMSE).
Figure 11. Steady state hydraulic head distribution in predicted in top of bedrock (layer 8 of the
model) in the Mayo model. Colored bars indicate difference between predicted and observed
hydraulic head at monitoring wells. Green = less than 4.2 ft (2.5% NRMSE); yellow = 4.2 to 9.5
ft (2.5% to 5% NRMSE); red = greater than 9.5 ft (5% NRMSE).
Figure 12. Steady state hydraulic head distribution in the vicinity of the ash basin (layer 4 of the
Mayo model). Colored bars indicate difference between predicted and observed hydraulic head
at monitoring wells. Green = less than 4.2 ft (2.5% NRMSE); yellow = 4.2 to 9.5 ft (2.5% to 5%
NRMSE); red = greater than 9.5 ft (5% NRMSE).
Figure 13. Concentrations of boron in ash used to specify concentrations in the Mayo model.
2
Laye r 5
Sapralite ` �I Layer 10
Fractured Rack
f
5\'i
tLayer 12 ,
r,x Ti
Fractured Rock
'i
Layer 7 , t
Transition zo e. �-
ejr
Layer 8 Layer 13
Fractured Rock Rock
R
Ilk,o
t
Figure 15a. Concentrations of boron predicted to occur in Oct 2015 in Layer 5 of the Mayo
model. Red dashed line is outline of inferred fracture zone in Layer 12, and white shaded region
with black dashed border is inferred zone of low hydraulic conductivity.
Figure 15b. Concentrations of boron predicted to occur in Oct 2015 in Layer 7. Red dashed line
is outline of inferred fracture zone in Layer 12, and white shaded region with black dashed
border is inferred zone of low hydraulic conductivity.
Figure 15c. Concentrations of boron predicted to occur in Oct 2015 in Layer 8. Red dashed line
is outline of inferred fracture zone in Layer 12, and white shaded region with black dashed
border is inferred zone of low hydraulic conductivity.
Figure 15d. Concentrations of boron predicted to occur in Oct 2015 in Layer 10. Red dashed
line is outline of inferred fracture zone in Layer 12, and white shaded region with black dashed
border is inferred zone of low hydraulic conductivity.
Figure 15e. Concentrations of boron predicted to occur in Oct 2015 in Layer 12. Red dashed
line is outline of inferred fracture zone in Layer 12, and white shaded region with black dashed
border is inferred zone of low hydraulic conductivity.
Figure 15f. Concentrations of boron predicted to occur in Oct 2015 in layer 13. Red dashed line
is outline of inferred fracture zone in Layer 12, and white shaded region with black dashed
border is inferred zone of low hydraulic conductivity.
CLOSURE OPTION 2 - CLOSURE IN PLACE PROFILE
t11n
m
am
+oa
s.ao
IQ�00
1'n07
m•aa
a3.ma 7MOp
35+00
�WW_"
CLOSURE OPTION 2 - CLOSURE IN PLACE CROSS SECTION
esa
Sao
ma
1
Sow
W 5q0
10
16
EN00
25
_17
]0'lY
3B
z
Figure 16. Map and cross -sections describing design of the Cap -in -Place design used to develop
the model of this scenario.
Figure 17. Distribution of recharge in simulations of corrective actions. Red is ambient recharge
of 0.0018 ft/d, blue is recharge that is negligible. The small region in Layer 5 containing ash is
shown where the ash was removed in the lower left panel. Color scale reversed compared to
Figure 5.
Figure 17a. Hydraulic head calculated in the vicinity of the ash basin for the different corrective
actions.
Figure 18a. Simulated October, 2015 boron concentrations (µg/L) in the model layer of the
saprolite (layer 5).
Figure 18b. Simulated October, 2015 boron concentrations (µg/L) in the model layer of the
transition zone (layer 7).
Figure 18c. Simulated October, 2015 boron concentrations (µg/L) in the model layer of the upper
fractured rock (layer 10).
Figure 18d. Simulated October, 2015 arsenic concentrations (µg/L) in the model layer of the
saprolite (layer 5).
Figure 18e. Simulated October, 2015 arsenic concentrations (µg/L) in the model layer of the
transition zone (layer 7).
Figure 19. Simulated October, 2015 arsenic concentrations (µg/L) in the model layer of the
upper fractured rock (layer 10).
Figure 20. Simulated October, 2020 boron concentrations (µg/L) in the model layer of the
saprolite (layer 5) for CAP I.
Figure 21. Simulated October, 2020 boron concentrations (µg/L) in the model layer of the
transition zone (layer 7) for CAP 1.
Figure 22. Simulated October, 2020 boron concentrations (µg/L) in the model layer of the upper
fractured rock (layer 10) for CAP 1.
Figure 23. Simulated October, 2020 arsenic concentrations (µg/L) in the model layer of the
saprolite (layer 5) for CAP I.
Figure 24. Simulated October, 2020 arsenic concentrations (µg/L) in the model layer of the
transition zone (layer 7) for CAP 1.
Figure 25. Simulated October, 2020 arsenic concentrations (µg/L) in the model layer of the
upper fractured rock (layer 10) for CAP 1.
Figure 26. Simulated October, 2030 boron concentrations (µg/L) in the model layer of the
saprolite (layer 5) for CAP I.
Figure 27. Simulated October, 2030 boron concentrations (µg/L) in the model layer of the
transition zone (layer 7) for CAP 1.
Figure 28. Simulated October, 2030 boron concentrations (µg/L) in the model layer of the upper
fractured rock (layer 10) for CAP 1.
Figure 29. Simulated October, 2030 arsenic concentrations (µg/L) in the model layer of the
saprolite (layer 5) for CAP I.
Figure 30. Simulated October, 2030 arsenic concentrations (µg/L) in the model layer of the
transition zone (layer 7) for CAP 1.
Figure 31. Simulated October, 2030 arsenic concentrations (µg/L) in the model layer of the
upper fractured rock (layer 10) for CAP 1.
Figure 32. Simulated October, 2045 boron concentrations (µg/L) in the model layer of the
saprolite (layer 5) for CAP I.
Figure 33. Simulated October, 2045 boron concentrations (µg/L) in the model layer of the
transition zone (layer 7) for CAP 1.
Figure 34. Simulated October, 2045 boron concentrations (µg/L) in the model layer of the upper
fractured rock (layer 10) for CAP 1.
Figure 35. Simulated October, 2045 arsenic concentrations (µg/L) in the model layer of the
saprolite (layer 5) for CAP I.
Figure 36. Simulated October, 2045 arsenic concentrations (µg/L) in the model layer of the
transition zone (layer 7) for CAP 1.
Figure 37. Simulated October, 2045 arsenic concentrations (µg/L) in the model layer of the
upper fractured rock (layer 10) for CAP 1.
Figure 38. Simulated October, 2020 boron concentrations (µg/L) in the model layer of the
saprolite (layer 5) for CAP2.
Figure 39. Simulated October, 2020 boron concentrations (µg/L) in the model layer of the
transition zone (layer 7) for CAP2.
Figure 40. Simulated October, 2020 boron concentrations (µg/L) in the model layer of the upper
fractured rock (layer 10) for CAP2.
arsenic : 101812020 12:i]CI�QO;AM
1000.0
100
In
+ fk-I'a'
Figure 41. Simulated October, 2020 arsenic concentrations (µg/L) in the model layer of the
saprolite (layer 5) for CAP2.
Figure 42. Simulated October, 2020 arsenic concentrations (µg/L) in the model layer of the
transition zone (layer 7) for CAP2.
Figure 43. Simulated October, 2020 arsenic concentrations (µg/L) in the model layer of the
upper fractured rock (layer 10) for CAP2.
Figure 44. Simulated October, 2030 boron concentrations (µg/L) in the model layer of the
saprolite (layer 5) for CAP2.
Figure 45. Simulated October, 2030 boron concentrations (µg/L) in the model layer of the
transition zone (layer 7) for CAP2.
Figure 46. Simulated October, 2030 boron concentrations (µg/L) in the model layer of the upper
fractured rock (layer 10) for CAP2.
arsenic : 101612030 1
1000.0
100
o
Figure 47. Simulated October, 2030 arsenic concentrations (µg/L) in the model layer of the
saprolite (layer 5) for CAP2.
Figure 48. Simulated October, 2030 arsenic concentrations (µg/L) in the model layer of the
transition zone (layer 7) for CAP2.
Figure 49. Simulated October, 2030 arsenic concentrations (µg/L) in the model layer of the
upper fractured rock (layer 10) for CAP2.
Figure 50. Simulated October, 2045 boron concentrations (µg/L) in the model layer of the
saprolite (layer 5) for CAP2.
Figure 51. Simulated October, 2045 boron concentrations (µg/L) in the model layer of the
transition zone (layer 7) for CAP2.
Figure 52. Simulated October, 2045 boron concentrations (µg/L) in the model layer of the upper
fractured rock (layer 10) for CAP2.
arsenic : 1011012045 12
1000.0
100
Figure 53. Simulated October, 2045 arsenic concentrations (µg/L) in the model layer of the
saprolite (layer 5) for CAP2.
Figure 54. Simulated October, 2045 arsenic concentrations (µg/L) in the model layer of the
transition zone (layer 7) for CAP2.
Figure 55. Simulated October, 2045 arsenic concentrations (µg/L) in the model layer of the
upper fractured rock (layer 10) for CAP2.
Figure 56. Simulated October, 2020 boron concentrations (µg/L) in the model layer of the
saprolite (layer 5) for CAP3.
AM
Figure 57. Simulated October, 2020 boron concentrations (µg/L) in the model layer of the
transition zone (layer 7) for CAP3.
AM
Figure 58. Simulated October, 2020 boron concentrations (µg/L) in the model layer of the upper
fractured rock (layer 10) for CAP3.
Figure 59. Simulated October, 2020 arsenic concentrations (µg/L) in the model layer of the
saprolite (layer 5) for CAP3.
! s �MMEEK c -
Figure 60. Simulated October, 2020 arsenic concentrations (µg/L) in the model layer of the
transition zone (layer 7) for CA1 3.
! t '41MMEEK- L -
Figure 61. Simulated October, 2020 arsenic concentrations (µg/L) in the model layer of the
upper fractured rock (layer 10) for CAP3.
AM
Figure 62. Simulated October, 2030 boron concentrations (µg/L) in the model layer of the
saprolite (layer 5) for CAPS.
AM
Figure 63. Simulated October, 2030 boron concentrations (µg/L) in the model layer of the
transition zone (layer 7) for CAP3.
r ---- ..-� r ter. - 7-.1- --.`3r
Figure 64. Simulated October, 2030 boron concentrations (µg/L) in the model layer of the upper
fractured rock (layer 10) for CAP3.
0 �
Figure 65. Simulated October, 2030 arsenic concentrations (µg/L) in the model layer of the
saprolite (layer 5) for CAPS.
17m� - -
Figure 66. Simulated October, 2030 arsenic concentrations (µg/L) in the model layer of the
transition zone (layer 7) for CAP3.
Figure 67. Simulated October, 2030 arsenic concentrations (µg/L) in the model layer of the
upper fractured rock (layer 10) for CAP3.
Figure 68. Simulated October, 2045 boron concentrations (µg/L) in the model layer of the
saprolite (layer 5) for CAP3.
Figure 69. Simulated October, 2045 boron concentrations (µg/L) in the model layer of the
transition zone (layer 7) for CAP3.
Figure 70. Simulated October, 2045 boron concentrations (µg/L) in the model layer of the upper
fractured rock (layer 10) for CAP3.
Figure 71. Simulated October, 2045 arsenic concentrations (µg/L) in the model layer of the
saprolite (layer 5) for CAP3.
Figure 72. Simulated October, 2045 arsenic concentrations (µg/L) in the model layer of the
transition zone (layer 7) for CAP3.
Figure 73. Simulated October, 2045 arsenic concentrations (µg/L) in the model layer of the
upper fractured rock (layer 10) for CAP3.
UPDATE SIMULATIONS FOR GROUNDWATER FLOW AND
TRANSPORT REPORT FOR MAYO STEAM ELECTRIC PLANT,
ROXBORO,NC
February 2, 2016
Prepared for
SynTerra
148 River Street
Greenville, SC 29601
Investigators
Lawrence C. Murdoch, Ph.D.
Regina Graziano, M.S.
Scott E. Brame, M.S.
Ronald W. Falta, Ph.D.
This report is a continuation of the Groundwater Flow and Transport Modeling Report for
Mayo Steam Electric Plant (Murdoch et. al., 2015) [Attachment E in CAP 1 report]. The purpose of
this report is to simulate manganese transport at the Mayo Steam Electric Plant for the No Action and
the Cap -in -Place scenarios. In addition, the time projection for scenarios No Action and Cap -in -Place
is increased from 30 to 100 years for boron transport in groundwater.
Simulation of Manganese Transport in Ground Water
This section describes the analysis of manganese transport in the vicinity of the ash basin at the
Mayo Plant. During summer and fall 2015, a groundwater flow model was developed based on
geologic data obtained during drilling of a series of wells in the vicinity of the Plant, and the hydraulic
conductivity of the hydrogeologic units was estimated by comparing model predictions of hydraulic
heads to values observed in the wells. Transport simulations were conducted using the calibrated flow
model to evaluate the migration of boron and arsenic. Concentrations observed in monitoring wells in
the ash basin were used as initial conditions, and estimated Kd values were used to estimate the
distribution of those compounds in the ground water in the vicinity of the ash basin. After the transport
model was calibrated, it was used to estimate the performance of corrective actions, including No
Action, Cap -In -Place, Ash Removal, and a Hybrid option. These analyses were described in the
Groundwater Flow and Transport Modeling Report for Mayo Steam Electric Plant (Murdoch et. al.,
2015).
After the original report (Murdoch et. al., 2015) was prepared, it was decided that manganese
was of sufficient concern as a constituent to warrant including it in the transport analyses. The
procedure was to add manganese as a constituent and simulate transport from 1983 when the basin
became operational until 2015. It was included as initial conditions at concentrations ranging from 800
ppb to 2000 ppb in the ash basin based on analyses from monitoring wells. The manganese transport in
the Cap -in -Place corrective action scenario was simulated because it is the currently viewed as the most
likely scenario to be implemented.
The distribution coefficient, Kd, was estimated to vary as a function of pH and the
concentrations of other competing compounds, according to analyses by Brian Powell. The average
observed pH was 6.5, and this indicates that the average Kd value is 0.1 1/kg, according to Figure 1.
The simulation of the Cap -in -Place scenario used for this evaluation was patterned after the
scenario described in the CAP 1 report, with several changes. The most notable change is in the
conditions used to simulate concentrations in the ash. In the No Action scenario, the concentration
distribution from 2015 was used as initial conditions, but other constraints were lifted. In particular,
the concentrations in the ash were fixed at the measured concentrations from 1983-2015, but this
constraint was lifted in the No Action scenario for simulations of the corrective action. This allowed
ambient water to infiltrate the ash and reduce the concentration of the pore water.
In this analysis, the constraint that the concentrations in the ash are constant and equal to the
observed concentrations was maintained throughout the simulation, from 1983-2045. This approach
will cause the concentrations predicted for future corrective actions to exceed the previous approach,
which considered reduction of concentrations by dilution in the ash. This change in approach was
made to be more conservative. In reality, infiltration will dilute the concentrations in the ash to some
extent, so by ignoring this dilution effect the approach used here will tend to give an upper bound on
the expected concentrations.
Results
According to the simulation, manganese is transported downward out of the ash basin, beneath
the ash basin dam, and manganese forms a plume extending along Crutchfield Branch (Figs 2-9). The
analysis indicates that manganese follows the same flow paths as boron, and indeed, the distribution of
manganese predicted by the model is remarkably similar to that of boron. The maximum extent of the
plume in 2015 is in the vicinity of the compliance boundary, according to the simulations (Figs 2-9).
The simulation is able to approximate some of the observed concentrations in the vicinity of
Crutchfield Branch and beneath the ash basin. There are six wells where both the observed and
simulated concentrations are above the 2L standard (Table 1). The simulations indicate that some or all
of the manganese observed in water from these wells can be explained as having a source from within
the ash basin. Concentrations at 19 wells are above the 2L standard and the simulations indicate a
concentration of 0. The manganese in the water from these 19 wells cannot be explained as having a
source from within the ash basin. Concentrations in water from seven wells are predicted to be below
the 2L standard, and this is consistent with observations (Table 1).
The simulations indicate that approximately 3/4 of the wells where manganese exceeded the 2L
standard represent naturally occurring manganese that is not included in the model. The other '/4 of the
affected wells are located in a region influenced by the manganese plume in the simulations.
The most striking change in the ground water is a reduction in concentration of manganese
following the Cap -in -Place corrective action. This is apparent by comparing the location and extent of
the 1100 ppb contours in the same layer at different times (e.g compare Figure 3 to Figure 14, or Figure
4 to Figure 15). This reduction in concentration is also apparent in cross-section (compare Figure 9 to
Figure 19).
The leading edge of the plume marked by the 2L contour receded slightly 30 years after
corrective action (in year 2045 in the simulation). For example, the leading edge was approximately 50
ft beyond the compliance boundary along Crutchfield Branch at 2015 (Fig. 5), but the maximum extent
was at or within the compliance boundary in the 2045 simulation (Fig. 16). While encouraging, this
change is small and within a reasonable expectation of uncertainty in the location of the leading edge in
the simulations.
The concentrations of the plume decreased significantly and the plume decreased in size 100
years after the Cap -in -Place corrective action (in year 2115 in the simulation). This simulation predicts
the extent of the plume 100 ft away from the compliance boundary, which is farther away from
compliance boundary than the 2045 boron simulation prediction.
Simulation of Boron Transport in Groundwater —100 year projection
A 100 year projection for the No Action and the Cap -in -Place scenarios for boron transport in
groundwater were simulated. Originally, the groundwater modeling report (Murdoch et. al., 2015)
projected 5, 15, and 30 years. The dates for those simulations are referred to in the groundwater
modeling report (Murdoch et. al., 2015) as 2020, 2030, and 2045 respectively. In this report, the boron
simulations are projected 100 years after the corrective action plans take place. The date for the
simulations is 2115. Details of the No Action and Cap -in -Place modeling method can be found in the
Groundwater Flow and Transport Modeling Report for Mayo Steam Electric Plant (Murdoch et. al.,
2015).
No Action
Figures 47 through 49 depict the transport of boron for the No Action scenarios within model
layers 5 (saprolite), 7 (transition), and 10 (upper fractured bedrock). The simulated 2115 boron
concentrations delineated by the 2L contour within the saprolite is similar to the 2045 simulation
(Murdoch et. al., 2015). The 2115 boron plumes within the transition zone and bedrock are slightly
bigger than the 2045 simulations. The simulated 2115 boron bedrock plume has two locations where
the edge of the plume has migrated slightly north of Mayo Lake Road, which is the property and
compliance boundary. The leading edge of the 2115 boron plume within the saprolite and transition
zone are just south of Mayo Lake Road, which is similar to the 2045 simulations.
Cap -in -Place
Figures 50 through 53 depict the transport of boron for the Cap -in -Place scenarios within model
layers 5, 7, and 10. The boron concentrations decreases from 2045 to 2115 and is noticeable in the
saprolite, transition zone, and fractured bedrock layers. The leading edge of the boron plume also
recedes from 2045 to 2115. By 2115, the leading edge in some areas has receded by 200 to 300 ft in
the vicinity of Mayo Lake Road.
References
Langley, W.G., J. Daniels, and S. Oza, 2015, Sorption Evaluation of the. Roxboro Steam Electric Plant.
Charlotte Department of Civil and Environmental Engineering, report prepared for SynTerra,
McDonald, M.G. and A.W. Harbaugh, 1988, A Modular Three -Dimensional Finite -Difference Ground -
Water Flow Model, U.S. Geological Survey Techniques of Water Resources Investigations,
book 6, 586 p.
Murdoch, L.C., S. E. Brame, R. W. Falta, and R. A.Graziano, November 2015. Groundwater Flow and
Transport Modeling Report for Mayo Steam Electric Plant, Roxboro, NC.
Niswonger, R.G.,S. Panday, and I. Motomu, 2011, MODFLOW-NWT, A Newton formulation for
MODFLOW-2005, U.S. Geological Survey Techniques and Methods 6-A37, 44-.
SynTerra, 2015, Comprehensive Site Assessment Report, Roxboro Steam Electric Plant, Semora, NC.
September 2, 2015.
Powell, Brian, November 2015. Analysis of Geochemical Phenomena Controlling Mobility of Ions from
Coal Ash Basins at the Duke Energy Mayo Steam Electric Plant. Pendleton, SC.
Zheng, C. and P.P. Wang, 1999, MT3DMS: A Modular Three -Dimensional Multi -Species Model for
Simulation of Advection, Dispersion and Chemical Reactions of Contaminants in Groundwater
Systems: Documentation and User's Guide, SERDP-99-1, U.S. Army Engineer Research and
Development Center, Vicksburg, MS.
Mn Kd values versus pH in three G\%* silnulants
100E-01
1.00E-00
100E41
100E-0'
1 3 S 6 11 6 9 10
1H
*%%Kd. \SIB (', t' Vxhxs
•Mh Kd. AVGGn' Fives
& %% Kd. NLAA GR' Fives
Figure 1. Distribution coefficient for manganese as a function of
pH and relative concentration of competing ions. Figure from
report by Brian Powell.
Table 1. Concentrations of manganese
observed in monitoring wells and simulated
using the model. Rows shaded brown were
fixed concentrations in the model. The 6 rows
shaded yellow indicate concentrations that are
above 2L in both the field observations and the
simulations. The 19 rows shaded green
indicate field observations that are above 2L,
but the simulated values are less than 2L. The
7 rows shaded pink are where both the
observations and the simulations are less than
2L.
observed simulated
well name
Mn
Mn
A B M W -01
2000
2000
A B M W -02
1000
1000
ABMW-02BR
149
1
A B M W -03
1000
1000
ABMW-03S
478
203
A B M W -04
800
800
ABMW-04D
3640
337
ABMW-04BR
407
0
MW-03BR
731
0
MW-OSBR
1740
0
MW-07D
188
0
MW-07BR
263
0
MW-08BR
5040
0
MW-09BR
809
0
MW-10BR
894
0
MW-11BR
67
0
MW-12S
267
0
MW-12D
683
0
MW-13BR
474
0
MW-14BR
544
0
MW-16S
260
0
MW-16D
269
0
MW-16BR
30
0
BG-1
35
0
BG-2
150
0
CW-1
27
0
CW-1D
6
0
CW-2
170
250
CW-2D
123
394
CW-3
9
0
CW-4
0
0
CW-5
696
0
CW-6
1300
0
MW-2
833
7
M W -3
880
400
M W -4
15
0
Figure 2. Simulated 2015 manganese concentrations (µg/L) in model layer 4 (ash).
Figure 3. Simulated 2015 manganese concentrations (µg/L) in model layer 5 (saprolite).
Figure 4. Simulated 2015 manganese concentrations (µg/L) in model layer 7 (transition zone).
Figure 5. Simulated 2015 manganese concentrations (µg/L) in model layer 9 (upper rock).
Figure 6. Simulated 2015 manganese concentrations (µg/L) in model layer 11 (rock).
Figure 8. Simulated 2015 manganese concentrations (µg/L) in model layer 13 (rock).
manganese: 10/10/2015 12:01:24AM
Figure 9. Simulated 2015 manganese concentrations (µg/L) in cross-section trending along
Crutchfield Branch. Purple line shows top of rock.
Figure 10. Simulated 2045 manganese concentrations (µg/L) in model layer 1 (ash) for No Action.
Figure 11. Simulated 2045 manganese concentrations (µg/L) in model layer 2 (ash) for No Action.
Figure 12. Simulated 2045 manganese concentrations (µg/L) in model layer 3 (ash) for No Action.
Figure 13. Simulated 2045 manganese concentrations (µg/L) in model layer 4 (ash) for No Action.
Figure 14. Simulated 2045 manganese concentrations (µg/L) in model layer 5 (saprolite) for No Action.
Figure 15. Simulated 2045 manganese concentrations (µg/L) in model layer 7 (transition zone) for No
Action.
Figure 16. Simulated 2045 manganese concentrations (µg/L) in model layer 9 (shallow rock) for No
Action.
Figure 17. Simulated 2045 manganese concentrations (µg/L) in model layer 11 (rock) for No
Action.
Figure 18. Simulated 2045 manganese concentrations (µg/L) in model layer 13 (rock) for No
Action.
Figure 19. Simulated 2115 manganese concentrations (µg/L) in model layer 1 (ash) for No Action.
Figure 20. Simulated 2115 manganese concentrations (µg/L) in model layer 2 (ash) for No Action.
Figure 21. Simulated 2115 manganese concentrations (µg/L) in model layer 3 (ash) for No Action.
Figure 22. Simulated 2115 manganese concentrations (µg/L) in model layer 4 (ash) for No Action.
Figure 23. Simulated 2115 manganese concentrations (µg/L) in model layer 5 (saprolite) for No Action.
Figure 24. Simulated 2115 manganese concentrations (µg/L) in model layer 7 (transition zone) for No
Action.
Figure 25. Simulated 2115 manganese concentrations (µg/L) in model layer 9 (shallow rock) for No
Action.
Figure 26. Simulated 2115 manganese concentrations (µg/L) in model layer 11 (rock) for No
Action.
Figure 27. Simulated 2115 manganese concentrations (µg/L) in model layer 13 (rock). Assumes
Cap -in -Place corrective action.
Figure 28. Simulated 2045 manganese concentrations (µg/L) in model layer 1 (ash). Assumes
Cap -in -Place corrective action.
Figure 29. Simulated 2045 manganese concentrations (µg/L) in model layer 2 (ash). Assumes
Cap -in -Place corrective action.
Figure 30. Simulated 2045 manganese concentrations (µg/L) in model layer 3 (ash). Assumes
Cap -in -Place corrective action.
Figure 31. Simulated 2045 manganese concentrations (µg/L) in model layer 4 (ash). Assumes
Cap -in -Place corrective action.
Figure 32. Simulated 2045 manganese concentrations (µg/L) in model layer 5 (saprolite).
Assumes Cap -in -Place corrective action.
Figure 33. Simulated 2045 manganese concentrations (µg/L) in model layer 7 (transition zone).
Assumes Cap -in -Place corrective action.
Figure 34. Simulated 2045 manganese concentrations (µg/L) in model layer 9 (shallow rock).
Assumes Cap -in -Place corrective action.
Figure 35. Simulated 2045 manganese concentrations (µg/L) in model layer 11 (rock). Assumes
Cap -in -Place corrective action.
Figure 36. Simulated 2045 manganese concentrations (µg/L) in model layer 13 (rock). Assumes
Cap -in -Place corrective action.
manganese : M9I2045 11:58:36 PM
Figure 37. Simulated 2045 manganese concentrations (µg/L) in cross-section trending along
Crutchfield Branch. Assuming Cap -in -Place corrective action.
Figure 38. Simulated 2115 manganese concentrations (µg/L) in model layer 1 (ash). Assumes Cap-m-
Place corrective action.
Figure 39. Simulated October, 2115 manganese concentrations (µg/L) in model layer 2 (ash).
Assumes Cap -in -Place corrective action.
Figure 40. Simulated October, 2115 manganese concentrations (µg/L) in model layer 3 (ash).
Assumes Cap -in -Place corrective action.
Figure 41. Simulated October, 2115 manganese concentrations (µg/L) in model layer 4 (ash).
Assumes Cap -in -Place corrective action.
Figure 42. Simulated October, 2115 manganese concentrations (µg/L) in model layer 5
(saprolite). Assumes Cap -in -Place corrective action.
Figure 43. Simulated October, 2115 manganese concentrations (µg/L) in model layer 7
(transition zone). Assumes Cap -in -Place corrective action.
Figure 44. Simulated October, 2115 manganese concentrations (µg/L) in model layer 9 (shallow
rock). Assumes Cap -in -Place corrective action.
Figure 45. Simulated October, 2115 manganese concentrations (µg fl in model layer 11 (rock).
Assumes Cap -in -Place corrective action.
Figure 46. Simulated October, 2115 manganese concentrations (µg/L) in model layer 13 (rock).
Assumes Cap -in -Place corrective action.
Figure 47. Simulated October, 2115 boron concentrations (µg/L) in the model layer of the saprolite
(layer 5) for No Action.
Figure 48. Simulated October, 2115 boron concentrations (µg/L) in the model layer of the transition
zone (layer 7) for No Action.
Figure 49. Simulated October, 2115 boron concentrations (µg/L) in the model layer of the upper
fractured rock (layer 10) for No Action.
Figure 50. Simulated October, 2115 boron concentrations (µg/L) in the model layer of the saprolite
(layer 5). Assumes Cap -in -Place corrective action.
Figure 51. Simulated 2115 boron concentrations (µg/L) in the model layer of the transition zone (layer
7). Assumes Cap -in -Place corrective action.
Figure 52. Simulated 2115 boron concentrations (µg/L) in the model layer of the upper fractured rock
(layer 10). Assumes Cap -in -Place corrective action.