Chromium may be present naturally in groundwater as reduced trivalent chromium, Cr(III), or as oxidized Cr(VI); Rai and Zachara, 1984; Guertin and others, 2004). Trivalent chromium is the predominate form of chromium in acidic, reduced (oxygen absent) groundwater (Rai and Zachara, 1984; Ball and Nordstrom, 1998), and Cr(III) concentrations can exceed several hundred micrograms per liter (µg/L) in reduced geothermal water (Stefánsson and others, 2015; Kaasalainen and others, 2015). However, for the chemical conditions found in most uncontaminated surface and groundwater, Cr(III) concentrations are limited by dissolution of chromite, Cr(OH)₃, which has low solubility (Ball and Nordstrom, 1998). Hexavalent chromium is not present in reduced groundwater, but is the predominate form of chromium in alkaline, oxic (oxygen present) groundwater (Rai and Zachara, 1984; Ball and Nordstrom, 1998).

In the past, Cr(VI) in groundwater was considered evidence of an anthropogenic (human-made) source (Hem, 1985), but Cr(VI) has recently been recognized as naturally occurring in alkaline, oxic groundwater (Robertson 1975, 1991; Ball and Izbicki, 2004; Izbicki and others, 2008a). Hexavalent chromium concentrations in alkaline, oxic groundwater are commonly controlled by pH-dependent sorption (Rai and Zachara, 1984; Ball and Nordstrom, 1998; Izbicki and others, 2015; Xie and others, 2015), and natural Cr(VI) concentrations in saline groundwater in arid regions can exceed 1,000 µg/L (Eriksen, 1983).

Trivalent chromium is an essential micronutrient. Hexavalent chromium is a carcinogen if inhaled (Daugherty, 1992; Agency for Toxic Substances and Disease Registry, 2012) and may be a carcinogen if ingested (Sedman and others, 2006; Beaumont and others, 2008). The U.S. Environmental Protection Agency (EPA; U.S. Environmental Protection Agency, 2019) does not have a maximum contaminant level (MCL) for Cr(VI) in drinking water but does have an MCL for total dissolved chromium, Cr(t), of 100 µg/L. The California MCL for Cr(t) is 50 µg/L (State Water Resources Control Board, 2018a). In 2014, California established an MCL for Cr(VI) of 10 µg/L (State Water Resources Control Board, 2017). The cost of treatment prior to use as a source of public supply was not considered during development of the 2014 California MCL for Cr(VI), and the MCL was withdrawn in August 2017. The California MCL for Cr(t) is being used for regulation of drinking water supplies until a new California MCL for Cr(VI) is developed (State Water Resources Control Board, 2017).

Natural Cr(VI) concentrations in alkaline, oxic groundwater pumped for public supply can exceed MCLs developed for the protection of public health in some geologic and hydrologic settings within California (Chung and others, 2001; Ball and Izbicki, 2004; Dawson and others, 2008; Izbicki and others, 2008a, 2012, 2015; Morrison and others, 2009; Mills and others 2011; Manning and others, 2015; McClain and others, 2016; Hausladen and others, 2018), the southwestern United States (Robertson, 1975, 1991), and elsewhere in the world (Oze and others, 2007; Kazakis and others, 2015) where chromium abundance in rock and soils is high. In addition to chromium abundance in geologic material, natural Cr(VI) concentrations in groundwater are influenced by a combination of processes, including (1) the mineralogy and weathering rates of chromium-containing minerals, (2) accumulation of chromium weathered from minerals within surface coatings on mineral grains, (3) oxidation of accumulated Cr(III) associated with the surface coatings on mineral grains to Cr(VI), and (4) pH-dependent desorption of chromium from coatings on the surfaces of mineral grains into groundwater under suitable aqueous geochemical conditions (Richard and Bourg, 1991; Kotaś and Stasicka, 2000; Izbicki and others, 2008a; Ščančar and Milačič, 2014; fig. E.1). After chromium has weathered from mineral grains, it typically must oxidize to Cr(VI) before it can desorb and enter groundwater. During natural conditions, chromium oxidation commonly occurs in the presence of manganese oxides (Mn oxides), including Mn(III) (Nico and Zasoski, 2000), Mn(III)/(IV) (Oze and others, 2007), and Mn(IV) oxides (Schroeder and Lee, 1975). Oxidation of Cr(III) to Cr(VI) in the presence of Mn oxides reaches a maximum near pH 5.6, and rates decrease as pH increases (Oze and others, 2007), although oxidation persists in alkaline water (pH greater than 7.0). Oxidation of Cr(III) to Cr(VI) is enhanced in fine-textured aquifer material and in older, slow-moving groundwater, where diffusive transport of reactants facilitates oxidation (Hausladen and others, 2019). In the absence of Mn oxides, oxidation of Cr(III) to Cr(VI) also can occur at slower rates in the presence of dissolved oxygen (Schroeder and Lee, 1975) or in the presence of peroxide during serpentinization (Oze and others, 2016).

Once oxidized, Cr(VI) on the surfaces of mineral grains is potentially mobile into oxic groundwater under alkaline pH. Desorption of Cr(VI) from mineral grains increases in alkaline water until pH 9.0, when almost all Cr(VI) is desorbed (Xie and others, 2015). In settings having an abundance of easily weathered chromium-containing minerals and reactive Mn oxides, Cr(VI) may occur naturally in groundwater at concentrations of concern for public health (Izbicki and others, 2008a; Morrison and others, 2009; Mills and others, 2011). In areas having comparatively low geologic abundance of chromium or chromium-containing minerals that are resistive to weathering, Cr(VI) may be present in groundwater at concentrations of concern as contact time with aquifer materials, weathering of minerals within those materials, oxidation of Cr(III) to Cr(VI), and pH increases along groundwater-flow paths—as long as oxic conditions persist within groundwater (Izbicki and others, 2008a, 2015; Manning and others, 2015).

Microbially mediated reduction of Cr(VI) to Cr(III) may occur in the presence of organic material (Oze and others, 2007) and may be inhibited in the presence of nitrate (Izbicki and others, 2008b). Abiotic reduction of Cr(VI) to Cr(III) also can occur in the presence of ferrous iron, Fe(II), at a pH greater than 5.6, or in the presence of hydrogen sulfide (Fendorf and others, 2000).

The Pacific Gas and Electric Company Hinkley compressor station, in the Mojave Desert about 80 miles (mi) northeast of Los Angeles, California, is used to compress natural gas as it is transported through a pipeline from Texas to California. Between 1952 and 1964, cooling water used at the Hinkley compressor station was treated with a compound containing Cr(VI) to prevent corrosion of machinery within the compressor station. Cooling wastewater was discharged to unlined ponds resulting in release to soil and groundwater in the underlying unconsolidated aquifer (Lahontan Regional Water Quality Control Board, 2013a). Although regulatory data have been collected at the site since the late 1980s (Ecology and Environment, Inc., 1988), the extent of anthropogenic Cr(VI) in groundwater downgradient from the Hinkley compressor station remains uncertain, in part because of uncertainty regarding background Cr(VI) concentrations in groundwater in the area. This uncertainty was not resolved by a previous background Cr(VI) study (CH2M Hill, 2007), and the USGS was requested to complete an updated Cr(VI) background study in Hinkley and Water Valleys beginning in January 2015 (Izbicki and Groover, 2016, 2018).

Between March 2015 and November 2017, the USGS collected water samples for analyses of chemical constituents (table E.1) from more than 100 wells in Hinkley and Water Valleys, California (fig. E.2). Most sampled wells were selected from monitoring wells installed by PG&E for regulatory purposes near the margins of the October–December 2015 (Q4 2015) regulatory Cr(VI) plume. Samples also were collected from (1) 12 monitoring wells installed by PG&E at 6 locations upgradient from the Hinkley compressor station as part of the USGS Cr(VI) background study, (2) 7 domestic wells in areas where monitoring wells were not available, and (3) 5 selected production wells. Chemical analyses of water from the surface discharge of temporary pumps installed within the production wells are presented in this chapter; unpumped and pumped well-bore flow data from those production wells and depth-specific water chemistry and isotopic data collected during pumping conditions are discussed in chapter H within this professional paper. Water from sampled wells was analyzed for chemical constituents (table E.1) and for isotopic and age-dating constituents are discussed in chapter F within this professional paper. Additional data, collected in areas where monitoring wells were not available, were collected from more than 70 domestic wells sampled between January 25 and 31, 2016; these samples were analyzed for a smaller number of constituents that included field parameters (pH, specific conductance, and dissolved oxygen) and selected trace elements including Cr(VI) and Cr(t). Porewater samples were collected at 11 locations in Hinkley Valley to evaluate Cr(VI) and selected trace-element concentrations in porewater within fine-grained materials that do not contribute water freely to wells. Water chemistry and isotopic data are available in appendix E.1 (table E.1.1) and also from the USGS National Water Information System (NWIS) online database (U.S. Geological Survey, 2021).

Sampled wells were selected by the USGS in collaboration with a technical working group (TWG) composed of Hinkley community members, the Independent Review Panel (IRP) Manager (Project Navigator, Ltd.), the Lahontan RWQCB, PG&E, and consultants for PG&E. The selected wells represent a mutually agreed upon, spatially distributed set of wells covering a range of geologic, hydrologic, and geochemical settings within and near the regulatory Cr(VI) plume. Most data were collected in three sample rounds between March 2015 and March 2017 to allow preliminary interpretation of earlier data that guided the collection of later data. An additional four wells were sampled in November 2017 to fill data gaps. Collectively, wells sampled as part of this study represent the most complete, non-regulatory set of water-quality data available near the Cr(VI) plume.

Most geologic materials in Hinkley and Water Valleys are low in chromium (Smith and others, 2014), although chromium is locally present in hornblende diorite in Iron Mountain, in basalt in Water Valley, and in local alluvium eroded from those materials (chapter B). Chromium concentrations in unconsolidated materials are higher in fine-textured silt and clay and lower in coarse-textured sand and gravel (chapter B). Chromite is not present, and most chromium is substituted within magnetite, which is relatively resistant to weathering. Some more easily weathered chromium-containing minerals also are present, including actinolite in older alluvium deposited by the ancestral Mojave River, hornblende in weathered bedrock, and local alluvium eroded from Iron Mountain (chapter C).

Hinkley Valley is about 62 square miles (mi²) and contains about 36 mi² of unconsolidated deposits that were saturated under predevelopment (pre-1930) conditions. Aquifers of interest in Hinkley and Water Valleys are composed primarily of unconsolidated deposits consisting of alluvium and lake-margin deposits sourced from the Mojave River (Miller and others, 2018, 2020), and are referred to as “Mojave-type” deposits (chapter A, table A.1). Locally derived alluvium, lacustrine deposits, and weathered bedrock also are important aquifers in some areas.

On the basis of differences in geology and hydrology, the study area was divided into eastern, western, and northern subareas within Hinkley Valley and Water Valley (fig. E.2). The eastern subarea is closest to recharge areas along the Mojave River. Mojave-type deposits in this area compose the upper aquifer, which overlies fine-textured lacustrine (lake) deposits generally described as “blue clay” at a depth of about 160 feet (ft) below land surface (bls; ARCADIS and CH2M Hill, 2011; Jacobs Engineering Group, Inc., 2019). Fine-textured deposits, generally described as “brown clay,” are interspersed throughout unconsolidated deposits, and in places, this brown clay separates the upper aquifer into shallow and deep zones (ARCADIS and CH2M Hill, 2011; Jacobs Engineering Group, Inc., 2019). Mudflat/playa deposits sourced from the Mojave River are present at land surface near Mount General and at depth within the eastern subarea, where these deposits also are commonly described as brown clay. The western subarea consists of Mojave-type deposits overlying groundwater-discharge deposits and weathered bedrock (CH2M Hill, 2013; Miller and others, 2018, 2020). The northern subarea consists of Mojave River alluvium overlying fine-textured lacustrine and mudflat/playa deposits sourced from the Mojave River and local materials (Stantec, 2013; Miller and others, 2018, 2020). Aquifers within Water Valley consist of lake-margin deposits sourced from the ancestral Mojave River along the margins of Harper (dry) Lake that overlie and interfinger with locally derived alluvium (Miller and others, 2018, 2020).

Groundwater recharge is primarily from intermittent flows in the Mojave River (fig. E.2) that occur on average once every 5–7 years (Lines, 1996; Stamos and others, 2001; Seymour, 2016). During predevelopment conditions, groundwater flow was from the Mojave River northward toward Hinkley Gap and Water Valley where groundwater discharged by evaporation along the margins of Harper (dry) Lake (Thompson, 1929). The Lockhart fault is an impediment to groundwater flow in Hinkley Valley (Stamos and others, 2001; Stantec, 2013). Predevelopment (pre-1930) and 2018 water-level maps are provided in chapter H within this professional paper (fig. H.8).

Water-level declines resulting from agricultural pumping since the early 1950s, have been as much as 60 ft (Stone, 1957; California Department of Water Resources, 1967; Seymour and Izbicki, 2018). As a consequence of water-level declines, saturated alluvium in much of the western subarea downgradient (northeast) from the Lockhart fault, is a thin veneer commonly less than 10 ft thick overlying groundwater-discharge deposits and weathered bedrock (CH2M Hill, 2013; Miller and others, 2018, 2020), and saturated alluvium in much of the northern subarea is a thin veneer overlying fine-textured lacustrine and mudflat/playa deposits (Stantec, 2013; Miller and others, 2018, 2020). Many monitoring wells in the western subarea are completed partly or entirely in weathered bedrock aquifers, and many monitoring wells in the northern subarea are completed partly or entirely in fine-textured mudflat/playa or lacustrine deposits. Lake-margin deposits in much of Water Valley, formerly pumped for agricultural water supply, were largely above the water table at the time of this study (2015–18; Stamos and others, 2001; Miller and others, 2018, 2020).

The Hinkley compressor station and much of the Q4 2015 regulatory Cr(VI) plume are located in the eastern subarea of Hinkley Valley (fig. E.2). In Q4 2015, the regulatory Cr(VI) plume extended 3.0 mi downgradient from the release location within the Hinkley compressor station (ARCADIS, 2016), and the highest Cr(VI) concentrations in groundwater remained less than 3,000 ft downgradient from the release location within the Hinkley compressor station. However, the actual extent of the Cr(VI) release was uncertain and Cr(VI) concentrations greater than the interim regulatory background of 3.1 µg/L (CH2M Hill, 2007; Lahontan Regional Water Quality Control Board, 2008) were present in water from wells within Water Valley, more than 8 mi downgradient from the Hinkley compressor station (fig. E.2; ARCADIS, 2016). Remediation of Cr(VI) released from the Hinkley compressor station began in 1992, and in 2010 site cleanup was projected to require 10–95 years (Haley and Aldrich, Inc., 2010; Pacific Gas and Electric Company, 2011).

Monitoring wells installed for regulatory purposes by PG&E were commonly identified by the prefix MW, with sites numbered sequentially in the order they were drilled (ARCADIS, 2016). Shallow wells, commonly screened across or just below the water table, were identified with the suffix S or S1 (ARCADIS, 2016). Deeper wells were identified with the suffix D, D1, or D2, or with the suffix S2 or S3 if a hydrologically important clay layer was not present; older monitoring wells were identified with the suffix A or B for shallower or deeper wells, respectively (ARCADIS, 2016). The suffix C was used for wells completed in consolidated rock, and the suffix R was added if the well was a replacement for a well that was destroyed (ARCADIS, 2016). Wells installed by PG&E as part of the USGS Cr(VI) background study were identified with the prefix BG; the sites were numbered sequentially in the order they were permitted, and BG wells were identified from shallowest to deepest with the suffix A, B, or C. Although drilling methods changed through time and in response to site conditions, most monitoring wells were drilled with auger rigs. Core material, archived by PG&E, was available for most wells installed after 2011 from near the water table to below the depth of the deepest well.

The purpose of this chapter is to evaluate the chemical composition and Cr(VI) concentrations in water from selected wells sampled as part of the USGS Cr(VI) background study in Hinkley and Water Valleys, California. Scope of the work included (1) evaluating the quality of PG&E and USGS data; (2) evaluating the chemistry of groundwater in Hinkley and Water Valleys, including the spatial distribution of Cr(VI); (3) evaluating geochemical controls on Cr(VI), including redox and pH-dependent sorption processes, and cooccurrence of Cr(VI) with selected trace elements; (4) evaluating Cr(VI) concentrations in porewater extracted from fine-textured material within the study area, including the impact of porewater fluids on aquifer Cr(VI) concentrations; and (5) evaluating Cr(VI) concentrations in water from domestic wells sampled within the study area.

Interpretations derived from Cr(VI), pH, and trace-element data were used within a summative-scale analysis (SSA) to determine the Cr(VI) plume extent in the upper aquifer system underlying Hinkley and Water Valleys. Although data collected as part of this study between March 2015 and November 2017 were used to define the summative scale Cr(VI) plume extent, these data were not used to calculate Cr(VI) background concentrations in chapter G within this professional paper. Instead, Cr(VI) background concentrations were calculated from regulatory data collected from selected wells between April 2017 and January 2018.

Samples for Cr(VI) and Cr(t) were field filtered, using a 0.45-micrometer (μm) pore-sized filter, into a 1-liter (L) plastic bottle. Water in the bottle was gently swirled to ensure complete mixing, and then subsampled into appropriate containers for analyses by various methods. About 20 milliliters (mL) of sample water was subsampled from the 1-L bottle and field speciated using cation exchange resins according to procedures described by Ball and McCleskey (2003). Field speciated and unspeciated samples were collected in duplicate in color-coded, 2-mL plastic vials, preserved with nitric acid using a micropipette, and shipped at the end of the sample collection period to the USGS Redox Chemistry Laboratory in Boulder, Colorado, for analyses within the Cr(t) holding time. When using this method, Cr(III) is removed by sorption onto an anion-exchange resin, and only Cr(VI) remains in the field-speciated sample for laboratory analyses, while both Cr(VI) and Cr(III) are present in the unspeciated sample vial. The remaining water in the 1-L bottle was subsampled into 500- and 250-mL bottles to be analyzed for Cr(VI) and Cr(t), respectively. Samples to be analyzed for Cr(VI) were adjusted to a pH of 9–9.5 with a concentrated buffer solution (U.S. Environmental Protection Agency, 1994a), stored on ice, and shipped from the field either by courier or overnight express to the same commercial laboratory used by PG&E for Cr(VI) regulatory data for analyses within 24 hours of collection. Samples for Cr(t) were acidified to a pH less than 2 with nitric acid, stored on ice, and shipped with Cr(VI) samples, although analyses for Cr(t) were not necessarily completed within 24 hours.

Hexavalent chromium within each duplicate field-speciated sample vial was analyzed as Cr(t) by Zeeman-corrected graphite furnace atomic absorption spectroscopy (GFAAS) using EPA Method 7010 (U.S. Environmental Protection Agency, 2007) at the USGS Redox Chemistry Laboratory. Total dissolved chromium within each duplicate unspeciated sample vial was analyzed using the same method. The laboratory reporting level (LRL) for Cr(t) analyses by GFAAS is 0.06 µg/L. Results of each duplicate field-speciated and unspeciated vial were averaged and reported; Cr(III) was calculated as the difference between reported Cr(t) and Cr(VI).

Hexavalent chromium was analyzed by ion chromatography using EPA Method 218.6 (U.S. Environmental Protection Agency, 1994a) at the commercial laboratory used for samples collected for regulatory purposes by PG&E. The EPA Method 218.6 (U.S. Environmental Protection Agency, 1994a) has a LRL for Cr(VI) of 0.2 µg/L; the method was modified by the commercial laboratory to provide an LRL of 0.06 µg/L. Total dissolved-chromium samples were analyzed at the commercial laboratory by inductively coupled plasma-mass spectrometry (ICP-MS) using EPA Method 200.8 (U.S. Environmental Protection Agency, 1994b) with a LRL of 0.20 µg/L.

Performance of laboratory analytical methods for Cr(VI) and Cr(t) was evaluated for more than 100 samples collected from wells outside the mapped Q4 2015 regulatory plume between March 2015 and November 2017 (fig. E.3). Hexavalent chromium concentrations ranged from less than the reporting level of 0.06 µg/L to about 11 µg/L. Comparison of Cr(VI) analyses by ion chromatography (EPA Method 218.6) with field speciation for Cr(VI) (Ball and McCleskey, 2003) and analyses by GFAAS (EPA Method 7010) yielded a least-squares regression line having a slope of 0.97, an intercept of 0, and a coefficient of determination (R²) of 0.97 (fig. E.3A). The slope of the regression line was not statistically different from 1 on the basis of the t-test (Neter and Wasserman, 1974) with a significance criterion of α=0.05; Cr(VI) results from the two methods were statistically similar over the range of data tested. In contrast, comparison of more than 100 Cr(t) analyses by ICP-MS (EPA Method 200.8) with Cr(t) analyses by GFAAS yielded a least-squared regression line having a slope of 1.04, an intercept of 0, and an R² of 0.99. The slope of this line was significantly different from 1, indicating Cr(t) analyses by ICP-MS differed slightly from analyses by GFAAS (not shown on fig. E.3).

Comparison of Cr(VI) and Cr(t) data analyzed at the commercial laboratory by EPA Methods 218.6 and 200.8, respectively, yielded a least-squared regression line having a slope of 0.91, an intercept of 0, and an R² of 0.90 (fig. E.3B). Similar comparison of field-speciated Cr(VI) and unspeciated Cr(t) samples with subsequent analyses by GFAAS using EPA Method 7010 (fig. E.3C) yielded a slope of 0.87, an intercept of 0, and an R² of 0.97. Both results are consistent with data from previous studies of Cr(VI) in groundwater elsewhere in California that show about 90 percent of the Cr(t) present as Cr(VI) (Izbicki and others, 2008a, 2015). However, the slopes of these two regression lines are significantly different on the basis of the t-test (Neter and Wasserman, 1974) at a significance criterion of α=0.05; that difference, about 4 percent, corresponds to the difference between the analytical methods for Cr(t). The higher R² for Cr(VI) as a function of Cr(t) for the field-speciated data (fig. E.3C) compared to Cr(VI) as a function of Cr(t) for the commercial laboratory data (fig. E.3B) indicates more consistent performance and greater precision for the field speciation data.

Samples also were analyzed for Cr(t) by ICP-MS at the USGS National Water Quality Laboratory (NWQL) as part of the suite of trace elements analyzed as part of this study (table E.1). These samples were not subsampled from the same bottle as Cr(VI) samples but were from a different bottle used for other trace elements analyzed as part of the study by ICP-MS. Samples analyzed for Cr(t) and other trace elements at NWQL were filtered in the field, acidified with nitric acid, and shipped weekly to the NWQL. The LRL for Cr(t) from the NWQL was 0.3 µg/L in 2015 and 2016 but was raised to 0.5 µg/L in 2017.

Equipment and field blanks for Cr(VI) and Cr(t) were prepared from freshly opened, reagent-grade, certified inorganic-free blank water having Cr(t) concentrations less than 0.02 µg/L. Equipment blanks were prepared in the USGS San Diego Office for equipment, other than sample pumps and dedicated tubing, in contact with sample water. Equipment blanks were analyzed prior to the March 2015, 2016, and 2017 field trips; Cr(VI) and Cr(t) were analyzed at the USGS Redox Chemistry Laboratory and other chemical constituents were analyzed at the NWQL. Field blanks were prepared in the field at randomly selected wells by each sample-collection team at approximately weekly intervals. Field blanks were prepared by pumping certified inorganic-free blank water through sample pumps, a short length of clean plastic tubing, and equipment in contact with sample water; plastic tubing dedicated to wells for purging and sample collection was not used in the preparation of field blanks. Field blanks were processed in the field in the same manner as environmental samples, and they were preserved and shipped blind to respective laboratories in the same manner as environmental samples. Field blanks were analyzed for Cr(VI) and Cr(t) at the USGS Redox Chemistry Laboratory in Boulder, Colo., and at the commercial laboratory. Field blanks were analyzed for other chemical constituents at the NWQL. Field-blank data are provided in appendix E.1 (table E.1.2).

Concentrations in equipment blanks were generally less than the Redox Chemistry Laboratory reporting levels for Cr(VI) and Cr(t) of 0.06 and 0.2 µg/L, respectively, and did not show increases in Cr(VI) or Cr(t) concentrations associated with sampling equipment. An exception to that generalization is that prior to the March 2015 sample collection, equipment blanks pumped through fittings to be attached to the sample pump showed detectable Cr(VI) and Cr(t). The fittings were designed to split flow from the pump into plastic sample lines used to collect water for chemical analyses and into copper tubing used to collect water for dissolved gases. As a consequence, the fittings were not used during the March 2015 sample collection. The fittings were redesigned to eliminate welds identified as the source of Cr(VI) and Cr(t). After successful analyses of subsequent equipment blanks, the fittings were used for the March 2016 and 2017 sample collection—decreasing sample collection time and expediting sample collection. The fittings were not used in the November 2017 sample collection because dissolved-gas samples requiring copper tubing were not collected.

Eight of 17 field blanks collected between March 2015 and March 2017 and analyzed at the commercial laboratory using EPA Method 218.6 had detectable Cr(VI); most Cr(VI) concentrations were 0.10 µg/L or less, although one field blank had a concentration of 0.28 µg/L. Similar results were obtained for blank samples collected for regulatory purposes by PG&E between March 2015 and March 2017 (data available at https://www.waterboards.ca.gov/lahontan/water_issues/projects/pge/), with 116 of 266 blank samples collected during this period having detectable Cr(VI), with concentrations as high as 0.21 µg/L in a small number of blank samples. Frequency of Cr(VI) detection was greater and concentrations were higher in field-blank samples analyzed by the commercial laboratory during March 2016, when five of six samples had detectable Cr(VI). Five of 15 blank samples analyzed using the field-speciation technique had detectable Cr(VI) with a maximum concentration of 0.09 µg/L. The frequency of Cr(VI) detection from each laboratory differed during each sample round, and the differences did not appear to be systematic because Cr(VI) detections in field blanks analyzed by one laboratory were not generally associated with Cr(VI) detections in the paired field blank that was analyzed by the other laboratory.

Six of 17 field blanks collected between March 2015 and March 2017 and analyzed for Cr(t) by the commercial laboratory using Method 200.8 had detectable Cr(t) with concentrations as high as 0.25 µg/L. Seven of 15 field blanks analyzed by GFAAS using Method 7010 at the USGS Redox Chemistry Laboratory had detectable Cr(t) at concentrations as high as 0.16 µg/L. None of the 12 field blanks analyzed at the NWQL had detections greater than the NWQL LRL for Cr(t) of 0.3 µg/L (or 0.5 µg/L in 2017).

The study reporting level (SRL) was estimated as the upper 90 percent quantile of the field-blank data. A SRL of 0.1 µg/L was assigned for Cr(VI) data analyzed by the commercial laboratory using EPA Method 218.6, and an SRL of 0.08 µg/L was assigned for the field-speciation data analyzed by the USGS Redox Chemistry Laboratory. A SRL of 0.25 µg/L was assigned for Cr(t) data from the commercial lab and 0.11 µg/L for Cr(t) data from the USGS Redox Chemistry Laboratory. The SRL for 266 Cr(VI) blank samples collected by PG&E for regulatory purposes between March 2015 and March 2017 and analyzed by the contract laboratory using EPA Method 218.6, was 0.12 µg/L—which is similar to the SRL for Cr(VI) of 0.1 µg/L estimated from blank data collected as part of this study. The binomial distribution 90/90 (BD-90/90) approach (Davis and others, 2014) used to estimate the SRL for regulatory data collected by PG&E (chapter D) was not used for data collected as part of the background study because of the smaller number of field blanks.

Analyses of field-blank data indicate that although the SRLs for Cr(VI) and Cr(t) were higher than the respective LRLs at both laboratories, there were no systematic issues associated with sample collection, handling, and analytical procedures used in this study. Field-blank data showed that SRLs assigned for Cr(VI) and Cr(t) are sufficiently low for the purposes of this study.

Between March 2015 and March 2017, 30 pairs of sequential-replicate samples were collected for Cr(VI) and Cr(t) determinations. Pumps were pulled from wells, and water levels were allowed to recover between collection of the replicate sample suite. Replicate samples were then collected, processed, and shipped to respective laboratories for analysis in the same manner as environmental data. Replicate data are provided in appendix E.1 (table E.1.2).

Sequential replicates are a suite of samples collected as close in time as possible after a suite of environmental samples were collected (Wilde, 2006). Sequential-replicate data provide information on variability introduced during collection, field processing, and shipping of samples, in addition to variability introduced during laboratory handling and analyses of samples (Wilde, 2006; Geboy and Engle, 2011). In contrast, concurrent-replicate data are a suite of samples where multiple bottles for individual analyses are collected one after the other as part of a suite of environmental samples. Concurrent-replicate data provide information on variability introduced during field processing, laboratory handling, and analysis of samples (Wilde, 2006; Geboy and Engle, 2011). Hexavalent chromium samples analyzed by the commercial laboratory and by the USGS Redox Chemistry Laboratory (fig. E.4A) are examples of concurrent replicates used to evaluate differences in laboratory analytical methods. For the purposes of this study, Cr(VI) replicate data collected by PG&E for regulatory purposes (chapter D) are concurrent-replicate data.

Paired sequential-replicate values for Cr(VI) analyzed at the commercial laboratory by EPA Method 218.6 compared favorably, with an R² greater than 0.99 (fig. E.4A), and a mean square error (MSE) of 0.06 consistent with a precision of about 6 percent at the mean value of 4.2 µg/L. Paired sequential-replicate data for field-speciated Cr(VI) samples analyzed by Method 7010 had an R² greater than 0.99 and an MSE of 0.32—consistent with a precision of about 12 percent at the mean value. Both precision values estimated from sequential-replicate data are poorer than the precision of 3 percent estimated from paired Cr(VI) regulatory data (chapter D) and reflect increased variability introduced during sample collection, field processing, and shipment of samples. On the basis of the t-test with a significance criterion of α=0.05 (Neter and Wasserman, 1974), the slope of the regression line through the paired sequential data analyzed at the commercial laboratory was statistically different from one (fig. E.4), which indicated that although the results from EPA Method 218.6 were more precise, they were less accurate over the range of values analyzed than results from the field-speciated samples; this may result from differences in the analytical procedures or from changes in dissolved-chromium species during shipment to the laboratory prior to analyses.

Samples of National Institute of Standards and Technology (NIST) traceable water, having known concentrations of Cr(VI) and Cr(t), were shipped to the field within 24 hours of preparation by a commercial vendor during the March 2015, 2016, and 2017 sample collections (table E.2). Because of the small sample volume, reference waters were not passed through sample pumps, but were otherwise processed in the field using the same techniques as field blanks. Samples were shipped blind to respective laboratories and analyzed with environmental samples. On the basis of least-squares regression analysis (Neter and Wasserman, 1974), measured Cr(VI) concentrations were similar to NIST-traceable concentrations analyzed by the contract laboratory (R² of 0.98, fig. E.4A) and by the U.S. Geological Survey Redox Chemistry Laboratory (R² of 0.97 fig. E.4B) despite the laboratories using different methods (R² values for NIST-traceable sample comparisons not shown on fig. E.4). Least-squares regression intercepts and slopes were not significantly different from zero and one, respectively, on the basis of the t-test (Neter and Wasserman, 1974) with a significance criterion of α=0.05. Regression statistics for reference waters are not shown on figures E.4A, B.

U.S. Geological Survey data were collected over a period of several years and, by design, were not a synoptic (single point in time) representation of Cr(VI) concentrations in Hinkley and Water Valleys. Hexavalent chromium concentrations in water from 14 wells collected in March 2015 were compared with concentrations in the same wells collected in March 2017 (appendix E.1, table E.1.1); there was a small, but statistically significant, decrease in Cr(VI) concentrations of 4 percent during the sample collection period (on the basis of the slope of 0.96 for the least-squares regression line fit through the data, evaluated using the t-test [Neter and Wasserman, 1974] at a significance criterion of α=0.05, not shown on figures). Only water from well MW-97S showed a statistically significant increase in Cr(VI) concentrations during this period. These data contrast with Cr(VI) concentration trend analyses of PG&E data that show generally increasing Cr(VI) concentration trends outside the mapped regulatory plume between July 2012 and June 2017 (chapter D).

To ensure comparable Cr(VI) data, the sample collection methods, equipment, and analytical procedures used in the USGS Cr(VI) background study were as similar as possible, given study constraints, to methods and equipment used by PG&E to collect data for regulatory purposes. However, a modified low-flow method was used for regulatory purposes in which only one casing volume was commonly removed from wells prior to sample collection (chapter D). Purging three casing volumes and the larger number of analyses done on samples collected by the USGS as part of the Cr(VI) background study required a greater volume of water to be pumped from monitoring wells than the volume of water pumped during regulatory sample collection. In addition, to ensure collection of representative data for dissolved gases and other constituents, water levels (where possible) were not drawn down into the screened interval of the well during purging and sample collection—this occasionally resulted in lower pumping rates and longer pumping times for USGS samples compared to samples collected by PG&E for regulatory purposes.

To address possible differences in data quality related to sample collection, USGS data for Cr(VI) were compared to PG&E regulatory data for Cr(VI) collected during the quarter prior to USGS sample collection. To provide a benchmark for this comparison, PG&E regulatory data for Cr(VI) collected the quarter prior to USGS sample collection were compared with PG&E regulatory data for Cr(VI) collected two quarters prior to USGS sample collection. It is assumed as part of this analyses that temporal trends in Cr(VI) concentrations were not large and that the quarterly samples were effectively sequential replicates collected 3 months apart. This assumption is supported by Cr(VI) concentration trend data (chapter D) and replicate data collection from wells sampled in March 2015 and March 2017 that show only small changes in Cr(VI) concentrations over time.

Hexavalent chromium concentrations in regulatory data collected from wells the quarter prior to USGS sample collection were similar to data collected from those same wells two quarters prior to USGS sample collection (fig. E.5A). The least-squares regression line fit between the two datasets had a slope of 1.02 and was not significantly different from 1, on the basis of the t-test (Neter and Wasserman, 1974) with a significance criterion of α=0.05. These results indicate that for the purpose of this discussion, Cr(VI) concentrations in water from most wells can be treated as sequential replicates collected 3 months apart. However, consistent with the greater timespan, samples collected 3 months apart showed greater variation than sequential-replicate samples collected immediately after collection of environmental samples (figs. E.4A,B)—as shown by the decrease in R² values from greater than 0.99 to 0.8 (figs. E.4A,E.5A). Increases in Cr(VI) concentrations in water from wells MW-203D, MW-153S, MW-137S3, and MW-207S1 and the decrease in Cr(VI) concentration in water from well MW-154S1, were significantly different over the 3-month sample period (falling outside the 90-percent predictive interval around the least-squares regression line shown on figure E.5A). Monotonic increasing and decreasing Cr(VI) concentration trends measured in these wells between July 2012 and June 2017 (chapter D, table D.1) were not large enough to explain changes in these wells over the 3-month sample period. Wells MW-203D, MW-153S, and MW-154S1 were difficult to sample because of low yields and measured differences likely reflect variability associated with sample collection issues in low-yielding wells.

In contrast, Cr(VI) concentrations in USGS samples differed from Cr(VI) concentrations in regulatory samples in water from the same wells sampled the quarter prior to USGS sample collection (fig. E.5B). The least-squares regression line between the two datasets had similar variability as sets of PG&E regulatory samples collected one quarter apart, with an R² of 0.83 and an MSE of 0.58; however, the slope of the line, 0.91, was significantly less than 1, on the basis of the t-test (Neter and Wasserman, 1974) with a significance criterion of α=0.05. These results indicate that the larger volume of water pumped from wells, coupled with slower pumping rates, may have resulted in Cr(VI) concentrations that were on average about 9 percent lower in USGS samples compared to PG&E samples collected for regulatory purposes. Significantly lower Cr(VI) concentrations were measured by the USGS in water from wells MW-207S1, MW-128S1 near the leading edge of the mapped Q4 2015 regulatory Cr(VI) plume, and BW-01S near the mapped plume margin by the Hinkley compressor station (falling outside the 90-percent predictive interval around the least-squares regression line shown on figure E.5B). Although not statistically significant, large decreases in Cr(VI) concentrations also were measured in USGS samples of water from wells MW-137S3, MW-97S, and MW-153S compared to PG&E regulatory data from the previous quarter (fig. E.5B). Only a few wells had increases in Cr(VI) concentrations in USGS data compared to PG&E regulatory data from the previous quarter. Significant increases were measured in water from well MW-203D between March 2015 and March 2017 (falling outside the 90-percent predictive interval around the least-squares regression line shown on figure E.5B), and although not statistically significant, large increases were measured in water from wells MW-133S1 and MW-146S (fig. E.5B). Sample collection in these wells was complicated by low well yields.

Overall Cr(VI) concentrations in USGS data were about 9 percent lower than in regulatory data from the same well collected in the previous quarter; however, in most wells the effects on Cr(VI) concentrations associated with pumping volume and rate were small. Maintaining water levels above the well screen interval during purging and sample collection necessitated lower pumping rates and longer pumping times, and may have resulted in lower Cr(VI) concentrations in samples collected by USGS compared to regulatory samples. Where present, effects on Cr(VI) concentrations associated with pumping volumes and rates varied on a well-by-well basis, and sample collection effects on Cr(VI) concentrations were more important in samples from low-yielding wells.

Laboratory methods used by the USGS NWQL for chemical constituents other than Cr(VI) are available in Fishman (1993), Fishman and Friedman (1989), and Garbarino and others (2006), and laboratory reporting levels are provided in table E.1. Equipment blanks, field blanks, and sequential replicates were collected in a similar manner as quality control data for Cr(VI) discussed previously.

Low-level detections in field-blank samples measured at the NWQL resulted in slightly larger SRLs compared to LRLs for 11 of 31 constituents measured as part of this study (table E.3). Most detections in blank samples appeared random, although low-level detections were present in about two-thirds of blank samples analyzed for calcium, strontium, and manganese—indicating systematic low-level contamination for these constituents of 0.07 milligrams per liter (mg/L), 70 µg/L, and 2.2 µg/L, respectively (table E.3). The SRL was estimated as the highest concentration in the blank data; this differs from the approach discussed previously to estimate SRLs for Cr(VI) because of the fewer number of blanks analyzed at the NWQL. Differences between the NWQL’s LRL and the SRL calculated for this study for other constituents were generally small and did not affect interpretation of the data. No constituents measured at the NWQL as part of this study were detected in equipment blanks.

Analyses of replicate data showed most constituents measured either in the field or at the NWQL had a precision, calculated as the root mean squared difference (RMSD) divided by the mean concentration expressed as a percent (Hyslop and White, 2009), at the overall mean replicate concentration of plus or minus 5 percent or less. In contrast, aluminum and iron had precision values of plus or minus 260 and 220 percent, respectively. Aluminum and iron can form colloids that are not filtered using 0.45-μm pore-sized filters; when present, these colloids can cause poor precision in analytical data. Recalculation of the RMSD after removing the two sets of paired replicate samples that had the greatest difference and were potentially affected by colloids within the water, resulted in a precision of plus or minus 3.8 and 9.1 percent for these constituents, respectively. Aluminum and iron concentrations in water from wells should be interpreted with the understanding that values for these constituents can have higher variability than other constituents measured. Precision was not calculated for constituents such as bromide, iodide, ammonia, nitrite, antimony, lead, and cadmium, which either had few concentrations greater than the LRL or replicate concentrations near the SRL.

Porewater from fresh core material, collected using an auger rig, was pressure extracted using a hydraulic device designed by Manheim and others (1994). Drilling fluids were not used, and contamination of porewater by drilling fluids was not an issue. In general, porewater was extracted in the field as soon as possible after collection under a pressure of about 5,000 pounds per square inch (psi) using techniques described by Izbicki and others (2008a). Each extraction produced between 2 and 10 mL of sample water depending on the material. An individual extraction required from 1 hour to as long as 24 hours to complete, with clay-textured materials requiring more time and yielding less water than silt-textured material. To avoid damage to the equipment, porewater was not pressure extracted from coarse-textured sand and gravel material. Most porewater samples were extracted from saturated material, although two samples were extracted from unsaturated material above the water table. Sample water from individual extractions collected at the same site and approximate depth were consolidated into a single sample to obtain about 30 mL of water required for sample processing and analyses. Although porewater samples were extracted and processed in the field at the time of core-material collection between September and December 2015, some porewater samples collected between March and May 2018 were extracted and processed off-site after extractions from clay-textured materials lagged behind drilling and core-material collection.

Porewater samples were analyzed in the field for pH and specific conductance using meters equipped with probes designed for low-volume samples. Samples for dissolved Cr(VI) and Cr(t) were filtered, field speciated, and preserved at the time of collection, using techniques described by Ball and McCleskey (2003), for analysis by GFAAS using EPA Method 7010 (U.S. Environmental Protection Agency, 2007) at the USGS Redox Chemistry Laboratory. Field speciation for Cr(VI) and Cr(t) with subsequent analyses by GFAAS requires less water than Cr(VI) analyses by ion chromatography and Cr(t) analyses by ICP-MS. Porewater samples were filtered in the field and analyzed for selected trace elements including iron, manganese, arsenic, chromium as Cr(t), vanadium, and uranium by ICP-MS at the NWQL. These samples were identified for special handling because of their low volume. Samples for the stable isotopes of oxygen and hydrogen in the water molecule (discussed in chapter F within this professional paper) were analyzed by mass spectroscopy at the Reston Stable Isotope Laboratory (RSIL) in Reston, Virginia, using methods described by Révész and Coplen (2008a, b).

Replicate porewater samples extracted from splits of core material had concentrations that agreed to within 0.06 µg/L for Cr(VI), and within 0.4 µg/L for manganese, arsenic, vanadium, and uranium. Replicate samples for iron agreed to within 6 µg/L.

The sample container and hydraulically driven piston used to pressure-extract porewater were made of stainless steel, which contains chromium. Blank samples for porewater extractions, intended to determine if equipment contributed chromium to sample water, were prepared by saturating fine-textured (60-mesh) silica sand with trace-metal grade inorganic-free blank water, certified to be chromium free. Prior to use as a blank, the silica sand was cleaned with acidified reagent-grade blank water and rinsed repeatedly with reagent-grade, inorganic-free blank water. Hexavalent chromium concentrations in porewater extracted from two silica sand blanks in the same manner as environmental samples ranged from 0.1 to 0.2 µg/L. In contrast, Cr(VI) concentrations in water extracted from silica-sand blanks using a centrifuge were less than the LRL of 0.06 µg/L. Although blank data show higher Cr(VI) concentrations associated with the pressure-extraction procedure compared to the centrifuge-extraction procedure, the differences were small relative to environmental concentrations and do not affect interpretation of porewater data. In contrast to Cr(VI), Cr(t) concentrations in the two silica blanks were 1.3 µg/L. This concentration is greater than Cr(t) concentrations in about 60 percent of sampled porewater. Manganese, arsenic, vanadium, and uranium concentrations in pressure-extracted blanks and centrifuged blank water were less than their respective reporting levels. Iron concentrations in pressure-extracted blank water ranged from 17 to 20 µg/L; iron in centrifuged blank water was less than the reporting level of 5 µg/L.

During January 27–31, 2016, more than 70 domestic and agricultural wells were sampled and analyzed for field parameters (including pH, specific conductance, and dissolved oxygen), selected anions (including chloride, sulphate, fluoride, and nitrate), selected trace elements (including arsenic, Cr(VI), Cr(t), iron, manganese, uranium, and vanadium), and the stable isotopes of oxygen and hydrogen in the water molecule (discussed in chapter F within this professional paper). Domestic wells were purged and sampled using the installed pump. Water chemistry and isotope data (including field-replicate and field-blank data) for sampled domestic wells are provided in appendix E.1 (table E.1.3).

Hexavalent chromium was analyzed by ion chromatography using EPA Method 218.6 (U.S. Environmental Protection Agency, 1994a) by the USGS in a mobile field laboratory shortly after the time of collection. Analysis of Cr(VI) in the mobile field laboratory enabled interested residents to follow water collected from their wells to the mobile lab, where it was analyzed in their presence—minimizing residents’ concerns associated with delays in laboratory analyses and tracking analytical data. In addition, Cr(VI) and Cr(t) were speciated in the field by anion exchange (Ball and McCleskey, 2003) with subsequent analyses done at the USGS Redox Chemistry Laboratory in Boulder, Colo., using EPA Method 7010 (U.S. Environmental Protection Agency, 2007). Field-blank and replicate samples were collected using techniques reported previously.

To expedite collection of a large number of samples in a short period of time, sample collection chambers used to process other samples collected as part of this study were not deployed to process and preserve samples from these domestic wells. Field-blank and replicate data for Cr(VI) and Cr(t) for field-speciated samples collected during windy conditions on January 30–31, 2016, showed increased Cr(VI) concentrations and variability compared to other blank and replicate samples collected as part of this study. Field speciated samples collected in 2-mL plastic vials appear vulnerable to contamination by wind-blown particles. The larger volume field blanks and replicate samples collected for analyses using EPA Method 218.6 in the mobile laboratory were not similarly affected.

An SRL of 0.1 µg/L was assigned for Cr(VI) data analyzed by ion chromatography using EPA Method 218.6 at the commercial laboratory used by PG&E for regulatory data collection; an SRL of 0.08 µg/L was assigned for Cr(VI) data prepared using field speciation with analyses by GFAAS at the USGS Redox Chemistry Laboratory. Hexavalent chromium results analyzed by ion chromatography and by field speciation with analyses by GFAAS were statistically similar throughout the range of data tested. Hexavalent chromium SRLs for data collected as part of this study were similar to the Cr(VI) SRL of 0.12 µg/L for regulatory data collected between 2011 and 2017 (chapter D, fig. D.4). No systematic quality concerns were identified with sample collection, handling, or analytical procedures used in this study, and Cr(VI) and Cr(t) data were suitable for the purposes of this study.

For consistency with regulatory data, Cr(VI) data analyzed by ion chromatography at the commercial laboratory are most commonly reported and discussed in this chapter. However, field-speciation data for Cr(VI) and Cr(t) were used to calculate redox potential because of the lower SRLs for Cr(VI) and Cr(t). Field-speciation data with analyses by GFAAS were reported for Cr(VI) in porewater analyses because sample volume was insufficient for analyses of porewater by ion chromatography. Hexavalent chromium data analyzed by ion chromatography using EPA Method 218.6, on site in a mobile USGS field laboratory, are reported for samples from domestic wells collected in January 2016 because samples for analyses of Cr(VI) at the commercial laboratory were not collected from those wells.

Analyses of replicate data showed most constituents measured at the NWQL had a precision of plus or minus 5 percent or less. Although, aluminum and iron had poorer precision and should be interpreted with the understanding that concentrations of these constituents can have greater variability than other constituents measured.

To ensure comparable Cr(VI) data, sample collection methods, equipment, and analytical procedures used in the USGS Cr(VI) background study were similar to methods and equipment used by PG&E to collect data for regulatory purposes. However, a modified low-flow method in which one casing volume was purged from monitoring wells prior to sample collection was used for regulatory data collection. In contrast, three casing volumes were purged from wells prior to sample collection, and a larger volume of water was withdrawn for sample collection as part of this study than was collected for regulatory purposes. In addition, water levels were monitored during purging and USGS sample collection and pumping rates adjusted to ensure (where possible) water levels remained above the screened interval of the sampled well—resulting in lower pumping rates and longer pumping time in some wells. Differences in well purging and sample collection protocols, including adjusting pumping rates to maintain water levels above the well screen interval during purging and sample collection, may have resulted in lower Cr(VI) concentrations in samples collected by the USGS compared to PG&E regulatory samples; overall, this difference was less than 9 percent. Differences in Cr(VI) concentrations associated with well purging and sample collection protocols likely vary on a well-by-well basis but were commonly greater in samples from low-yielding wells and most wells showed little difference in Cr(VI) concentrations resulting from sample collection methods. Additionally, samples of water from 14 selected wells collected in 2015 and 2017 showed a small, but statistically significant, decrease in Cr(VI) concentrations of about 2 percent per year.

By design, Cr(VI) data collected as part of the USGS Cr(VI) background study between March 2015 and November 2017 were used within a SSA to define the Cr(VI) plume extent. Hexavalent chromium data collected quarterly between April 2017 and March 2018 (discussed in chapter G within this professional paper) using PG&E sample collection protocols were used to calculate Cr(VI) background concentrations. Data used to calculate Cr(VI) background provide a 1-year snapshot in time, and the data were not impacted by pumping of larger volumes of water for purging and sample collection in the same manner as samples collected between March 2015 and November 2017.

Most statistics in this chapter were calculated using the computer program Statistical Analysis System (SAS; SAS Institute, Cary, North Carolina). Regression analysis including slope, intercept, R², and MSE was done using the method of least-squares (Neter and Wasserman. 1974). The significance of slope and intercept values relative to expected values of 1 and 0, respectively, was evaluated on the basis of the t-test (Neter and Wasserman, 1974) at a significance criterion (significance level) of α=0.05 unless otherwise stated. The 90-percent and 95-percent prediction intervals around regression lines were calculated according to methods described by Neter and Wasserman (1974) using Microsoft Excel (Microsoft, Redmond, Washington). Correlation coefficients were evaluated using Kendall’s Tau β correlation coefficient (Kendall, 1938; Helsel and Hirsch, 2002; Helsel and others, 2020). Comparison of median values were done on the basis of the median test (Neter and Wasserman, 1974). Results of statistical tests presented within the chapter were considered statistically significant at a significance criterion of α=0.05, unless otherwise stated. Probability values (p-values) for individual tests are not provided.

Principal component analysis is a multivariate technique that uses matrix algebra to transform potentially correlated and potentially non-linear data (in this case, concentrations of arsenic, Cr(VI), uranium, vanadium, iron, and manganese in groundwater) into new variables, which are uncorrelated linear combinations of the original data (Hotelling, 1933; Wold and others, 1987). The new variables are called principal components, and there is one principal component for each variable in the original dataset. The magnitude of the principal components for each measurement within the dataset is called a score; scores are calculated from eigenvectors, ranging from ˗1 to 1, that describe the contribution of each measured concentration to the principal component score. Principal component scores were used in this chapter to identify the source of Cr(VI) in water from wells. Principal component analysis was done using the computer program SAS (SAS Institute, Cary, North Carolina) and is discussed in greater detail in chapter B within this professional paper.

Specific conductance, a measure of the ability of water to conduct electricity that is related to its dissolved-solids concentration (McCleskey and others, 2011U.S. Geological Survey, 2019), ranged from 385 to 4,790 microsiemens per centimeter (µS/cm). Specific conductance was generally less than 600 µS/cm in wells near recharge areas along the Mojave River and increased as water flowed downgradient through agricultural areas in Hinkley Valley, with the highest specific conductance in water from shallower monitoring wells underlying agricultural land use. Specific conductance less than 500 µS/cm was measured in water from deeper wells in the eastern and northern subareas within Hinkley Valley and from domestic wells in locally derived alluvial-fan deposits along the flanks of Mount General (fig. E.6B). Specific conductance exceeded 1,000 µS/cm in water from some deeper monitoring wells and domestic wells in Water Valley and exceeded 1,500 µS/cm in water from wells MW-197S1, S2, and S3 in the northern subarea of Hinkley Valley (fig. E.6). Specific-conductance data collected by the USGS between March 2015 and November 2017 compare favorably with specific-conductance data collected by PG&E during Q4 2015 (fig. E.7), with an R² of 0.91, and values were spatially distributed in a similar manner.

Agricultural land use and irrigation return to groundwater in Hinkley Valley increased in the early 1950s (Stamos and others, 2001; Jacobs Engineering Group, Inc., 2019), about the same time as Cr(VI) releases began from the Hinkley compressor station. Specific conductance (along with delta oxygen-18 and delta deuterium isotope data, δ¹⁸O and δD, respectively; chapter F) provide a measure of irrigation return in water from wells. High specific-conductance values, greater than 1,000 µS/cm, consistent with irrigation return were present in water from wells downgradient from the mapped Q4 2015 regulatory Cr(VI) plume as far as wells MW-105S and MW-123S1 (fig. E.6B), but these high values were not widely present within the northern subarea. The data are consistent with water-level data (Stone, 1957; California Department of Water Resources, 1967) that show groundwater pumping reversed the water-level gradient between the eastern and northern subarea within Hinkley Valley between 1953 and 1982, thereby limiting the movement of water containing irrigation return to the north. This reversal of the water-level gradients also likely limited movement of Cr(VI) released from the Hinkley compressor station into the northern subarea during this period.

Specific conductance was not correlated with Cr(VI). Irrigation return water commonly has low Cr(VI) concentrations, likely as a result of Cr(VI) reduction to Cr(III) by biological activity within the crop root zone (Salt and others, 1995; Guertin and others, 2004). Land application of Cr(VI)-containing groundwater is used by PG&E to remove Cr(VI) at agricultural land treatment units (LTUs) operated by PG&E in Hinkley Valley (chapter A; Guertin and others, 2004; ARCADIS, 2017; Bell and others, 2019). Given the scale of agricultural pumping in Hinkley Valley and the widespread presence of irrigation return water within the mapped Q4 2015 regulatory Cr(VI) plume, it is likely that at least some anthropogenic Cr(VI) released from the Hinkley compressor station was inadvertently reduced to Cr(III) by irrigation of agricultural fields.

Nitrate in groundwater is commonly associated with irrigation return, other agricultural land uses, and septic return. Although not related to anthropogenic Cr(VI) from the Hinkley compressor station, nitrate concentrations in water from wells in Hinkley and Water Valleys are a public health concern (Lahontan Regional Water Quality Control Board, 2015). Nitrate concentrations were greater than the MCL of 10 mg/L as nitrogen (nitrate as N) in water from 20 percent of wells sampled as part of the USGS Cr(VI) background study, with the highest concentration of 36 mg/L nitrate as N in water from MW-126S2. Specific conductance in water from wells was significantly correlated with nitrate concentrations, with a Kendall’s correlation coefficient (r) of 0.71; if high specific conductance in water from deep wells in Water Valley (which are associated with geologic conditions rather than with irrigation return water) are omitted, the correlation coefficient increases to 0.77.

Major ions in water from wells include the cations calcium, magnesium, sodium, and potassium, and the anions bicarbonate (including carbonate in water having pH greater than 8.3), sulfate, and chloride. The major-ion composition of water from wells sampled as part of this study was evaluated using a trilinear diagram (Piper, 1944). The diagram shows the proportions of major cations and major anions on a charge-equivalent basis in which the total positive charge of the cations and the total negative charge of the anions theoretically sums to zero. Cations are plotted on the lower left triangle, anions (including nitrate, which has a high enough concentration to contribute to the charge balance) are plotted on the lower right triangle, and the data are integrated in the central diamond within the trilinear diagram (fig. E.8). Trilinear diagrams are useful for understanding hydrologic, geologic, geochemical, and anthropogenic processes (including irrigated agriculture) that affect groundwater chemistry. On the basis of the relative proportions of major cations and anions, water from sampled wells was divided into four groups (fig. E.8):

Major-ion composition varies within each group, and water samples from some wells have major-ion compositions that appear to represent transitions between different groups. Major-ion data are not routinely collected by PG&E from monitoring wells for regulatory purposes. The characteristics of each group are discussed in the following paragraphs.

Group 1 water, having a calcium/bicarbonate or mixed-cation/mixed-anion composition, was present in almost 40 percent of wells and was the most common composition of water from sampled wells in Hinkley and Water Valleys (fig. E.8). Calcium/bicarbonate water predominated near sources of recharge along the Mojave River, whereas mixed-cation/mixed-anion water predominated in shallow wells within the eastern and western subareas (fig. E.9). The Hinkley compressor station is located near the Mojave River, and Cr(VI) from the Hinkley compressor station would have been released into groundwater having a major-ion composition predominately within group 1. Water having a mixed-anion/mixed-cation composition was present in wells completed in Mojave River alluvium as far north as well MW-174S1 in Water Valley (fig. E.9). Water from this well is downgradient from high specific-conductance irrigation return water and may have been recharged prior to the onset of large-scale irrigated agriculture in Hinkley Valley beginning in the early 1950s (Stamos and others, 2001)—potentially predating Cr(VI) releases from the Hinkley compressor station. Water from wells in group 1 had specific-conductance values ranging from less than 600 µS/cm near recharge areas along the Mojave River to more than 2,000 µS/cm (fig. E.10A). Specific conductance increased as calcium, sulfate, and chloride proportions increased in downgradient areas underlying agricultural land use. Nitrate concentrations ranged from less than the reporting limit of 0.04 to 27 mg/L nitrate as N, with more than 20 percent of wells exceeding the MCL for nitrate (fig. E.10B). Specific conductance and nitrate concentrations were highly correlated within group 1, with a correlation coefficient of r=0.90. Water from wells within group 1 upgradient (southwest) of the Lockhart fault or within mapped strands of the fault had higher proportions of bicarbonate than water from wells on the downgradient side of the fault. Changes in major-ion composition across the Lockhart fault are consistent with water-level data (Stone, 1957; California Department of Water Resources, 1967; Stamos and others, 2001) that indicate the Lockhart fault is an impediment to groundwater flow.

Group 2 water, having a calcium/chloride-sulfate composition, was present in almost 20 percent of sampled wells in Hinkley and Water Valleys (fig. E.8). Calcium/chloride-sulfate water is associated with irrigation return to groundwater in areas underlying agricultural land use (fig. E.8) and was present in shallow wells and some deep wells in the eastern subarea and in wells MW-126S1, MW-126S2, and MW-123S1 in the northern subarea downgradient from the mapped Q4 2015 regulatory Cr(VI) plume (fig. E.9). Large-scale irrigated agriculture in Hinkley and Water Valleys began in the early 1950s (Stamos and others, 2001), near the time of Cr(VI) releases from the Hinkley compressor station. Water from wells in group 2 had specific-conductance values ranging from 680 to 4,790 µS/cm. Nitrate concentrations in group 2 water ranged from 1.1 to 36 mg/L as nitrogen (fig. E.10B). Nitrate concentrations exceeded the MCL of 10 mg/L nitrate as N in water from 50 percent of wells within group 2. Median specific-conductance values and nitrate concentrations within group 2 are significantly higher than median specific-conductance values and nitrate concentrations within group 1 (figs. E.10A, B).

Group 3 water, having a sodium/bicarbonate composition, was present in slightly more than 30 percent of sampled wells and was the second most common major-ion composition in water from sampled wells in Hinkley and Water Valleys (fig. E.8). Sodium/bicarbonate water in aquifers within the Mojave Desert (Izbicki and others, 1995; Izbicki and Michel, 2004) and elsewhere in California (Izbicki and Martin, 1997; Izbicki and others, 1995) indicates chemical changes in groundwater commonly driven by microbial respiration, consumption of oxygen, and production of bicarbonate, coupled with subsequent calcite precipitation and cation exchange within aquifers (fig. E.8). These chemical changes commonly require time to occur, and sodium/bicarbonate groundwaters tend to be isolated from surface sources of recharge by depth or by distance along long flow paths within an aquifer (fig. E.9). Water from wells in group 3 had specific-conductance values ranging from 385 to 2,020 µS/cm (fig. E.10). Low specific conductance in water is consistent with removal of calcium and bicarbonate through calcite precipitation, and the specific conductance of sodium/bicarbonate water from some wells in the northern subarea was lower than that of water near recharge sources along the Mojave River. Higher specific conductance in sodium/bicarbonate water from some wells in Water Valley results from differences in geologic material composing aquifers in that area. Although less common in the eastern subarea, strongly sodium/bicarbonate groundwater potentially isolated from surface sources of recharge was present in water from domestic wells 30E-01 and BGS-48 along the flanks of Mount General and in water from well MW-115D near mudflat/playa deposits near Mount General (fig. E.9). Nitrate concentrations in group 3 water ranged from less than the reporting limit of 0.04 to 14 mg/L nitrate as N (fig. E.10). Ninety-five percent of wells within group 3 had nitrate concentrations less than 1.2 mg/L as N, consistent with their depth within the aquifer and isolation from surface sources of recharge, including irrigation return water containing nitrate. The age (time since recharge) of sodium/bicarbonate water from sampled wells in Hinkley and Water Valleys likely predates large-scale irrigated agriculture and Cr(VI) releases from the Hinkley compressor station.

Group 4 water, having a sodium/chloride composition, was present in 10 percent of sampled wells in Hinkley and Water Valleys (fig. E.8). Sodium and chloride form salts with other major ions that are highly soluble, and sodium/chloride water may be present in low-permeability mudflat/playa deposits or other materials where water may have evaporated extensively during deposition. Water from well MW-197S1 in the northern subarea and wells MW-184S1 and 27N-01 in Water Valley, completed entirely or partly in lake margin and mudflat/playa deposits (Groover and Izbicki, 2018), have a sodium-chloride composition. Water having a sodium/chloride composition also was present in well MW-115S, near playa lake deposits near Mount General in the eastern subarea. High concentrations of sodium and chloride also are commonly associated with fluid inclusions within granitic and metamorphic bedrock; water from domestic wells BGS-34, 22-09, 22-63, and production well 27-38 completed entirely or partly in weathered bedrock or bedrock underlying the western subarea have a sodium/chloride composition. Water from wells in group 4 had specific-conductance values ranging from 640 to 2,380 µS/cm and nitrate concentrations ranging from 0.07 to 16 mg/L as nitrogen (fig. E.10). Specific conductance was inversely correlated with nitrate, correlation coefficient r=0.33, and only well 22-63, in a formerly residential area served by onsite (septic) treatment systems, had a nitrate concentration in excess of the MCL for nitrate. Similar to sodium/bicarbonate water, wells yielding sodium/chloride water tend to be isolated from surface sources of recharge, and the age of sodium/chloride water from sampled wells likely predates Cr(VI) releases from the Hinkley compressor station.

Differences in the major-ion composition of water from wells within Hinkley and Water Valleys result from differences in hydrology, geology, natural processes, and anthropogenic changes in groundwater chemistry (including agricultural activity and irrigation return) that occur along groundwater-flow paths through aquifers. Although qualitative, the differences in the major-ion composition of groundwater can be used to identify areas that are well connected to surface sources of recharge and areas where groundwater is more isolated and may predate Cr(VI) releases from the Hinkley compressor station (fig. E.8).

Wells near recharge areas along the Mojave River have water with a calcium/bicarbonate composition (group 1); Cr(VI) from the Hinkley compressor station likely would have been released into groundwater having a major-ion composition within group 1. Water with a calcium/chloride-sulfate composition (group 2), commonly having high specific conductance and nitrate concentrations, is affected by irrigation return water within Hinkley Valley and within the southern part of the northern subarea as far downgradient as well MW-123S1. The spatial distribution of water samples with mixed-cation/mixed-anion composition (group 1) showed that areas within the southern part of Water Valley as far downgradient as well MW-174S1 were recharged from the Mojave River, but recharge likely predated the onset of irrigated agriculture in Hinkley Valley and presumably predated Cr(VI) releases from the Hinkley compressor station. Sodium/bicarbonate water from wells (group 3) near the margins of the eastern and western subareas and within the northern subarea and Water Valley have reacted more extensively with aquifer materials; this water likely predates large-scale irrigated agriculture and Cr(VI) releases from the Hinkley compressor station. Similarly, sodium/chloride water from wells in bedrock and mudflat/playa deposits also have reacted extensively with aquifer materials and likely predate Cr(VI) releases from the Hinkley compressor station. Groundwater sources and ages (time since recharge) are addressed more quantitatively using chemical and isotopic tracers in chapter F within this professional paper.

Hexavalent chromium concentrations in water from wells sampled between March 2015 and November 2017 ranged from less than the SRL of 0.10 to 2,500 µg/L. The highest Cr(VI) concentrations were in water from wells within the Q4 2015 regulatory Cr(VI) plume downgradient from the Hinkley compressor station that were sampled as end-members representative of anthropogenic Cr(VI) in groundwater. Consistent with Cr(t) and Cr(VI) occurrence in alkaline, oxic groundwater elsewhere in California (Izbicki and others, 2015), approximately 90 percent of the Cr(t) outside the mapped Q4 2015 regulatory plume, was present as Cr(VI) (fig. E.3C). In contrast, approximately 97 percent of the chromium in groundwater within the Q4 2015 regulatory Cr(VI) plume was in the form of Cr(VI).

Hexavalent chromium concentrations in samples, collected by the USGS between March 2015 and November 2017, were commonly lower than Q4 2015 regulatory Cr(VI) concentrations with a least-squares regression slope of 0.76 (fig. E.7B) but otherwise were strongly correlated with an R² greater than 0.99. Lower Cr(VI) concentrations in USGS data may result from a combination of differences in sample collection methodology (fig. E.5) and changes in Cr(VI) concentrations caused by management and remediation activities within the regulatory Cr(VI) plume. If the two highest Cr(VI) concentrations within the regulatory Cr(VI) plume are excluded, then the slope of the refitted regression line increases to 0.91 (not shown on fig. E.7B) and is consistent with the 9 percent difference in Cr(VI) concentrations attributable to differences in sample collection methodology.

Hexavalent chromium concentrations in water from wells outside the Q4 2015 regulatory Cr(VI) plume were as high as 11 µg/L in water from well MW-154S1, completed in fine-textured, mudflat/playa deposits in the northern subarea (fig. E.11A). Hexavalent chromium concentrations as high as 10 µg/L were measured in water from well MW-163S, downgradient from the “western excavation site” (Lahontan Regional Water Quality Control Board, 2014), and concentrations as high as 8.9 µg/L were measured in water from well MW-203D, completed in Miocene deposits (5.3 to 23 million years old) underlying the western subarea. In addition, Cr(VI) concentrations exceeded the interim 3.1 µg/L regulatory Cr(VI) background concentration in water from some wells completed near mudflat/playa deposits near Mount General in the eastern subarea and downgradient from the Lockhart fault in the western subarea (fig. E.11B). Hexavalent chromium concentrations also exceeded the interim 3.1 µg/L regulatory background in some wells in the northern subarea and Water Valley (fig. E.11B), with Cr(VI) concentrations as high as 4.2 µg/L in water from well MW-193S1 in Water Valley. Median Cr(VI) concentrations in water from wells outside the regulatory Cr(VI) plume were significantly higher in the northern subarea and Water Valley (fig. E.12A) compared to other areas, and the median Cr(VI) concentration was significantly higher in water from wells having a sodium/bicarbonate composition (that is presumably isolated from surface sources of groundwater recharge) than other wells outside the Q4 2015 regulatory Cr(VI) plume (fig. E.12B).

Hexavalent chromium concentrations in water from almost 40 percent of wells outside the Q4 2015 regulatory Cr(VI) plume sampled by the USGS between March 2015 and November 2017 exceeded the 3.1 µg/L interim regulatory Cr(VI) background concentration (fig. E.13). In contrast, Cr(VI) concentrations in water from about 20 percent of wells outside the Q4 2015 regulatory Cr(VI) plume sampled for regulatory purposes by PG&E during Q4 2015 exceeded the 3.1 µg/L interim regulatory Cr(VI) background concentration (fig. E.13). The data are consistent with selection of wells having higher Cr(VI) concentrations for sample collection by the USGS in collaboration with the TWG. The differences in PG&E and USGS Cr(VI) data are more pronounced in the western subarea and less pronounced in Water Valley. However, both datasets illustrate that reevaluation of natural Cr(VI) occurrence in water from wells and the 3.1 µg/L interim regulatory Cr(VI) background concentration in Hinkley and Water Valleys is appropriate.

Chromium concentrations in aquifer material adjacent to the screened interval of sampled wells outside the Q4 2015 regulatory Cr(VI) plume were higher in fine-textured materials than in coarser-textured materials (fig. E.14A). However, there was no statistically significant difference in median Cr(VI) concentrations in water from sampled wells grouped by the texture of aquifer materials (fig. E.14B). The data indicate there is poor correspondence between chromium concentrations in aquifer material and Cr(VI) concentrations in water from wells. It is likely that Cr(VI) concentrations in water from wells are controlled by factors other than geochemical abundance within aquifer materials. These factors include weathering of chromium from primary mineral grains and sorption of chromium to aquifer solids (discussed in chapter C within this professional paper), coupled with aqueous geochemical factors including the redox state and pH of groundwater.