2016-01-19_HAI_ANOVA_Bkg-BTV_V-Cr6

VM
MEMORANDUM
19 January 2016
File No. 42058-002-01
TO: Duke Energy
Richard Baker
FROM: Haley & Aldrich, Inc.
Lisa JN Bradley, PhD, DABT
HALEY & ALDRICH, INC.
360 Quaker Street
Northbridge, MA 01534
978-846-3463
SUBJECT: Background Evaluation for Vanadium and Hexavalent Chromium
As requested by the North Carolina (NC) Department of Environmental Quality (DEQ), statistical analysis
of the background groundwater data available for private wells using groundwater in the vicinity of the
fourteen Duke Energy (Duke) facilities where ash basins are located. This evaluation focused on
vanadium and hexavalent chromium. An outlier test was conducted, an ANOVA evaluation was
conducted, data were grouped as indicated to be appropriate by the ANOVA, the difference in the
means was evaluated and was used to group the data as high or low, and based on these groupings,
background threshold values (BTVs) were calculated using a USEPA statistical software program, ProUCL
(USEPA, 2013).
Background Datasets
The background groundwater data set consists of data provided by DEQ and by Duke. Three datasets
were provided by the DEQ for private wells in the vicinity of Allen, Buck, and Marshall. Fourteen 14
datasets were provided by Duke for private wells located between 2 and 10 miles from each facility.
Thus there are 17 datasets evaluated here.
Outlier Testing
Outlier testing was performed using three standard deviations of the combined datasets error estimate.
Based on the use of the three standard deviations as the criteria, three outliers were identified:
• Vanadium at 112 ug/L for Roxboro
• Vanadium at 60.4 ug/L for Weatherspoon
Hexavalent chromium at 73.5 ug/L for Weatherspoon
www.haleyaldrich.com

Background Evaluation —Vanadium and Hexavalent Chromium
Duke Energy
19 January 2016
Page 2
ANOVA
The remaining data for vanadium and hexavalent chromium were used to conduct analysis of variance
computations for the two constituents. The statistical computations revealed that there are significant
differences in the mean concentrations amongst the 17 datasets for both the vanadium and the
chromium VI data.
Least Mean Squares
To identify which mean concentrations were significantly higher than the remaining mean values
another statistical method called "Least Significant Difference" (LSD) was used. A one -tailed test using a
5% probability value was used to compute the LSD for vanadium and for hexavalent chromium. The
calculated LSD value for vanadium is 2.5138. The calculated LSD value for hexavalent chromium is
0.5906.
The mean values for each of the plant sites for each constituent were organized in ascending order. The
lowest mean value was subtracted from each of the remaining mean values to obtain the difference.
This difference was compared with the computed LSD. If the difference was found to be larger than the
LSD then it is concluded that there is a significant difference in the mean value for that plant site. Table
1 summarizes the results of the LSD testing for vanadium and Table 2 summarizes the results of the LSD
testing for hexavalent chromium.
Based on the statistical evaluation of the using ANOVA and LSD statistics, it is concluded that for
vanadium and hexavalent chromium each can combine data in two groups, and a BTV can be developed
for each combined dataset.
Datasets
Based on the LSD analysis, the following four background datasets have been identified.
Vanadium Group A— Higher Background Concentrations
• Allen
• Allen -DEQ
• Buck
• Buck -DEQ
• Marshall
• Mayo
• Riverbend
Vanadium Group B — Lower Background Concentrations
• Asheville
• Belews Creek
• Cape Fear
i
�1 I ■ 2TO

Background Evaluation —Vanadium and Hexavalent Chromium
Duke Energy
19 January 2016
Page 3
• Cliffside
• Dan River
• Lee
• Marshall -DEQ
• Roxboro
• Sutton
• Weatherspoon
Hexavalent Chromium Group A — Higher Background Concentrations
• Allen
• Allen -DEQ
• Buck
• Buck -DEQ
• Marshall
• Marshall -DEQ
• Mayo
Hexavalent Chromium Group B — Lower Background Concentrations
• Asheville
• Belews Creek
• Cape Fear
• Cliffside
• Dan River
• Lee
• Riverbend
• Roxboro
• Sutton
• Weatherspoon
BTV Calculations
ProUCL (USEPA, 2013) was used to calculate BTVs for each of the four datasets. The results are shown
on Table 3.
The BTVs are:
• Vanadium Group A — Higher Background Concentrations — BTV = 42 ug/L
• Vanadium Group B — Lower Background Concentrations — BTV = 5.8 ug/L
• Hexavalent Chromium Group A — Higher Background Concentrations — BTV = 7.4 ug/L
• Hexavalent Chromium Group B — Lower Background Concentrations — BTV = 3 ug/L
i
�1 I ■ 2TO

Background Evaluation —Vanadium and Hexavalent Chromium
Duke Energy
19 January 2016
Page 4
These values can be used to further our understanding of the private well results collected by DEQ
under the Coal Ash Management Act (CAMA), and help inform decision-making for the Duke facilities.
Reference
USEPA. 2013a. Statistical Software ProUCL 5.0.00 for Environmental Applications for Data Sets with and
without Nondetect Observations. Software: http://www2.epa.gov/land-research/proucl-software, and
User's Guide: http://www2.epa.gov/sites/production/files/2015-03/documents/proucl v5.0 user.pdf
iF,,�I I =�
,�1 I ■ 2TO

Background Evaluation —Vanadium and Hexavalent Chromium
Duke Energy
19 January 2016
Page 5
Table 1 - Summary of the LSD testing results for the mean values for vanadium
* Denotes that there significant differences in the means with Ashville mean
concentration.
** Riverbend mean difference is only slightly below the LSD, so was grouped with those
results with a significant difference.
LSD = Least Significant Difference
Number
of
Results
Mean
Difference in
Means
Significance
(LSD = 2.5138)
Asheville
13
0.6261
Dan River
8
0.7038
0.0777
Sutton
13
0.8549
0.2288
Lee
11
0.9364
0.3103
Weatherspoon
1
1
Belews Creek
11
1.377
0.7509
Marshall -DEQ
10
1.7
1.0739
Roxboro
25
1.7204
1.0943
Cliffside
9
1.6347
1.0086
Cape Fear
18
1.697
1.0709
Riverbend
11
2.8881
2.262
**
Mayo
14
3.1889
2.5628
Marshall
29
3.9021
3.276
Buck
17
5.4053
4.7792
Allen -DEQ
7
6.0571
5.431
Allen
16
7.2512
6.6251
Buck -DEQ
7
7.9286
7.3025
* Denotes that there significant differences in the means with Ashville mean
concentration.
** Riverbend mean difference is only slightly below the LSD, so was grouped with those
results with a significant difference.
LSD = Least Significant Difference

Background Evaluation —Vanadium and Hexavalent Chromium
Duke Energy
19 January 2016
Page 6
Table 2 - Summary of the LSD testing results for the mean values for hexavalent chromium
* Denotes that there significant differences in the means with Sutton mean
concentration
LSD = Least Significant Difference
i
�1 I ■ 2TO
Number
of
Results
Mean
Difference in
Means
Significance
(LSD = 0.5906)
Weatherspoon
1
0.03
Only one
point
Sutton
13
0.03
Lee
10
0.0515
0.0215
Asheville
7
0.0876
0.0576
Cape Fear
18
0.14472
0.11472
Dan River
3
0.2227
0.1927
Cliffside
1
0.27
0.24
Riverbend
5
0.282
0.252
Roxboro
16
0.3917
0.3617
Belews Creek
6
0.56683
0.53683
Marshall
11
0.6481
0.6181
Mayo
13
0.683
0.653
Marshall -DEQ
10
0.721
0.691
Buck
17
0.7718
0.7418
Buck -DEQ
7
0.7943
0.7643
Allen -DEQ
7
1.1571
1.1271
Allen
6
1.5323
1.5023
* Denotes that there significant differences in the means with Sutton mean
concentration
LSD = Least Significant Difference
i
�1 I ■ 2TO

Background Evaluation—Vanadium and Hexavalent Chromium
Duke Energy
19 January 2016
Page 7
TABLE 3
BACKGROUND THRESHOLD VALUES FOR VANADIUM AND HEXAVALENT CHROMIUM
Page 1 of 1
Vanadium
Hexavalent Chromium
Statistically significant difference = Group A (higher background concentrations), no statistically significant difference = Group B (lower background concentration).
(b) - BTV values calculated using ProUCL v. 5.0.00 (USEPA, 2013).
Range of
distributions is very unstable
(d) - Vanadium Group A: Vanadium Group B: Hexavalent Chromium Group A:
Hexavalent Chromium Group B:
Allen DEQ Asheville Allen DEQ
Asheville
Range of
Belews Creek
Buck Cape Fear Buck
Statistical
Buck DEQ Cliffside Buck DEQ
Frequency of
Detected
Mean
Mayo Lee Marshall DEQ
Statistical
River Bend Marshall DEQ Mayo
Frequency of
Detected
Mean
Sutton
Grouping
Frequency of
Detection
Concentrations
Detect
Grouping
Frequency of
Detection
Concentrations
Detect
(a, d)
Detection
Percent
(ug/L)
(ug/L)
Background Threshold Value (BTV) (b) (ug/L)
(a, d)
Detection
Percent
(ug/L)
(ug/L)
Background Threshold Value (BTV) (b) (ug/L)
Group A
79 / 101
78%
0.318 - 23.7
6.072
42
Gamma
95% WH KM Gamma USL
Group A
51 / 71
72%
0.039 - 5.8
0.992
7.434
Gamma
95% WH KM Gamma USL
Group B
45 / 119
38%
0.346 - 13.1
2.302
5.776
Lognormal
95% UTL (c)
Group B
32 / 80
40%
0.033 - 3
0.292
3
Nonparametric
95% USL
Notes:
BTV - Background Threshold Value. KM - Kaplan -Meier.
DEQ- Department of Environmental Quality. USL- Upper Simultaneous Limit.
NC - North Carolina. UTL - Upper Tolerance Limit.
ug/L - micrograms/liter. WH - Wilson H ilferty.
(a) - Based on the statistical evaluation of the 17 background data sets (14 data sets from Duke, one for each plant; and 3 from DEC, for Allen, Buck and Marshll)
data for vanadium and hexavalent chromium were each divided into two groups following the conclusion of statistical significance between means.
Statistically significant difference = Group A (higher background concentrations), no statistically significant difference = Group B (lower background concentration).
(b) - BTV values calculated using ProUCL v. 5.0.00 (USEPA, 2013).
(c) - For constituents with lognormal distributions, the 95% UTL was selected as the BTV as the USL for lognormal
distributions is very unstable
(d) - Vanadium Group A: Vanadium Group B: Hexavalent Chromium Group A:
Hexavalent Chromium Group B:
Allen DEQ Asheville Allen DEQ
Asheville
Allen Belews Creek Allen
Belews Creek
Buck Cape Fear Buck
Cape Fear
Buck DEQ Cliffside Buck DEQ
Cliffside
Marshall Dan River Marshall
Dan River
Mayo Lee Marshall DEQ
Lee
River Bend Marshall DEQ Mayo
Riverbend
Roxboro
Roxboro
Sutton
Sutton
Weatherspoon
Weatherspoon
Haley & Aldrich, Inc. 1/19/2016
2016 -0119 -Vanadium Hex Chrome BTVs.xlsx