Characterizing water quality background …...Characterizing water quality background concentrations...

transcript

Characterizing water quality background concentrations of aluminum, PCBs, and radioactivity on the arid Pajarito Plateau, New Mexico

Datasets

Publicly available data were downloaded from LANL’s Intellus database:

• More than 750 samples collected by LANL or the NewMexico Environment Department (NMED) from 85locations within 11 canyons (and 23 individualsub-watersheds) from 2005 to 2017 (Figure 1) wereincluded in the dataset.

• Data collected prior to 2005 were deemed unreliable byLANL and excluded.

• Data from July 4, 2011, to January 1, 2014, wereinfluenced by the Las Conchas Fire and were excludedper LANL guidance. (fire-affected watersheds only).

• Data for 28 constituents—corresponding to theconstituents regulated by LANL’s Individual Permit (IP)

Background Characterization Framework

The BCF is a process for evaluating background storm water data, accounting for dependencies (i.e., spatial or temporal factors affecting concentrations), and calculating BTVs. The process follows five steps (Figures 4a and 4b):

• Step 1. Identify sufficient IID populations within the dataset.

• Step 2. Explore and describe dependencies within the dataset.

• Step 3. If dependencies exist, split the dataset into subpopulations ornormalize the dataset as appropriate to meet stability requirements.

• Step 4. Calculate BTVs.

• Step 5. Characterize uncertainty.

Brian G. Church,1* Scott Tobiason,1 Amanda B. White,2 Emily M. Day,2 Steve Veenis,2 Armand R. Groffman,2 Don J. Carlson III2 1 Windward Environmental, Seattle, WA2 Los Alamos National Laboratory, Los Alamos, NM* Poster presenter

Figure 2. Conceptual site model for background storm water on Pajarito Plateau

Figure 3. Example streams on Pajarito Plateau during the early monsoonal season

Figure 4a. Background characterization framework, Steps 1–3

Figure 4b. Background characterization framework, Steps 4–5

Figure 9. Time plot used during BCF Step 2.3

Figure 10. Q-Q plots used during BCF Steps 4.2 and 4.3

Figure 11. Comparison of 2018 BTVs and existing LANL BTVs for unfiltered (total) aluminum

Figure 12. Comparison of 2018 BTVs and existing LANL BTVs for gross alpha radiation

as well as total aluminum, total recoverable cyanide, and dissolved selenium—were included.

– Several constituents regulated by LANL’s IP were not detected in anysample, including 2,3,7,8-tetrachlorodibenzo-p-dioxin, cyanide,benzo(a)pyrene, and silver.

LANL assigned spatial categories to each sampling location to define the major watershed (canyon), minor watershed, and “location grouping” variables. Location groupings corresponded to subareas of interest, such as undeveloped landscapes to the west of the Laboratory property (Western Reference) or undeveloped landscapes to the north of the town of Los Alamos (Northern Reference) (see Figure 1).

Suspended sediment concentration (SSC) data were paired with as many samples as possible using concurrent (same-day, co-located) samples. If only total suspended solids (TSS) data were available for a sample, SSC was estimated from TSS using a log-log linear regression (p < 0.05, r2 ~ 0.7) developed from samples with paired SSC and TSS data.

ReferencesEPA. 2016. Statistical software 5.1.00 for environmental applications for data sets with and without nondetect

observations [online]. Office of Research and Development, US Environmental Protection Agency, Washington, DC. Updated June 20, 2016. Available from: https://www.epa.gov/land-research/proucl-software.

ITRC. 2013. Groundwater statistics and monitoring compliance. Statistical tools for the project life cycle [online]. Interstate Technology & Regulatory Council, Washington, DC. Updated December 2013. Available from: https://www.itrcweb.org/gsmc-1/.

LANL. 2012. Polychlorinated biphenyls in precipitation and stormwater within the Upper Rio Grande Watershed. Los Alamos National Laboratory, Los Alamos, NM.

LANL. 2013. Background metals concentrations and radioactivity in storm water on the Pajarito Plateau, northern New Mexico. Los Alamos National Laboratory, Los Alamos, NM.

Q-Q plot, alldata: multiple distributions?

Step 1.2

SSC/chemical scatter plot: is there a relationship?

Step 2.1

Calculate BTVs

Evaluate SSC-normalized

Evaluate raw data

Step 4 (Figure 4b)

Steps 2.2 & 2.3

Boxplots: are there spatial differences?

(K-W+Dunn test)

Time plots : are there changes over time?(Theil-Sen/LOESS)

If SSC/chemical relationship unclear, does normalization improve stability?

Step 3

Would subset be sufficient for BCF

Assessment after

splitting?

Sufficient data for BCF

Assessment?

Step 1.1

YesUnclear

PossibleNo

Split dataset into multiple subsets

Remove small subpopulation to preserve sample size

Run ProUCL software (UTL, UPL, and USL); calculate percentiles, maximum

Step 4.1

Step 2

Step 3Step 4.2

Q-Q plot, all data: are thereextreme values?

Q-Q plot, all data: are thedistribution assumptionsvalid? If so, how many?

Yes, 1No Yes, >1

Step 5.2

Exclude extreme values and recalculate UTL, UPL,

and USL using ProUCL; compare to initial results

(from Step 4.3)

Step 1

Steps 1-3 (Figure 4a)

Report non-parametric BTVs

Report BTVs for appropriate distribution

Q-Q plot, all data:distribution withclearly better fit?

Step 4.3

Step 5.1

Report BTVs for best-fit distribution

Report more conservative BTVs

Step 5.3

Follow Steps 4.2 and 4.3 for revised dataset and

compare with Step 4 result

State any other

uncertainties

Step 5.4

●●●

●●

●●●

●●

● ●

●●

Theil−Sen, p=0.001Local regressionFire affected periodBandelier ReferenceNorthern ReferenceSEP ReferenceWestern ReferenceDetectNon−detect

Aluminum, UF, UndevelopedSEP and Western Reference

Gray dashed line is 2010 MTAL (750 µg/L), and dotted black lines are the range of watershed-specific (hardness-adjusted) 2015 draft MTALs (274 to 4,122 µg/L). Blue bars are 2018 BTVs, and orange bars are 2013 BTVs (95-95 UTLs). Whiskers and blue bars represent a likely range of back-transformed BTVs calculated by multiplying the SSC-normalized BTVs by the 25th, 50th, and 75th percentile SSC values for the respective landscape type (developed or natural). Back-transformation was used herein to allow for comparison to 2013 BTVs and permit limits.

Gray dashed line is 2010 ATAL (15 pCi/L). Blue bars are 2018 BTVs, and orange bars are 2013 BTVs (95-95 UTLs). Whiskers and blue bars represent a likely range of back-transformed BTVs calculated by multiplying the SSC-normalized BTVs by the 25th, 50th, and 75th percentile SSC values for the respective landscape type (developed or natural). Back-transformation was used herein to allow for comparison to 2013 BTVs and permit limit.

Figure 13. Comparison of 2018 BTVs and existing LANL BTVs for total PCBs

Minor watersheds other than South Fork Acid Northern and Western Reference

Developed Background Baseline

Background Threshold Values

Dashed line is 2010 ATAL (0.64 ng/L) based on human health exposure to fish in the Rio Grande. Blue bars are 2018 BTVs, and orange bars are 2012 BTVs (95-90 UTLs).

10,000

All locations All locations

Developed Background Natural Background

10,000

100,000

1,000,000

All locations Northern andBandelierReference

SEP andWestern Reference

Developed Background Natural Background

!(!( !(

!( !(!(

!(!(!(!(

!(!(!(

")")")

Guaje Canyon

Water Canyon

Sandia Canyon

Pueblo Canyon

Frijoles Canyon

Bayo Canyon

Pajarito Canyon

Chupaderos Canyon

Mortandad Canyon

Ancho Canyon

Garcia Canyon

Los Alamos Canyon

Canada del Buey

Potrillo Canyon

Canada de las Marias

Two-Mile Canyon

Fence Canyon

Cedro Canyon

Canada de las Latas

North Ancho Canyon

Vallecitos Canyon

Agua Prieta Canyon

Quemazon Canyon

Corral Canyon

Bulldog Gulch

Barrancas Canyon

Canon de Valle

Rendija Canyon

Barrancas Canyon

Chaquehui Canyon

Three-Mile Canyon

Cabra C

Indio Canyon

StarmersGulch

106°9'0"W

106°10'30"W

106°12'0"W

106°13'30"W

106°15'0"W

106°16'30"W

106°18'0"W

106°19'30"W

106°21'0"W

106°22'30"W

F o r e s t S e r v i c e

P r i v a t e

V a l l e s C a l d e r a N a t i o n a l P r e s e r v e

269136

106°29'20"W

106°30'40"W

106°32'0"W

106°33'20"W

106°34'40"W

106°36'0"W

106°37'20"W

106°38'40"W

106°40'0"W

White Rock

Los AlamosTownsite

Los Alamos0 10.5 Miles¯New Mexico State Plane Coordinate System Central Zone (3002)North American Datum, 1983 (NAD 83)

US Survey Ft

GIS: Jeffrey LyonCreated: 27-February-2018Last Opened: Friday, April 13, 2018

FIle: Map_Plate2_18-009_SW_Developed_Undeveloped_Background_04112018

DISCLAIMER: This map was created for work processes associated with ER-ES. All other uses for this map should be confirmed with LANL staff.

Project Extent

Lab Developed

Watercourse

Rio Grande

Los AlamosNational Lab

Bandelier ReferenceUndeveloped")

Northern ReferenceUndeveloped")

Paved road

Communities

Town Developed!(

SEP ReferenceUndeveloped")

Western ReferenceUndeveloped")

Figure 1. Map of sampling locations for Los Alamos National Laboratory background water quality datasetNote: “SEP” samples are associated with LANL’s Supplemental Environmental Projects

The Pajarito Plateau lies between the Jemez Mountains and the Rio Grande. The few perennial streams that exist on the plateau are located where effluent or natural springs provide sufficient water volume. Otherwise, streams are ephemeral or intermittent, only flowing in response to seasonal snowmelt or heavy rainfall. Storm water from the plateau infrequently reaches the Rio Grande. Figure 2 provides a conceptual site model for LANL

Introduction

The purpose of this study was to characterize background concentrations in storm water that runs off developed and undeveloped landscapes on the Pajarito Plateau, New Mexico, near Los Alamos National Laboratory (LANL) (Figure 1). LANL’s National Pollutant Discharge Elimination System (NPDES) Individual Permit (IP) target action levels (TALs) are set near or below background concentrations, and it is impracticable for LANL to meet those IP TALs.

Step 1 involved the initial screening out of datasets with n < 10 samples and detection frequencies < 20%. Quantile-quantile (Q-Q) plots (Figure 5) were also used to find possible subpopulations in the dataset for further evaluation.

In Step 2, spatial or temporal differences or correlations with SSC were assessed. The relationship between SSC and chemical concentration was quantified using linear regression (e.g., Figure 6); a significant slope (Student’s t-test, p < 0.05) and reasonably strong linear fit (r2 ≥ 0.5) indicated a dependencythat should be dealt with, whereas a significant but relatively weak slope wasevaluated further (e.g., Figure 7). SSC is a covariate of stream discharge and is auseful proxy for “storm intensity.” Spatial factors included major and minorwatersheds and major and minor location groupings. Differences weredetermined with box plots and nonparametric Kruskal-Wallace and post-hocConover-Inman tests (two-tailed, Bonferroni-corrected familywise alpha = 0.05)(e.g., Figure 8). Temporal trends were evaluated using Theil-Sen medianregression and local regression (LOESS) (e.g., Figure 9).

In Step 3, dependencies were removed, either by splitting the full datasets into smaller subsets of similar concentrations by spatial grouping(s) or by dividing storm water concentrations by sample-specific SSC (normalization). Temporal trends were found to be consistently related to differences in spatial sampling over time; thus, splitting data by spatial groupings also controlled for observed temporal trends. Data subsets and normalized values were re-evaluated using Steps 1 through 3.

storm water and how it relates to hydrology across the plateau, and Figure 3 provides photographic examples of typical streams on the plateau.

The key results of this study were background threshold values (BTVs), which quantify background concentrations in storm water. BTVs were developed using a background characterization framework (BCF), consistent with existing guidance for characterizing groundwater (ITRC 2013) and other media (EPA 2016). The BCF included multiple steps and decision points for developing a sufficient and “stable” or independently and identically distributed (IID) dataset, calculating BTVs, and evaluating uncertainties.

This poster focuses on key constituents of concern for LANL that have elevated background concentrations in undeveloped watersheds: aluminum, gross alpha radiation, and polychlorinated biphenyls (PCBs).

Abstract

If National Pollutant Discharge Elimination System (NPDES)

permit action levels are set near or below background

concentrations, achieving compliance is difficult or impossible

and might serve only to attempt to reduce naturally occurring

constituents. In regard to an NPDES stormwater permit for Los

Alamos National Laboratory (LANL), which is situated on the

arid Pajarito Plateau near Santa Fe, New Mexico, studies show

that aluminum and gross alpha radiation concentrations are

attributable to Bandelier tuff, the major geologic medium in the

area. Meanwhile, atmospheric deposition of polychlorinated

biphenyls (PCBs) contributes to an anthropogenic baseline,

measurable in reference watersheds. As a result, exceedances of

NPDES action levels and state water quality criteria (WQC) for

these constituents are erroneously attributed to LANL

discharges. To address this situation, a framework was

established to generate water quality background threshold

values (BTVs) that characterize natural background (NBG),

anthropogenic baseline, and developed background conditions

using statistical methods.

Water quality data were evaluated for potential spatial and

temporal dependencies, as well as relationships with suspended

sediment concentration (SSC), a parameter positively correlated

with storm water discharge. Data were subsetted or normalized

to SSC to address dependencies prior to calculating BTVs for

18 constituents, resulting in a total of 43 BTVs. A subset of

those BTVs are discussed herein.

Most recommended BTVs were calculated using ProUCL as

95-95 upper tolerance limits (UTLs) based on gamma,

lognormal, normal, or nonparametric methods. The resulting

anthropogenic baseline PCB BTV of 58 ng/L is 90 times higher

than the New Mexico human health WQC (0.64 ng/L). The

NBG gross alpha BTV is 190 pCi/g SSC, which, after back-

transformation using the 25th and 75th percentiles of NBG SSC,

ranges from 170 to 1,900 pCi/L, well above the New Mexico

WQC (15 pCi/L). Two NBG aluminum BTVs of 17 and 76

mg/g SSC were developed (for subareas of Pajarito Plateau),

which, after back-transformation, range from 15 to 1,700 mg/L

and 68 to 780 mg/L, respectively; these values are much higher

than the New Mexico hardness-based WQC (from 0.37 to 1.9

mg/L as total aluminum). Such results suggest that BTVs should

be taken into account before concluding that exceedances of

state WQC and related LANL NPDES permit action levels are

attributable to LANL discharges, since such conclusions can

lead to unwarranted actions such as engineered controls, 303(d)

listings, and developing total maximum daily loads (TMDLs).

In Step 4, BTVs were calculated. Various censored statistics (i.e., UTLs, upper prediction limits [UPLs], and upper simultaneous limits [USLs]) were generated using ProUCL (EPA 2016), and upper percentiles and maxima were calculated using R software. Parametric statistics were preferred, when valid based on goodness-of-fit test results and visual inspection of Q-Q plots (e.g., Figure 10).

In Step 5, uncertainties associated with the data and BTVs were summarized.

Results and Discussion

• Using the BCF, 43 sets of BTVs were developed for 18 constituents. BTVs for unfilteredaluminum, gross alpha, and total PCBs are presented in Figures 11, 12, and 13, along withLANL’s current and draft permit benchmarks (average or maximum target action levels[ATALs or MTALs]).

Figure 8. Box plot used during BCF Step 2.2

Major location grouping (undeveloped)

ab a b abn=9 n=21 n=30 n=9

Figure 7. Q-Q plot used during BCF Step 2.1

●●

●●●

●●

50 500

200000

500000

100 1000 5000 50000

−1 0 1 2

Log−normal Q−Q plotAluminum, UF, Undeveloped, SSC−normalized

Theoretical Quantiles

Bandelier ReferenceNorthern ReferenceSEP ReferenceWestern ReferenceDetectNon−detect

Figure 6. Scatter plot with linear regressionused during BCF Step 2.1

●●

●●●

● ●

●●

● ●

Suspended sediment concentration (mg/L)

Linear modelBandelier ReferenceNorthern ReferenceSEP ReferenceWestern ReferenceDetectNon−detect

p<0.001, r−square=0.616

Figure 5. Q-Q plot used during BCF Step 1.2

●●

●●●●

●●●

●●

−2 −1 0 1 2

Log−normal Q−Q plotAluminum, UF, Undeveloped

Theoretical Quantiles

Bandelier ReferenceNorthern ReferenceSEP ReferenceWestern ReferenceDetectNon−detect

• Few constituents were strongly related to SSC, although unfiltered aluminum and gross alpha radiation are exceptions (e.g., Figure 6).

• BTVs for natural conditions on the Pajarito Plateau exceed LANL’s TALs for aluminum, gross alpha, and total PCBs (Figures 11, 12, and 13).

• For the aluminum concentrations at Supplemental Environmental Projects (SEP) Reference and Western Reference locations (Figure 1), which represent the most spatially relevant undeveloped condition for streams that receive LANL storm water, 95-95 UTLs (based on back-calculation from 76 mg/g SSC) range from 68 to 780 mg/L, approximately two to three orders of magnitude (depending on SSC) greater than the 2010 IP TAL of 0.75 mg/L. This reflects a key problem with any total aluminum standard: Aluminum is the third most abundant element in Earth’s crust and a large component of any natural sediment; therefore, even low SSC will result in noncompliance with total aluminum standards.

• Gross alpha in natural waters of the Pajarito Plateau far exceeds the TAL of 15 pCi/L; estimates (based on back-calculation from SSC-normalized 95-95 UTL of 190 pCi/g SSC) range from 170 to 1,900 pCi/L, approximately one to two orders of magnitude greater than the TAL. There is not a significant spatial difference in gross alpha radiation levels across the Pajarito Plateau, which suggests that the common geology is the dominant source.

• Although PCBs are not part of the “natural” condition (PCBs are strictly man-made), they exist nonetheless in natural Pajarito Plateau waters at levels that exceed permit limits. The95-95 UTL for total PCBs of 58 ng/L is approximately two orders of magnitude greater than the 2010 IP TAL of 0.64 ng/L, which is based on human health (via diet). Previous studies by LANL have traced PCBs to aerial deposition (LANL 2012). Thus, there is an anthropogenic “baseline” level of PCBs that should be considered.

• Aluminum and gross alpha BTVs tended to be low in developed landscapes (Figures 11 and12) relative to BTVs for undeveloped landscapes; this is likely due to the greater amount of sediment measured in undeveloped watersheds relative to developed watersheds. Total PCBs concentrations were very similar between developed and undeveloped landscapes, possibly reflecting a similar regional source (i.e., precipitation).

• BTVs for copper and zinc from developed landscapes (not detailed herein) tended to be greater than both TALs and NBG BTVs, reflecting urban sources. Thus, urban sources should be also considered for storm water constituents when characterizing background.

Next Steps:

• LANL is continuing to collect data as part of its SEP program. Eventually, BTVs may be recalculated using a larger, updated dataset.

• BTVs will be used to evaluate LANL-influenced storm water data for LANL’s forthcoming IP application.

• BTVs may be proposed as site-specific standards or as replacements for TALs.

• LANL will work with the US Environmental Protection Agency (EPA) Region 6 and NMED on options for the use of PCB BTVs in the permitting process, given that PCBs are not naturally occurring.

• LANL will establish which statistic is acceptable to both regulators and LANL for defining background conditions (i.e., protective of aquatic life without requiring unnecessary effort to control background conditions potentially unrelated to historical LANL activities).

Figure 1. Sampling location map for background storm water datasets

−20 0 20 40 60

Normal(invalid)

Theoretical quantiles

● ● ● ● ● ●●●●●●●●●●

●●●

●●●●●●●

●●●●●

●●●●

● ●

1 2 3 4

Lognormal (valid)

●●

● ●●

●●●

●●●●●

●●●

●●●●

●●

●●●●●

●●●●

● ●●

0 20 40 60 80

Gamma (valid)

●●●●●●●●

●●●●●●●

●●●●●●●●

●●

●●●●●● ●

● ●● ●

2006 2008 2010 2012 2014 2016 2018

Characterizing water quality background …...Characterizing water quality background concentrations...

Documents