9-28-17 Zimmerman - Provisional Background Threshold Values Letter with ...

(' DUKE
ENERGY
Paul Draovitch
Senior Vice President
Environmental, Health & Safety
526 South Church Street
EC3XP
Charlotte, NC 28202
980.373.0408
September 29, 2017
Mr. S. Jay Zimmerman, P.G.
Director, Division of Water Resources
North Carolina Department of Environmental Quality
1611 Mail Service Center
Raleigh, NC 27699-1611
RE: Response to DEQ Letters Dated September 1, 2017
Provisional Background Threshold Values - Belews Creek, Mayo, Riverbend and Roxboro
Dear Mr. Zimmerman:
Duke Energy is in receipt of the above -referenced letters regarding provisional background threshold
values (PBTV) for soil and groundwater for four of our facilities. Attached to the letters were tables for
soil and groundwater which indicated whether or not the Division was in agreement with the PBTV
values submitted by Duke Energy. There was concurrence for the vast majority of the constituents in
soil and groundwater. When there was not agreement, the Division requested Duke Energy "provide
revised values for DWR to review and approve". Enclosed is Duke Energy's response. Duke Energy
requests a conference call or meeting with Division technical staff, Duke Energy and Duke Energy's
environmental consultant to discuss our response in more detail.
Duke Energy appreciates the opportunity for open technical discussion regarding these PBTVs. If you
have any questions or need any clarification regarding the information provided, please contact Ed
Sullivan at [e-mail] or at 980-373-3719 at your convenience.
Respectf ly submitted,
Paul Draovitch
Senior Vice President
Environmental, Health & Safety
cc/enc: Mr. Ed Sullivan — Duke Energy
Ms. Kathy Webb — SynTerra Corp.

Mr. Jay Zimmerman Enclosure for Provisional Background Threshold Values for Soil and Groundwater
September 29, 2017
GENERAL COMMENTS APPLICABLE TO ALL FOUR SITES
General Notes
• The distribution of constituent datasets should be determined by assessing the significance level
(P-value), test value, critical value, and censored and uncensored Q-Q GOF plots rather than sole
reliance on the text statement, provided in ProUCL goodness -of -fit (GOF) test results output. As
stated at the end of the second paragraph in section 4 (Identifying Data Distributions) on page 7
of the Revised Statistical Methods for Developing Reference Background Concentrations for
Groundwater and Soil at Coal Ash Facilities (May 2017) technical memorandum, "GOF tests will
be performed using a significance level of 0.05."
The phrase "approximate distribution" is sometimes observed in ProUCL GOF test results
output. For each GOF test, ProUCL determines dataset distributions based on two techniques.
For normal and lognormal GOF tests, either the Shapiro -Wilk (S-W) or Lilliefors test is used. For
gamma GOF tests, either the Anderson -Darling (A-D) or Kolmogorov-Smirnov (K-F) test is used.
Occasionally, one of two GOF tests may lead to the conclusion that data fit a given distribution.
When this occurs, ProUCL provides a text output stating that the dataset follows an
`approximate distribution'. A decision as to which GOF technique is appropriate to use is
provided in the statistics technical memorandum, which states, "The Shapiro -Wilk GOF Test is
applicable for data sets comprised of 50 or fewer samples, while the Lilliefors GOF test is
appropriate for data sets containing more than 50 samples. To evaluate if data are gamma
distributed, the Anderson -Darling or Kolmogorov-Smirnov GOF test will be utilized."
Furthermore, approximate distributions may also occur when either a significance level between
0.05 and 0.01 is observed, indicating a significant deviation from the distribution of interest, or a
test statistic value does not exceed the critical value for the specified significance level of 0.05.
Some approximate distributions do not adhere to the requirement set forth in the statistics
technical memorandum and should not be used to calculate BTVs. If normal, gamma, and
lognormal are all approximately distributed, then it may be assumed that the data are non -
parametric. UTLs calculated assuming different distributions can sometimes lead to
substantially different values; therefore, attention should be given to GOF output statistics
rather than solely rely on the text result.
• Outliers are values that are not representative of the population from which they were sampled
and whose presence can skew statistical results. Outliers can be data residing in either the
upper- or lower -tails of datasets. Screening datasets solely for outliers in the upper -tail of
dataset distributions lead to biased -low results. One of the primary goals for using statistics is to
reduce bias in the interpretation of results. To obtain unbiased results, the upper- and lower -
tails of datasets need to be screened for outliers.
• While screening background groundwater datasets for outliers, all available data should be
included. This includes autocorrelated values, but does not include samples with turbidity
greater than or equal to 10 NTU, a pH greater than 8.5 S.U., or no recorded turbidity or pH
measurements. The only stages in which autocorrelated values should be omitted are during
GOF tests and calculation of UTLs. This is because statistical derivation of UTLs makes an
underlying assumption that data are statistically independent (or autocorrelated) with respect
to time and space. Outlier tests, though, do not make this underlying assumption. Additionally,
1

Mr. Jay Zimmerman Enclosure for Provisional Background Threshold Values for Soil and Groundwater
September 29, 2017
retaining autocorrelated values during outlier screening can strengthen decision making by
helping determine if suspected outliers are a result of pure chance (i.e., very low probability of
occurrence) or are in fact representative of groundwater constituent concentrations at a given
point in time.
In a letter from DEQ to Duke Energy (Zimmerman to Draovitch, September 1, 2017) regarding
provisional background threshold values for soil and groundwater for Belews Creek Steam
Station, Mayo Steam Electric Plant, Riverbend Steam Station, and Roxboro Steam Electric Plant,
DEQ stated "Outliers are identified with three statistical lines of evidence; Box Plots, Q-Q Plots,
and 95% Significance Levels." Foremost, 'significance levels' should refer to 'confidence levels'.
A significance level of 95% would result in a probability -value (P-value) of 0.95 and would result
in a substantial increase the number of false positives. The appropriate significance level for a
confidence level of 95% is 0.05 (or 5%). Additionally, the use of 5% significance levels contradict
the procedure discussed in the second paragraph of section 3 on page 7 in the Revised Statistical
Methods for Developing Reference Background Concentrations for Groundwater and Soil at Coal
Ash Facilities (May 2017) technical memorandum, which states, "Extreme outliers are of
interest; therefore, outlier tests will be conducted using a significance level of 0.01" (or 1%). All
outlier tests should be performed in accordance with the technical memorandum.
Part IV, page 15, number 3, of the Revised Statistical Methods for Developing Reference
Background Concentrations for Groundwater and Soil at Coal Ash Facilities (May 2017) states,
"When data sets used for producing UTLs can be fitted to multiple distribution models, a specific
hierarchy preference is applied...
I. normal,
II. gamma,
III. lognormal, and;
IV. nonparametric.
The exception to the hierarchy is based on situations where the data set exhibits skewness that
is moderate to high (e.g., standard deviation of logged data is greater than 1) and sample size is
small (e.g., n < 30). In these situations, the nonparametric UTL is preferred over lognormal UTL."
Calculations should be conducted in accordance with the methodology presented in the
statistics technical memorandum.
In cases where the non -parametric UTL is calculated for a constituent, the coverage should be
manually changed to 85% as opposed to 95%. Section IV, on page 15, number 4, of the Revised
Statistical Methods for Developing Reference Background Concentrations for Groundwater and
Soil at Coal Ash Facilities (May 2017) states, "It has been demonstrated that if there are
insufficient sample sizes, the non -parametric UTL cannot achieve the desired confidence
coefficient of 95 percent. Depending on background sample size, a different order statistic is
selected to produce UTLs. For constituent data sets containing less than 59 samples, UTLs will
be produced using a coverage of 85 percent (i.e., the 85th percentile) and a confidence
coefficient of 95 percent. ... For data sets containing 59 or more samples, UTLs will be
produced using coverage of 95 percent and confidence coefficient of 95 percent."
• For constituent datasets containing 50% or more non -detects (NDs), UTLs should be calculated
using non -parametric statistics. Section 1 on page 15 of the Revised Statistical Methods for

Mr. Jay Zimmerman Enclosure for Provisional Background Threshold Values for Soil and Groundwater
September 29, 2017
Developing Reference Background Concentrations for Groundwater and Soil at Coal Ash Facilities
(May 2017) states, "Some constituent data sets may be represented by 50 percent or more NDs.
A large percentage of NDs makes it difficult to fit data to distribution modes. For data sets
containing 50 percent or more NDs, UTLs will be constructed utilizing nonparametric
techniques."

Mr. Jay Zimmerman Enclosure for Provisional Background Threshold Values for Soil and Groundwater
September 29, 2017
BELEWS CREEK COMMENTS
Belews Creek Groundwater Comments
Manganese in Background Deep Flow Zone
Discrepancies between the upper tolerance limit (UTL) for manganese in the deep flow zone appears to
be the result of the following:
DEQ commented on the BTV for manganese in the deep flow zone, stating, "Deep should be 13
(or 2L) because of five identified outliers."
Elevated manganese concentrations are observed in background monitoring well BG-1D. Five
out of nine manganese results from well 13G-1D are identified as outliers by DEQ in the deep
flow zone background dataset. Manganese concentrations in BG-1D exhibit moderate
correlation with pH and moderate to strong correlation with specific conductance. Data from
BG-1D suggest elevated and variable manganese concentrations may be a result of natural
changes in groundwater conditions at this location. Review of laboratory analytical reports
indicate the five outliers are valid and it is Duke Energy's position that the results should be
retained to incorporate background variability.
While attempting to replicate DEQ's results, it appears DEQ may have used non -parametric
statistics to calculate the UTL for manganese. The UTL should not be calculated using non -
parametric statistics because GOF tests indicate manganese data are gamma distributed. Part
IV, page 15, number 3, of the Revised Statistical Methods for Developing Reference Background
Concentrations for Groundwater and Soil at Coal Ash Facilities (May 2017), states, "When data
sets used for producing UTLs can be fitted to multiple distribution models, a specific hierarchy
preference is applied....
I. normal,
II. gamma,
III. lognormal, and;
IV. non parametric."
Therefore, non -parametric statistics should not have been used to calculate the UTL for
manganese. Instead, the UTL should be calculated assuming the gamma distribution.
Furthermore, if the non -parametric UTL was calculated for manganese, the coverage should
have been set to 85% as opposed to the 95% coverage that DEQ may have used. Section IV, on
page 15, number 4, of the Revised Statistical Methods for Developing Reference Background
Concentrations for Groundwater and Soil at Coal Ash Facilities (May 2017), states, "It has been
demonstrated that if there are insufficient sample sizes, the non -parametric UTL cannot achieve
the desired confidence coefficient of 95 percent.... For constituent data sets containing less
than 59 samples, UTLs will be produced using a coverage of 85 percent (i.e., the 85th percentile)
and a confidence coefficient of 95 percent.... For data sets containing 59 or more samples,
UTLs will be produced using a coverage of 95 percent and confidence coefficient of 95 percent."
4

Mr. Jay Zimmerman Enclosure for Provisional Background Threshold Values for Soil and Groundwater
September 29, 2017
As a result, the non -parametric UTL should not be calculated using a coverage of 95% because
the dataset consists of 30 samples.
The range of BTVs determined for manganese in the deep flow zone at facilities stationed in the
Piedmont region is 55 µg/L (the lowest calculated is for Belews Creek) to 580 µg/L. Duke Energy
requests continued use of 55 LLRA as the BTV for manganese in the deep flow zone at Belews
Creek.
Vanadium in Background Shallow Flow Zone
Discrepancies between the UTL for vanadium in the shallow flow zone appears to be the result of the
following:
DEQ commented on the BTV for vanadium in the shallow flow zone, suggesting that the value be
changed to, "1.33 (or 2L) because of 2.01 µg/L identified as an outlier." Quantile-quantile (Q-Q)
and box plots do not suggest concentration value 2.01 µg/L is an outlier. Duke Energy will retain
concentration value 2.01 µg/L for current calculations.
• Duke Energy recalculated the UTL for vanadium using normal statistics. With the inclusion of
2.01 µg/L, the UTL for vanadium was calculated to be 1.36 L.
In addition, it appears DEQ may have used non -parametric statistics to calculate the proposed
UTL for vanadium. The data for vanadium should not be assumed to be non -parametric because
the Shapiro -Wilk GOF test indicates the data are normally distributed. Part IV, page 15, number
3, of the Revised Statistical Methods for Developing Reference Background Concentrations for
Groundwater and Soil at Coal Ash Facilities (May 2017) states, "When data sets used for
producing UTLs can be fitted to multiple distribution models, a specific heirarchy preference is
applied... .
I. normal,
II. gamma,
III. lognormal, and;
IV. nonparametric."
For these reasons, non -parametric statistics should not have been used to calculate the UTL for
vanadium. Instead, the UTL should have been calculated using normal statistics.
Furthermore, if the non -parametric UTL was calculated for vanadium, the coverage should be
85% instead of 95%. Section IV, on page 15, number 4, of the Revised Statistical Methods for
Developing Reference Background Concentrations for Groundwater and Soil at Coal Ash Facilities
(May 2017) states, "It has been demonstrated that if there are insufficient sample sizes, the
non -parametric UTL cannot achieve the desired confidence coefficient of 95 percent.
Depending on background sample size, a different order statistics is selected to produce UTLs.
For constituent data sets containing less than 59 samples, UTLs will be produced using a
coverage of 85 percent (i.e., the 85th percentile) and a confidence coefficient of 95 percent... .
For data sets containing 59 or more samples, UTLs will be produced using coverage of 95
percent and confidence coefficient of 95 percent."

Mr. Jay Zimmerman Enclosure for Provisional Background Threshold Values for Soil and Groundwater
September 29, 2017
Belews Creek Soil Comments
Beryllium in Background Soil
DEQ indicated in conference on August 25, 2017 that the analytical results for beryllium and boron
provided in Table 1 of the Updated Background Threshold Values for Soil (August 30, 2017) technical
memorandum had been inadvertently switched.
The background soil dataset for beryllium has been corrected, and the updated UTL for
beryllium using the corrected dataset is calculated to be 1.65 me/ke. Please revise accordingly
from 19.3 mg/kg.
Boron in Background Soil
DEQ indicated in conference on August 25, 2017 that the analytical results for beryllium and boron
provided in Table 1 of the Updated Background Threshold Values for Soil (August 30, 2017) technical
memorandum had been inadvertently switched.
• The background soil dataset for boron has been corrected, and the updated UTL for boron using
the corrected dataset is calculated to be 22 m k . Please revise accordingly from 17 mg/kg.
Chromium in Background Soil
Discrepancy between the upper tolerance limit (UTL) for chromium in background soil appears to be the
result of the following:
• DEQ commented on the BTV for chromium in soil, suggesting that the, "PBTV for Chromium
should be 36 mg/kg instead of 41.1 mg/kg because 43 was statistically identified as an outlier."
Quantile-quantile (Q-Q) and box plots indicate that the concentration value 43 mg/kg is a
moderate outlier. Conversely, Rosner's outlier test indicated that no extreme outliers are
present within the dataset. As stated in the Revised Statistical Methods for Developing
Reference Background Concentrations for Groundwater and Soil at Coal Ash Facilities (May
2017), "Extreme outliers are of interest; therefore, outlier tests will be conducted using a
significance level of 0.01" (or 99 percent confidence level). All outlier tests were performed in
accordance with the technical memorandum. Therefore, Duke Energy recommends retaining
the concentration value 43 mg/kg, and the UTL for chromium should remain 41.1 mg/kg.
Thallium in Background Soil
Discrepancies between the UTL for thallium in background soil appears to be the result of the following:
DEQ commented on the BTV for thallium, indicating the, "PBTV for Thallium should be 0.69
mg/kg instead of 0.85."
It appears DEQ used normal statistics to calculate the UTL for thallium. Observation of the p-
value, rather than reliance on the text output, as well as Q-Q GOF plots indicates that the data
are gamma distributed instead of normally distributed. The Revised Statistical Methods for
Developing Reference Background Concentrations for Groundwater and Soil at Coal Ash Facilities
(May 2017) on page 8, section 4, states GOF tests will be performed using a significance level of
9

Mr. Jay Zimmerman Enclosure for Provisional Background Threshold Values for Soil and Groundwater
September 29, 2017
0.05. The p-value (significance level) for the normal GOF tests on thallium data is less than 0.05.
Please see General Notes section for more discussion on data distributions.
• The calculated UTL for thallium using gamma statistics is 0.85 ma/kg.
Vanadium in Background Soil
Discrepancy between the UTL for vanadium in background soil appears to be the result of the following:
DEQ commented on the BTV for vanadium in background soil, stating, "PBTV for Vanadium
should be 114 mg/kg instead of 127, because, in addition to 280, 141 was also statistically
identified as an outlier."
• Review of Q-Q and box plots indicate concentration value 280 mg/kg is the only outlier.
DEQ performed outlier tests using a significance level of 0.05 (or 95% confidence level) as
opposed to a significance level of 0.01(or 99% confidence level), as stated in the Revised
Statistical Methods for Developing Reference Background Concentrations for Groundwater and
Soil at Coal Ash Facilities (May 2017) on page 7, section 3, second paragraph, "Extreme outliers
are of interest; therefore, outlier tests will be conducted using a significance level of 0.01." The
September 1, 2017 letter from DEQ to Duke Energy incorrectly stated that "95% significance
levels" should be used. The use of a 0.05 significance level for outlier tests is not in accordance
with the approved statistical guidance document.
In addition, the value 141 mg/kg does not appear to be an outlier because elevated values are
characteristic of its location (BG-3S/ BGSB-3) and may indicate variability in background soils
across the site. Duke Energy recommends retaining the concentration value 141 mg/kg during
current calculations.
• The UTL for vanadium should remain 127 mg/kg.
7

Mr. Jay Zimmerman Enclosure for Provisional Background Threshold Values for Soil and Groundwater
September 29, 2017
MAYO COMMENTS
Mayo Soil Comments
Copper in Background Soil
Discrepancy between the upper tolerance limit (UTL) for copper in soil appears to be the result of the
following:
• It appears DEQ calculated the UTL for copper assuming data were normally distributed.
• Goodness -of -fit (GOF) tests indicate the data for copper are gamma distributed and not
normally distributed.
• Duke Energy calculated the UTL statistics for copper assuming gamma distribution.
• The UTL for copper should remain 124 mg/kg.
Iron in Background Soil
Discrepancies between the UTL for iron in soil appears to be the result of the following:
• DEQ determined the UTL for iron assuming data were approximately normally distributed.
Observation of the p-value (less than 0.05) for normal GOF tests indicates iron data are not
normally distributed. Please see General Notes section for further discussion.
• GOF tests indicate the data for iron are gamma distributed.
• Duke Energy calculated the UTL for iron using gamma statistics.
• The UTL for iron should remain 82,634 me/kg.
Manganese in Background Soil
Discrepancies between the UTL for manganese in soil appears to be the result of the following:
• DEQ calculated the UTL for manganese by omitting concentration values 3,100 mg/kg and 2,900
mg/kg and retained statistical outlier 22.1 mg/kg. DEQ also assumed data were normally
distributed and calculated the UTL for manganese using normal statistics.
• Outlier screening identified manganese concentration values 3,100 mg/kg and 2,900 mg/kg as
statistical outliers. These values represent data from a single boring (BGS13-9). Elevated
manganese concentration may be characteristic of this location and may represent variability
within background soils across the site.
Concentration value 22.1 mg/kg resides outside the majority of the data distribution and was
determined to be a low -end statistical outlier. This value was omitted prior to calculating the
UTL for manganese. Please see General Notes section as to why low -end outliers in addition to
high -end outliers should be considered.
H.

Mr. Jay Zimmerman Enclosure for Provisional Background Threshold Values for Soil and Groundwater
September 29, 2017
• GOF tests indicate manganese data are not normally distributed but are gamma distributed.
• Duke Energy calculated the UTL for manganese using gamma statistics.
• The UTL for manganese should remain 3,144 mg/kg.
Sodium in Background Soil
Discrepancies between the UTL for sodium in soil appears to be the result of the following:
It appears DEQ may have determined the distribution of sodium data by substituting % the
detection limit for non -detect concentrations as opposed to addressing non -detects using
imputation or modeling techniques. ProUCL software provides a mechanism for distinguishing
between detect and non -detect data. Justification for why distinctions should be made
between detected and non -detect concentrations during calculations is provided in the General
Notes section.
• It also appears DEQ may have used Lilliefors GOF test to assess the distribution of sodium data.
Lilliefors GOF test should be used when datasets contain more than 50 samples (per ProUCL
Technical Guidance Document and the Revised Statistical Methods for Developing Reference
Background Concentrations for Groundwater and Soil at Coal Ash Facilities (May 26, 2017).
• The Anderson -Darling (A-D) and Kolmogorov-Smirnov (K-S) GOF tests in addition to censored
and uncensored quantile-quantile (Q-Q) GOF test plots indicate data for sodium are gamma
distributed.
• The initial sodium UTL proposed by Duke Energy (1,111 mg/kg) was calculated using lognormal
statistics. Further evaluation of the distribution of sodium data indicates the data are gamma
distributed. As a result, the UTL for sodium was recalculated using gamma statistics.
• The updated UTL for sodium using gamma statistics should be 832.7 me/kg.
Thallium in Background Soil
Discrepancies between the UTL for thallium in soil appears to be the result of the following:
• It appears DEQ calculated the UTL for thallium by omitting the concentration value 0.79 mg/kg.
• Data for thallium follow an approximate gamma distribution. This is a result of the A-D GOF test
suggesting data are not gamma distributed, whereas the K-S GOF test indicates data are gamma
distributed. Evaluation of censored and uncensored Q-Q GOF test plots suggests assuming
thallium data are gamma distributed may not be appropriate.
• Additional GOF tests indicate data for thallium are lognormally distributed and have a logged
standard deviation of less than one (0.77).
• Duke Energy calculated the UTL for thallium by retaining concentration values 0.79 mg/kg and
0.33 kg/mg and using lognormal statistics.
9

Mr. Jay Zimmerman Enclosure for Provisional Background Threshold Values for Soil and Groundwater
September 29, 2017
• The UTL for thallium should remain 1.03 me/kiz.
• When using ProUCL to calculate UTLs, constituent datasets must be categorized based on the
presence or absence of non -detected concentrations. ProUCL calculates UTLs differently
depending on whether or not non -detects are present within datasets. The algorithms in
ProUCL use imputation or modeling techniques to address non -detects. ProUCL does not
encourage substituting non -detects with a constant (e.g., multiplying the detection limit by 1 or
1/2). Doing so introduces bias in the calculation of UTLs.
10

Mr. Jay Zimmerman Enclosure for Provisional Background Threshold Values for Soil and Groundwater
September 29, 2017
RIVERBEND COMMENTS
Riverbend Groundwater Comments
Antimony in Background Bedrock Flow Zone
Discrepancy between the upper tolerance limit (UTL) for antimony in the bedrock flow zone appears to
be the result of the following:
• The background dataset for the bedrock flow zone used for determination of background
threshold values (BTUs) contains only four valid samples for antimony. DEQ suggests that the
provisional background threshold value (PBTV) for antimony in the bedrock flow zone should be
the Interim Maximum Allowable Concentration (IMAC) of 1 µg/L (rather than 2.9 µg/L proposed
by Duke Energy) until additional data are available.
The proposed use of the IMAC as the PBTV for antimony is not in accordance with the Revised
Statistical Methods for Developing Reference Background Concentrations for Groundwater and
Soil at Coal Ash Facilities (May 2017) technical memorandum and the letter provided by DEQ to
Duke Energy (Zimmerman to Draovitch) dated April 28, 2017. Both documents state, "If less
than ten valid samples are available for this determination, no formal UTL statistics should be
run and the provisional background threshold value (PBTV) for a constituent and flow system
should be computed to be either: a) the highest value, or b) if the highest value is above an
order of magnitude greater than the geometric mean of all values, then the highest value should
be considered an outlier and removed from further use and the PBTV is computed to be the 2"d
highest value."
Duke Energy's proposed PBTV (2.9 L) for antimony was calculated in accordance with the
statistics technical memorandum approved by DEQ. Therefore, this value should continue to be
used as the PBTV for antimony in the bedrock flow zone.
Chromium in Background Shallow Flow Zone
Discrepancies between the UTL for chromium in the shallow flow zone appears to be the result of the
following:
DEQ commented on the BTV for chromium in the shallow flow zone, stating, "Shallow should be
5 (or 2L) because 14 was identified as an outlier."
The chromium concentration value of 14 µg/L is an outlier and will be omitted from current
calculations.
• Duke Energy agrees that the UTL for chromium should be 5 ue/L.
Cobalt in Background Shallow Flow Zone
Discrepancies between the UTL for cobalt in the shallow flow zone appears to be the result of the
following:
DEQ commented on the BTV for cobalt in the shallow flow zone, stating, "Shallow should be 2.3
because 11.3, 9.4, and 8.5 were identified as outliers."
11

Mr. Jay Zimmerman Enclosure for Provisional Background Threshold Values for Soil and Groundwater
September 29, 2017
Duke Energy does not agree that the concentration values 11.3 µg/L, 9.4 µg/L, and 8.5 µg/L
should be omitted from current calculations.
• Background monitoring well GWA-14S contains four valid samples for cobalt (2 µg/L, 8.5 µg/L,
11.3 µg/L, and 9.4 µg/L). These four values represent the four highest concentration values
within the background cobalt dataset. Review of quantile-quantile (Q-Q) and box plots indicate
concentration of cobalt in groundwater at this location is variable and may be characteristic of
this location in addition to documenting variability in background across the Site.
• Duke Energy suggests the UTL for cobalt in the shallow flow zone remain 11.64 L.
• Additionally, it appears DEQ did not calculate the UTL for cobalt (2.3 µg/L) in accordance with
the Revised Statistical Methods for Developing Reference Background Concentrations for
Groundwater and Soil at Coal Ash Facilities (May 2017) technical memorandum.
In an attempt to replicate DEQ's results, it appears DEQ may have omitted non -detected cobalt
concentrations of 1 µg/L during calculations. Unless non -parametric statistics are used to
calculate UTLs (which do not make a distinction between detects and non -detects), non -detect
values should be retained during calculations, unless values exceed 2L or IMAC. Removal of
non -detects can result in biased -high UTLs.
Iron in Background Shallow Flow Zone
Discrepancies between the UTL for iron in the shallow flow zone appears to be the result of the
following:
• DEQ suggests the UTL for iron in the shallow flow zone should be 950 µg/L, rather than the
1,278 µg/L obtained by Duke Energy, because the concentration value 1,890 µg/L should have
been identified as an outlier and omitted from current calculations.
• Duke Energy agrees that concentration value 1,890 µg/L is an outlier and should be omitted
from current calculations.
• In an attempt to replicate DEQ's results, it appears DEQ may not have calculated the UTL for iron
in accordance with the Revised Statistical Methods for Developing Reference Background
Concentrations for Groundwater and Soil at Coal Ash Facilities (May 2017) technical
memorandum by using gamma statistics rather than lognormal statistics, and did not make a
distinction between detected and non -detected values during calculations. Please refer to the
General Notes section below for a description on why a distinction should be made between
detected and non -detected values during calculations and why the UTL should be calculated
using lognormal statistics instead of gamma statistics.
• Duke Energy calculated the UTL for iron using lognormal statistics without the inclusion of
concentration value 1,890 µg/L. The UTL for iron should be 1,014 L.
12

Mr. Jay Zimmerman Enclosure for Provisional Background Threshold Values for Soil and Groundwater
September 29, 2017
Riverbend Soil Comments
Cobalt in Background Soil
Discrepancy between the upper tolerance limit (UTL) for cobalt in soil appears to be the result of the
following:
DEQ commented on the background threshold value (BTV) for cobalt in soil, stating, "PBTV for
Cobalt should be 46.5 without any identified outliers."
• Duke Energy agrees that no outliers are present within the cobalt dataset.
• Duke Energy was unable to replicate DEQ's calculations to obtain a value of 46.5 mg/kg.
Therefore, a BTV of 46.5 mg/kg for cobalt cannot be agreed upon until the specific steps taken
to calculate the UTL for cobalt can be reviewed.
13

Mr. Jay Zimmerman Enclosure for Provisional Background Threshold Values for Soil and Groundwater
September 29, 2017
ROXBORO COMMENTS
Roxboro Groundwater Comments
Duke Energy submitted via e-mail on August 31, 2017 to the Raleigh Regional Office information
requested during a conference call on August 26, 2017 related to the aluminum concentration in the
transition zone, chromium concentration in the transition zone and iron concentration in the bedrock
zone. However, it appears that the information submitted was too late to be factored into the DEQ
response dated September 1, 2017. Duke Energy requests that DEQ factor in the August 31, 2017
information and revisit the 'not acceptable' reply to the aluminum concentration in the transition zone,
the chromium concentration in the transition zone and the iron concentration in the bedrock zone.
Roxboro Soil Comments
Manganese in Background Soil
Discrepancy between the upper tolerance limit (UTL) for manganese in soil appears to be the result of
the following:
It appears DEQ may have assumed manganese data were normally distributed and used normal
statistics to calculate the UTL. The observed p-value for the normal goodness -of -fit (GOF) test
results is less than the 0.05 significance level, indicating a significant deviation from normality.
Therefore, the UTL should not be calculated using normal distribution statistics. Please see
General Notes section for further discussion.
• GOF tests indicate the data for manganese are gamma distributed, and the UTL should be
calculated using gamma statistics. Duke Energy calculated the UTL to be 532 mg/kg.
Nitrate (as N) in Background Soil
Discrepancies between the UTL for nitrate (as N) in soil appears to be the result of the following:
• It appears DEQ may have assumed nitrate (as N) data were approximately gamma distributed
and calculated the UTL using gamma statistics. The approximate gamma distribution is the
result of the Anderson -Darling (A-D) GOF test indicating the data significantly deviate from the
gamma distribution, whereas the Kolmogorov-Smirnov GOF test indicated data were gamma
distributed.
• To further assess the distribution of nitrate (as N) data, censored and uncensored quantile-
quantile (Q-Q) GOF plots were observed.
• Q-Q plots indicate that the data do significantly deviate from the gamma distribution and
instead are lognormally distributed.
• Duke Energy calculated the UTL for nitrate (as N) assuming data using lognormal statistics. The
logged standard deviation of the data was less than one (0.821).
• The UTL for nitrate (as N) should remain 0.945 mg/kg.
14