Purpose and Scope

This report describes a quality-assurance (QA) study in which the individual water samples of a paired set of samples collected in 2020 from 26 wells at INL were processed and handled differently. One of the samples in each set was chilled upon collection and during shipment to the laboratory for analysis of VOCs; this sample will be referred to in this report as “unpreserved.” The other sample in each set was acidified with HCl and then chilled upon collection; this sample will be referred to here as “preserved.” Results of the analyses were then compared in statistical tests to evaluate reproducibility between VOC concentrations in the unpreserved and preserved samples in each sample set. The data described in this study can be found in the USGS National Water Information System (U.S. Geological Survey, 2016).

Geohydrologic Setting

The INL is located on the west-central part of the eastern Snake River Plain, a northeast-trending structural basin about 200 miles (mi) long and 50–70 mi wide. The ESRP aquifer is one of the most productive aquifers in the United States (U.S. Geological Survey, 1985, p. 193). Groundwater generally moves from northeast to southwest and discharges to springs along the Snake River downstream from Twin Falls, Idaho, about 100 mi southwest of the INL. Groundwater moves horizontally through basalt interflow zones and vertically through joints and interfingering edges of basalt flows. Infiltration of surface water, pumping from wells completed in the aquifer, geohydrologic conditions, and seasonal fluxes of recharge and discharge locally affect the movement of groundwater (Garabedian, 1986). The aquifer is recharged primarily by infiltration of applied irrigation water, infiltration of streamflow, groundwater inflow from adjoining mountain drainage basins, and infiltration of precipitation (Ackerman and others, 2006). At the INL, depth to water in wells completed in the ESRP aquifer ranges from about 225 feet (ft) below land surface in the northern part of the INL to more than 900 ft below land surface in the southeastern part of the INL (Bartholomay and others, 2020, p. 20).

Previous Investigations

In the mid-1990s, the USGS conducted a study to investigate a suggested change to the protocol for collection and field processing of water samples for laboratory analyses of VOCs. The standard protocol at that time was to simply chill the raw samples for shipment to the laboratory. The suggested change was to add HCl to the samples as a preservative upon collection and then chilling them. A comparison of the results of analyses of paired sets of samples processed by the different field techniques indicated that the addition of HCl along with chilling of the samples did not have a statistically significant effect on the concentration of VOCs determined by the laboratory in samples from multiple sources, including groundwater, surface water, stormwater runoff, and sewage effluent. On the basis of those findings, the standard protocol for field processing of samples for VOC analyses was updated to include the addition of HCl along with chilling (app. 1; Connor and others, 1998).

In addition to the type of study mentioned above, the USGS INL Project Office has a history of performing investigations to assure the validity and accuracy of water-quality data collected at the INL. Investigations to determine variability and bias in laboratory analytical results (concentrations) of individual constituents were summarized by Wegner (1989), Williams (1996, 1997), and Rattray (2012, 2014). Additional QA studies that focus on sample collection and field processing techniques by USGS INL Project Office personnel include:

Sample Collection and Processing

To evaluate the effect of different sample preservation methods on the reproducibility of laboratory-determined VOC concentrations, 37 paired-samples were collected from 25 wells completed in the ESRP aquifer, and 1 well completed in a perched groundwater zone above the aquifer (figs. 1, 2). Discrete samples were pumped from wells equipped with permanently installed submersible pumps through in-line stainless-steel sampling pipes attached to the well head or spigots. The samples were collected according to procedures described in the USGS National Field Manual (with the exception of adding HCl as a preservative for VOC analyses) (U.S. Geological Survey, 2018; Connor and others, 1998, and further described in the INLPO field methods and quality-assurance plan (Bartholomay and others, 2014).

Both unpreserved and preserved samples for the analysis of VOCs were collected in clean 40-mL borosilicate amber vials (VOC vials). The historical and current VOC sample collection and processing method for the INL Project Office is to fill the vials from the bottom up and fill to overflowing, capping immediately without any headspace, and chilling (historically collected unpreserved samples). The preserved sample was processed in the mobile field laboratory by adding, dropwise from a Teflon squeeze bottle, a concentrated solution of (1:1) HCl supplied by the NWQL. In-house testing at the INLPO prior to the field sampling determined that approximately five drops of HCl were necessary to achieve a sample pH of 2. Both unpreserved and preserved samples were kept at 4 °C±2 °C by chilling immediately after collection, and during shipment to the NWQL. Samples are shipped by overnight-delivery mail in a sealed ice chest and are typically sent to the laboratory within five days of collection (Maimer and Bartholomay, 2016).

Laboratory Analyses

The current NWQL method for the analyses of VOCs is summarized in Rose and others (2016) and in references within. Samples are analyzed for ambient purgeable compounds to determine the suite of VOCs present (table 1). The NWQL verifies minimum detection limits or minimum reporting limits (MRL) for a rotating subset of analytes on a yearly basis, if applicable (Foreman and others, 2021). The VOCs for which samples are analyzed by NWQL, the MRL for each compound, and maximum contaminant levels (MCLs) recommended by the U.S. Environmental Protection Agency (2020) are listed in table 1.

Relative Percent Difference

The relative percent difference (RPD) is a direct measurement of the variability between two independent sets of data without considering analytical uncertainty. In this study, the datasets consisted of the concentrations of VOCs in a) unpreserved and b) preserved samples, respectively, of a co-located sample-pair. Only those sample-pairs in which VOCs were detected in both the unpreserved and the preserved samples from a well at concentrations above the MRL for each analyte were considered for evaluation (n = 100). In one paired-sample set, that from well USGS 120, the concentration of 1,1,1-trichloroethane exceeded the MRL in the unpreserved sample but was less than the MRL in the preserved sample; thus, the data for this sample set were not included in the evaluation. A typical data-quality criterion, applied to this set of paired-sample data, is that equivalence is demonstrated when the RPD is equal to or less than 20 percent (Taylor, 1987).

RPD is calculated using formula (1);

Sample pairs with VOC concentrations greater than the MRL and with a calculated RPD of less than 20 percent were considered statistically equivalent. Of the samples in which VOC concentrations exceeded their respective MRL, the RPDs for all paired-samples evaluated (n = 100) were less than 20 percent and thus are considered statistically equivalent (table 2).

Normalized Absolute Difference

A second approach to test for variability, the normalized absolute difference (NAD), was applied to the concentrations of VOCs in the paired-samples that were detected above the MRL to determine if these values are statistically equivalent with the incorporation of analytical precision. When analytical uncertainties, or standard deviations, are not provided for the analytical method used by the laboratory, estimations of the standard deviations for individual analytes can be used for statistical comparison (Bartholomay and Williams, 1996). The standard deviations used to estimate the variability in concentrations of VOCs in groundwater are summarized in Rose and others (2016). The standard deviations for VOC concentrations were determined for different matrices including blank source-water, surface water, and groundwater. The standard deviations for the VOC suite were determined from seven groundwater replicates, spiked at two concentrations, and preserved with HCl in the laboratory (table 14 in Rose and others, 2016). The standard deviation applied for each constituent detected at a concentration above the MRL (0.5 or 5 micrograms per liter) was chosen based on the value closest to the paired-sample concentrations being compared.

NAD is calculated using formula (2);

The paired-samples for each analyte detected at a concentration above the MRL were considered statistically equivalent at the 95-percent confidence level, or 2 standard deviations from the mean, if the NAD ≤ 1.96 (Volk, 1969). Given that this statistical test utilized estimated standard deviations (Rose and others, 2016), and the data are not normally distributed, the NAD equal or less than equal to 1.96 was used as a guide to test for equivalence (Rattray, 2012; Rattray, 2014). Based on this approach, 81 percent of the sample-pairs with VOC concentrations above the MRL are considered statistically equivalent using this statistical test for variability. The sample-pairs that were not considered statistically equivalent with the NAD method guidelines were those pairs that had the highest RPD (table 2). No bias was detected in the NAD results with respect to site locations, specific VOC constituents, or sample depth. While this approach to evaluate variability did not show the same level of statistical equivalence as the RPD approach, the NAD is used as a guide, given that the standard deviations are pooled and not specific to this suite of data. These results are consistent with those in the previous study by the USGS Office of Water Quality that evaluated the statistical equivalence of VOC concentrations and preservation technique on a wide range of sample types and VOC analytes (app. 1).

Rank-Sum Test

To evaluate the equivalence between unpreserved and preserved samples in terms of concentrations of individual VOC’s, the Mann-Whitney Wilcoxon rank-sum test was applied to analytical results from this study when concentration values above the MRL for an individual VOC were detected in more than six paired-samples (Helsel and others, 2020). Thus, the VOCs selected for further analysis included: 1,1,1-Trichloroethane, Tetrachloroethene, Tetrachloromethane, Trichloroethene, and Trichloromethane. The rank-sum test is a nonparametric test that evaluates the difference between two groups of data, the median difference between concentrations in VOCs in unpreserved and preserved samples. The null hypothesis (H_o) is that the median of the difference of the concentrations are equal to zero, or that the number of negative differences is approximately the same as the positive differences between the concentrations in unpreserved and preserved paired-samples at a specified level of significance. The alternative hypothesis (H_A) is that x > y, where x represents the unpreserved sample results and y represents the preserved sample results. This test, applied to the subset of data for individual VOCs, was applied with an α of 0.90. The null hypothesis was rejected if the p-value was ≤ 0.1 (Helsel and others, 2020).

Mann-Whitney Wilcoxon Rank-Sum is calculated using formula (3);

Results from the Mann-Whitney rank-sum (Wilcoxon) statistical test reveal that the medians of unpreserved and preserved VOC concentrations are not statistically different (table 3) even though the concentrations in the unpreserved samples were slightly higher than those in the preserved samples for all individual VOCs evaluated. The medians of the differences between the unpreserved and preserved VOC concentrations were less than 0.26 µg/L for the subset of individual VOCs evaluated. These results give further confidence that there is no statistical difference in the concentrations of individual VOCs detected in laboratory analyses regardless of the sample preservation method used at time of sample collection. This subset of data accounts for 92 percent of all VOC concentrations above the MRL measured for this study.

Bias

As a general practice, the potential for the introduction of environmental bias during groundwater sampling at the INL is periodically evaluated on the basis of analyses of quality-control (QC) samples collected during the collection of environmental samples. These QC samples include equipment, field, and source-solution blank samples, as appropriate. During all routine water-quality sampling at the INL in 2020, QC samples were collected to validate field-sample collection practices and protocols. Although in 2020 the QC samples were not collected and analyzed specifically for VOCs, an analysis of QC samples collected during previous periods of sampling specifically for VOCs from 1996 through 2018 (Rattray, 2012, 2014; Davis and others, 2013; Bartholomay and others, 2017, 2020) did not show any potential bias, field-equipment, or source-solution blank contribution that would indicate any bias introduced by field-collection techniques, sample preservation, or sample handling. This study did not evaluate the effects of the collection and preservation of “spiked” QC samples on the results of the analyses of those samples for VOCs. A “spiked” sample is prepared in a laboratory with a known concentration of an analyte to evaluate any potential sample preparation and analysis method bias at the laboratory. There is no consistent positive or negative bias to VOC concentration data determined for this suite of paired-samples from either preservation method.