Comment on “Characterization of Marcellus Shale Flowback Water”

Abstract

In their article (Abualfaraj et al., 2014), the authors conclude that chlorinated solvents and several other chemical constituents are commonly present in Marcellus Shale flowback water at average concentrations greater than 10 times federal drinking water standards. However, we are concerned that the authors' conclusions regarding chlorinated solvents, and likely other chemicals, appear to be driven by their approach for the handling of nondetect results. Specifically, our review of both the article and the original data source for most of the authors' data set (Hayes, 2009) shows that all of the vinyl chloride results and most of the other volatile organic chemical (VOC) results were originally reported as nondetectable (some with detection limits that are in excess of drinking water criteria). The authors appear to have interpreted these results as detections even though these compounds were not detected in the associated flowback water samples as reported in the original source report. On this basis, key conclusions presented in both the abstract and throughout the article appear to be incorrect. Specifically, the authors conclude that high concentrations of vinyl chloride and other chlorinated solvents are commonly present in flowback water. However, an evaluation of the study's data set that properly distinguishes between detections and nondetect results would yield the conclusion that chlorinated solvents included in the study are not present in Marcellus Shale flowback water. This issue would likely also affect the authors' interpretation of other chemical concentration results.

Evidence that VOC Results for Most Samples Were Nondetect

In the Analysis section, the authors state that they used a value of 1/2 the detection limit for nondetect results. Otherwise, the authors provide no information concerning the number of detect versus nondetect results obtained for individual chemicals. However, a review of the original data source for most of the authors' data set (Hayes, 2009) indicates that at least 86 of the 87 vinyl chloride results were nondetect. Evidence that VOC test results were mostly nondetect include the following:

• Identical median and maximum values: For 11 of the 12 VOCs reported in Table 6, the median concentrations are identical (5 μg/L). Identical medians are very unlikely to occur unless more than 50% of the analytical results for each of the VOCs were nondetect; consequently, the numerical values simply reflect the substitution of 1/2 of the detection limit (or the full detection limit) for the nondetect results. Similarly, the maximum values are identical (2,000 μg/L) for 10 of the 12 VOCs. The maximum concentration values for these 10 VOCs would not be identical unless these the values reflected nondetect results for a sample with elevated detection limits.

• Nondetect results in GTI report: We reviewed the publically available Gas Technology Institute (GTI) report (Hayes, 2009) that the authors identified as the source of 75% of their total data set. The GTI report contains vinyl chloride results for 86 flowback water samples, accounting for all but one of the 87 vinyl chloride results considered by the authors (Abualfaraj et al., 2014; Table 6). The GTI report indicates nondetect (U-flag) vinyl chloride results for all 86 samples. The median quantitation limit of 5 μg/L and the maximum quantitation limit of 2,000 μg/L in the GTI report are equal to the median concentration and the maximum concentration reported by the authors in Table 6. Note that the GTI report provides quantitation limits but does not provide detection limits. Thus, it appears that the authors treated all of the nondetect results reported in the GTI report as actual detections rather than nondetect results. In other words, the authors used a value equal to the reported quantitation limit rather than a value of 1/2 the detection limit as stated in the article for nondetect results.

When a sample contains elevated concentrations of one or more VOCs, the laboratory must dilute the sample to obtain a concentration that falls within the calibrated linear range of the analytical instrument (thus allowing for accurate quantification of the VOC concentration). These dilutions result in elevated detection limits for all of the remaining VOCs that are not detectable in the sample. The elevated detection limits for the nondetected VOCs do not provide any evidence that the nondetected VOCs are actually present in the sample at some concentration below the elevated detection limit. Instead, the elevated detection limits merely reflect the dilution that was required to accurately measure the specific VOCs that are present in the sample at high concentrations. For example, the flowback water sample in the GTI report with a vinyl chloride quantitation limit of 2,000 μg/L had a carbon disulfide concentration of 54,000 μg/L. The dilution required to quantify the carbon disulfide, resulted in elevated quantitation limits for vinyl chloride and all of the other VOCs reported by the laboratory.

As reflected in the GTI report, when the laboratory was able to detect individual VOCs at concentrations below the reported quantitation limit, these results were reported as estimated (J-flag) concentrations. Therefore, results reported as nondetect do, in fact, indicate an absence of the target VOC in the sample except possibly at concentrations well below the reported quantitation limit.

• Chlorinated solvents are not known additives to fracturing fluid: As acknowledged by the authors, chlorinated solvents are not known to be used in hydraulic fracturing fluids (FracFocus, 2013). Consequently, these chemicals are not expected to be present in flowback water. This is further evidence that the high concentrations of chlorinated solvents reported by the authors are an artifact of their handling of nondetect results rather than indicative of the actual presence of these chemicals in flowback water.

Effect of the Apparent Misinterpretation of Nondetect Results on the Conclusions of the Article

In interpreting their data set, the authors have placed the same meaning on samples with nondetect results and samples with actual detected concentrations of chemicals. In the abstract and throughout the article, the authors indicate that their data set shows “high concentrations of chlorinated solvents” and other constituents in flowback water. For chlorinated solvents, this finding appears to be driven exclusively or primarily by nondetect results in samples with elevated detection limits. However, based on the information provided in the article, we cannot determine a complete list of other chemicals affected by this error. Proper consideration of detected versus nondetected results could also significantly change the interpretation for other classes of organics such as polyaromatic hydrocarbons, disinfectants, and petroleum hydrocarbons.

Understanding the chemical composition of flowback water is important for designing treatment systems and for understanding potential environmental impacts of flowback water. In this case, we believe that the misinterpretation of nondetect results has resulted in incorrect conclusions regarding the chemical characteristics of flowback water. Given the importance of this issue, we encourage the authors to prepare an alternate analysis of their data set that distinguishes between detected and nondetected results.

Footnotes

Author Disclosure Statement

No competing financial interests exist.

References

Abualfaraj

, Gurian

P.L.

, and Olson

M.S.

(2014). Characterization of Marcellus Shale flowback water. Environ. Eng. Sci., 31, 514.

FracFocus. (2013). Chemical Disclosure Registry. Available at: http://fracfocus.org (accessed August 24, 2015).

Hayes

(2009). Sampling and Analysis of Water Streams Associated with the Development of Marcellus Shale Gas. The Marcellus Shale Coalition Report by Gas Technology Institute, Des Plaines, IL.