PPBC Tech Memo for DEC sites (2)

REVISED PROPOSAL PROVISIONAL BACKGROUND CONCENTRATIONS
DUKE SITES ALLEN, BELEWS, BUCK, CLIFFSIDE, DAN RIVER, MARSHALL, RIVERBEND
During the Comprehensive Site Assessment (CSA), background wells were established in accessible
upgradient locations based on information available regarding topography and likely groundwater flow
direction. Only one sampling event was performed for the CSA reports, due to Coal Ash Management
Act (CAMA) imposed reporting timelines. As a result for the CSA reports, Duke Energy utilized the
approved background wells from the National Pollutant Discharge Elimination System (NPDES)
compliance monitoring well system, to develop the initial set of site-specific proposed provisional
background concentrations (PPBCs) presented in Chapter 10 of each CSA report. The methods and
results of statistical analysis were presented in Appendix G of each CSA report.
Following submittal of the CSAs, North Carolina Department of Environmental Quality (NCDEQ)
requested the removal of all sample observations from the data set where the sample had > 10
nephelometric turbidity units (NTU) turbidity. New site-specific PPBCs were developed with this
truncated data set. For some sites, NCDEQ requested the inclusion of groundwater sampling data from
on-site background landfill wells. Thus, during the Corrective Action Plan (CAP) Part 1 report submittal,
Duke Energy reviewed and re -performed statistical analysis for backgrounds for the facilities with the
following changes:
Y Locations of all background wells were reviewed to determine if they remained suitable for
background (at Allen, for instance, well AB -1 R was removed because of increasing
concentrations of some Constituents of Interest [COls]). The evaluations were presented in
Appendix B of each CAP Part 1 report.
Y All sample data from wells with > 10 NTU turbidity were removed from the data set. Where
applicable, data from background wells from landfills at the sites were used to add to the data set.
Y Tests for outliers, tests for normality and parametric and non -parametric prediction limits were re-
performed.
Y Each site was reviewed to assess whether sufficient data were available to calculate a new Upper
Prediction Limit (UPL).
At the request of NCDEQ on December 14, 2015, all CSA background data from all three rounds of
sampling have been included in the assessment of PPBCs. Round 3 data became available (post -data
validation and upload to a database for output of tables) on December 14, 2015. Due to the extremely
short turn -around time, re -running PPBC statistics using ProUCL was not possible. This will be done as
Round 4 sampling results become available in early 2016. Questions that will be assessed as Round 4
data is incorporated in the statistical results include:
Y Do the data all come from the same distribution? Mixing the compliance and voluntary well data,
landfill data and CSA background well data creates both spatial and temporal distribution issues
that will be addressed once four rounds of CSA background well data became available.
Y With new data added, if appropriate, is the distribution of data for each constituent normal or non-
normal?
Y Once the distribution of the data is know, review and determine the most appropriate UPL method
to use.
At this time, the assumption is that all the data remain non -normal, and therefore the UPL is a non-
parametric limit which is the maximum value in the data set, per the U.S. Environmental Protection
Agency (USEPA) Unified Guidance (2009, Section 18.3). This reflects standard scientific method for
groundwater investigations.

Table 1 provides revised PPBCs based on the first three rounds of CSA sampling events.
SAMPLE SIZES
Samples have been collected from the following number of background locations for one or more
constituents at each site. The data sets have some minor differences due to variability in requested
analyses over time
Allen: 4 locations
Belews: 9 locations
Buck: 9 locations
Cliffside: 13 locations
Dan River: 5 locations
Marshall: 9 locations
Riverbend: 10 locations
CSA wells generally have 3 rounds of sampling data; some of the background ash basin compliance and
landfill groundwater monitoring wells have been sampled for more than 5 years.
The USEPA Unified Statistical Guidance (2009) (Chapter 18, Section 18.1.1) specifies that a minimum of
eight independent background measurements be available to conduct parametric prediction limit tests.
USEPA recommends a "much higher number of observations" be collected if data require non -parametric
prediction limit tests. Groundwater data commonly do not meet tests of normal distribution, thus requiring
the development of non -parametric prediction limits. Without this number of tests, the prediction limits
developed have very little confidence associated with them.
In addition to the lack of sufficient sample size for which statistics can be performed, CAMA timelines do
not allow the collection of four or more sampling rounds distributed over the course of a year or more in
order to determine if seasonal trends are present in the concentrations of Cols in background wells. At
this time, Duke Energy has conducted three sampling rounds (June, September, and November 2015)
and is, at NCDEQ's request, currently collecting a fourth round of background samples (December 2015).
While this will allow a re-evaluation of background data, it will not allow any evaluation of seasonality.
Four observations from a population are rarely enough to adequately characterize its statistical features;
statisticians generally consider sample sizes of n <_ 4 to be insufficient for good statistical analysis. A key
implication of the independent and identically distributed assumption is that a series of sample
measurements should be stationary overtime (i.e., stable in mean level and variance).
Also, according to the USEPA Unified Guidance, with newly -installed background wells, it may be
necessary to discard initially collected observations and to wait several months for aquifer disturbances
due to well construction to stabilize.
For all these reasons, these PPBCs will be re-evaluated in early 2016 once the fourth round of CSA
background well data are collected.
2