Duke Energy Letter May 2017 Background Soil & Groundwater

(� DUKE
ENERGY
May 2b, 2U17
Mr. S. Jay Zimmerman, P.v.
Director, Division of Water Resources
North Carolina Department of Envirojrmental Quality
1b11 Mail Service Center
Raleigh, NC 27699-1611
Peal Drauvitch
yCIu.r Vi,e President
Environmental. Health tic Satery
52b Suuth Unurcrt Street
tC3XP
11C 2�21,12
9 0.373.U4Ub
RE: Response to DEQ Letter Dated April 28, Zu1/
Request for Additional Intormation Regarding Statistical Methods for Developing Reference
Background Concentrations for Groundwater mid Soil at Coal Ash Facilities (HDR E„g;nee„ng,
Inc. and Syn Terra Corporation, January 2017)
Dear Mr. Zimmerman:
Duke Energy is in receipt of the above-reterencCd letter trom the North Carolina Depaffinemt of
Environmental Quality (DEQ) which included comments on the Technical Memorandum (TM) titled,
"Statistical Methods for Developing Heference Background Concentrations for Groundwater and Soil at
Coal Ash Fac;l;r;es (Background Merhodology)", prepared by HDR Engineering, Inc. and SynTerra
Corporation, dated January 2uT7. Although Conditional Approval was granted to the January TM, DEQ
requested a revised TM along with updated data sets for background groundwater concentrations and
soil concentrations.
Duke Energy encloses herein the revised TM which incorporated DEQ comment along with updated data
sets for Background groundwater concentrations and Background soil concentrations for each of the 14
present and former coal-fired facilities in North Carolina. Below are DEQ's comments from the April 28,
2U17 letter and Duke Energy's response in italics.
Datasets
"Provide up-to-date digital spreadsheets of raw background groundwater data for each facility by May
26, 2017, and include, within the raw Background data spreadsheet as'strike-throughs', the individual
monitoring data results that Duke Energy believes should be omitted trom the Background databet doe
to (a) fiigh pH, (b) high turbidity, (c) auto- correlation (see comment below), (d) outlier designation, (e)
non -detect values that are above 2L/IMAC, or (t) other reasons.'
Enclosed are Excel® spreadsheets for each of the 14 Duke Energy facilities which provide the updated
background groundwater dara for each facility. 'Srrike-rhroughs" were used to identify which analytical
results would be excluded from the initial background data set for statistical evaluurion doe rro e;rher of
the following criteria pH greater than $_S, turbidity greater than 1U NTU, less than a bu-day interval
between viable sample events, method derecrion limits (MDLs) greater than the ZL/IMAC for non -detect
samples, or other documented rationale_ the enclosed site-specific Technical Memorandums „ote where
additional sampling results are required to add to the existing data sets to provide a sufficient pool of
dura to begin zrar;.5r;cal rntuly.,1.5.

Pir. Jay Zimmtl rman Cotter
May 26, 2017
"Provide up-to-date digital spreadsheets of raw bacRground soil data for each tacility By May 26, 2017.
Any soil data collected since submittal of the Comprehensive Site Assessments should be included in the
table and existing soil data should Be reviewed to identify any quality control issues (i.e., sampling
intervals, corresponding Boring sample ID, etc.) along with identiticaiion of outliers and revised
accordingly."
Enclosed are EAcel m spreadsheers for each of rhr 14 facilities which provi-de the updated background soil
data for each facility. 'strike-rhroughs" were used to idenrify which analytical results would be
excludedrrom the initial background data set for statistical evaluation. The enclosed site-specific
Technical Memorandums ►cote where additional soil sampling results are necessary to add ra the emiariny
data sets to provide a sufficient pool of data to begin statistical analysis.
Introduction. Page 1. "Thememorandum states that'For the purpose otestaftshing 5acRground
threshold values atthistime, the value whish reps eseritsthe upper threshold value trom tfie upper
tail of the data distributionfor a given constituent will be considered the value representative of a
naturally occurring concentration,orthe proposed provisional Background threshold value.' The
wuida the--uppertail vt' ,hould be removed. A»rated in EPA/600/R-07/041, PivUCLTecfinical
ouide, 1uIs, page 17, The objective is to compute background statistics based upon the majority of
the dataset representing the main dominant bacRground population, and not to accommodate a few
low probability high outliers (e.g. coming trom eAtrume tails otthe data dish ibotion) that may also be
present inthe background dataset.' I he memorandum should reference the Prouc.L page 17
language and be updated accordingly."
The text within the revised draft of the TM has been modified accordingly to reflect the suggested
removal of "rhe upper rail of". A5`dirionally, rhe rexr has been modified ro include reference to the
Pro UCL Technical Guide.
Part II, Page 7. Autocorrelation. "Duke Energy states that autocorrelation will be evaluated using a
'sample autocorrelation tunction'. The Division position is to use a minimum bu-day interval Between
sample events. Dake may, at any time, submit for consideration evidence that data trom additional
sample events spaced closer together in time are not autocorrelated, but for purposes of the raw
Background dataset and Background determinations required at this time, the minimum b0 -day interval
should Be used. The memorandum should Be updated accordingly."
The text within the revised draft of the TM has been updated in response to this comment, i he following
text was added in Part Il:
"For purposes of the initial r aw backgr ound dataset and development of PPBTVs, a r►rinimum 60 -day
interval between sample events will be used. In the event samples are collected at intervals shorter than
60 days (e.g., for catchup sampling at problematic locations, site conditions, etc.), autocorrelation
evaluations will be performed and may be provided to the Division of Water Resources as lines of
evidence to confirm the samples are not autocorrelated and carr be included irr the background 5`010
seri."
Part IV, Page 14. Step 3. "The memorandum states that *if data are normally and lognormally
disrr ibured, bur ,or gamma disrribured preference should be given to upper tolerance levels (UTLs)

Mr. Jay zimmc, man Cettci
May 26, 2017
produced assuming data are lognormally distributed provided the logged data have a standard
deviation less than or equal to one'. The Diviaivii dues nut concur. Assuming data tollow the
corresponding U I L distribution type, the following u 11 computation preference should be used: nog mal
UTL, then gamma UTC, then lognormal UTL, then nonparametric UTL. However, as stated in ProULL,
because lognormal distributions can result in unrealistically high UTL computations, lognormal UTL
Should be avoided when skewness is high (e.g. standard deviation of logged data is - 1 or 1.5) and
sample bice is small (e.g., < 20 - 30). In This case, nonparametric UTL should be preterred over lognormal
u 11. I he memorandum should be updated accordingly."
The text within the revised draft of rhe TM has been updared as follows:
"When Elora sets used for producing UTLs can be fitted to multiple distribution models, a specific
hierarchy preference is applied. calculation of a specific u►L will follow the distribution hierarchy
preference below, with the noted exceptions:
1. normal,
2. gamma,
3. lognormal, and
4. nonparamerric.
The exception to the hierarchy is based on situations where the dura ser eAhibirs skewness rhar is
mode, ate urrd higher (e. y. sruridard deviarion 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."
Part Iv, Page 14. Step 5. "The memorandum states that 'A minimum of eight valid Background
groundwater samples should Be obtained prior to producing background threshold values (BTVs) for
each cumiitaent in each tlow layeF. Eight samples should Be changed to ten samples. The
memorandum also states that 'In addition, a minimum of eight additional samples should be obtained
prior to evaluating it new Background data should be combined with previous data to produce revised
BTVs.' In practice, time trames needed to collect ten additional Background samples may not Be
available given LAMA and other deadlines. In these cases, DEQ will dCtCrininu wheat data are
appropriate for inclusion in a comprehensive background dataset based on all relevant considerations.
The memorandum should Be updated accordingly."
The text within the revised draft of the TM has been updated as follows to reflect a minimum of ten
samples will be obtained prior to development of background threshold values:
"In addition, the allocated tune fr ume riecebsary ro collecr an additional ten samples for further
evaluarion of background may not be available given the assessment deadlines and autocorrelation
restrictions. In evaluating the need for inclusion of additional background data to produce revised BTVs,
DEQ will determine what data are appropriate for inclusion it, a comprehensive background dara set
based on relevant considerations."
Pari IV, Page 14. Step 5. "The memorandum states that 'If it is deemed necessary to produce ii 1 vs
prior to obtaining eight valid samples, NCDEQ will be consulted and the maximum observation may
need to Be used as a BTV. BTVs are expected to be produced for all constituents and flow bybtCms
whether ten valid samples are available. It less than ten valid samples are available for this
3

Mr. jay LImmerman Cotter
May 26, 2017
determination, no tormal 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 6C cun�idered a. � outliEr and , emoved true , , turther use and the PBTV
is computed to be the znd highest value. The memorandum should be updated accordingly."
The rexr w;rh;„ the r evi.5ed dr aft of rhe TM hay been updated ru clur;fy haw aurrrples will be handled ;f
less than 10 valid samples I he text states the following:
A minimum of ten valid background groundwater samples should be obtained prior to producing BTVs
for each corrst;tuent;rr each flow layer. If it is deemed necessary to produce BTVs prior to obtaining ten
valid samples, UTLs will not be calculated and the PPBTV for a constituent within a flow layer will be
estimated to be either:
rhe highest value, or
• if the highesr value ;s an order of magnitude grearer rhan the geometric mean of all
values, then the highest value will be considered an outlier and the second highest value
Will be utilized as the PPBTV.
In situations where there are non -detects and less than ten valid samples, the geometric mean, which is
the product of all values (including the censored values) taken to the root of n, may not be representative
of the central tendency of thar sample. The median may be a better reference value from which to
deter r►r;rre;f the highest value is an acceptable estimate for the PPBTV, and rrray be ut;l;zed ;f deter rrrirred
appropriate. "
Cu, ,mCnta on Doke Energy's January 20, 2017 "Responses to DEQ'b Nuvember 22, 2016 informal
comments"
"Duke Energy Rebpoiue #3. Duke Eneigy'brcz)ponbE ducz, "m addre» the EAclubiun of
autocorrelated samples that have been collected too close in time. DELt's protocol included a
criterion that samples not be collected too close in time and selected bU days as a reasonable
treuuency at whish aato co,relatiun would not be EAPELTed. Duke Energy should ack, owledge and
address this issue."
The rexr within the r ev;ped dr aft of rhe TM hug beer, updured ;n Parr Il, Section b ro address rh;s
comment. The text states:
"For purposes of the initial raw background dataset and development of PPB 1 Vs, a minimum 60 -day
inrerval berween v;able sample events will be used. In the evenr samples are collected or intervals
shorter than 6u days (e.g., for catchup sampling at problematic locations site conditions, etc_),
outocorrelation evaluations will be performed and provided as lines of evidence to confirm the samples
are ,or ourocor, alured and uan be ;nrladed it, rhe barky, ound darn sers."
"Duke Energy Response #11. The summary spreadsheets to which Duke Energy reters do not, in
many cases, include an indication of outlies or autocorrelated data. Thebe values bhoald be clearly
indicated within the summary spreadsheets."
The enclosed EncelO spreadsheets for each of the 14 fac;l;t;es have been updated to show outl;er3 ;f they
are present prior to performing the preliminary data analysis as described in the revised draft TIVI. Note
4

Mr. Jay cimmurr„ar, Lcttcr
May 26, 2017
that autocorrelated data (i.e., less than bU days between sample events) and data excludedror being
uar„de rhe pH and turbidity rai,gc� a, a Oetirif;ed using ' Srr;ke-rh, ough�” H, the updated spreadsheers as
described above and in the revised draft 1101.
"Non -detect values above the 2L/IMAC standards should be deleted front use in the raw Background
dataset."
The enclosed Excel' spreadsheets for each of the 14 facilities have been updated accordingly. The text in
Part 1 of the revised draft TM Teas been revised to reflect the comment above as well_
"PaFT I, Page 4. The mer riorandam btaiCb that 'Duke EirCrgy and NCDEQ have agreed that soil
samples can be pooled from multiple depth intervals. A statistical evaluation of the soil data sets
will Be pertormed to contirm the approach is appropriate'. The memorandum references USEPA's
Guidance for Comparing Background and Chemical Concentrations inSoil for CERCLA Sires (USEPA
2uu2) but does not specify the statistical evaluations that are proposed to evaluate the appliLa6ility
of pooling soil data. Tse memorandum should be updated with methodology to:
confirm that pooling multiple depth intervals at a site is appropriate.
cuntirm that pooling Anil data truer different geologic turmativm, at a aitC
appropriate.
statistically evaluate soil results that are below a detection limit when that
detection limit is above the protection of groundwater threshold."
The text it, Parr 1 of rhe i ev;.,3ed dr aft TM ha.) been mad;f;ed ro reflect the commenr above. The rexr
states:
'...Only constituent concentrations from samples collected above the water table will be utilized for
praduc;ng BTVs. To allow fa, compar;aon of,eziulra from soil samples collected from differenr depth
intervals and locations across the site to the BTVs, background soil samples will be pooled from multiple
depth intervals and non -impacted locations. Non -detect sample results with a method detection limit
above the North CaroGnu Prorecr;ori of G, Fu,,dvvurer Prel;m;nary Soil Remediarion Goal (PSRG) will be
excluded from the background soil dataset.
So;l dara are suscepr;ble to eAh;bir spatial variation (by depth and geology), and as such preliminary data
analysis methods will be used to evaluate the soil data set. To aid in identifying outliers, visual
assessments will be performed using box -and -whisker plots and quantitative assessments will be used to
reZ,r for d;fferences ;n mean or med;an concenrrar;on across depth inrervals or geologic formations.
Results from the statistical analysis of soil data sets will allow for decisions to be made if pooling of soil
data across multiple depth intervals or geologic formations is appropriate."
Dufce Energy, along with our consultants, and DEQ statt discussed the above -referenced letter during a
meeting at the Winzotun-SalCnr Rtcgiunal Office ort May 18, 2017. It was Duke Energy's understanding
from that meeting that DEQ will review the enclosed data sets within two weeks. Duke Energy will then
have 30 days to complete the statistical analysis for background soil and groundwater determinations
for each site.
5

Mr. Jay Zimmerman Leiter
May Z6, ZUl/
If you have any questions or need any clarification regarding the information plovided, NlCabe Lcpntact
Ed Sullivan at ed.sullivanCuduke-energy.com or at 98U-373-3719 at your convenience.
Respe tfully submitte
Pa I Draoviich
SVP — Environmental, Realth & safety
cc: Mr. Ed Sullivan — Duke Energy
IDIr. Lhad Hearn — HDR
Ms. Kathy We66 — Syn I erra
Enclosures:
Rcvlscd Draft iCchni,al Plurn.randum, dated Play 26, 2017 — "Statistical Methods for Developing Reterence 6ackgrouna
Concentrations for Groundwater and Soil at Coal Asfi Facilities"
Upaated Site Soil and Groundwater 1 eChniLal Mernurandums
Updated Site suit and Ciruundwater Excel® Spreadsheet Data Sets
0