Total Chromium and Hexavalent Chromium Occurrence Analysis

Subject Area: Water Quality

Web Report #4414

Total Chromium and Hexavalent Chromium Occurrence Analysis

Total Chromium and Hexavalent Chromium Occurrence Analysis

©2012 Water Research Foundation. ALL RIGHTS RESERVED.

About the Water Research Foundation The Water Research Foundation is a member-supported, international, 501(c)3 nonprofit organization that sponsors research that enables water utilities, public health agencies, and other professionals to provide safe and affordable drinking water to consumers. The Foundation’s mission is to advance the science of water to improve the quality of life. To achieve this mission, the Foundation sponsors studies on all aspects of drinking water, including resources, treatment, and distribution. Nearly 1,000 water utilities, consulting firms, and manufacturers in North America and abroad contribute subscription payments to support the Foundation’s work. Additional funding comes from collaborative partnerships with other national and international organizations and the U.S. federal government, allowing for resources to be leveraged, expertise to be shared, and broad-based knowledge to be developed and disseminated. From its headquarters in Denver, Colorado, the Foundation’s staff directs and supports the efforts of more than 800 volunteers who serve on the board of trustees and various committees. These volunteers represent many facets of the water industry, and contribute their expertise to select and monitor research studies that benefit the entire drinking water community. Research results are disseminated through a number of channels, including reports, the Website, Webcasts, workshops, and periodicals. The Foundation serves as a cooperative program providing subscribers the opportunity to pool their resources and build upon each others’ expertise. By applying Foundation research findings, subscribers can save substantial costs and stay on the leading edge of drinking water science and technology. Since its inception, the Foundation has supplied the water community with more than $460 million in applied research value. More information about the Foundation and how to become a subscriber is available at www.WaterRF.org.


Total Chromium and Hexavalent Chromium Occurrence Analysis Prepared by: Chad J. Seidel and Christopher J. Corwin Jacobs Engineering, Denver, CO 80202

Rajiv Khera U.S. Environmental Protection Agency Office of Ground Water and Drinking Water, Washington, DC 20460 Sponsored by: Water Research Foundation 6666 West Quincy Avenue Denver, CO 80235 Published by the


DISCLAIMER

This study was funded by the Water Research Foundation (Foundation). The Foundation assumes no responsibility for the content of the research study reported in this publication or for the opinions or statements of fact expressed in the report. The mention of trade names for commercial products does not represent or imply the approval or endorsement of the

Foundation. This report is presented solely for informational purposes.

Copyright © 2012 by Water Research Foundation

ALL RIGHTS RESERVED.

No part of this publication may be copied, reproduced or otherwise utilized without permission.

Printed in the U.S.A.


v

CONTENTS 1. Introduction ............................................................................................................................. 1

2. Description of Data ................................................................................................................. 1

2.1. National Chromium and Boron Occurrence Survey (NCBOS) ....................................... 2 2.2. EPA’s Second Six Year Review ICR Contaminants Monitoring Data (6YR2) .............. 2 2.3. CDPH Data ....................................................................................................................... 6

3. Analysis of Total and Hexavalent Chromium Occurrence ..................................................... 8

3.1. Comparison of Chromium Occurrence between Datasets ............................................... 8 3.2. Analysis of Chromium Occurrence by Source Type, Region, and System Size ........... 10 3.3. Analysis of Paired Total and Hexavalent Chromium Samples ...................................... 14

4. Spatial Distribution of Chromium Occurrence ..................................................................... 16

5. Co-occurrence with Chromium ............................................................................................ 19

6. Estimated Extent of Potential Future Regulation ................................................................. 20

7. Conclusions ........................................................................................................................... 23 8. References ............................................................................................................................. 24


vi

LIST OF TABLES

1. Summary of occurrence data sources ..................................................................................1 2. Summary of non-detect results in the 6YR2 dataset ............................................................2 3. Summary of non-detect results in the CDPH dataset ...........................................................7 4. Percent of entry points with 75th percentile total chromium concentration above the

threshold values (6YR2 data, non-detects set to one-quarter of the MRL) ...........18 5. Correlation of co-occurring contaminants with total chromium in the 6YR2 dataset ..........................................................................................................20 6. Estimation of the national CWS entry point inventory ......................................................21 7. Number of CWS entry points above the threshold concentration at various threshold concentrations ........................................................................................22


vii

LIST OF FIGURES 1. Analysis of non-detect MRLs of the 6YR2 dataset (all non-detect samples). .....................3 2. Cumulative probability distribution of median total chromium concentration by entry point of the 6YR2 dataset with non-detects set to zero, one quarter of the MRL, half the MRL, and at the MRL ......................................................................4 3. Cumulative probability distribution of median total chromium concentration by entry point of the 6YR2 dataset with non-detects set to zero, and one quarter of the MRL, compared to the NCBOS dataset and the Bayesian estimate ..............5 4. Cumulative probability distribution of total chromium concentration by entry point of the 6YR2 dataset (non-detects set to one-quarter of the MRL) .................6 5. Cumulative probability distribution of chromium in CDPH for raw versus treated by sample location (75th percentile concentration per each sample

location) ...................................................................................................................8 6. Cumulative probability distribution of total chromium data from the three datasets (6YR2 and CDPH are the 75th percentile concentration per sample location) ........9 7. Cumulative probability distribution of hexavalent chromium NCBOS & CDPH raw data (CDPH shows the 75th percentile concentration per sample location, non-detects set to zero) ..........................................................................................10 8. Cumulative probability distributions of total chromium in the 6YR2 dataset by source water type (75th percentile concentration per sample location, non-detects set to one-quarter of the MRL) ...........................................................11 9. Cumulative probability distributions of CDPH hexavalent chromium data in ground and surface waters divided into raw and treated samples (75th percentile concentration per sample location, non-detects set to zero) .................................12 10. Map of EPA Regions .........................................................................................................13 11. Cumulative probability distribution of 6YR2 total chromium data by EPA region (75th percentile concentration per sample location, non-detects set to one-quarter of the MRL) ........................................................................................13 12. Cumulative probability distributions of 6YR2 total chromium data by system size category (75th percentile concentration per sample location, non-detects set to one-quarter of the MRL) ..............................................................................14 13. CDPH paired sample analysis of ground waters and surface waters .................................15 14. CDPH paired sample analysis of ground waters grouped by raw and treated waters .......15 15. CDPH paired sample analysis of surface waters grouped by raw and treated waters .......16 16. National pin map of 75th percentile total chromium concentration from the 6YR2 dataset by system (for systems with more than two samples) ...............................17 17. California pin map of 75th percentile hexavalent chromium concentration from the CDPH database by system ...............................................................................19 18. Estimation of total number of CWS entry points with total chromium above threshold concentrations ........................................................................................23


1

1. Introduction

The purpose of this study is to update the chromium occurrence analysis originally presented in the Water Research Foundation (WaterRF) Report #2759, “Occurrence Survey of Hexavalent Chromium and Boron,” (Frey et al., 2004) with new data from the U.S. Environmental Protection Agency (EPA) and the State of California. New data were obtained from two primary sources: the California Department of Public Health (CDPH) Water Quality Analysis Database, and the EPA’s second six year review database. While these occurrence data are already available, the objective of this assessment was to identify general occurrence trends of chromium and capture regional occurrence patterns and assess the geographical coverage and extent of impact of a potential future revision of total chromium or a new hexavalent chromium regulation on the nation’s water utilities.

2. Description of Data

Chromium occurrence information was gathered from available sources and the previous WaterRF survey to guide the development of this updated national occurrence survey. Three primary datasets are used in this analysis: WaterRF Report #2759, EPA’s second six year review, and CDPH Water Quality Analysis Database (WQAD). A summary of the datasets is given in Table 1. Other data sources considered but found not to be useful were: USGS National Water Information System, National Water-Quality Assessment, and Water Quality Monitoring Networks databases, US EPA STORET and voluntary water utility monitoring data.

Table 1. Summary of occurrence data sources. Data Source WaterRF #2759 USEPA CDPH Database NCBOS 6YR2 WQAD

Geographic Coverage

41 States 46 States and 2 Tribes California

Period of Record 2003 1998 - 2005 1995 - 2011 Source Water Sampled

Raw Drinking Water Treated Drinking Water Raw and Treated Drinking

Water

Total Chromium Data

MRL=0.6 µg/L n=342, 39%<MRL

MRL varies from 0.02 -100 µg/L

n=162,823, 84%<MRL

MRL varies from 1 -50 µg/L

n=108,782, 49%<MRL

Hexavalent Chromium Data

MRL=0.2 µg/L n=341, 57%<MRL

None MRL=1 µg/L

or less n=27,507, 13%<MRL

Strengths Low and consistent

MRL Number and geographic distribution of samples

Number of samples for total and hex chromium

Weaknesses Number and

geographic distribution of samples

Some high MRL data, no hexavalent chromium

data

Some high MRL data, limited geographical

distribution

MRL = minimum reporting level


2

2.1. National Chromium and Boron Occurrence Survey (NCBOS)

The National Chromium and Boron Occurrence Survey (NCBOS) was compiled, analyzed, and reported in WaterRF Report #2759 (Frey et al., 2004). The survey utilities were recruited on a voluntary basis, rather than a designed survey to adequately represent regional, system size, and source water occurrence. Field samples were collected in 2003 and transported to a central laboratory where the analysis of total and hexavalent chromium was performed. Reporting limits were 0.6 µg/L for total chromium and 0.2 µg/L for hexavalent chromium. A total of 342 total chromium and 341 hexavalent chromium raw water samples were reported. The samples were taken at the same relative location for each of 342 different sources, from 189 different systems in 41 states. Samples from the treatment system profiles included with the NCBOS were not included in this analysis.

2.2. EPA’s Second Six Year Review ICR Contaminants Monitoring Data (6YR2)

The 6YR2 dataset consists of nationwide entry point samples for total chromium. The EPA collected the 6YR2 data by a voluntary call-in of compliance data from state and primacy agencies for samples collected between 1998 and 2005 (EPA, 2010). The samples were analyzed by many different labs and methods resulting in reporting levels from 0.02 – 100 µg/L. A total of 162,823 samples were included in the dataset after applying a data quality auditing procedure. The EPA’s data auditing involved standardizing units of concentration for the contaminants and ensuring that parameters such as the system type were consistent with data in federal safe drinking water information system (SDWIS). Systems classified as a groundwater include only ground water sources. However, systems classified as surface waters may include groundwater and groundwater under the direct influence (GWUDI) of surface water among its sources. Some of the data submissions included source water samples. Where possible the source water data points were removed by the EPA. Also, each sampling point was inferred to be an entry point to the distribution system.

Table 2 summarizes the non-detections in the 6YR2 dataset. The results were calculated considering all the samples in the dataset, then grouping the data by entry point. Thus, a non-detect result at an entry point means that chromium was never detected in that source.

Table 2. Summary of non-detect results in the 6YR2 dataset. Total count Non-detect count Percent non-detects All Samples - Cr 162,823 136,826 84% Entry Point - Cr 74,672 58,010 78%

Of the 162,823 samples in the 6YR2 database, 136,826 were reported as non-detects with an associated minimum reporting level (MRL). During the EPA’s data auditing procedure, non-detects that did not include a MRL were assigned the mode MRL value for that state where other data in the state were available (EPA, 2010). If no other MRL value was available for that state, the national mode MRL value was substituted. Figure 1 shows the distribution of the non-detect results from the 6YR2 dataset. Figure 1 indicates 47% of non-detect results have a MRL of 10 µg/L, which many of these presumably resulting from samples that did not include an MRL. Another 15% of the non-detects have a MRL of 5 µg/L.


3

Figure 1. Analysis of non-detect MRLs of the 6YR2 dataset (all non-detect samples). Panel A shows the histogram of the non-detect MRLs. Panel B shows the cumulative probability distribution of non-detect MRL concentrations.

Non-detect samples can have a profound effect on the analysis of the data. Because the concentration of a non-detect sample can be any value from zero to the MRL, cumulative probability distributions of the dataset were computed under four conditions to characterize the effect of the non-detect results:

1. zero was substituted for a non-detect result,

2. one-quarter of the MRL was substituted for the non-detect result,

3. half of the MRL was substituted for the non-detect result, and

4. the MRL was substituted for the non-detect result.

Substituting zero for the non-detects results in bias towards lower concentrations because some of the non-detects would have had detectable chromium if it were measured by a method with a lower MRL. Similarly, substituting the MRL for a non-detect results in bias towards higher concentrations because the actual sample concentrations were likely lower than the MRL in most cases. Substituting one-quarter and one-half of the MRL for the non-detect attempts to balance these biases, but the bias is unknown. See Raucher et al. (1995) for more details of non-detect handling.

To create the cumulative probability distributions, the samples were grouped by entry point to analyze the results on a drinking water source basis (all water source types for both community water systems (CWS) and non-transient, non-community water systems (NTNCWS)). Next, the median, 75th percentile, 95th percentile, and maximum value concentrations were calculated for each of the 74,672 entry points. Figure 2 presents the cumulative probability curves of median total chromium concentration for the 6YR2 data by the four non-detect substitution conditions. The median value varies from zero to 10 µg/L as

0%

20%

40%

60%

80%

100%

0 10 20

Cumulative

Probability

Total Chromium MRL (µg/L)

0

10,000

20,000

30,000

40,000

50,000

60,000

70,000

0 10 20

Number of N

on‐Detects in 6YR

2

Total Chromium MRL (µg/L)

A) B)


4

influenced by the non-detects with an MRL of 10 µg/L, again many of which would have been assigned a value of 10 µg/L because no MRL was reported with the data. The vertical portions in each curve represent a large number of non-detect results at a particular MRL. For instance, the vertical lines from about the 40th to 80th percentile are indicative of the non-detects with an MRL of 10 µg/L.

Figure 2. Cumulative probability distribution of median total chromium concentration by entry point of the 6YR2 dataset with non-detects set to zero, one quarter of the MRL, half the MRL, and at the MRL.

Another method to analyze the impact of the non-detect results is a Bayesian estimate of the

posterior cumulative probability. A Bayesian analysis of the non-detects from total chromium concentrations between 0.5 and 15 µg/L was prepared using the WinBUGS software at the mean concentration of each entry point location. The Bayesian analysis uses the detect data and a defined distribution to estimate the cumulative probability curve.

Figure 3 shows the cumulative probability of the 6YR2 dataset with non-detects substituted with zero and one-quarter of the MRL with the cumulative distributions of the NCBOS dataset and the Bayesian estimate. The NCBOS and Bayesian curves show similar results. This result is somewhat unexpected as the NCBOS dataset was based upon voluntary participation of water utilities that were willing to sample their source waters and therefore considered to be biased towards utilities where chromium was suspected or known to occur. NCBOS was generated on a sample-by-sample basis because only one sample was taken per sample location, and were raw water samples. But notably, the NCBOS data shows the smooth, nearly log-normal curve to the

0%

20%

40%

60%

80%

100%

0 10 20 30 40

Cumulative

Probability

Total Chromium (µg/L)

n.d.=0

n.d.=0.25∙MRL

n.d.=0.5∙MRL

n.d.=MRL


5

origin resulting from the low reporting limits used in the study. Raucher et al. (1995) observed that occurrence data often display a log-normal behavior. Substitution of one-quarter of the MRL in the 6YR2 data is close to the NCBOS and EPA Bayesian estimates while still being conservative, particularly above 5 µg/L. Therefore, substitution of one-quarter of the MRL for non-detects was used for all further analyses of the 6YR2 data, except where noted. Continuing analyses with the Bayesian approach are not practical because the non-detect values are not re-assigned, thus each data sub-set would require its own Bayesian analysis.

Figure 3. Cumulative probability distribution of median total chromium concentration by

entry point of the 6YR2 dataset with non-detects set to zero, and one quarter of the MRL, compared to the NCBOS dataset and the Bayesian estimate.

Figure 4 presents the cumulative probabilities of total chromium from the 6YR2 data by the median (50th percentile), 75th percentile, 95th percentile, and maximum concentration at each entry point. Using percentiles to represent the data lessens the impact of non-detect handling over averaging; therefore, the average at each entry point was not included in this analysis. The 75th percentile, 95th percentile, and maximum value yield very similar curves, with the 95th percentile and the max value curves being indistinguishable. The median value was eliminated from consideration because it may not be inclusive enough of systems with variable chromium concentrations. For instance, a utility that has source water with a variable chromium concentration, may have a median concentration below a threshold; however, if a regulation were in place at that threshold the utility would be in violation at times when the chromium concentration was above the median concentration. The similarity of the 75th, 95th, and

0%

25%

50%

75%

100%

0 5 10 15 20

Cumulative

Probability


6YR2 (n.d.=0)

NCBOS

EPA Bayesian Estimate

6YR2 (n.d.=0.25∙MRL)


6

maximum curves occurs because the sample size of most entry points is low (162,823 samples divided by 74,672 entry points equals an average of 2.2 samples per entry point). The 75th percentile has been chosen to be consistent with the USGS analysis of arsenic national occurrence data (USGS, 2012), and its similarity to the 95th percentile results.

Figure 4. Cumulative probability distribution of total chromium concentration by entry point of the 6YR2 dataset (non-detects set to one-quarter of the MRL).

2.3. CDPH Data

The California Department of Public Health (CDPH) maintains a water quality analysis database, which includes total and hexavalent chromium in the state of California collected generally between 1995 and 2011. Data are representative of raw and treated waters from throughout the state. The database included 136,286 results, 108,779 for total chromium and 27,507 for hexavalent chromium as of November 18, 2011. Because CDPH provided data for the USEPA 6YR2 database, there is an overlap of up to 23,397 total chromium samples between the CDPH data and 6YR2 datasets for California (22% of the CDPH database, 14% of the 6YR2 database). CDPH data came classified as raw, treated, and unknown for treatment type. Samples also came classified as groundwater, surface water, or non-source water.

Table 3 summarizes the CDPH data and shows a similar percentage of non-detect results on a sample basis, but a lower percentage of non-detects on a source location basis. The lower percentage of non-detects on a source location basis can be attributed to two factors. First, the CDPH data is weighted on more recent data with lower detection limits. Second, from a spatial

0%

25%

50%

75%

100%

0 5 10 15 20

Cumulative

Probability


median

75th percentile

95th percentile & max


7

distribution perspective, California has some of the highest levels of chromium (demonstrated in Section 3).

Table 3. Summary of non-detect results in the CDPH dataset. Total count Non-detect

count Percent non-

detects All Samples – Total Cr 108,779 90,560 83% All Samples - Cr(VI) 27,507 11,387 41% Sampling Location – Total Cr 17,846 11,104 62% Sampling Location - Cr(VI) 7,880 3,096 39%

The CDPH total chromium dataset includes 35,968 non-detect samples results at an MRL of 10 µg/L (29% of all non-detects), and another 40,255 had no reported MRLs and were therefore set to the mode value of the CDPH dataset of 10 µg/L (bringing the total to 73% of all non-detects). Thus, the CDPH total chromium dataset is similarly influenced by the non-detect results as the 6YR2 dataset. For comparable analysis with the 6YR2 data, the CDPH chromium data have been analyzed with non-detects substituted with one-quarter of the MRL and missing MRL values set to the mode value of 10 µg/L. The MRL for the CDPH hexavalent chromium is generally 1 µg/L or less. The lower MRL adds very little bias to the analysis of hexavalent chromium in the range of 1 to 50 µg/L of interest in this study. Therefore, non-detect values were set to zero for the CDPH dataset for hexavalent chromium.

Before further analysis can be made, it must be determined whether distribution of the data differs between raw and treated waters. Figure 5 plots the cumulative probability for raw and treated waters for both total and hexavalent chromium. The analysis was performed with the 75th percentile concentration on a sample location basis.

The data shown in Figure 5 indicate more occurrence of total chromium than hexavalent chromium. Higher total chromium is expected because hexavalent chromium is only a portion of the total chromium measurement. Treated water data in Figure 5 show a lower occurrence of both total and hexavalent chromium compared to raw waters. This could be due to treatment removing some chromium, but a more likely explanation is that sources with high concentrations of chromium are not used for drinking water except where they cannot be avoided. Also, the disparity in the number of samples (treated waters only account for about 5% of the data) could also result in some selection bias. Because the distributions of raw and treated water differ, they will be analyzed separately throughout the remainder of this report.


8

Figure 5. Cumulative probability distribution of chromium in CDPH for raw versus treated by sample location (75th percentile concentration per each sample location).

3. Analysis of Total and Hexavalent Chromium Occurrence

3.1. Comparison of Chromium Occurrence between Datasets

The NCBOS, 6YR2, and CDPH datasets all include total chromium. However, the datasets differ in the number of samples and sample type (raw versus treated). The 6YR2 and CDPH data were analyzed with various MRLs. CDPH data include raw and treated waters, where the NCBOS contain only raw, and 6YR2 contain only data from treated drinking waters. The datasets are compared by plotting the cumulative probability function of each in Figure 6. The cumulative probabilities were calculated from the 75th percentile concentration at each sample location.

The data in Figure 6 indicates the CDPH raw waters exhibited the most occurrence and the NCBOS exhibited the least occurrence. The 6YR2 and CDPH treated waters showed roughly equivalent occurrence. The difference observed between these two curves is due to the 6YR2 dataset having a proportionately higher number of non-detects with an MRL of 20 µg/L (the vertical step in the black line at 5 µg/L).

0%

25%

50%

75%

100%

0 5 10 15 20

Cumulative

Probability

Concentration (µg/L)

Total Cr Raw (n=14,735)

Total Cr Treated (n=1,003)

Hex. Cr Raw (n=6,547)

Hex. Cr Treated (n=419)


9

Figure 6. Cumulative probability distribution of total chromium data from the three datasets (6YR2 and CDPH are the 75th percentile concentration per sample location).

Figure 7 compares the cumulative probability distributions of hexavalent chromium from the CDPH raw waters and the NCBOS dataset. The analysis was performed with the 75th percentile concentration of hexavalent chromium on a sample location basis. The California data suggested higher occurrence than nationally, as expected.

0%

25%

50%

75%

100%

0 5 10 15 20

Cumulative

Probability


NCBOS (n=342)

CDPH Treated (n=1,003)

6YR2 (n=74,672)

CDPH Raw (n=14,735)


10

Figure 7. Cumulative probability distribution of hexavalent chromium NCBOS & CDPH raw data (CDPH shows the 75th percentile concentration per sample location, non-detects set to zero).

3.2. Analysis of Chromium Occurrence by Source Type, Region, and System Size

Figure 8 displays the cumulative probability distributions of the 6YR2 total chromium groundwater data versus surface waters. The concentrations were calculated from the 75th

percentile at each sample location and non-detects were substituted with one-quarter of the MRL. Results of the analysis indicate that distributions are similar, with a maximum of about 2% difference between them. As noted previously, some groundwater sources may be labeled as surface waters if they have any surface water, or GWUDI sources in the same system. Frey et al. (2004) also found no significant difference in distribution of total chromium for ground waters versus surface waters. Thus, the analysis of total chromium data will not be separated by source water type. A similar analysis was performed for water system type. No observable difference was found between CWS and NTNCWS. Therefore, all data has been included in the analysis (rather than limit the data to CWS data only).

0%

25%

50%

75%

100%

0 5 10 15 20

Cumulative

Probability

Hexavalent Chromium (µg/L)

NCBOS (n=341)

CDPH Raw (n=6,547)


11

Figure 8. Cumulative probability distributions of total chromium in the 6YR2 dataset by source water type (75th percentile concentration per sample location, non-detects set to one-quarter of the MRL).

Figure 9 shows the cumulative probability distributions for hexavalent chromium by ground and surface water from the CDPH database. The occurrence in ground waters is higher than for surface waters. Frey et al. (2004) found similar results. Figure 9 also separates the raw waters from the treated waters. Not much difference in distribution is observed in the surface water data as occurrence was only observed at low levels. The groundwater data; however, indicates higher occurrences in the treated waters below 6 µg/L and higher occurrences in raw waters above 6 µg/L. An increase in hexavalent chromium could result from oxidation of trivalent chromium during disinfection. A decrease in hexavalent chromium could be due to treatment provided at some sites. However, the explanation for the crossover in Figure 9 cannot be determined with any certainty, and it could also be due to the disparity in the number of samples (1,242 treated samples vs. 14,448 raw samples).

0%

25%

50%

75%

100%

0 5 10 15 20

Cumulative Probability


Ground Water (n=67,269)

Surface Water (n=7,403)


12

Figure 9. Cumulative probability distributions of CDPH hexavalent chromium data in ground and surface waters divided into raw and treated samples (75th percentile concentration per sample location, non-detects set to zero).

The regional distribution of total chromium was investigated using EPA regions (shown in

Figure 10). Figure 11 presents the cumulative probability distributions for total chromium grouped by EPA region. Some variability based on region was observed, but the uncertainty surrounding the non-detect data complicates the analysis. Non-detect handling complicates analysis of Figure 11; however, it can be seen that Region 9 (California, Nevada, and Arizona) show the highest concentration occurrences. The vertical steps in the curves represent the re-occurring non-detect values. The Region 9 curve shows the highest concentrations towards the end of the curves where the influence of non-detects no longer influences the analysis.

0%

25%

50%

75%

100%

0 5 10 15 20

Cumulative

Probability

Hexavalent Chromium (µg/L)

GW Raw (n=14,448)

GW Treated (n=1,242)

SW Raw (n=628)

SW Treated (n=370)


13

Figure 10. Map of EPA Regions.

Figure 11. Cumulative probability distribution of 6YR2 total chromium data by EPA region (75th percentile concentration per sample location, non-detects set to one-quarter of the MRL).

0%

25%

50%

75%

100%

0 5 10 15 20

Cumulative Probability


Region 1 (n=6,677)Region 2 (n=4,808)Region 3 (n=4,601)Region 4 (n=10,143)Region 5 (n=14,680)Region 6 (n=8,916)Region 7 (n=5,703)Region 8 (n=5,367)Region 9 (n=10,996)Region 10 (n=2,781)


14

System size was investigated in the same manner as EPA region. Figure 12 presents the cumulative distribution curves for the 6YR2 data by system size category for all systems included in the 6YR2 dataset (CWS and NTNCWS). Non-detect handling makes interpretation of the 6YR2 data based on system size capricious. For instance, the higher concentration distribution curve for systems serving a population of 1,000,000 results from a significant number of samples analyzed at an MRL of 50 µg/L.

Figure 12. Cumulative probability distributions of 6YR2 total chromium data by system size category (75th percentile concentration per sample location, non-detects set to one-quarter of the MRL).

3.3. Analysis of Paired Total and Hexavalent Chromium Samples

The CDPH data were also used to study the speciation of chromium. A total of 5,976 samples with both total and hexavalent chromium results were paired by matching sample location and date. Figures 12-14 display the results by plotting hexavalent chromium versus total chromium by ranges from 0-100 µg/L and 0-10 µg/L by water type and raw versus treated. In Figures 12-14 there are a number of samples that indicate a higher hexavalent chromium concentration than total chromium, which is not truly possible. These points result from analytical issues that are well described and explained by Parks et al. (2004).

0%

25%

50%

75%

100%

0 5 10 15 20

Cumulative

Probability


25‐100 (n=18,566)

101‐500 (n=21,909)

501‐1,000 (n=7,423)

1,001‐3,300 (n=9,397)

3,301‐10,000 (n=6,906)

10,001‐50,000 (n=6,497)

50,001‐100,000 (n=1,631)

100,001‐1,000,000 (n=2,053)

>1,000,000 (n=290)

Population Range (count)


15

Analysis of the ground versus surface water paired samples shown in Figure 13 indicates surface waters have generally more trivalent chromium than hexavalent chromium. The ground waters tend to have much of their total chromium as hexavalent chromium, consistent with Frey et al. (2004). Despite some scatter in the data the hexavalent chromium concentration is highly correlated with the total chromium concentration (r=0.98).

Figure 13. CDPH paired sample analysis of ground waters and surface waters.

Figure 14 displays groundwater paired samples similarly divided into raw and treated categories. The data in Figure 13 also indicate treated waters have lower concentrations of chromium. This may be due to treatment, but could also be due to sampling bias created by high chromium sources not being used for drinking water. The disparity in sample sizes between raw and treated waters may also contribute to bias.

The treated samples visually appear to be more heavily speciated towards hexavalent chromium, but this is not supported by statistical analysis. Statistically, the central tendency of the raw waters is more heavily speciated in hexavalent chromium. However, there is not a large difference between the raw and treated waters.

Figure 14. CDPH paired sample analysis of ground waters grouped by raw and treated waters.

0

10

20

30

40

50

60

70

80

90

100

0 10 20 30 40 50 60 70 80 90 100

Hexavalen

t Chromium (µg/L)


GW (n=4,809)

SW (n=1,167)

0

1

2

3

4

5

6

7

8

9

10

0 1 2 3 4 5 6 7 8 9 10Hexavalen

t Chromium (µg/L)


GW (n=3,434)

SW (n=1,147)

0102030405060708090100

0 10 20 30 40 50 60 70 80 90 100

Hexavalent Chromium (µ

g/L)


GW Raw (n=3,943)

GW Treated (n=646)

012345678910

0 1 2 3 4 5 6 7 8 9 10

Hexavalen

t Chromium (µg/L)


GW Raw (n=2,798)

GW Treated (n=473)


16

Figure 15 indicates surface waters tended to have little hexavalent chromium in raw or treated waters.

Figure 15. CDPH paired sample analysis of surface waters grouped by raw and treated waters.

4. Spatial Distribution of Chromium Occurrence

The 6YR2 dataset was used to calculate the 75th percentile concentration of total chromium at each entry point. The highest 75th percentile entry point concentration per system was mapped by latitude and longitude identifiers assigned to each system from SDWIS data. No data were available in the states of Washington, Pennsylvania, Louisiana, or Mississippi. Figure 16 shows a national pin map of total chromium concentrations between 5 and 10 µg/L, 10 and 20 µg/L, and above 20 µg/L. The figure also shows how total chromium is distributed across the country. The greatest number of high chromium concentrations appears to be in the southwest US from California to Texas. The region surrounding the Great Lakes also appears to have widespread occurrence.

0102030405060708090100

0 10 20 30 40 50 60 70 80 90 100

Hexavalent Chromium (µ

g/L)


SW Raw (n=676)

SW Treated (n=473)

012345678910

0 1 2 3 4 5 6 7 8 9 10Hexavalent Chromium (µg/L)


SW Raw (n=661)

SW Treated (n=468)


Figure 1

Table

The 6YRpercent oµg/L. TMuch ofdetectionTable 4 i

6. Natio6YR2

e 4 providesR2 (with noof entry poin

The results if this is duns out of 2,3indicates ess

nal pin ma2 dataset by

s a tabular anon-detects sents that excindicate thatue to non-de58 samples, entially all e

ap of 75th pesystem (for

nalysis of thet to one-quceed total cht the percenetect handlinbut because

entry points

17

ercentile tor systems wi

he spatial disuarter of thehromium trent of systemng. For in

e the MRLs exceed 5 µg

otal chromiuith more th

stribution ofe MRL) dateatment goal

ms impacted nstance Geoof the Georg

g/L.

um concentan two sam

f total chromta was usedls of 1, 2, 5rises steeply

orgia samplegia data wer

tration frommples).

mium occurrd to calculat5, 10, 20, any below 5 µes only hadre relatively

m the

rence. te the nd 50 µg/L.

d two high,


18

Table 4. Percent of entry points with 75th percentile total chromium concentration above the threshold values (6YR2 data, non-detects set to one-quarter of the MRL).

Threshold 1 µg/L 2 µg/L 5 µg/L 10 µg/L 20 µg/L 50 µg/L All EPs 84.3% 69.0% 21.6% 7.6% 1.7% 0.2% Alaska 24.0% 17.9% 7.7% 1.3% 0.4% 0.2% Alabama 100.0% 99.8% 2.0% 1.0% 0.0% 0.0% Arkansas 100.0% 99.4% 0.2% 0.2% 0.0% 0.0% Arizona 95.1% 85.6% 55.9% 29.1% 12.3% 1.2% California 94.6% 92.4% 15.2% 10.6% 3.2% 0.2% Colorado 66.9% 56.8% 41.1% 9.5% 2.5% 0.2% Connecticut 82.4% 74.5% 16.7% 12.4% 1.9% 1.1% Delaware 53.7% 34.3% 7.8% 1.0% 0.2% 0.0% Florida 90.6% 84.6% 12.1% 6.6% 1.8% 0.1% Georgia 100.0% 99.9% 99.9% 0.1% 0.1% 0.0% Hawaii 34.8% 34.2% 8.9% 2.3% 0.0% 0.0% Iowa 97.8% 66.3% 8.7% 1.2% 0.8% 0.2% Idaho 21.0% 20.0% 10.9% 4.6% 0.5% 0.1% Illinois 99.8% 10.9% 7.6% 2.9% 0.7% 0.1% Indiana 97.3% 42.3% 28.9% 16.8% 7.0% 0.2% Kansas 89.5% 62.8% 16.4% 2.1% 0.5% 0.1% Kentucky 26.5% 19.3% 3.0% 2.7% 0.4% 0.0% Massachusetts 85.6% 46.4% 10.0% 6.0% 0.1% 0.0% Maryland 96.5% 94.1% 72.9% 66.1% 2.8% 0.3% Maine 29.4% 11.4% 1.3% 0.1% 0.1% 0.0% Michigan 100.0% 93.3% 9.2% 7.6% 1.1% 0.1% Minnesota 100.0% 100.0% 0.1% 0.1% 0.1% 0.0% Missouri 12.9% 7.0% 0.4% 0.1% 0.0% 0.0% Montana 84.6% 83.4% 5.2% 1.4% 0.4% 0.1% North Carolina 99.2% 93.7% 87.3% 1.4% 0.6% 0.1% North Dakota 66.9% 52.0% 30.6% 12.9% 2.4% 0.4% Nebraska 62.1% 46.5% 18.4% 4.6% 0.0% 0.0% New Hampshire 100.0% 99.9% 1.3% 0.5% 0.3% 0.1% New Jersey 78.6% 62.9% 9.9% 3.2% 0.8% 0.2% New Mexico 71.2% 60.9% 28.6% 9.7% 1.2% 0.1% Nevada 95.2% 63.2% 34.7% 16.7% 4.4% 0.6% New York 94.4% 14.6% 8.4% 6.1% 0.2% 0.0% Ohio 87.0% 52.7% 8.5% 6.9% 1.1% 0.0% Oklahoma 90.7% 89.8% 13.0% 12.7% 8.6% 2.8% Oregon 11.0% 5.3% 1.7% 0.6% 0.2% 0.0% Rhode Island 100.0% 100.0% 17.6% 5.3% 0.0% 0.0% South Carolina 100.0% 100.0% 1.8% 1.7% 0.9% 0.6% South Dakota 97.8% 90.7% 18.3% 6.5% 0.6% 0.0% Tennessee 91.0% 86.4% 2.7% 2.2% 0.3% 0.3% Texas 97.8% 96.6% 22.3% 6.9% 1.9% 0.0% Utah 99.1% 25.7% 17.4% 6.5% 1.8% 0.2% Virginia 99.0% 96.8% 2.4% 1.6% 0.2% 0.0% Vermont 90.6% 77.0% 3.9% 2.6% 2.1% 0.6% Wisconsin 37.4% 19.0% 8.7% 4.9% 2.3% 0.3% West Virginia 98.7% 75.3% 3.8% 1.5% 0.4% 0.0% Wyoming 99.6% 98.0% 73.8% 72.3% 0.4% 0.2%


Figure 17lower det10 - 20 µacross th

Figure 1

5. Co-o

The included total chrowith the meaning variationnaturallyhowever,between results achromium6YR2 da

7 shows the tection limit

µg/L and aboe state.

7. Califofrom

occurrenc

6YR2 databin the 6YR2

omium and thighest corrthat only 5

n in total chroy occurring, , the relationtetrachloroe

re consistenm, except beataset and co

CDPH occut for hexavalove 20 µg/L

ornia pin mthe CDPH d

e with Ch

base was qu2 dataset for the other 68relations. T5% of the vomium concand may benships are vethylene andnt with the etween total ould not be i

urrence of helent chromiuare shown.

map of 75thdatabase by

romium

ueried for theach system

8 contaminanThe maximumvariation concentration. Te suspected tery weak. Fd trichloroetfinding of chromium ainvestigated

19

exavalent chrum concentra

Hexavalent

h percentiley system.

he maximumm. As a prelints was calcum correlationntaminant beThe most higto be solublFor comparithylene fromSeidel (200

and iron in sufurther. In

romium in Cation ranges chromium a

e hexavalen

m concentraiminary analulated. Tabn is seleniumeing comparghly correlatle under simison purposem this same06) which fourface water

n our opinion

California. Bof 1 - 5 µg/L

appears to oc

nt chromium

ation of eachlysis, the cor

ble 5 shows m with a cored can be ted compoun

milar geochemes, the corree analysis wound no co-rs. Iron is non, these resu

Because of thL, 5 - 10 µg/ccur at locati

m concentr

h of 69 anarrelation betthe contami

oefficient of explained bnds on the limical condit

elation coeffiwas 0.29. T-occurrenceot included i

ults do not ju

he /L, ions

ation

alytes tween inants 0.05, y the st are tions;

ficient These

with in the ustify


20

further investigation into co-occurrence using higher statistical techniques such as the Chi Squared approach.

Table 5. Correlation of co-occurring contaminants with total chromium in the 6YR2

dataset. Contaminant r Selenium 0.05 Arsenic 0.04 Fluoride 0.04 Beryllium 0.03 Antimony 0.03 Cadmium 0.03 Thallium 0.02 Beta Particles (Gross beta) 0.02 Tetrachloroethylene (PCE) 0.02 Nitrate 0.02 Barium 0.02 Uranium 0.02 Trichloroethylene (TCE) 0.02 Chlordane 0.02 1,1,2-Trichloroethane 0.02 cis-1,2,-Dichloroethylene 0.02 Alpha Particles 0.01 trans-1,2-Dichloroethylene 0.01 1,1-Dichloroethylene 0.01 Ethylene Dibromide (EDB) 0.01 Combined Radium- 226 & 228 0.01

6. Estimated Extent of Potential Future Regulation

One of the primary objectives of this occurrence survey is to determine the extent a potential future regulation may affect drinking water systems. Thus, the number of entry points greater than a given threshold concentration were estimated for CWSs. Because the 6YR2 database does not represent a census of all nationwide entry points, the total number of entry points was estimated. The number of nationwide CWS entry points was estimated by comparing the 6YR2 data to the SDWIS database of CWSs (retrieved in December of 2011), broken down by system size. Table 6 shows the results of the estimation. Columns 2 and 3 show the number of systems in the SDWIS and 6YR2 databases, respectively. Table 6 indicates that the 6YR2 database includes between about 60-90% of the SDWIS totals. To project this to entry points, the number of entry points in the 6YR2 database is divided by the fraction of systems represented, which is presented in column 6 of Table 6. Last, the average number of entry points per system is calculated. The numbers generally increase with system size from about 2 to 14.5 for the largest systems.


21

Table 6. Estimation of the national CWS entry point inventory. 1 2 3 4 5 6 7

System Population

Served

SDWIS CWS

PWSID Census

6YR2 PWSID

Inventory6YR2 / SDWIS

6YR2 Entry Point

Inventory

Projected Number of Entry Points

Average Number of

Entry Points per PWSID

All 50,936 32,810 64% 74,672 115,925 2.3 0-100 11,460 8,217 72% 20,116 28,055 2.4

101-500 16,333 10,509 64% 20,977 32,602 2.0 500-1,000 5,707 3,432 60% 7,085 11,781 2.1

1,001-3,300 8,259 4,719 57% 9,189 16,082 1.9 3,301-10,000 4,946 3,092 63% 6,868 10,986 2.2 10,001-50,000 3,254 2,122 65% 6,509 9,981 3.1 50,001-100,000 555 386 70% 1,585 2,279 4.1 100,001-1,000,000 400 313 78% 2,053 2,624 6.6

>1,000,000 22 20 91% 290 319 14.5

The number of entry points greater than the threshold values was estimated by multiplying the total number of entry points by the percent of entry points exceeding the potential regulation. Table 7 shows these results for total chromium concentrations of 1, 2, 5, 10, 20, and 50 µg/L. The number of entry points above the threshold concentration have been estimated based on three conditions:

1. Using the 6YR2 database with non-detects set to one-quarter of the MRL 2. Using the 6YR2 database with non-detects set to zero 3. Using the NCBOS database with non-detects set to zero


22

Table 7. Number of CWS entry points above the threshold concentration at various threshold concentrations.

The results of Table 7 indicate the number of entry points impacted increases steeply below treatment goals of 10 µg/L. Again this is largely due to the uncertainty associated with non-detect handling. Figure 18 shows this trend graphically aggregating all system sizes. For instance, the estimated number of entry points with concentrations above 10 µg/L range from 1,971 (NCBOS) to 8,829 (6YR2 with non-detects set to one-quarter of the MRL). Thus, even at a concentration above the influence of the non-detect results, over a four-fold difference is estimated depending on which data set is selected.

System Population

Served

Threshold Concentration

1 µg/L 2 µg/L 5 µg/L 10 µg/L 20 µg/L 50 µg/L

All Entry Points

6YR2 n.d.=0.25·MRL 97,693 79,953 25,044 8,829 1,976 217 6YR2 n.d.=0 23,447 18,581 10,417 6,013 1,826 217

NCBOS 50,543 29,793 10,897 1,971 1,043 0

0-100 6YR2 n.d.=0.25·MRL 23,022 19,135 5,957 2,178 432 42

6YR2 n.d.=0 5,245 4,048 2,332 1,408 400 42 NCBOS 12,232 7,210 2,637 477 252 0

101-500 6YR2 n.d.=0.25·MRL 27,254 22,268 7,248 2,539 516 71

6YR2 n.d.=0 6,655 5,242 2,827 1,680 461 71 NCBOS 14,215 8,379 3,065 554 293 0

500-1,000 6YR2 n.d.=0.25·MRL 9,969 7,817 2,554 825 197 27

6YR2 n.d.=0 2,325 1,827 1,032 544 181 27 NCBOS 5,137 3,028 1,107 200 106 0

1,001-3,300

6YR2 n.d.=0.25·MRL 13,553 10,811 3,101 881 181 15 6YR2 n.d.=0 3,262 2,552 1,289 589 164 15

NCBOS 7,012 4,133 1,512 273 145 0

3,301-10,000

6YR2 n.d.=0.25·MRL 9,424 7,733 2,144 724 208 17 6YR2 n.d.=0 2,319 1,839 966 550 192 17

NCBOS 4,790 2,823 1,033 187 99 0

10,001-50,000

6YR2 n.d.=0.25·MRL 8,769 7,136 2,231 800 221 15 6YR2 n.d.=0 2,089 1,742 1,072 605 218 15

NCBOS 4,352 2,565 938 170 90 0

50,001-100,000

6YR2 n.d.=0.25·MRL 2,055 1,787 476 245 88 14 6YR2 n.d.=0 489 422 303 210 87 14

NCBOS 994 586 214 39 21 0

100,001-1,000,000

6YR2 n.d.=0.25·MRL 2,276 2,001 749 299 77 6 6YR2 n.d.=0 705 602 394 272 69 6

NCBOS 1,144 674 247 45 24 0

>1,000,000 6YR2 n.d.=0.25·MRL 304 289 149 132 13 2

6YR2 n.d.=0 76 68 37 24 13 2 NCBOS 139 82 30 5 3 0


23

Figure 18. Estimation of national number of CWS entry points with total chromium above threshold concentrations.

7. Conclusions

The purpose of this study was to update the NCBOS chromium occurrence analysis originally presented in the WaterRF Report #2759 with new data from the EPA 6YR2 and CDPH WQAD datasets. The comparisons also validate key findings already presented in WaterRF Report #2759, including:

1. no significant difference in distribution of total chromium for ground waters versus surface waters;

2. ground waters tend to have much of their total chromium as hexavalent chromium;

3. total chromium occurrence is distributed across the country; and

4. the highest chromium concentration occurrences were observed in EPA Region 9 (California, Nevada, and Arizona).

The total chromium data is limited by MRLs, making predictions of systems impacted subject to a high degree of uncertainty below 10 µg/L. A suitable dataset is not available for predicting hexavalent chromium occurrence on a national basis. Preliminary analysis of systems impacted at different chromium concentrations indicates the potential for a very large number of systems to be impacted by regulation.

0

20,000

40,000

60,000

80,000

100,000

0 10 20 30 40 50

Number of Entry Points

Above

Threshold

Total Chromium Threshold Concentration (µg/L)

6YR2 n.d.=0.25∙MRL

NCBOS

6YR2 n.d.=0


24

8. References

EPA. (2010) “The Analysis of Regulated Contaminant Occurrence Data from Public Water Systems in the Support of the Second Six-Year review of national Primary Drinking Water Regulations,” US Environmental Protection Agency report EPA-815-B-09-006.

Frey, M.M., Seidel, C.J., Edwards, M., Parks, J.L., McNeill, L. (2004) Occurrence Survey of Boron and Hexavalent Chromium, Water Research Foundation, Water Research Foundation Report #2759 Denver, CO.

Parks, J. L., McNeill, L., Frey, M., Eaton, A. D., Haghani, A., Ramirez, L., Edwards, M. (2004) “Determination of total chromium in environmental water samples,” Water Research, 38(12): 2827-2838.

Raucher, R.S., Castillo, E.T., Dixon, A., Breffle, W., Waldman, D., and Drago, J.A. (1995) Estimating the Cost of Compliance with Drinking Water Standards: A User’s Guide. Water Research Foundation, Denver, CO.

Seidel, C.J. (2006) “Investigation of Hexavalent Chromium as a Future Regulated Drinking Water Contaminant,” Doctoral Thesis, University of Colorado at Boulder.

USGS. (2012) Arsenic in Groundwater of the United States. http://water.usgs.gov/nawqa/trace/arsenic/ accessed March 1, 2012.


Documents

Total Chromium and Hexavalent Chromium Occurrence Analysis