Domestic Wells

The GAMA-PBP sampled 153 domestic wells in the southeastern San Joaquin Valley during 2013–15 (Bennett and others, 2017; Shelton and Fram, 2017). The wells were selected using a grid-based design to ensure that the selected wells were distributed across the entire area. Each study area was divided into equal-area grid cells (Scott, 1990), and the grid cells were used to select wells for sampling and to aggregate data in areas with greater density of sampled wells.

The grids were designed to encompass areas where domestic wells were used for drinking-water supply. Grids were constructed to cover the entire Madera-Chowchilla, Kings, Kaweah, and Tule study areas because domestic wells were distributed across nearly the entire study area (Johnson and Belitz, 2015). Few domestic wells were located in the southern part of the Tulare Lake study area, and to avoid having grid cells without wells available for sampling, the gridded area was defined as a subset of the study area. This subset was defined as the aggregate of all 1-square mile (mi²) public land survey sections estimated to contain a non-zero number of domestic wells by Johnson and Belitz (2015). Figure 5 shows the estimated locations of people using domestic wells in 2010 (Johnson and others, 2019). Some parts of the Tulare Lake study area that were included in the gridded area, based on estimated locations of domestic wells by Johnson and Belitz (2015), were estimated to have no domestic wells in 2010. The Madera-Chowchilla and Kings study areas were divided into 80-km² grid cells, and the Kaweah, Tule, and Tulare Lake study areas were divided into 60-km² grid cells (fig. 8). Grid cells may be composed of non-contiguous pieces, and the number of cells in each study area ranged from 49 to 18 (table 1).

Wells were selected from lists of domestic wells compiled from three sources: (1) wells used by small water systems registered with County health departments or with the SWRCB-DDW, (2) domestic wells previously sampled by the USGS or the SWRCB GAMA program, and (3) domestic wells with well-completion reports filed with the California Department of Water Resources.

Water systems serving fewer than 25 people and having fewer than 15 service connections are classified as “State small systems” (California State Water Resources Control Board, 2021), and these systems were not included in the earlier GAMA-PBP assessments of groundwater resources used by public-supply wells. State small systems generally consist of one well serving several households, and the wells generally have depths similar to the depths of domestic wells serving individual households in the same area. The absence of water-quality data for State small water systems was identified as a critical information gap by the SWRCB (California State Water Resources Control Board, 2013). Wells belonging to public systems serving fewer than about 300 people also were considered as potential targets if the wells had depths similar to nearby domestic wells. Domestic wells previously sampled in 2006 by the SWRCB for a study of water quality in domestic wells in Tulare County (California State Water Resources Control Board, 2016) and domestic wells with previous water-quality or water-level data in the USGS National Water Information System (NWIS; U.S. Geological Survey, 2023b) were included as potential targets because we assumed that previously sampled wells would likely meet the criteria for sampling in this study. Domestic wells with well-completion reports in the California Department of Water Resources library of scanned images of well-completion reports (Online System for Well Completion Reports [OSWCR]; California Department of Water Resources, 2023b) provided several potential target wells. The OSWCR system locates wells at the centroid of the 1-mi² township-range-sections; for this study, many of the wells were assigned point locations using well addresses or other location information from the well-completion report (Stork and others, 2019).

Sites with potential wells on the target list were visited by USGS field personnel, beginning with the well nearest to a randomly selected location in the grid cell to ensure random selection of wells. Door-to-door canvassing continued until permission was obtained to sample a well meeting the criteria for the study or until all potential target wells on the list for that cell were exhausted. The criteria were as follows: (1) water is used for domestic drinking-water supply, (2) a well-completion report or other documentation of well depth is available, (3) samples could be collected upstream from treatment systems or tanks, and (4) the well has the capacity to pump for long enough to properly purge the well and collect the sample (Bennett and others, 2017; Shelton and Fram, 2017175).

The 5 study areas contained a total of 157 cells, and wells in 140 of the cells were sampled (fig. 8; appendix 1). For several cells that did not have available wells, a well less than 1 mile outside of the cell was selected to represent the cell (Bennett and others, 2017; Shelton and Fram, 2017). However, for consistency with Belitz and others (2015), these wells were reassigned to the cell in which they were located for this report. The 153 wells sampled for this study included 139 private domestic wells, 5 State small system or small public-supply system wells, 3 irrigation wells, and 6 monitoring wells (Stork and Fram, 2021; Balkan and others, 2024).

Irrigation and monitoring wells were sampled only when permission could not be obtained for domestic wells and when the well depths were similar to the depths of domestic wells within the township-range sections closest to the well. Depths of wells in nearby sections were obtained from Stork and others (2019). Two irrigation wells originally sampled as grid wells representing the domestic-well aquifer in the Tule study area (TLE01 and TLE08; Bennett and others, 2017) were not counted as grid wells in this report. Well TLE08 is more than 1,000 ft deeper than nearby domestic wells, and well TLE01 is in a grid cell without domestic wells, indicating that the resource was not used for domestic drinking-water supply. In addition, TLE01 and TLE08 were sampled as grid wells, TULE-11 and TULE-12, respectively, for the GAMA-PBP public-supply aquifer assessment (Burton and Belitz, 2008; Burton and others, 2012). Two of the State small system wells sampled as grid wells for the domestic-well study also were sampled as grid wells in the GAMA-PBP public-supply aquifer assessment study (TLE02 and TLE16 are the same as TULE-15 and TULE-13, respectively). However, nearby public-supply and domestic wells had similar depths; therefore, these wells were retained as grid wells for public-supply and domestic-supply aquifer assessments.

Wells were named with an alphanumeric GAMA identification code that consisted of a prefix identifying the type of GAMA-PBP study, the study unit, the study area the well was originally sampled, and a numeric suffix (appendix 1). For logistical purposes, the five study areas in this report were sampled as part of two study units in successive years, S3_MACK in 2013–14 and S4_TUSK in 2014–15. Because the study unit names are not relevant to this report, the study unit prefixes are omitted when sites are labeled on figures or discussed in this report.

Wells in the Madera-Chowchilla and Kings study areas were sampled between August 2013 and April 2014, and wells in the Kaweah, Tule, and Tulare Lake study areas were sampled between November 2014 and August 2015. The 2014–15 study also included sampling of 20 wells in the Sierra Nevada foothills east of the Kaweah and Tule study areas, referred to as the “Highlands study area” in Bennett and others (2017). Only the three sites with mapped locations within the boundary of the Kaweah study area are included in this report.

The details of methods used to collect and analyze the samples are described in Shelton and Fram (2017) and Bennett and others (2017). Field sample collection methods followed the USGS National Field Manual (U.S. Geological Survey, variously dated), with minor modifications to protocols. The primary modification was that samples were collected close to the well, using sampling lines that were approximately 0.5 meter (m) in length, and sampling chambers were not used. The USGS National Field Manual describes sampling lines approximately 10–20 m in length that are routed from the well into a sampling chamber inside a mobile laboratory. The purpose of the sampling chamber is ostensibly to protect the sample from contamination from the atmosphere during sampling, particularly for trace element and volatile organic carbon (VOC) samples. For VOCs, Fram and others (2012) determined that sampling with short lines and no sampling chambers resulted in less contamination of field blanks with VOCs than sampling with long lines and sampling chambers. Field blanks collected by the GAMA-PBP had substantially lower detection frequencies of contamination with VOCs than field blanks collected contemporaneously by other USGS projects that required use of long sampling lines and sampling chambers (Fram and others, 2012). Detection frequencies of trace elements in field blanks collected by the GAMA-PBP are consistently low and indicative of contact with metal fittings in the sampling equipment being the primary source of contamination, not airborne particles (Olsen and others, 2010; Bennett, 2020).

To increase the spatial density of data, 45 additional domestic wells sampled by the USGS NAWQA project between April 2013 and August 2015 were included in the dataset for this report, increasing the number of cells with at least 1 well to 142 cells. This inclusion increased the number of wells per cell for 32 cells that already had wells sampled for the GAMA-PBP (fig. 8; appendix 1, tables 1.1–1.5). These additional wells are part of the NAWQA networks representing the regional San Joaquin Valley aquifer system (SANJSUS1; Burow and others, 1998a; Arnold and others, 2018), orchard, vineyard, and row crop land use in the San Joaquin Valley (SANJLUSCR1A, SANJLUSOR1A; Burow and others, 1998b; Arnold and others, 2016), and a regional flow path across the Kings groundwater subbasin (CVALFPS1, CVALFPS2; appendix 1). The well selection process and methods of groundwater sample collection and analysis used by NAWQA were similar to those used by GAMA-PBP. Wells were named with an alphanumeric code that consisted of the network name and a sequence number. Because the network names are not relevant to this report, the NAWQA sites were assigned short names for this report: the prefix “NQ” and a sequence number. The full and short site names are listed in appendix tables 1.1–1.5 (appendix 1) and in Balkan and others (2024).

Public-Supply Wells

Two sources of data for public-supply wells were used for this study: (1) data from wells sampled by the USGS for the GAMA-PBP or NAWQA projects and (2) data from wells sampled for regulatory compliance by water agencies and submitted to the SWRCB-DDW. The GAMA-PBP sampled 119 public-supply wells for assessment studies in these areas during 2005–08 (Burton and others, 2012; Shelton and others, 2013) and resampled approximately 20 percent of the wells during 2008–17 (Jurgens and others, 2018). Another five public-supply wells were sampled as part of the NAWQA CVALFPS2 network in 2014 (appendix 1). The SWRCB-DDW database contained data for 69,283 samples from 1,681 public-supply wells sampled during October 2005–October 2018 (California State Water Resources Control Board, 2023b; California State Water Resources Control Board Division of Drinking Water, 2022d). Wells with data in both databases were matched using the SWRCB-DDW public supply code (PSCODE). All wells were assigned to cells based on latitude and longitude. There were 1,701 public-supply wells with water-quality data distributed across 135 cells with 1–112 wells per cell (median is 7 wells per cell; Balkan and others, 2024). Well-construction information for public-supply wells were compiled by Levy and Borkovich (2022).

Water-Quality Data

Samples collected by the USGS were analyzed for approximately 375 constituents, including field water-quality parameters, major ions, trace elements, nutrients, VOCs, pesticides, a set of isotopic and groundwater age dating tracers, and other constituents of particular interest in the State of California (Bennett and others, 2017; Shelton and Fram, 2017). The results for organic constituents are based on 183 domestic wells rather than the 198 domestic wells used for the results on inorganic constituents because some of the wells sampled for the NAWQA studies were not analyzed for organic constituents.

Samples collected at public-supply wells by water agencies for regulatory compliance data reported to the SWRCB-DDW were analyzed for fewer constituents. Regulatory compliance sampling primarily is focused on constituents for which concentrations are regulated in public drinking-water supplies; analyses of unregulated constituents and geochemical and age-dating tracers are not systematically included.

All published and quality-assured water-quality data collected by the USGS are available online from multiple sources: the USGS GAMA-PBP website (Jurgens and others, 2018), the SWRCB GAMA Groundwater Information System (California State Water Resources Control Board, 2023b), the USGS NWIS web interface (U.S. Geological Survey, 2023b), and published reports (Arnold and others, 2016, 2018; Bennett and others, 2017; Shelton and Fram, 2017). The authoritative source of USGS data is NWIS (U.S. Geological Survey, 2023b); published reports represent a snapshot of data in NWIS at the time of publication. All data compiled from the SWRCB-DDW are available online from SWRCB GAMA Groundwater Information System (California State Water Resources Control Board, 2023b) and directly from SWRCB-DDW (California State Water Resources Control Board Division of Drinking Water, 2022d).

U.S. Geological Survey data for organic constituents were censored to meet GAMA-PBP data-quality objectives. Detections that were reported to NWIS that have concentrations less than the method detection level were re-coded as nondetections (Fram and Stork, 2019; Fram, 2020; Bexfield and others, 2021, 2022). Because sampled collection dates spanned 3 years, detection levels for some constituents changed, and the highest detection level used during the period was used to censor all the data. Note that NWIS reports nondetections as less than the reporting level, and the reporting level is commonly twice the detection level (Foreman and others, 2021). For example, for 1,2,3-trichloropropane (1,2,3-TCP), the reporting level at the time the domestic well samples were analyzed was 0.006 µg/L, and the method detection level was 0.003 µg/L. For this study, the USGS data were censored such that detections with concentrations less than 0.003 microgram per liter (µg/L) were recoded as nondetections. Before 2014, some USGS samples were analyzed for 1,2,3-TCP with a method that had a reporting level of 0.18 µg/L; nondetections in samples analyzed by this method were removed from the dataset.

Many wells have been sampled more than once, and because evaluation of trends in water quality was not part of this study, one sample was selected to represent each well. For the 198 wells representing the domestic-supply aquifer, each well was sampled once between July 2013 and August 2015, and that sample was used to represent the well. For the 1,701 wells representing the public-supply aquifer, a “pseudosample” was constructed for each well (Levy and Fram, 2021); for each constituent, the analysis from the date closest to June 26, 2014 (the midpoint of the GAMA-PBP domestic well sampling period), was selected to represent the well. Sample collection was during October 2005 through October 2018. The earliest date of October 2005 was selected to include the data collected for Burton and others (2012). The pseudosample may consist of data from samples collected on different dates because samples may only be analyzed for a subset of constituents at each sampling date. About 60 percent of the data in the pseudosamples for public-supply wells were from samples collected within the same 2-year time interval as the domestic well sampling, and 85 percent of the data were collected within an 8-year window (2010–18) centered on the midpoint date of the GAMA-PBP domestic well sampling. Most of the data derived from samples outside of this 8-year window were from public-supply wells that did not have more recent data for all constituents in the SWRCB-DDW database and were sampled by the USGS during 2008–10.

Data collected by the USGS and data compiled from SWRCB-DDW are reported with different reporting levels, which affects the comparisons that can be made between water quality in the domestic-supply aquifer (based on only USGS data) and the public-supply aquifer (based on a combination of USGS and SWRCB-DDW data). For all inorganic constituents, reporting levels used by USGS laboratories are lower than the reporting levels in the SWRCB-DDW database (Burton and others, 2012). However, for all inorganic constituents except perchlorate, the USGS and SWRCB-DDW data selected for public-supply well pseudosamples had reporting levels with concentrations lower than the boundary between moderate and low concentrations for the constituent. Nondetections that were reported relative to raised reporting levels with concentrations greater than the boundary were removed from the data compilation before selection of data for the pseudosamples. Because the data analysis in this study does not distinguish between low concentrations and nondetections for inorganic constituents, the difference between USGS and SWRCB-DDW reporting levels did not affect use of inorganic constituent data, except for perchlorate data.

The differences in reporting levels for perchlorate limited comparisons. The SWRCB-DDW reporting level for perchlorate generally was 4 µg/L, which is sufficient to identify high concentrations above the MCL-CA of 6 µg/L but not to distinguish between moderate and low concentrations. The USGS reporting level was 0.1 µg/L for the domestic-well samples and either 0.1 or 0.5 µg/L for the public-supply well samples. Aquifer-scale proportions of high, moderate, and low concentrations and detection frequencies are compared among the domestic-supply aquifer study areas, but the only comparisons made between the domestic-supply and public-supply aquifer study areas are for the proportion of high concentrations. Review of the SWRCB-DDW data for perchlorate identified potential errors in reporting nondetections where some nondetections appeared to have been reported as detections at the reporting level of 4 µg/L. This type of data-reporting error was suspected because the dataset includes several instances where multiple wells belonging to a single water agency sampled during the same week all had results of detection at a concentration of 4 µg/L, and previous and subsequent samples from the same wells all had results of nondetection reported with a concentration of 0 µg/L. This type of reporting error would cause overestimation of the prevalence of moderate concentrations of perchlorate. Limiting the use of the SWRCB-DDW perchlorate data to only identify high concentrations eliminated the need to further investigate this potential data-reporting issue.

Uranium can be measured using either mass units (µg/L) or in radioactivity units (picocuries per liter, pCi/L). The uranium data collected by USGS and used in this report were measured in mass units. The SWRCB-DDW dataset includes both types of uranium data; uranium in units of pCi/L was converted to mass units using the conversion factor 0.67 picocuries per liter per microgram (pCi/µg).

The SWRCB-DDW reporting levels for most organic constituents were several orders of magnitude higher than the USGS detections levels, and the SWRCB-DDW reporting levels for many organic constituents have concentrations greater than the boundary between low and moderate concentrations. Thus, aquifer-scale proportions of high, moderate, and low concentrations and detection frequencies are compared among the domestic-supply aquifer study areas, but the only comparisons made between the domestic-supply and public-supply aquifer study areas were for the proportion of high concentrations. Review of the SWRCB-DDW data for organic constituents identified potential errors in reporting nondetections, where some nondetections appeared to have been reported as detections at the reporting level. This type of data-reporting error was suspected because the dataset includes many instances where a single sample was reported as having detections of multiple organic constituents, with all detections having concentrations equal to the reporting level. In addition, previous and subsequent samples from the same well had results of nondetection reported with a concentration of 0 µg/L for all of the constituents. Because the SWRCB-DDW reporting levels for many organic constituents are in the range of moderate concentrations, this type of reporting error would cause over-estimation of the prevalence of moderate concentrations of organic constituents. Limiting the use of the SWRCB-DDW dataset to only identifying high concentrations eliminated the need to further investigate this potential data-reporting issue.

The following criteria were used to select constituents for discussion in this report. Of the approximately 375 constituents analyzed, approximately half had a comparison benchmark and were therefore included in the assessment of water quality. Assessment results are presented for the 34 constituents that were detected at moderate or high concentrations in at least 1 domestic well or for organic constituents detected at a frequency of greater than 10 percent in domestic wells in any of the 5 study areas (table 2). Constituents detected only at low concentrations in samples from domestic wells and detected constituents lacking benchmarks are listed in tables 3 and 4. Constituents analyzed but not detected are listed in Bennett and others (2017) and Shelton and Fram (2017). A few pesticides and pesticide degradate constituents reported as detected in one or two wells in Bennett and others (2017) or Shelton and Fram (2017) are not reported as detected in this study because the concentrations were below the highest detection level used during 2013–15. A subset of the constituents listed in table 2 were selected for identification of natural and human factors that may affect concentrations of the constituents in samples from domestic wells. This subset included most constituents present at high concentrations in greater than approximately 5 percent of the domestic-supply aquifer.

The regulatory benchmarks for microbial indicator constituents specify evaluation of results from repeat sampling, which was not done for this study. Results are presented for microbial indicator constituents detected in at least one domestic well in the one-time sampling done by GAMA-PBP (table 2). The field parameter specific conductance (SC) was detected at values above the SMCL-CA benchmark, but results are not presented. All the domestic wells had data for SC and TDS. All samples with concentrations of TDS above the upper SMCL-CA benchmark for TDS (high) also had values of SC above the upper SMCL-CA benchmark for SC; all samples with concentrations of TDS between the recommended and upper SMCL-CA benchmark for TDS (moderate) also had values of SC between the recommended and upper SMCL-CA benchmark for SC (Bennett and others, 2017; Shelton and Fram, 2017). A few public-supply wells had SC data but lacked TDS data; these wells were assigned to the categories of high, moderate, or low TDS based on the SC values. Values of pH outside of the SMCL-CA range of 6.5–8.5 standard units were detected, but the framework of high, moderate, and low values relative to a benchmark was not applicable to a benchmark that is defined as a range of values. In this report, pH is used as a potential explanatory factor rather than as a water-quality constituent.

Nutrients

Nutrients have natural and anthropogenic sources in groundwater (Dubrovsky and others, 2010). Natural sources include atmospheric deposition, animal waste, and dissolution of organic material in soils. Anthropogenic sources include fertilizer application, livestock and human waste, sewage and septic effluents, and combustion of fossil fuels (emits nitrogen oxides to the atmosphere). Three nitrogenous nutrients have benchmarks: nitrate and nitrite have MCL-US benchmarks (table 2), and ammonia has a HAL-US benchmark (table 3). Orthophosphate also is a nutrient that was analyzed in samples collected for this study; however, orthophosphate does not have a drinking-water benchmark (table 3) and is not included in the status assessment. The most common forms of dissolved nitrogen in groundwater are nitrate, nitrite, ammonia or ammonium, and dissolved nitrogen gas. The form of nitrogen that dominates depends on redox conditions (McMahon and Chapelle, 2008). Nitrate is the most oxidized form of nitrogen and is the most common form in oxic groundwater systems. Although evapotranspiration from shallow groundwater can increase concentrations of nutrients in groundwater, concentrations of nitrate that are greater than about 2 mg/L (relative concentration of 0.2) generally result from anthropogenic inputs (Nolan and others, 2002; Dubrovsky and others, 2010).

Nutrients were present at high concentrations in 27 percent of the SESJV-D and at moderate concentrations in 25 percent of the SESJV-D (table 5). Nitrate was the only nutrient present at high concentrations (greater than the MCL-US of 10 mg/L) and also accounted for most of the moderate concentrations of nutrients (fig. 18A). The proportion of SESJV-D with high concentrations of nitrate ranged from 8.9 percent in Tulare Lake to 40 percent in Kaweah, and the proportions of high concentrations were greater in the Kings and Kaweah study areas than in the Madera-Chowchilla and Tulare Lake study areas (tables 6–10; fig. 19). Nitrite was present at moderate concentrations (greater than 0.5 mg/L and less than or equal to 1 mg/L) in the Kaweah and Tulare Lake study areas (tables 6–10; fig. 18A). None of the samples with moderate concentrations of nitrite had moderate or high concentrations of nitrate (Bennett and others, 2017).

Nitrate also was the only nutrient present at high concentrations in the public-supply aquifer (table 11). The proportion of the SESJV-P with high concentrations of nitrate (10 percent) was significantly lower than the proportion in the SESJV-D (27 percent; table 11). The proportion with high concentrations of nitrate was significantly greater in the domestic-supply aquifer than in the public-supply aquifer in the Kings, Kaweah, Tule, and Tulare Lake study areas (tables 6–10; fig. 19).

The spatial distribution of elevated nitrate concentrations in San Joaquin Valley groundwater primarily is explained by sources of nitrate from fertilizer inputs in agricultural areas and by redox processes that cause degradation of nitrate in anoxic groundwater (Burow and others, 2008; Landon and others, 2011; Nolan and others, 2014; Ransom and others, 2017). The application of nitrogenous fertilizers increased substantially beginning in the 1950s, and nitrogenous fertilizers have been applied to agricultural lands across the southeastern San Joaquin Valley (Burow and others, 2013). In SESJV-D, nitrate was not significantly correlated with percentage of agricultural land use (Spearman test, p=0.201, rho=0.09). This apparently counter-intuitive result indicates that nitrate behaved as a widespread non-point source contaminant in the southeastern San Joaquin Valley and that factors in addition to land use affect nitrate concentrations in groundwater. Nearly all the domestic-supply wells were in areas with agricultural land use (figs. 10, 20): the median percentage of agricultural land use with 500-m of the wells was 88 percent with interquartile range of 70 percent to 93 percent agricultural land use (Balkan and others, 2024). The amount of nitrogenous fertilizer applied to agricultural land varies with crop type and cultivation practices, therefore although presence of agricultural land use generally indicates use of nitrogenous fertilizers at the land surface, the percentage of agricultural land use does not necessary correlate with the magnitude of use of nitrogenous fertilizers near the well. Detailed spatial analysis of nitrogenous fertilizer use in the southeastern San Joaquin Valley was outside of the scope of this report. In the absence of such an analysis, the variable used in this report as a proxy for nitrogenous fertilizer sources at the land surface, percentage of agricultural land use, is an imprecise estimate, hence lack of statistically significant correlation between nitrate concentrations in groundwater and percentage of agricultural land use in areas dominated by agricultural land use was not unexpected.

High and moderate nitrate concentrations were detected throughout the central and eastern parts of the southeastern San Joaquin Valley on the proximal and intermediate areas of the alluvial fans and in interfan areas (fig. 20). Nitrate concentrations were lowest along the western edge of the southeastern San Joaquin Valley (fig. 20), on the distal parts of the alluvial fans at the ends of regional flow paths near the center of the basin where anoxic groundwater conditions are most prevalent (fig. 14). This spatial distribution of elevated nitrate concentrations in domestic wells was consistent with Ransom and others (2017) prediction of the widespread distribution of elevated nitrate concentrations at the depth zone used by domestic wells in the southeastern San Joaquin Valley. The spatial distribution of high concentrations of nitrate in the SESJV-D was similar to the spatial distribution of high and moderate concentrations of nitrate in the SESJV-P (compare SESJV-D data shown on fig. 20 to SESJV-P data shown on fig. 17G in Burton and others, 2012 and fig. 14 in Shelton and others, 2013).

The spatial distribution of elevated nitrate concentrations largely matched distributions of redox conditions that affect nitrate stability and ages of groundwater accessed by the wells. Nitrate concentrations in oxic groundwater were greater than nitrate concentrations in anoxic or mixed-redox groundwater (Wilcoxon, p<0.001). Nitrate concentrations also were greatest in modern groundwater and least in premodern groundwater (Kruskall-Wallis, p<0.001). Both oxic conditions and presence of at least some modern groundwater were needed for elevated nitrate concentrations; modern or mixed groundwater with anoxic or mixed redox conditions and premodern groundwater with oxic redox conditions had lower nitrate concentrations than modern or mixed groundwater with oxic redox conditions (fig. 21). Anoxic conditions were most prevalent in wells in the basin deposits along the western edge of the southeastern San Joaquin Valley (figs. 13, 14) where nitrate concentrations were low (fig. 20). Anoxic conditions dominated the Tulare Lake study area; therefore, elevated nitrate concentrations were absent even though the Tulare Lake study area had shallow wells that tapped modern-age or mixed-age groundwater (fig. 11) in areas with sources of agricultural nitrate. Nitrate isotope data indicated that denitrification reduced nitrate concentrations in anoxic groundwater from wells in the San Joaquin Valley basin deposits and Tulare Lakebed area (Landon and others, 2010; Fram, 2017a).

The Madera-Chowchilla study area was dominated by oxic conditions that typically favor the preservation of nitrate (figs. 13, 14). However, the Madera-Chowchilla study area had lower prevalence of high nitrate concentrations than the Kings and Kaweah study areas (fig. 19), even though it had a greater prevalence of oxic conditions (figs. 13, 14). The lower prevalence of high nitrate concentrations in the Madera-Chowchilla study could be due to the age of the groundwater and land use. Agricultural nitrate is primarily associated with modern groundwater (Burow and others, 2013), and the Madera-Chowchilla study area had a substantially greater proportion of wells with premodern or mixed-age groundwater than the Kings and Kaweah study areas (fig. 12). The Madera-Chowchilla study area also had the greatest proportion of natural land use (fig. 10), with a corresponding lower proportion of agricultural land use.

Greater prevalence of high nitrate concentrations in SESJV-D compared to SESJV-P was consistent with the observed differences in hydrologic and geochemical characteristics. In all study areas, except Madera-Chowchilla, the domestic wells were in areas with higher percentages of agricultural land use than the public-supply wells (fig. 10), and greater nitrate loading at the land surface generally is associated with agricultural land use. Domestic wells were significantly shallower than public-supply wells in all five study areas (fig. 9) and thus were more likely to access groundwater closer to the land surface where nitrogen fertilizers were applied. In all five study areas, a higher proportion of domestic wells tapped modern groundwater that was more likely to contain nitrate, and in all study areas except Kaweah, a lower proportion of domestic wells tapped premodern groundwater that was unlikely to contain elevated nitrate (fig. 12).

Trace Elements

The constituent class of trace elements included a variety of metallic and non-metallic constituents that typically are present in groundwater at concentrations less than 1 mg/L. At least one trace element with an MCL was detected at high RCs in 30 percent and at moderate RCs in 18 percent of SESJV-D (table 5). The aquifer scale proportion of the SESJV-D high RCs for any trace element with an MCL ranged from 10 percent in Madera-Chowchilla to 60 percent in Tulare Lake (table 5). Arsenic and uranium were present at high and moderate concentrations in all five study areas (tables 6–10; fig. 18B) and are discussed in more detail later in the text. Fluoride was present at high concentrations in less than 1 percent of SESJV-D, and barium was only detected at moderate concentrations (table 11; fig. 18B).

Five trace elements with non-regulatory health-based benchmarks—boron, molybdenum, vanadium, strontium, and manganese—were present at high or moderate concentrations (table 11; fig. 18C). Manganese has a HAL-US benchmark of 300 µg/L and an SMCL-CA benchmark of 50 µg/L, and both benchmarks are used in this report. Manganese was present at concentrations above the HAL-US in 4.9 percent of SESJV-D and is discussed in the “Constituents with Secondary Maximum Contaminant Level Benchmarks” section. Molybdenum was present at high concentrations in 3.4 percent of SESJV-D and is discussed in more detail in the next paragraph. Boron, vanadium, and strontium were only detected at moderate concentrations (table 11; fig. 18C).

The prevalence of high concentrations of boron, molybdenum, vanadium, and strontium in SESJV-D could not be compared to the prevalence of high concentrations in SESJV-P because the available data were insufficient for these four constituents in the public-supply well dataset used for this study. The SWRCB-DDW dataset did not include a lot of data for these constituents because they do not have regulatory benchmarks. GAMA-PBP sampling of public-supply wells in the Madera-Chowchilla study area (Shelton and others, 2009) included analysis of trace element in all wells, but sampling in the Kings, Kaweah, Tule, and Tulare Lake study areas (Burton and Belitz, 2008) only included analysis of trace elements at a subset of wells. For this study, aquifer-scale proportions were only calculated for constituents and study areas for which at least half of the grid cells had data for the constituent (tables 6–11).

Boron was detected at concentrations greater than the NL-CA of 1,000 µg/L in four wells at the western edge of the Tulare Lake and Kings study areas (not shown; Bennett and others, 2017), which was similar to locations of public-supply wells with boron concentrations exceeding the NL-CA (fig. 17C in Burton and others, 2012). Concentrations greater than the NL-CA were defined as moderate in this report (table 2) but were defined as high in all reports for the GAMA-PBP statewide assessments of water quality in groundwater resources used for public drinking-water supply, including the ones covering areas of the southeastern San Joaquin Valley (Burton and others, 2012; Shelton and others, 2013; Belitz and others, 2015; Fram, 2017a). The public-supply assessment for the Westside study area in the western San Joaquin Valley adjacent to the Tulare Lake and Kings study areas indicated that 45 percent of the public-supply aquifer had boron concentrations greater than the NL-CA and that elevated boron concentrations generally were associated with groundwater recharge derived from Coast Ranges watersheds that contained boron derived from marine shales or from geothermal sources (Fram, 2017a). Elevated boron concentrations at the western edge of the southeastern San Joaquin Valley may indicate mixing of groundwater recharge derived from the Coast Ranges to the west and Sierra Nevada to the east in the center of the San Joaquin Valley.

Vanadium was detected at concentrations greater than the NL-CA of 50 µg/L in four wells in the interfan region between the Kings and Kaweah River fans (not shown; Bennett and others, 2017; Shelton and Fram, 2017). Concentrations greater than the NL-CA are defined as moderate in this report (table 2) but were defined as high in Burton and others (2012), Shelton and others (2013), Belitz and others (2015), and Fram (2017a). Wright and Belitz (2010) documented that in a statewide dataset, vanadium concentrations were positively correlated with pH values and that higher concentrations of vanadium were associated with locations likely to have aquifer sediments derived from mafic materials and oxic groundwater conditions. However, the relation between vanadium and pH for SESJV-D samples was not statistically significant (Spearman test p=0.182, rho= −0.10) and Burton and others (2012) and Shelton and others (2013) determined that the relation between vanadium and pH was not significant for SESJV-P samples. This lack of significant relation between vanadium and pH could indicate that local variability in sediment composition is an important factor in determining vanadium concentrations in groundwater in the southeastern San Joaquin Valley.

Arsenic

An estimated 8 percent of groundwater resources used for drinking water in the United States have high concentrations of arsenic (Focazio and others, 2000; Welch and others, 2000), and high concentrations of arsenic in groundwater used for drinking water are a worldwide concern (Smedley and Kinniburgh, 2002; Welch and others, 2006). The EPA lowered the MCL-US for arsenic from 50 to 10 μg/L in 2001 (U.S. Environmental Protection Agency, 2001). Natural sources of arsenic in groundwater include dissolution of arsenic-bearing minerals, desorption of arsenic from mineral surfaces, and mixing with hydrothermal fluids (Smedley and Kinniburgh, 2002; Welch and Stollenwerk, 2003). Potential anthropogenic sources of arsenic can include copper ore smelting, coal combustion, arsenical pesticides, and wood preservatives (Welch and Stollenwerk, 2003). In addition, mining for copper, gold, and other metals can increase the rate of dissolution of natural arsenic-bearing minerals (Smedley and Kinniburgh, 2002).

Arsenic was present at high concentrations (above the MCL-US of 10 µg/L) in 11 percent and at moderate concentrations in 5.4 percent of SESJV-D (table 11). The aquifer-scale proportions of high concentrations of arsenic for the five SESJV-D study areas fell into three groups that meet the criteria for significant difference (see discussion of equation 3 in the “Statistical and Graphical Methods” subsection of the “Methods” section). The proportions of the Madera-Chowchilla, Kings, and Kaweah study areas with high concentrations of arsenic were 3.7, 6.8, and 4.8 percent, respectively, but these proportions were not significantly different from each other because of the overlap of the values with the 90-percent confidence intervals (all three had similar confidence intervals of approximately 1.2–13 percent; tables 6–10). The proportion of the Tule study area with high concentrations of arsenic was significantly greater than the proportion in the Madera-Chowchilla, Kings, or Kaweah study areas because the proportion in Tule (18 percent) was greater than the upper confidence limit for the others (13 percent), and the proportions in the others, 3.7–6.8 percent, were lower than the lower confidence limit for Tule (9.2 percent). The proportion of the Tulare Lake study area with high concentrations of arsenic (49 percent, with confidence interval of 31–67 percent) was significantly greater than the proportion in the Tule study area (18 percent, with confidence interval of 9.2–30 percent; tables 6–10; fig. 22). Most of the wells with high or moderate concentrations of arsenic were along the western edge of the southeastern San Joaquin Valley in the basin and lakebed deposits and in the distal parts of the Kings River fan (fig. 23).

The aquifer-scale proportions of high concentrations of arsenic and the spatial distributions of wells with high arsenic concentrations in the SESJV-D and SESJV-P were compared to help infer the processes controlling the presence of high arsenic concentrations in groundwater. The proportion of SESJV-P with high concentrations of arsenic, 16 percent, was greater than the proportion of the SESJV-D with high concentrations (11 percent; table 11; fig. 22). The proportions with high concentrations were greater in the SESJV-P in all five study areas, although the difference was only significant in the Kings study area (tables 6–10). The locations of wells with high concentrations of arsenic in the Kings, Kaweah, Tule, and Tulare Lake study areas were similar (compare SESJV-D data shown on fig. 23 to SESJV-P data shown on fig. 17A in Burton and others, 2012), but locations in the Madera-Chowchilla study area were different (compare SESJV-D data shown on fig. 23 to SESJV-P data shown on fig. 17 in Shelton and others, 2013). The greater proportion of high arsenic concentrations in the SESJV-P and the difference in spatial distribution of wells with high arsenic concentrations can be explained by redox and sorption processes controlling arsenic solubility, the differences in well depth and groundwater age between the SESJV-P and SESJV-D wells, and the spatial variations in the compositions of sediments in the southeastern San Joaquin Valley.

Redox and sorption processes were dominant controls on arsenic solubility in groundwater in the southeastern San Joaquin Valley. Arsenic mobilization under iron- and manganese-reducing conditions has been attributed to reductive dissolution of iron and manganese oxyhydroxides, a process by which arsenic that is sorbed or co-precipitated with iron and manganese oxyhydroxides is released (Smedley and Kinniburgh, 2002; Belitz and others, 2003; Welch and others, 2006). Reducing conditions may also increase arsenic mobilization by reducing arsenate that is sorbed on mineral surfaces to arsenite, which is less strongly sorbed (Zobrist and others, 2000).

Sorption of arsenic decreases as pH increases; at pH values greater than about 7.8, the primary arsenic species is hydrogen arsenate, which is relatively soluble (Smedley and Kinniburgh, 2002). The highest median arsenic concentrations were measured in groundwater that had pH values greater than 7.7 and had either mixed or anoxic redox conditions (fig. 24). The maximum arsenic concentration in groundwater with pH less than 7.7 and oxic conditions was 5.5 µg/L. Groundwater with high pH and anoxic conditions was most prevalent near the Tulare Lake Bed (figs. 13, 14, 15). Welch and others (1997) and Welch and Lico (1998) inferred that high arsenic concentrations in Carson Desert (not shown) groundwater were caused by interaction between lacustrine and riverine sediments and groundwater with high pH and low dissolved oxygen concentrations.

Fujii and Swain (1995) determined that evaporative concentration contributed to increased arsenic concentrations in shallow groundwater in the Tulare Lake Bed region; however, the groundwater in domestic wells was not extensively evaporated (see discussion in the “Total Dissolved Solids” section), indicating evaporation was not an important mechanism for increasing concentration of constituents in domestic wells.

Contribution of arsenic from legacy use of arsenical pesticides cannot be ruled out; however, relations between arsenic concentrations and indicators of contamination from agricultural sources indicate legacy use of arsenical pesticides are not the primary source of arsenic. Arsenic concentrations were inversely correlated with concentrations of fumigants (Spearman test, DBCP, p=0.002, rho= −0.23; 1,2,3-TCP p=0.013, rho= −0.18) and pesticides (Spearman test, atrazine+simazine, p<0.001, rho= −0.37, total pesticides and degradates p<0.001, rho= −0.29).

The geochemical conditions favorable to arsenic solubility commonly are associated with hydrologic conditions that may enhance accumulation of arsenic in groundwater. Long contact times between groundwater and aquifer materials increase the reaction time with aquifer materials; therefore, elevated arsenic concentrations are often associated with older groundwater ages, decreased rates of flushing and recharge, and increased distance along groundwater flow paths (Smedley and Kinniburgh, 2002; Anning and others, 2012).The greater prevalence of high arsenic concentrations in the SESJV-P compared to the prevalence of high arsenic concentrations in the SESJV-D in the Madera-Chowchilla, Kings, Tule, and Tulare Lake study areas (fig. 22) is consistent with the greater prevalence of premodern groundwater in the SESJV-P compared to the SESJV-D for those study areas (fig. 12).

Molybdenum

Molybdenum is a metallic trace element used in high-strength steel alloys, dry lubricants, and other industrial products. High concentrations of molybdenum are present in organic-rich sediments and sedimentary rocks deposited in sulfide-reducing environments (Crusius and others, 1996). However, most molybdenum ore deposits are associated with porphyry granite or quartz monzonite plutons, and the primary ore mineral is molybdenite (MoS₂; Misra, 2000). Molybdenum has a non-regulatory, health-based benchmark (table 2) but is listed on EPA Contaminant Candidate Lists 3, 4, and 5 (U.S. Environmental Protection Agency, 2023). Under the Safe Drinking Water Act, the EPA (1) lists constituents that are not subject to any proposed regulations at the time the list is compiled but are known or anticipated to be present in public water systems at concentrations that might be of concern, (2) collects information about their presence and potential health effects, and (3) determines whether or not to start a process to evaluate setting an MCL for a constituent (U.S. Environmental Protection Agency, 2023).

Molybdenum was present at high concentrations (greater than the HAL-US of 40 µg/L) in 3.4 percent and at moderate concentrations in 2.0 percent of the SESJV-D (table 11). The proportions of the five study areas with high concentrations of molybdenum fell into three categories: the proportions in Madera-Chowchilla, Kaweah, and Kings study areas ranged from 0 to 2.4 percent and were not statistically significantly different from each other; the proportion was 7.1 percent in the Tule study area and 13 percent in the Tulare Lake study area (tables 6–10). The prevalence of high concentrations of molybdenum in the SESJV-D could not be compared to the prevalence in the SESJV-P because there were insufficient data for molybdenum concentrations in most of the SESJV-P (tables 6–11).

Molybdenum concentrations in groundwater primarily are controlled by groundwater redox conditions, pH, and availability of molybdenum from aquifer materials (Evans and Barabash, 2010; Harkness and others, 2017; Smedley and Kinniburgh, 2017). Molybdenum concentrations had a strong positive correlation with arsenic concentrations (Spearman test, p<0.001, rho=0.58). Similar to arsenic, two primary mechanisms for enhanced solubility of molybdenum in groundwater are (1) reductive dissolution of oxides or sulfides and (2) inhibition of sorption of molybdenum to mineral surfaces at higher pH (Evans and Barabash, 2010; Harkness and others, 2017; Smedley and Kinniburgh, 2017180).

Most of the domestic wells with moderate or high concentrations of molybdenum were near the margin of the Tulare Lake Bed (fig. 25) where anoxic conditions and higher pH conditions are more prevalent (figs. 13, 14, 15). Shallow groundwater from monitoring wells in the Tulare Lake Bed had molybdenum concentrations as high as 15,000 µg/L (Fujii and Swain, 1995). The greater prevalence of elevated molybdenum concentrations in groundwater near the Tulare Lake Bed compared to other areas of the southeastern San Joaquin Valley that have geochemical conditions favorable for molybdenum solubility (higher pH, older groundwater, reduced conditions) is likely related to the distribution of sediments that contain higher concentrations of molybdenum. Sediments near the Tulare Lake Bed include a higher proportion of sediments derived from marine sedimentary sequences in the Coast Ranges (Fujii and Swain, 1995). Sediments eroded from the Coast Ranges interfinger with sediments eroded from the Sierra Nevada along the trough of the San Joaquin Valley (Belitz and Heimes, 1990), and sediments derived from organic-rich shales in the Coast Ranges are a source of boron, molybdenum, selenium, TDS, and other inorganic constituents to groundwater in the western San Joaquin Valley (Deverel and Millard, 1988; Dubrovsky and others, 1991; Fujii and Swain, 1995; Fram, 2017a).

Uranium

Radioactive decay of uranium and thorium and their radioactive progeny are the main sources of natural radioactivity in groundwater (Hem, 1985). Ingestion of uranium may cause adverse health effects due to the radioactivity of uranium, but the primary health effect leading to establishment of the MCL-US of 30 µg/L is noncancer kidney toxicity (U.S. Environmental Protection Agency, 2000).

Uranium was present at high concentrations in 20 percent and at moderate concentrations in 13 percent of the SESJV-D (table 11). The proportions of the five study areas with high concentrations of uranium fell into three categories: (1) the proportions in Madera-Chowchilla and Tule were 6.5 and 7.1 percent, respectively, and were not statistically significantly different from each other; (2) the proportions in Kaweah and Tulare Lake were 19 and 22 percent, respectively, and were not statistically different form each other; and (3) the proportion in Kings was 39 percent (tables 6–10; fig. 26). High concentrations of uranium were detected primarily in wells on the San Joaquin, Kings, and Kaweah River fans in the intermediate and distal areas of the fans (fig. 27).

The aquifer-scale proportions of high concentrations of uranium and the spatial distributions of wells with high uranium concentrations in the SESJV-D and SESJV-P were compared to help infer the processes controlling the occurrence of high uranium concentrations in groundwater. The proportion of the SESJV-D with high concentrations of uranium, 20 percent, was higher than the proportion in the SESJV-P, 10 percent (table 11; fig. 26). The largest differences were observed in the Kings and Kaweah study areas (tables 6–10; fig. 26). In the Kings and Kaweah study areas, the area dominated by high concentrations of uranium in SESJV-D had a mixture of high, moderate, and low concentrations in SESJV-P (compare SESJV-D data shown on fig. 27 to SESJV-P data shown on fig. 17F in Burton and others, 2012 and fig. 21 in Shelton and others, 2013).

The spatial distribution of high concentrations of uranium in groundwater in the study area can be related, in part, to the spatial distribution of high concentrations of uranium in surficial rocks and sediment (Rosen and others, 2019). The highest concentrations of uranium in surficial rocks and sediment were related to the source of the sediment in the Sierra Nevada that was deposited in the alluvial fans and valleys. Alluvial fans deposited by the San Joaquin, Kings, and Kaweah Rivers drain large, glaciated watersheds in the Sierra Nevada that are dominated by granitic rocks (Weissmann and others, 2005). Granitic rocks of the Sierra Nevada contain uranium-bearing accessory minerals (Wollenberg and Smith, 1968; Ague and Brimhall, 1988; Thomas and others, 1993). Estimated uranium concentrations in surficial sediments (Hill and others, 2009) were higher in the San Joaquin, Kings, and Kaweah River fans than in the Chowchilla and Fresno River fans or the interfan areas (fig. 27; Rosen and others, 2019). Leaching experiments indicated that a substantial part of the uranium on southeastern San Joaquin Valley sediments was sorbed to mineral surfaces and not readily soluble in natural conditions (Jurgens and others, 2008).

Differences in the prevalence of high uranium concentrations between SESJV-D and SESJV-P and in the spatial distribution of elevated uranium concentrations in the SESJV-D were expected based on the conceptual models of Jurgens and others (2010), Burow and others (2017), and Rosen and others (2019). In SESJV-D, uranium concentrations were positively correlated with bicarbonate (fig. 28), TDS, and nitrate concentrations (Spearman p<0.001, rho=0.62, 0.58, and 0.46, respectively). Uranium concentrations also were greater in modern- and mixed-age groundwater compared to premodern-age groundwater (Kruskall-Wallis, p<0.001, Dunn’s p<0.001, and p=0.007, respectively; fig. 28). Uranium also was inversely correlated with depth to top of screened interval in the SESJV-D (Spearman, p=0.007, rho=0.22) and SESJV-P (Burton and others, 2012; Shelton and others, 2013). Jurgens and others (2010) and Burow and others (2017) documented that higher concentrations of uranium were associated with modern, oxic groundwater sourced from agricultural irrigation return flow, and higher concentrations of uranium were associated with higher concentrations of other constituents linked to agricultural irrigation, including bicarbonate, TDS, and nitrate.

Laboratory experiments (Jurgens and others, 2008) and geochemical modeling (Jurgens and others, 2010) indicated that the correlation between uranium and bicarbonate was due to formation of soluble uranium-bicarbonate complexes from uranium that had been sorbed on iron-oxyhydroxide coatings, sediment grains, clay mineral edges, or to organic matter in aquifer sediments. The correlation between uranium and nitrate did not indicate oxidation of reduced uranium minerals by agricultural nitrate as proposed by Nolan and Weber (2015); rather, uranium and nitrate were correlated because both have elevated concentrations in groundwater containing recharge from agricultural irrigation return flow (Rosen and others, 2019). The irrigation-return flows contain nitrate from application of fertilizers and uranium mobilized from aquifer sediments by the elevated bicarbonate produced by plant productivity under irrigated conditions. The greater proportions of high concentrations of nitrate and uranium in the SESJV-D compared to the proportions in the SESJV-P (figs. 19, 26) indicate that recharge from irrigation-return flows has reached the depth zone used by domestic-supply wells in many areas of the southeastern San Joaquin Valley. Trends of increasing nitrate and uranium concentrations in public-supply wells indicate that this front of groundwater containing elevated nitrate and uranium concentrations is gradually reaching the greater depth zones used by public supply wells (Jurgens and others, 2010, 2020).

Some of the highest uranium concentrations are in anoxic or mixed-redox groundwater at the distal ends of the alluvial fans (figs. 27, 28; Rosen and others, 2019). At a given bicarbonate concentration, these anoxic and mixed redox wells have higher uranium concentrations than oxic wells (fig. 28), indicating that there is an additional source of uranium contributing to anoxic and mixed wells that does not contribute to oxic wells. Rosen and others (2019) observed that these wells with anoxic or mixed redox groundwater and higher uranium concentrations (fig. 28) are located near the boundary of the old discharge zone of the groundwater system (see fig. 10 in Rosen and others, 2019). This old discharge zone had artesian conditions in the early 1900s before extensive pumping began (Mendenhall, 1908; Mendenhall and others, 1916). This boundary between recharge and discharge zones also was a redox boundary, and uranium may have accumulated in the sediment at the boundary by processes similar to formation of uranium roll-front mineral deposits (Rosen and others, 2019). Uranium roll-front deposits form when oxic groundwater with dissolved uranium crosses from oxic to anoxic zones along a groundwater flow path in the aquifer, and there is sufficient uranium to precipitate as uraninite or other reduced uranium minerals (Harshman, 1975; Abzalov, 2012). If the redox conditions in the groundwater at the boundary later change, then the uranium minerals may dissolve. Rosen and others (2019) observed that the domestic wells with anoxic or mixed redox conditions and high uranium concentrations all have modern or mixed age groundwater and geochemistry consistent with being mixtures of younger oxic water and older anoxic water. Rosen and others (2019) concluded that the anomalously high uranium concentrations in these wells may be the result of dissolution of uraninite at the old redox boundary in the aquifer system by modern recharge that is undersaturated with uraninite because it is oxic and has high bicarbonate concentrations. The spatial pattern of recharge in the San Joaquin Valley has changed from a natural system where recharge primarily entered the groundwater system at the sides of the San Joaquin Valley near the mountains and along the major rivers and flowed toward discharge zones in the San Joaquin Valley trough area, to a modern system where agricultural irrigation return flow enhances recharge across the entire San Joaquin Valley and the combination of groundwater pumping and recharge of applied irrigation water results in much greater rate of downward movement of modern water into the aquifer system (Williamson and others, 1989; Burow and others, 2004; Phillips and others, 2007; Faunt, 2009; Jurgens and others, 2016). Haugen and others (2021) examined trends in arsenic concentrations in public-supply wells in the San Joaquin Valley and concluded that decreasing trends in arsenic concentrations in wells from the San Joaquin Valley trough area may be caused by decreased solubility of arsenic as modern, oxic groundwater moves into the capture zones of wells that previously accessed only older groundwater.

Radioactive Constituents

Most of the radioactivity in groundwater comes from the decay of naturally occurring uranium and thorium in the rocks or sediments that compose the aquifers (Hem, 1985; Appelo and Postma, 2005). Radioactive decay of uranium and thorium isotopes produces a long series of radioactive daughter products, including isotopes of radium, uranium, and radon. These elements have different chemical properties, and their solubility in groundwater varies with geochemical conditions, water chemistry, and aquifer mineralogy (Hem, 1985). In this study, we assessed data for adjusted gross alpha and adjusted gross beta particle activities, which is consistent with how the benchmarks for gross alpha and adjusted gross beta are defined.

Adjusted gross alpha particle activity is the measured gross alpha particle activity minus the alpha activity accounted for by the concentration of uranium in the sample. Uranium concentrations were converted to activities using a conversion factor of 0.67 pCi/µg. Adjusted gross beta particle activity is the measured gross beta particle activity minus the beta activity attributed to potassium-40. Potassium concentrations were converted to activities using a conversion factor of 0.82 pCi/mg (Welch and others, 1995). The measured values of gross alpha and beta depend on the initial composition of the water sample and on the time intervals between sample collection, sample preparation, and analysis. The differences among measured values at various combinations of time intervals can be used to estimate the relative abundance of isotopes contributing to the alpha or beta emissions. Measured values change with time because of radioactive decay and ingrowth during hold time as a water sample or after preparation of the solid residue for analysis.

Samples collected by the USGS were commonly analyzed after hold times of 72 hours and 30 days between collection and analysis, and the analytical method excludes isotopes that were gases at the time samples were prepared for analysis. Data from samples analyzed after the 30-day-hold time were used in this study because the longer hold time was more comparable to the hold times used in regulatory compliance sampling of public-supply wells. The primary contributors to adjusted gross beta in most groundwater samples were radium-228 and its progeny actinium-228 and the beta-emitting progeny of uranium-238, protactinium-234, and thorium-234 (Welch and others, 1995). The primary contributors to adjusted gross alpha are alpha-emitting isotopes on the decay series for radium-224, radium-226, radium-228, lead-210, and polonium-210 (Arndt, 2010).

Radioactive constituents with MCL benchmarks were present at high concentrations in 7.7 percent and at moderate concentrations in 8.1 percent of the SESJV-D (table 5). The proportions of the five study areas with high concentrations of radioactive constituents with MCL benchmarks fell into three categories: (1) the proportions in Madera-Chowchilla and Tule were 0 percent; (2) the proportion in Kaweah was 4.8 percent; and (3) the proportions in Kings and Tulare Lake were 16 and 17 percent, respectively (table 5). High concentrations of adjusted gross beta particle activities were more prevalent in SESJV-D than high concentrations of adjusted gross alpha particle activities (table 11; fig. 18B).

The SESJV-P dataset contained insufficient data to calculate aquifer-scale proportions for adjusted gross beta particle activities. Adjusted gross beta is regulated, but analysis is not required unless a system is classified as potentially vulnerable to having elevated gross beta activities (California State Water Resources Control Board, 2017b). In addition, sampling requirements are lower for systems with previously measured gross beta particle activities less than the 50-pCi/L trigger level (California State Water Resources Control Board, 2017b).

Most SESJV-D wells with moderate or high concentrations of adjusted gross beta particle activities were on the intermediate and distal parts of the San Joaquin and Kings River fans and had high concentrations of uranium (compare fig. 27 to fig. 29). Adjusted gross beta particle activities were positively correlated with uranium concentrations (Spearman test, p<0.001, rho=0.80; fig. 30). Beta particle activity from ingrowth of thorium-234 and protactinium-234 from decay of uranium-238 during the 30-day-hold time before sample analysis was estimated using the half-lives for thorium-234 and protactinium-234 and the concentration of uranium (Welch and others, 1995). The estimated activity from decay of thorium-234 and protactinium-234 formed by uranium present in samples accounts for nearly all the adjusted gross beta particle activity in 23 of the 24 wells with high or moderate adjusted gross beta particle activities. Beta activity from decay of radium-228 (and its progeny actinium-228) may be an important contributor to adjusted gross beta activity in samples for which the decay series of uranium-238 during sample holding time does not account for the measured adjusted gross beta activities (Welch and others, 1995).

Most of the unadjusted gross alpha particle activity in SESJV-D groundwater samples was accounted for by uranium. For the 170 SESJV-D samples with data for gross alpha particle activity and uranium concentration, calculated uranium activity equaled a median of 93 percent (interquartile range of 56 to 117 percent) of the unadjusted gross alpha particle activity (Arnold and others, 2016, 20186; Bennett and others, 2017; Shelton and Fram, 2017). A total of 41 percent of samples had unadjusted gross alpha particle activities greater than 7.5 pCi/L, but only 7 percent of samples had adjusted gross alpha particle activities greater than 7.5 pCi/L, corresponding to moderate or high relative concentrations. The uncertainty in gross alpha particle activity measurements can be large (Arndt, 2010); therefore, adjusted gross alpha particle activities, particularly in samples with large unadjusted gross alpha particle activities, may have large uncertainties.

Adjusted gross alpha activities were not correlated with uranium concentrations (Spearman test, p=0.158, rho= −0.10; fig. 30). The lack of correlation with uranium may indicate that the adjusted gross alpha particle activity is primarily from of alpha-emitting isotopes on the decay chains of radioactive isotopes other than uranium present in the groundwater at the time of sample collection. Inference of isotopes contributing to the adjusted gross alpha activity was not attempted because of the uncertainty in gross alpha particle activity measurements (Arndt, 2010) and because samples for direct analysis of radium isotopes, lead-210, and polonium-210 were not collected for this study.

Constituents with Secondary Maximum Contaminant Level Benchmarks

The class of constituents with SMCL benchmarks includes salinity indicators (chloride, sulfate, and TDS) and the trace elements iron and manganese. These constituents can affect the aesthetic properties of water, such as taste, color, and odor, or create technical problems, such as scaling and staining. Secondary maximum contaminant level benchmarks are based on these aesthetic and technical concerns and are not health-based benchmarks (California State Water Resources Control Board Division of Drinking Water, 2022b).

Constituents with SMCL benchmarks were present at high concentrations in 14 percent and at moderate concentrations in 26 percent of the SESJV-D (table 5). Salinity indicators (chloride, sulfate, or TDS) were present at a high concentration in 4.3 percent, and iron or manganese were present at a high concentration in 12 percent of the SESJV-D. The constituents most commonly responsible for the high concentrations were TDS and manganese, which were present at high and moderate concentrations in all five study areas (tables 6–10; fig. 18D). The proportion of SESJV-P with high concentrations of salinity indicators (4.0 percent) was not different from the proportion in the SESJV-D (4.3 percent, confidence interval 2.4–7.2 percent), although the proportion with moderate concentrations was lower in the SESJV-P (12 percent) compared to SESJV-D (30 percent; table 5). The proportion of the SESJV-P with high concentrations of trace elements with SMCLs (17 percent), was higher than the proportion in the SESJV-D (12 percent; table 5). The only trace elements with SMCLs present at moderate or high concentrations were manganese and iron (tables 2–4).

Manganese and Iron

Natural sources of manganese and iron to groundwater include weathering and dissolution of minerals in soils, sediments, and rocks. Manganese- and iron-oxyhydroxide minerals commonly coat sediment grains, and manganese commonly substitutes for iron in silicate and oxide minerals in the iron-bearing silicate, sulfide, or oxide minerals present in most rocks and sediments. The solubilities of manganese and iron are strongly dependent on redox conditions; more reduced species are generally more soluble (Hem, 1985).

Manganese was present at concentrations above the HAL-US in 4.9 percent and at concentrations between the SMCL-CA and HAL-US in 7.2 percent of the SESJV-D (table 11; fig. 31). Wells with concentrations above the SMCL-CA were present in all five study areas, and concentrations above the HAL-US were present in Kings, Kaweah, and Tulare Lake study areas (tables 6–10; figs. 18, 32). All wells with high concentrations of iron also had concentrations of manganese above the SMCL-CA (Bennett and others, 2017; Shelton and Fram, 2017), and manganese and iron were positively correlated (Spearman’t test, p<0.001, rho=0.67). Most wells with concentrations of manganese above benchmarks were along the western edge of the SESJV-D, primarily in the Kings and Tulare Lake study areas (fig. 32). The locations of domestic wells with high concentrations of manganese were expected based on the locations of groundwater with anoxic or mixed redox conditions in the hydrologic system of the SESJV-D. Anoxic and mixed redox conditions were most prevalent along the western edge of the SESJV-D (fig. 14), and most wells with high manganese concentrations were along the western edge of the SESJV-D (fig. 32). The locations of domestic wells with manganese concentrations greater than the SMCL-CA and HAL-US benchmarks were consistent with the maps of Rosecrans and others (2017, 2018) which showed that the highest probabilities of occurrence of elevated manganese concentrations were in the San Joaquin Valley trough with sparse scattered zones of slightly elevated probability near the eastern margin of the valley.

Manganese was present at concentrations above the HAL-US in 2.6 percent and at concentrations between the SMCL-CA and HAL-US in 8.7 percent of the SESJV-P (table 11; fig. 31). As in SESJV-D, wells in SESJV-P with concentrations of manganese above benchmarks were primarily detected along the western edge of the southeastern San Joaquin Valley, in the Kings and Tulare Lake study areas, although a few were along the eastern edge close to the boundary between the Central Valley and the Sierra Nevada in the Tule, Kaweah, and Kings study areas (compare SESJV-D data shown on fig. 32 to SESJV-P data shown on fig. 17H in Burton and others, 2012). The SESJV-P wells with high concentrations of manganese along the eastern side of the southeastern San Joaquin Valley are represented entirely by SWRCB-DDW data in the dataset used for this report. The wells do not have dissolved oxygen concentration data to corroborate anoxic conditions and most do not have depth information. Elevated manganese concentrations in some samples in the SWRCB-DDW dataset may be due to sample collection protocols. All USGS data are for filtered samples, whereas SWRCB-DDW data may be for filtered or unfiltered samples. Agreement between USGS and SWRCB-DDW data for samples collected at the same well near the same time was poorest for constituents like manganese that may be associated with particulates in groundwater samples (appendix B in Fram and Shelton, 2015).

Manganese concentrations were significantly greater in premodern groundwater compared to modern groundwater in the SESJV-P (Burton and others, 2012) but were not significantly different in the SESJV-D (Kruskall-Wallis p=0.006, Dunn’s p=0.070). Manganese concentrations in the SESJV-D also did not have a significant correlation with well depth (Spearman, p=0.588, rho=0.04). Wells above the Corcoran Clay in the Kings and Tulare Lake study areas had anoxic conditions because of their location in the basin and Tulare Lake Bed deposits, but these wells are shallow and can have modern, mixed, or premodern groundwater (fig. 11). McMahon and others (2019) observed that elevated manganese concentrations were associated with shallow, anoxic water tables and soils enriched in organic carbon. Public-supply wells in the same area also were anoxic because of the sediments but were deep and had premodern or mixed-age groundwater (Burton and others, 2012).

Total Dissolved Solids

Groundwater salinity may be affected by many natural and anthropogenic factors (Hem, 1985; Appelo and Postma, 2005). The list of potential factors that may be relevant in the southeastern San Joaquin Valley includes, but is not limited to, the following: sources and quality of stream water recharging and contributing salts to the shallow groundwater; sources of depositional sediments in alluvial fans; use of water for irrigation and recharge of drainage water; evaporation and evapotranspiration; mixing with hydrothermal fluids; mixing with connate fluids from marine or lacustrine sediments; contributions from fertilizers; discharges from treatment plants, septic systems, and agricultural operations; geochemical processes, such as ion exchange, mineral dissolution and precipitation; and biological reactions that affect redox sensitive elements (Hem, 1985; Appelo and Postma, 2005). Furthermore, use of water for irrigation in the San Joaquin Valley has caused changes to groundwater chemistry that contribute to increased TDS concentrations in groundwater (Hansen and others, 2018).

Total dissolved solids is an indicator of salinity that was present at high concentrations (above the upper SMCL-CA of 1,000 mg/L) in 4.3 percent and at moderate concentrations (between the recommended SMCL-CA of 500 mg/L and the upper SMCL-CA) in 30 percent of SESJV-D (table 11). The proportions of the five study areas with high concentrations fell into two groups: (1) Madera-Chowchilla, Kaweah, and Tule study areas had high concentrations in 1.9, 1.8, and 0 percent of the study area, respectively, and (2) Kings and Tulare Lake had high concentrations of TDS in 8.0 and 8.9 percent of the study area, respectively (tables 6–10; fig. 33). High concentrations of TDS were detected in the distal parts of the alluvial fans near the center of the basin, and moderate concentrations were distributed across much of the SESJV-D area (fig. 34). Moderate concentrations were detected in all study areas (figs. 18, 33). The spatial distribution of TDS concentrations in the SESJV-D was similar to the spatial distribution predicted for the San Joaquin Valley using a larger dataset (Luhdorff and Scalmanini Consulting Engineers, 2016). Chloride and sulfate were present at high concentrations in 0.3 and 1.6 percent of the SESJV-D, respectively (table 11), and all wells with high concentrations of chloride or sulfate also had high concentrations of TDS (Bennett and others, 2017; Shelton and Fram, 2017).

The proportions of the Madera-Chowchilla, Kings, Kaweah, and Tule study areas that had high concentrations of TDS were similar in the SESJV-D and SESJV-P; the proportion of the Tulare Lake study area with high concentrations of TDS was greater in the SESJV-D than the SESJV-P (table 11; fig. 33). However, the SESJV-D had a substantially greater proportion with moderate concentrations of TDS (30 percent) compared to the SESJV-P (12 percent; table 11), and the proportion with moderate concentrations was greater in the SESJV-D than SESJV-P for all five study areas (tables 6–10; fig. 33). Differences in the spatial distribution of wells with high or moderate concentrations of TDS between the SESJV-D and the SESJV-P were greatest in the Tulare Lake study area and the central part of the Kings study area (compare SESJV-D data on fig. 34 to SESJV-P data on fig. 17J in Burton and others, 2012).

The diversity of major ion compositions of groundwater and the lack of a consistent relation between major ion composition and TDS concentration (fig. 35) indicated that multiple sources and processes affect TDS concentrations in SESJV-D (Mendenhall and others, 1916; Page, 1973; Planert and Williams, 1995; Larry Walker and Associates, 2010; Fram, 2017a; Hansen and others, 2018; Pauloo and others, 2021; Central Valley Salinity Alternatives for Long-term Sustainability, 2023). Because of this diversity of sources and processes, simple statistical tests for associations between water quality and potential explanatory factors were not effective for identifying factors controlling TDS concentrations (Fram, 2017a), and a more in-depth evaluation of the sources of salinity was beyond the scope of this report.

Selected processes and sources that may affect salinity in the SESJV-D are briefly described in this section. Hansen and others (2018) determined that the increases in TDS in the southeastern San Joaquin Valley during the past century can largely be explained by enhanced dissolution of minerals promoted by the increased bicarbonate generated by irrigated agriculture. Uranium and TDS were positively correlated in the SESJV-D (Spearman p<0.001, rho=0.58). As discussed in the “Uranium” section, uranium also was more prevalent at concentrations above benchmarks in the SESJV-D than in the SESJV-P, especially in the Kings study area. The positive statistical correlation between TDS and uranium supports previous findings that TDS and uranium are affected by agriculture.

Pauloo and others (2021) modeled salt accumulation in the Tulare Basin (including the Kings, Kaweah, Tule, and Tulare Lake study areas) that resulted from basin closure. Salts from imported surface water, pumped groundwater, and concentration from evapotranspiration and water-rock interaction result in increasing salinity (first in the shallow aquifer and later in deeper aquifers). Fram (2017a) used isotopic and major ion data to track sources of salinity in the western San Joaquin Valley. These sources of salts may affect groundwater in the SESJV-D along the western margin of the southeastern San Joaquin Valley.

Similar to the findings of Visser and others (2018) for public-supply groundwater in the eastern San Joaquin Valley and Fram (2017a) for the western San Joaquin Valley, extensive amounts of evaporative concentration were not observed in groundwater in the SESJV-D. The SESJV-D samples have stable isotopic compositions close to the local meteoric water line (LMWL) defined by Visser and others (2018; fig. 36). In contrast, the shallow groundwater from monitoring wells in the Tulare Lake Bed includes samples with stable isotope compositions plotting on evaporation lines extending from the LMWL, and the relation between TDS concentrations and stable isotope composition indicated evaporative concentration was an important mechanism contributing to high TDS concentrations in shallow groundwater (Fujii and Swain, 1995). The absence of evaporation trend lines in the stable isotope diagram for the SESJV-D (fig. 36) indicates that the extensively evaporated water in the shallow groundwater system identified by Fujii and Swain (1995) in the Tulare Lake Bed region apparently had not penetrated in large amounts to the depth zones used by domestic- or public-supply wells at the time of sampling in this study.

The range in stable isotopic compositions in the SESJV-D indicated sources of recharge ranging from local precipitation at relatively low elevations in the San Joaquin Valley and nearby foothills (heavier delta oxygen-18 values greater than about −9 per mil) to large rivers predominantly fed by high-elevation precipitation (fig. 36). The spatial pattern of stable isotopic composition lighter delta oxygen-18 values less than about −12 per mil) was similar to the spatial pattern observed in public-supply wells by Visser and others (2018; compare fig. 37 to fig. 4 in Visser and others, 2018). Wells in the interfan areas along the eastern side of the southeastern San Joaquin Valley commonly had stable isotope compositions greater than −10 per mil, indicating a source from low-elevation precipitation. Wells in the fans of the Chowchilla and Fresno Rivers also had stable isotope compositions in the range from −11 to greater than −8 per mil, consistent with watersheds predominantly in the lower-elevation foothills of the Sierra Nevada. The Kings and Kaweah River fans and parts of San Joaquin River fan generally have stable isotope compositions from −11 to less than −13 per mil that indicate high-elevation sources of water from the headwaters of the rivers.

The main difference between the spatial pattern on figure 37 and the pattern observed for public-supply wells by Visser and others (2018) was that the region dominated by high elevation recharge in the Kings River fan is larger in the SESJV-D. In the Kings and Tulare Lake study areas, where the Kings River fan is located, the SESJV-D has a greater proportion of modern groundwater than the SESJV-D (fig. 12). The larger region dominated by recharge from high-elevation sources in the SESJV-D compared to the SESJV-P indicates that imported surface water routed to the area for irrigation is recharging the aquifer system and is reaching the depths of domestic-supply wells, but not yet the depths of public-supply wells.

Perchlorate

Perchlorate is an inorganic anion with natural and anthropogenic sources that is formed naturally in the atmosphere and is present in precipitation (Dasgupta and others, 2005; Parker and others, 2008; Rajagopalan and others, 2009). Perchlorate can accumulate in salts that precipitate in soils and unsaturated zones of aquifer systems in arid areas (Walvoord and others, 2003; Rao and others, 2007; Jackson and others, 2010). The probability of detecting perchlorate in groundwater during natural conditions increases as conditions become more arid (Fram and Belitz, 2011). Perchlorate salts are a primary ingredient in solid rocket fuels and are used in explosives, safety flares, and fireworks; thus, manufacture, use, or disposal of those products are potential sources of perchlorate contamination to groundwater (Interstate Technology and Regulatory Council, 2005; U.S. Environmental Protection Agency, 2014; California State Water Resources Control Board, 2022b). Perchlorate also is present as a minor salt in the Chilean nitrate fertilizer that was used extensively in California before industrial sources of nitrate fertilizer were plentiful (Dasgupta and others, 2006; Böhlke and others, 2009).

Perchlorate was detected at high concentrations (above the MCL-CA of 6 µg/L) in 0.9 percent and at moderate concentrations in 3.3 percent of the SESJV-D (table 11). The prevalence of high concentrations in SESJV-P was the same (0.9 percent; table 11). The computed prevalence of moderate concentrations in the SESJV-P (0.9 percent; table 11) was a minimum estimate because the reporting limit used for most perchlorate data in the SWRCB-DDW database is 4 µg/L (California State Water Resources Control Board Division of Drinking Water, 2022d; California State Water Resources Control Board, 2023b).

Logistic regression models of Fram and Belitz (2011) were used to estimate if the perchlorate detected in the SESJV-D was primarily natural or anthropogenic in origin. This estimation was only done for the SESJV-D because all perchlorate data had reporting levels of 0.1 µg/L. The proportion of the SESJV-D with detection of perchlorate at concentrations greater than the detection limit of 0.1 µg/L was 70 percent (66 percent at low concentrations, 3.3 percent at moderate, and 0.9 percent at high; table 12).

Detection frequencies of perchlorate were compared to predicted probabilities of perchlorate detection in groundwater for the aridity conditions in the southeastern San Joaquin Valley. Because perchlorate degrades under redox conditions similar to conditions for degradation of nitrate (Coates and Achenbach, 2004), the SESJV-D dataset was trimmed to remove 26 samples for which the nondetection of perchlorate could either have been due to absence of perchlorate input to the sample or to degradation of perchlorate that had been present in the sample. These 26 samples were anoxic or mixed redox samples that had nondetections of perchlorate and had nitrate concentrations less than 0.5 mg/L. Of the 15 Tulare Lake samples, 8 were removed, leaving an insufficient number of samples with which to evaluate perchlorate detections in the Tulare Lake study area by this method. Detection frequencies of perchlorate in the remaining 126 samples relative to threshold levels of 0.1, 0.5, and 1.0 µg/L were compared to predicted probabilities of detecting perchlorate in groundwater at concentrations above those thresholds under natural conditions at the study-area scale (fig. 38; Fram and Belitz, 2011). Error bars for the measured detection frequencies were estimated as the 90-percent Jeffreys confidence interval (equation 3; Brown and others, 2001; Belitz and others, 2010); ranges for the predicted probabilities of detection were estimated as the predicted probability values for a range in aridity index of plus or minus two standard deviations from the mean aridity index for the samples.

The observed detection frequencies of perchlorate relative to threshold levels of 0.1, 0.5, and 1.0 µg/L were all greater than the predicted probabilities of detection for the Madera-Chowchilla, Kings, Kaweah, and Tule study areas. All measured detection frequencies were greater than predicted probabilities of detection (fig. 38), indicating presence of some anthropogenic sources of perchlorate. The differences between predicted and observed detection frequencies were greatest for the higher concentrations (greater than 0.5 and 1.0 µg/L) and in the Kings, Kaweah, and Tule study areas. Perchlorate concentration was positively correlated with nitrate concentration (Spearman, p<0.001, rho=0.81), and 94 percent of samples with perchlorate greater than 0.5 µg/L also had nitrate greater than 5 mg/L, indicating the contribution of anthropogenic sources of nitrate. Nitrate concentrations in groundwater greater than 2 mg/L generally indicate contributions from anthropogenic sources of nitrate (Dubrovsky and others, 2010). We used a more conservative threshold of 5 mg/L (Burow and others, 2013) to increase the certainty that the elevated nitrate concentrations were indicative of anthropogenic sources of nitrate. Because both nitrate and perchlorate have agricultural sources, if elevated concentrations of both nitrate and perchlorate are detected, and agricultural sources of nitrate contribute to the nitrate load in the groundwater, it is likely that anthropogenic sources contributed to the perchlorate load in the groundwater. The relation between nitrate and perchlorate concentrations indicates that most perchlorate detections with concentrations greater than 0.5 µg/L are consistent with contribution from anthropogenic sources of perchlorate.

The predicted probabilities of detection relative to a threshold of 0.1 µg/L under natural conditions were 60–70 percent (fig. 38), indicating that most of the perchlorate detections at concentrations between 0.1 and 0.5 µg/L likely were due to natural sources. Only 25 percent of samples with perchlorate concentrations less than 0.5 µg/L also had nitrate greater than 5 mg/L. The two agricultural sources are nitrate fertilizers that contain perchlorate and perchlorate from the unsaturated zone swept by increased recharge from irrigation (Walvoord and others, 2003; Dasgupta and others, 2006; Rao and others, 2007; Böhlke and others, 2009; Jackson and others, 2010; Fram and Belitz, 2011); differentiating between the two agricultural sources is beyond the scope of this report.

Fumigants

Soil fumigants can be used to kill or manage pests in urban and agricultural settings and are of particular interest in California because of their widespread historical use in certain agricultural settings, including in the San Joaquin Valley. Nine VOCs used as fumigant pesticides or included in fumigant mixtures were analyzed (Bennett and others, 2017; Shelton and Fram, 2017), and four were detected: 1,2-dibromo-3-chloropropane (DBCP), 1,2-dibromoethane (EDB), 1,2-dichloropropane (1,2-DCP), and 1,2,3-trichloropropane (1,2,3-TCP; table 2; fig. 18E). Use of the soil fumigants DBCP, EDB, and 1,2-DCP was banned in California beginning in the late 1970s and early 1980s. The fumigant DBCP was used from 1955 to 1977 primarily to control nematodes on vineyard crops but also on some orchard and vegetable crops (Peoples and others, 1980; Domagalski and Dubrovsky, 1992; Kloos, 1996; Burow and others, 1999). The fumigant EDB was used from about 1948 to 1983, and it was also used as a gasoline additive beginning in the 1920s (Kloos, 1996). Fumigant mixtures of 1,2-DCP and 1,3-dichloropropene isomers were used from 1942 to 1984 (Deeley and others, 1991; Loague and others, 1998). The VOC 1,2,3-TCP is used as a solvent but also was present in fumigant mixtures containing 1,2-DCP (Oki and Giambelluca, 1987; Zebarth and others, 1998; California State Water Resources Control Board, 2017a).

Fumigants were present at high concentrations in 19 percent, at moderate concentrations in 2.0 percent, and at low concentrations in 9.0 percent of the SESJV-D. Fumigants were not detected at concentrations greater than analytical detection limits in 70 percent of the SESJV-D (table 12). High concentrations of fumigants were detected in all five study areas, ranging from 3.6 percent in Kaweah study area to 30 percent in Madera-Chowchilla study area.

Most of the detections at high concentrations throughout all study areas were for 1,2,3-TCP (fig. 18E). The health-based benchmarks for the fumigants have different magnitudes. The MCL-CA for 1,2,3-TCP is 40 times less than the MCL-US for DBCP and one thousand times less than the MCL-US for 1,2-DCP (table 2). The median detected concentration of 1,2,3-TCP was significantly greater than the median detected concentration of 1,2-DCP (Wilcoxon test, p=0.023) but not significantly different from the median detected concentrations of DBCP or EDB (Wilcoxon test, p=0.203 and p=0.096, respectively). However, because the benchmark for 1,2,3-TCP is less than the benchmarks for the other fumigants, detections of 1,2,3-TCP were more likely to have concentrations greater than the benchmark than were detections of the other fumigants.

The compound 1,2,3-TCP was detected at high concentrations in 15 percent of the SESJV-D (table 12). Because the detection level of 0.003 µg/L (half of the laboratory reporting level of 0.006 µg/L) was close to the MCL-CA, the presence of 1,2,3-TCP at moderate and low RCs could not be determined. There were high concentrations detected in all five study areas (fig. 39), with the proportion of high concentrations falling into two groups: (1) 3.6 and 6.7 percent, respectively, in the Kaweah and Tulare Lake study areas and (2) 19, 18, and 21 percent, respectively, in the Madera-Chowchilla, Kings, and Tule study areas (table 12; Burow and others, 2019) detected 1,2,3-TCP at concentrations above the MCL-CA in about 8 percent of domestic wells statewide and in 18 percent of domestic wells in the San Joaquin Valley. The fumigant DBCP was detected in the Madera-Chowchilla, Kings, Kaweah, and Tule study areas (fig. 40), corresponding to 16 percent of the SESJV-D. DBCP, and was present at high and moderate concentrations in 4.8 and 7.0 percent of the area, respectively (table 12). The proportion of high concentrations ranged from 0 percent in Tulare Lake and Kaweah to 12 percent in Madera-Chowchilla (table 12). Burow and others (2019) observed that samples with detections of 1,2,3-TCP commonly had detections of other fumigant compounds. All samples from SESJV-D domestic wells that had a detection of 1,2,3-TCP also had detections of at least one other fumigant (Bennett and others, 2017; Shelton and Fram, 2017; Fram and Shelton, 2018), and concentrations of 1,2,3-TCP were positively correlated with concentrations of DBCP (Spearman test, p<0.001, rho 0.47) and 1,2-DCP concentrations (Spearman test, p<0.001, rho 0.68).

Together, two explanatory factors, percentage of orchard and vineyard land use and groundwater age, described the occurrence pattern of DBCP and 1,2,3-TCP in SESJV-D wells. Because DBCP and fumigants containing 1,2,3-TCP were both used on orchard and vineyard crops, it was not surprising that wells with detections of 1,2,3-TCP and DBCP were primarily in areas of orchard and vineyard land use (figs. 39, 40). The detection frequency of DBCP or 123-TCP in SESJV-D samples was greater in modern (34 percent) and mixed (16 percent) groundwater compared to the detection frequency in premodern groundwater (4 percent; not shown). Although DBCP and 1,2,3-TCP are human-made compounds and therefore generally not present in groundwater that does not contain a component of modern recharge, the presence of modern recharge alone is not sufficient to predict presence of DBCP or 1,2,3-TCP in a well. When the dataset was stratified by land use, the relation between groundwater age and detection frequency became stronger. The detection frequency of DBCP or 123-TCP in SESJV-D samples from wells surrounded by greater than 35-percent orchard or vineyard land use and having modern groundwater was 53 percent compared to a detection frequency of 8.5 percent in wells having modern groundwater but surrounded by less than 35-percent orchard or vineyard land use (fig. 41).

Burow and others (2008) reported that DBCP was detected in 50 percent of domestic wells beneath vineyards or almond orchards in the eastern San Joaquin Valley during 2001–02. Concentrations were above the MCL in 32 percent of the wells, and although concentrations near the water table decreased through time, concentrations at the depth of domestic wells increased between the 1990s and 2000s as DBCP moved deeper in the groundwater system (Burow and others, 2007, 2008). Because of the long half-lives of DBCP and 1,2,3-TCP, both compounds are expected to persist in groundwater for a long time (Burow and others, 2019). Concentrations of DBCP and 1,2,3-TCP were not correlated to well depths in the SESJV-D (Spearman test, p=0.915, rho= −0.01 and Spearman test p=0.787, rho= 0.02, respectively). Studies of public-supply wells in the southeastern San Joaquin Valley reported either weak, inverse correlations between DBCP or 1,2,3-TCP and well depth or no significant correlations (Kloos, 1996; Burton and others, 2012; Shelton and others, 2013). This lack of significant relation to depth may be due to limitations of using the Spearman’s test for correlations on a dataset with many nondetections (Helsel and others, 2020), the uneven distribution of fumigant applications at the land surface, or the complex relations between well depth and location in the southeastern San Joaquin Valley (fig. 7) that were not considered in a simple test of correlation between well depth and fumigant concentration.

Herbicides and Herbicide Degradates

A total of 125 herbicides or select herbicide degradates were analyzed (Bennett and others, 2017; Shelton and Fram, 2017175), and 29 of these constituents were detected in at least 1 of the 180 domestic-supply wells for which pesticide data were available. Of the 12 herbicides detected, 3 had MCL benchmarks and 8 had non-regulatory health-based benchmarks. Of the 17 herbicide degradates detected, 2 had non-regulatory benchmarks and the rest did not have benchmarks (tables 2, 4). Most the degradates detected were degradation products of herbicides.

Herbicides were not detected at concentrations greater than MCLs, and the only herbicide detected at moderate concentrations (greater than one-tenth of the MCL) was simazine, which was detected in one well in the Madera-Chowchilla study area (fig. 18E). Most detections of herbicides and herbicide degradates with benchmarks were detected at concentrations less than one-one hundredth of the benchmark concentrations (fig. 18E, F), which is similar to what was observed in a national study of the same herbicides using the same methods (Bexfield and others, 2021).

Comparison between SESJV-D and SESJV-P required censoring data collected by the USGS to the same reporting level as used for data in the SWRCB-DDW regulatory compliance database (California State Water Resources Control Board Division of Drinking Water, 2022d; California State Water Resources Control Board, 2023b). The one SESJV-D well with a moderate concentration of simazine (M23 in the Madera-Chowchilla study area) had a concentration of 0.64 µg/L, which is less than the SWRCB-DDW reporting level of 1 µg/L; therefore, this sample was considered a nondetection for the comparison of SESJV-D and SESJV-P. All three pesticides with MCLs detected in SESJV-D were herbicides (atrazine, simazine, and bentazon; tables 2, 4). Bentazon was only detected in one sample (Bennett and others, 2017). Simazine, atrazine, or both were detected in all five study areas (figs. 18, 42). The proportions of SESJV-D and SESJV-P with moderate or high concentrations of pesticides with MCLs was zero (table 5). Detection frequencies relative to the SWRCB-DDW reporting levels also were zero in the SESJV-P and SESJV-D.

Herbicides with MCLs were detected at low concentrations in 30 percent and at moderate concentrations in 0.8 percent of the SESJV-D (table 12). The detection frequency ranged from 13 percent in Tulare Lake to 46 percent in Kings (table 12). Herbicides have been widely used in the San Joaquin Valley for roadside weed control and on agricultural crops (California Department of Pesticide Regulation, 2023). The patterns of herbicide detections in this study were similar to the results shown for sampling of domestic wells in 2012 in parts of the Kings, Kaweah, and Tule study areas by the California Department of Pesticide Regulation (CDPR; Troiano and others, 2013). The CDPR regulates applications of pesticides and regularly monitors domestic wells to determine if regulated pesticides or their degradation products are reaching drinking-water supply wells (Troiano and others, 2001). Low concentrations of simazine, degradation products of simazine, diuron, and bromacil were the most commonly detected herbicide constituents in domestic wells in SESJV-D (table 12; Troiano and others, 2013).

Simazine was the most frequently detected herbicide with an MCL; it was detected at low concentrations in 27 percent and at moderate concentrations in 0.8 percent of the SESJV-D (table 12). Didealkylatrazine (CAAT) is a degradate of both simazine and atrazine (Scribner and others, 2000) and has a non-regulatory health-based benchmark (table 2). CAAT was detected at low concentrations in 37 percent and at moderate concentrations in 5.8 percent of the SESJV-D (table 12). Of samples with detections of atrazine or simazine, 80 percent also had detections of CAAT; of samples with detections of CAAT, 60 percent also had detections of atrazine or simazine (Bennett and others, 2017; Shelton and Fram, 2017).

Detections of simazine and CAAT were related to groundwater age. This finding was expected because the first products containing simazine and atrazine were registered in 1958 (U.S. Environmental Protection Agency, 2019, 2020); therefore, simazine and CAAT can only be present in groundwater that contains water recharged after 1958. For this study, modern groundwater is defined as water recharged after 1952; premodern groundwater as recharged before 1952, including water that may have recharged hundreds or thousands of years ago; and mixed as groundwater containing both modern and premodern recharge. As expected, given that simazine and CAAT should only be present in groundwater containing a component of modern recharge, the detection frequencies of simazine and CAAT are significantly greater in modern groundwater than in mixed groundwater and significantly greater in mixed groundwater than in premodern groundwater (fig. 43). As described in the “Methods” section of this report, detection frequencies are defined as significantly different if the Jeffrey’s 90-percent confidence intervals do not overlap (fig. 43).

Burow and others (2008) observed that simazine concentrations in San Joaquin Valley groundwater generally increased from the 1970s to the 2000s. Troiano and others (2013) observed that concentrations of simazine, diuron, bromacil, and their degradates in San Joaquin Valley groundwater decreased between 2001 and 2012. Troiano and others (2013) concluded that the decrease in herbicide concentrations in groundwater likely represents the effects of regulation of these herbicides and the time required for recharge to reach the depths of the wells. Regulation of herbicide applications began around 1990, and groundwater age modeling indicates that it takes 3–33 years for recharge at the land surface to reach the depths of the wells (Spurlock and others, 2000; Troiano and others, 2013).

Simazine and CAAT concentrations in the SESJV-D were negatively correlated with well depth (Spearman test, p<0.001, rho= −0.33 and −0.31, respectively). Although the relations between simazine and CAAT and well depth were statistically significant, the low absolute values of the rho values indicate that substantial amounts of the variations in concentrations are not explained by the relation with well depth. A non-monotonic change in simazine input with time due to regulation (Troiano and others, 2013), uneven application of simazine across the landscape and with time, and degradation of simazine and CAAT during different aquifer conditions with time (Burow and others, 2008) likely contribute to the amount of variabilities in concentrations of simazine and CAAT that are not explained by relations with well depth.

Volatile Organic Compounds Other than Fumigants

A total of 81 VOCs that were not fumigants were analyzed (Bennett and others, 2017; Shelton and Fram, 2017175), and 13 of these VOCs were detected in at least 1 of the 198 SESJV-D domestic-supply wells. Of the 13 VOCs detected, 9 had MCL benchmarks, 3 had non-regulatory health-based benchmarks, and 1 did not have a benchmark (tables 2, 4). Non-fumigant VOCs with MCLs were not detected at high concentrations in the SESJV-D, and most detections had concentrations less than one-tenth of benchmark concentrations (fig. 18G). Detected VOCs included disinfection byproducts, solvents, gasoline oxygenates, refrigerants, and naturally occurring VOCs.

Well owners are commonly advised to use hypochlorite solutions (bleach) to disinfect well water used for drinking water and other household uses in domestic, municipal, and community systems (Centers for Disease Control and Prevention and U.S. Department of Housing and Urban Development, 2006). However, hypochlorite reacts with organic matter to produce trihalomethanes and other chlorinated and brominated disinfection byproducts. The trihalomethane trichloromethane was detected at low concentrations in 10 percent of the SESJV-D (table 12). Another trihalomethane, bromodichloromethane, was detected in some samples in which trichloromethane was detected (Bennett and others, 2017; Shelton and Fram, 2017). Trichloromethane is the most frequently detected VOC nationwide (Zogorski and others, 2006; Bexfield and others, 2022).

Solvents with MCL benchmarks were detected at moderate concentrations in 1.1 percent and at low concentrations in 4.7 percent of the SESJV-D (table 12). The most frequently detected solvent was tetrachloroethene (PCE), which also was the most frequently detected solvent in groundwater nationally (Zogorski and others, 2006; Bexfield and others, 2022). Tetrachloroethene also was the only solvent with an MCL detected at moderate concentrations in the SESJV-D. Tetrachloroethene typically is used for dry-cleaning of fabrics and degreasing metal parts and is an ingredient in a wide range of products, including paint removers, polishes, printing inks, lubricants, and adhesives (Zogorski and others, 2006).

The gasoline oxygenate methyl tert-butyl ether (MTBE) was detected at low concentrations in 11 percent of the SESJV-D (table 12). Methyl tert-butyl ether was used as a gasoline oxygenate from 1979 until it was banned in California at the start of 2004, with peak usage in the late 1990s (California State Water Resources Control Board, 2017c). None of the wells with detections of MTBE had detections of hydrocarbons associated with gasoline, and MTBE concentrations were not significantly related to underground storage tank density (Spearman test, p=0.089, rho=0.13). These relations indicate that leaks from fuel tanks were not the primary source of MTBE in SESJV-D wells.

However, concentrations of MTBE detected in SESJV-D wells were in a range that would be consistent with an atmospheric source for the MTBE. The median MTBE concentration in the atmosphere in Fresno during 1996–2000 of about 1.5 parts per billion by volume (ppbV; California Air Resources Board, 2023) would be in equilibrium with precipitation containing a concentration of approximately 0.33 µg/L at 20 degrees Celsius (Baehr and others, 1999). Concentrations of MTBE detected in SESJV-D samples ranged from 0.009 to 0.17 ug/L, with a median detected concentration of 0.017 µg/L (Bennett and others, 2017; Shelton and Fram, 2017).

Even shorter-screened wells like many of the SESJV-D wells generally capture groundwater that consists of a mixture of groundwater ages. For this report, groundwater age was summarized by classification as modern, mixed, or premodern groundwater; 14 of the SESJV-D wells with MTBE detections were modern, and 4 were mixed. The MTBE concentrations detected in the groundwater samples could be produced from a mixture consisting of 3–51 percent water recharged during the time MTBE was used (1979–2004; California State Water Resources Control Board, 2017c), and 97–49 percent water recharged before or after that time. Faulkner and Jurgens (2022) presented results of lumped parameter models for groundwater age that proved more resolution of groundwater ages than the simple classification used in this report. However, that level of analysis is beyond the scope of this report.

Carbon disulfide was detected at low concentrations in 8.7 percent of the SESJV-D, and detection frequencies ranged from 0 percent in the Madera and Kaweah study areas to 42 percent in Tulare Lake study area (table 12). Of the wells with carbon disulfide detections, 82 percent had anoxic or mixed redox conditions, and 72 percent did not have other VOCs detected in the samples (Bennett and others, 2017; Shelton and Fram, 2017). These detection results were consistent with natural sources of carbon disulfide (U.S. Department of Health and Human Services, Agency for Toxic Substances and Disease Registry, 1996; Fram and others, 2012; Bexfield and others, 2022).