Groundwater Pesticide Sampling—National Groundwater Network

The national groundwater network decadal pesticide analysis focused on samples collected from the USGS NWQN-GW. The current study presents pesticide data from 59 well networks (Lindsey and Kingsbury, 2024). The groundwater quality monitoring networks are defined as groups of wells with similar characteristics, and each network was originally designed with 20 to 30 wells using spatially distributed randomized sampling (Scott, 1990; fig. 1). Samples were collected from each network about every 10 years between 1993 and 2022. The alphanumeric network identifier is an abbreviation indicating the drainage basin name and the network type, such as ccptlusag2b for the Central Columbia Plateau (ccpt) land use study network (lus) focused on agricultural land use (ag) (table 1). The first decade of groundwater pesticide sampling was 1993–2001, the second was 2002–12, the third was 2013–22, and the fourth decade of sampling started in 2023. Herein, the term “decadal sampling” refers to the frequency of water quality sampling in this network. The “National Groundwater Network Decadal Pesticide Analysis” examines data from decades 1, 2, and 3, but the “DBCP in the Central Valley, Calif. change analysis” examines data from decades 1-4.

During each year in decades 1 and 2, pesticide samples were collected from 20 to 30 wells in 8 to 10 networks, resulting in all networks being sampled within a 10 year period. Because groundwater pesticide concentrations in exceedance of HHBs were not detected in the first two decades of monitoring and analysis (Toccalino and others, 2014), the number of monitored wells where groundwater pesticide samples were collected was reduced. In the third decade, the rotating schedule of network sampling was maintained, but the collection of pesticide samples was reduced to 8 wells within each network. Pesticide samples from groundwater were collected from these locations following protocols established by the USGS; raw samples were collected (before any treatment for domestic well water), and they were passed through 0.7-micrometer filters (Lapham and others, 1995; Koterba, 1998).

The objective of this monitoring framework was to evaluate the quality of groundwater underlying selected agricultural (number of well networks [n]=24) and urban (n=15) settings or groundwater used for domestic supply (n=20) within 25 principal aquifers of the United States (Gilliom and others, 1995). Wells in the agricultural and urban land use networks are typically shallower (average well depth, 6 meters [m] below surface) than wells in the domestic supply networks (average well depth, 43 m below surface), but well depths can vary by aquifers (Lindsey and Rupert, 2012) (table 1). For the purposes of discussion, wells in agricultural networks are referred to as “agricultural wells.” Comparable language is used for wells in urban networks. Previous studies have referred to the well networks representing the depth zone used for domestic supply as major aquifer studies (Lindsey and others, 2023). Those networks are referred to herein as “the domestic supply network type.”

The well networks are distributed across seven of the eight aggregated ecoregions (Pacific Northwest, Pacific Coast, Arid West, Semiarid West, Midcontinent, South Atlantic Gulf, and Northeast), which are regions of the conterminous United States with similar soils, climate, and landforms. There were no wells with pesticide monitoring in the Mountain West. These aggregated ecoregions have been used in prior groundwater-quality trend reports (Lindsey and others, 2023). The aggregated ecoregions combine multiple factors that affect groundwater quality such as soils, climate, and landforms into large, relatively homogenous units (Commission for Environmental Cooperation, 1997), with the latest ecoregions reflecting a regrouping of the EPA Level 3 ecoregions (Commission for Environmental Cooperation, 2021; Lindsey and others, 2023). The boundaries of the 59 well networks shown in figure 1 fall within the aggregated ecoregions. The boundaries of the aggregated ecoregions cross the boundaries of the principal aquifers shown in figure 1.

Groundwater Pesticide Sampling—1,2-Dibromo-3-Chloropropane (DBCP) in the Central Valley, California

The 1,2-dibromo-3-chloropropane (DBCP) in the Central Valley, Calif. change analysis relied on two sets of sampling data. Decadal samples of DBCP were collected from one well network in the San Joaquin-Tulare River Basin; refer to the sanjlusor1a network in figure 1 and table 1. The decadal concentration samples were collected from 36 wells in 1993, 2001, 2013, and 2023. The USGS also evaluated changes in DBCP concentrations at shorter time scales through a type of NWQN-GW well network known as an Enhanced Trends Network (ETN; Mathany and others, 2019). The ETN in the Central Valley aquifer of California collected bimonthly DBCP samples from a 98-m deep supply well and a 71-m deep monitoring well from 2014 to 2019 (Saraceno and others, 2018). Unfiltered DBCP samples were collected before any treatment for domestic well water (Lapham and others, 1995; Koterba, 1998).

Laboratory Pesticide Analysis

Water quality samples were analyzed for pesticide concentrations at the USGS National Water Quality Laboratory (NWQL) in Denver, Colorado. Water quality samples for the national groundwater network decadal pesticide analysis were analyzed using two different laboratory methods. Samples collected in decades 1 and 2 were analyzed using gas chromatography/mass spectrometry (Zaugg and others, 1995) or high-performance liquid chromatography (Werner and others, 1996; Furlong and others, 2001), also known as NWQL Laboratory Schedule 2033. Decade 3 samples were analyzed at the NWQL using a broad-spectrum liquid chromatography-tandem mass spectrometry method (Sandstrom and others, 2015), also known as NWQL Laboratory Schedule 2437. The change in laboratory analytical methods between decades 2 and 3 has the potential to introduce bias into the analysis of datasets that span all three decades. For this report, the relative comparability of the older and newer concentration data was confirmed based on the findings of a prior USGS report from Martin and others (2017), which compared the data from the 2033 method to the 2437 method. Martin and others (2017) focused on data collected from surface waters and incorporated adjustments for changes in laboratory recovery as assessed through laboratory spikes for 21 pesticide compounds included in this report. Martin and others (2017) found that the relation between the atrazine concentrations based on the older and newer NWQL methods was weak at concentrations below 0.01 micrograms per liter (µg/L). For concentrations greater than 0.1 µg/L, the relation between concentrations from both methods was close to a 1:1 line (fig. 2). The results of the two different analytical methods are comparable at concentrations relevant to reporting on potential risk to human health, that is, at concentrations greater than 10 percent of the human-health benchmark. A relation similar to the one found for atrazine was also observed for the other pesticides that have been commonly detected in groundwater (metolachlor, prometon, and simazine). Therefore, for this report, we treat pesticide data from decades 1 and 2 as comparable to decade 3.

Water quality samples for the DBCP concentration change analysis in the Central Valley, Calif. relied on a different laboratory analysis because DBCP is a volatile organic compound. In decade 1, the USGS DBCP laboratory analysis method had a reporting level of 0.03 µg/L (Fishman, 1993). An enhanced method, to measure concentrations for more compounds, at lower concentrations, was implemented, which resulted in a lower reporting level of 0.02 µg/L (Rose and others, 2016).

Similar to prior USGS pesticide (Toccalino and others, 2014) and DBCP analysis (Rowe and others, 2007) studies, this report analyzed data across multiple laboratory schedules and laboratory reporting levels (LRL). Results that are greater than the LRL are reported, and results that are less the LRL are reported as nondetects (Bonn, 2008). The highest magnitude LRL for all decades was used in our analysis and this value is included in table 2.

Quality-assurance and quality-control (QAQC) samples were collected for all 22 pesticides. These QAQC field blank samples were used to evaluate the potential for sample collection, handling, and analysis to introduce contamination to water quality samples. The QAQC field replicate samples are used to characterize the variability of analytical results caused by random measurement error, and to estimate any positive or negative bias that might result from method performance. QAQC matrix spikes prepared in the field or laboratory samples were used to assess the effects of the sample matrix and (or) analyte degradation between sampling and analysis (Rowe and others, 2007, Toccalino and others, 2014). We did not identify any QAQC issues that would affect the results of our analysis. The accepted holding time for pesticide samples is 14 days (Sandstrom and others, 2015; Rose and others, 2016); we did not delete any data because of holding time issues. The quality of pesticide data collected in the third decade of sampling is summarized in Bexfield and others (2020), and QAQC of DBCP data for the public supply wells is the same as it is for other groundwater wells and is summarized in Bexfield and others (2022a; 2022b). The data analyzed in this study are available at Lindsey and Stackpoole (2025).

Human Health Benchmarks for Potential Toxicity

The national groundwater network and the DBCP in the Central Valley, Calif. change analyses examined pesticide concentrations relative to HHBs, defined here as relative concentrations. HHBs were used to classify groundwater pesticide concentrations into one of four categories. Pesticide concentrations above an HHB were defined as high. Concentrations that exceeded 0.10 of the HHB but were lower than or equal to the HHB were moderate. Concentrations that exceeded 0.05 of the HHB but were lower than or equal to 0.10 of the HHB were defined as low-moderate. Concentrations lower than or equal to 0.05 of the HHB were low (fig. 3A). These definitions are modified from the definitions used in prior USGS groundwater quality assessment studies (Belitz and others, 2022), with the addition of the low-moderate category.

The HHBs used for this study were the U.S. Environmental Protection Agency (EPA) maximum contaminant levels (MCLs) (U.S. Environmental Protection Agency, 2018). MCLs, established through the Safe Drinking Water Act of 1974 (42 U.S.C. 300(f)), are legally enforceable drinking-water standards for public water supplies. If MCLs were not available, then the EPA human health benchmarks for pesticides (HHBPs) (U.S. Environmental Protection Agency, 2021) or the USGS Health-Based Screening Levels (HBSLs) (U.S. Geological Survey, 2024) were used. HHBPs are nonenforceable drinking water levels that offer information about adverse health effects from drinking water exposure to contaminants that have no drinking water standards or health advisories, and they include noncancer benchmarks for acute (one-day) and chronic (lifetime) drinking water exposures. HBSLs are also nonenforceable guidelines for pesticides for which the EPA has not issued a drinking water health advisory or set an enforceable Federal drinking water standard. The HBSLs are determined only for those compounds that do not already have a MCL or a HHBP. HBSLs are typically determined using the same methodology that the EPA uses for chronic non-cancer HHBPs, with the exception of permethrin, for which the HHBP non cancer acute benchmark was used; the USEPA determined that there is no lower concentration level to protect against chronic or carcinogenic toxicity (U.S. Geological Survey, 2024).

Only pesticides with an HHB were included in the multidecadal pesticide change analysis. Similar to other groundwater pesticide assessments, the assumption of equimolar toxicity described in Bexfield and others (2021) was used for pesticide parent products and their degradates in this analysis. The results represent the number of wells with an exceedance for at least one pesticide; if multiple pesticides exceeded a HHB in a well, only one HHB exceedance was counted for that well. As a result, changes in the potential combined risk of mixtures of pesticides are not addressed in this study.

For most of the compounds analyzed in this study, the magnitude of the HHBs was at least 100 times the magnitude of the highest LRL; therefore, the changing LRLs over time did not affect the application of the method of reporting concentrations relative to HHBs for these compounds. However, for 2 compounds, terbufos and methyl-parathion, the highest LRL was greater than both 10 percent and 5 percent of the HHB. These two pesticides are less likely than the others to be quantified when they are present in groundwater at concentrations near their HHBs, meaning these compounds may be reported as nondetects, when they are actually in the sample (Bexfield and others, 2020). Terbufos and methyl-parathion were only evaluated for potentially exceeding a HHB, not the moderate or low-moderate categories. The highest LRL that we used for DBCP was greater than both 10 percent and 5 percent of the HHB, but not higher than the HHB. We only focused on reporting changes in concentration greater than the HHB for this compound.

National Groundwater Network Decadal Pesticide Analysis

The change in relative concentrations of 21 pesticides sampled across all 59 networks over three decades is presented by category (high, moderate, low-moderate, and low) for each network type and aggregated ecoregion. Toccalino and others (2014) used a different method, the Wilcoxon-Pratt signed-rank test, to determine changes in groundwater pesticide concentrations between decades 1 and 2. The Wilcoxon-Pratt signed-rank test (Pratt, 1959) is a nonparametric test that can be used for analysis of change between matched pairs of wells and two sampling periods. This test was used because 20 to 30 matched-pair trend tests within a well network were available for both decades 1 and 2. This study employed a more conservative, semi-quantitative approach because fewer wells were represented in decade 3, as compared to decades 1 and 2. In the third decade, the rotating schedule of network sampling was maintained, but the collection of pesticide samples was reduced to 8 wells within each network.

DBCP in the Central Valley, California Change Analysis

DBCP concentrations were obtained once per decade from 36 wells over four decades from one network in the Central Valley, Calif. The Wilcoxon-Pratt signed-rank test was used for the analysis of change between two sampling periods, and the Regional Kendall test was used to evaluate changes in DBCP concentration across all four sampling periods (Frans and Helsel, 2005; Helsel and others, 2006). The Regional Kendall test is an adaptation of the Seasonal Kendall test, where the wells were used instead of season of year as the “season” variable.

Bimonthly concentrations from two monitoring wells from 2014 to 2019 to the MCL (0.2 µg/L) were also collected. A Mann-Kendall trend test was used to evaluate temporal changes in these bimonthly data for the period 2014–19 (Helsel and others, 2020). The Mann-Kendall trend test can be used to test for monotonic trends, either a decrease or an increase in concentrations over time, and it does not assume that the data are normally distributed. The statistical analyses performed for this study were considered significant at a probability value (p-value) of less than or equal to 0.1.

National Groundwater Network Decadal Pesticide Analysis

No pesticides were detected at high concentrations, and five pesticides were detected at moderate concentrations (alachlor, atrazine, deethylatrazine [DEA], prometon, and simazine). The percentage of all wells that had pesticide concentrations in the moderate category decreased each decade, from 7 percent in decade 1 to 2 percent in decade 3 (fig. 3A). Alachlor and prometon were detected at moderate concentrations in decade 1. Simazine was detected at moderate concentrations in decades 1 and 3. Atrazine or DEA were detected at moderate concentrations in decades 1, 2, and 3.

The agricultural wells were the well type that had the highest percentages of moderate concentrations, and these percentages decreased each decade (fig. 3B). Moderate concentrations were generally less frequent in the urban as compared to the agricultural wells. Moderate concentrations in the urban wells decreased between decades 1 and 2 from 4 percent to 0 percent and remained at 0 percent in decade 3. The percentage of domestic wells with moderate concentrations (1 percent) was the lowest of all the network types and did not change across the three decades.

Four pesticides (alachlor, atrazine, DEA, and simazine) were detected at low-moderate concentrations. The percentage of wells with detections in that category showed little change across all three decades (fig. 3A). The highest percentages of low-moderate concentrations occurred in the agricultural wells and remained at about 8 percent for each decade (fig. 3B). The percentage of wells with low-moderate concentrations also had little change across the decades in the urban and domestic supply wells.

Pesticides were detected at moderate or low-moderate concentrations during at least one decade in 6 of the 8 aggregated ecoregions. The Northeast, Midcontinent, and the Semiarid West aggregated ecoregions had wells with moderate concentrations in all three decades (figs. 4A and 4B). Atrazine or DEA occurred at moderate concentrations in 6 of the 8 aggregated ecoregions in at least one of the three decades. Moderate concentrations of alachlor, prometon, and simazine were also found in groundwater from some NWQN-GW wells in decades 1 and 3. The Pacific Coast was the only ecoregion where groundwater pesticide monitoring took place in all three decades, and no pesticides were detected in this ecoregion at moderate concentrations. No pesticide monitoring occurred in the Mountain West ecoregion.

DBCP in the Central Valley, California Change Analysis

One network in the Central Valley had DBCP concentrations that exceeded the 0.2 µg/L HHB in the first and second decades of sampling, so continued monitoring of DBCP concentrations in 20–30 wells in this network continued into the third and fourth decades of sampling. The results of the Wilcoxon-Pratt signed rank test indicated statistically significant decreases from the first to second decade (p=0.058) and the second to third decade (p=0.002), but there was no statistically significant change from the third to the fourth decade (p=0.34) (fig. 5). The Regional Kendall test for changes at the network level indicated a statistically significant decrease in DBCP concentrations from 1993 to 2024 (p=0.0006). The percentage of samples exceeding the HHB dropped from 50 percent to 15 percent, most concentrations were less than the MCL in the fourth decade of sampling.

The Mann-Kendall trend test for DBCP concentrations in seasonal samples from the two wells sampled from 2014 to 2019 in the ETN indicated a significant negative trend for each well (monitoring well, tau=−0.6061, and p-value=0.0061; supply well, tau=−0.6095, p-value=0.0001). In the shallower monitoring well, (71-m deep), DBCP concentrations were greater than the HHB for the entire sampling period. In the deeper supply well, (98-m deep), all measured DBCP concentrations were less than the HHB by 2016 and remained less than the HHB to the most recent sampling in 2019 (fig. 6).

National Groundwater Network Decadal Pesticide Analysis

The proportion of pesticide concentrations in the moderate category (greater than 10 percent of the HHB but less than or equal to HHB) decreased over time. Five of the 21 pesticides were found at moderate concentrations: alachalor, atrazine, DEA, simazine, and prometon. These compounds had the highest pesticide concentrations found, and they are the focus of our discussion. In our study, atrazine was detected at moderate concentrations in 6 different ecoregions (Pacific Northwest, Arid West, Semiarid West, Midcontinent, South Atlantic Gulf, and Northeast) in decade 1, but by decade 3, it was only detected in 1, the Midcontinent ecoregion. We hypothesize that one of the processes that may have contributed to the reduction of pesticide concentrations in groundwater include degradation through abiotic or biotic processes in soils or groundwater (Fenner and others, 2013; Aisopou and others, 2015; Kundu and others, 2019). Alternative explanations for the decrease include a reduction in pesticide use or a change in the transport of pesticides to groundwater over time. Changes in pesticide use and soil management, coupled with changes in precipitation and temperature, can vary over time, affecting pesticide transport to groundwater (Arias-Estévez and others, 2008; Bexfield and others, 2021).

Atrazine and its degradate, DEA, were the pesticides that were most commonly observed at moderate concentrations across all three decades in the Semiarid West, Midcontinent, and Northeast aggregated ecoregions. In the United States, atrazine is still one of the most heavily used agricultural pesticides (Wieben, 2025). For roughly the same period of record encompassed in this report, atrazine and DEA were frequently detected at elevated concentrations in surface water for the period from 1993 to 2017, (Larson and others, 1999; Stone and others, 2014; Nowell and others, 2018; Stackpoole and others, 2021). In the prior NGWN-GW groundwater pesticide trends assessment, these two compounds were frequently detected in groundwater in decades 1 and 2. Atrazine concentrations decreased in agricultural wells, and DEA concentrations increased in the urban and domestic drinking water supply wells (Toccalino and others, 2014).

The specific aquifers where atrazine and DEA detections occurred across all three decades include the South Platte River Basin Alluvial aquifer in the Semiarid West, the Western Lake Michigan Drainages Glacial aquifer in the Midcontinent, and the Potomac River Basin Valley and Ridge aquifer in the Northeast. In these specific geographic locations, if atrazine or DEA were detected at moderate concentrations in decade 1, they were also detected at moderate concentrations in decades 2 and 3. We also found geographic locations where either atrazine or DEA were only found in decade 3; this occurred in the Midcontinent ecoregion, the Glacial Aquifer system in the Eastern Iowa Basin, and in the Semiarid West ecoregion, the Alluvial Aquifer system in the High Plains of Texas. Identification of these new locations with moderate concentrations of atrazine or DEA would not have been possible without continued monitoring of this pesticide and its degradate in the environment across all three decades. However, we are unable to determine the drivers of these decadal patterns in the groundwater pesticide concentration data because we lack information about pesticide use and management and the soil conditions and precipitation, key factors affecting transport and degradation rates of pesticides near these aquifer locations.

Moderate concentrations of alachlor, prometon, and simazine were also found in groundwater from some NWQN-GW wells in decades 1 and 3. In 1990, alachlor was a heavily used herbicide, especially where corn was commonly a dominant agricultural crop. However, because of the introduction of acetochlor for corn and glyphosate-resistant soybeans, the use of alachlor has steadily declined since 1994 (Vecchia and others, 2009). Alachlor is not a pesticide that has frequently been detected in prior USGS national-scale groundwater reports, which were focused on decades 1 and 2 of sampling (Kolpin and others, 1998; Toccalino and others, 2014). Prometon was detected at moderate concentrations in the Semiarid West ecoregion in decade 1. Simazine was detected in moderate concentrations in the Arid West and the Northeast in decades 1 and 3. In prior USGS studies prometon and simazine were commonly detected in groundwater (Kolpin and others, 1998; Toccalino and others, 201472).

There were differences in groundwater pesticide trend patterns across the different well types. The agricultural and urban networks capture the quality of recently recharged groundwater underlying selected agricultural land uses, whereas the domestic drinking-water supply wells are generally deeper than the land use networks and reflect a mixed land use signal. A larger number of moderate concentrations were detected in the agricultural as compared to the urban or domestic use networks, but the percentage of agricultural wells with moderate concentrations decreased. In contrast, the percentage of urban and domestic wells with moderate concentrations was lower than the agricultural wells. Moreover, the percentage of moderate pesticide concentrations in urban wells decreased over the three decades, but the percentage remained the same in the domestic supply wells.

Groundwater from individual wells in the domestic-well networks is generally used by rural residents as a primary drinking-water source (U.S. Environmental Protection Agency, 1977; Degnan and others, 202118; Lindsey and others, 2021). In our study, pesticides were detected at moderate concentrations in domestic supply wells in three aggregated ecoregions: the Arid West, Northeast, and Semiarid West. The domestic-well networks within the NWQN-GW cover areas that supply groundwater to more than 6 million people, or about 13 percent of the total number of people relying on domestic supply in the United States (Dieter and others, 2018; Johnson and others, 2019), and these networks cover at least part of the principal aquifers that together represent 99 percent of the withdrawals for domestic supply (Maupin and Arnold, 2010). Domestic-supply wells are not regulated by federal or state laws, and individuals are responsible for the maintenance and any monitoring of these drinking-water sources (Leistra and Boesten, 1989; DeSimone, 2009; Tesoriero and others, 2024).

The rarity of HHB exceedances and the national-scale decrease in moderate concentrations of pesticides can be viewed as encouraging results from a human-health standpoint. However, continued monitoring and assessment of groundwater pesticides is warranted, as many negative human-health effects have been linked to pesticide exposure (Montiel-León and others, 2019; de Souza and others, 2020; Stradtman and Freeman, 2021; McGinley and others, 2023), and these negative effects can occur when pesticide concentrations are below the human health benchmarks used in this study (Ackerman, 2007; Kim and others, 2017; Remigio and others, 2024). Additionally, multiple geogenic constituents (arsenic, lithium, and strontium), as well as nitrate and salt, are common and widely distributed at concentrations exceeding current regulatory thresholds in aquifers underlying the hydrologic regions across the United States (Lindsey and others, 2023; Erickson and others, 2024). Currently, groundwater contaminants from either geogenic or anthropogenic origin are very likely to coexist in groundwater, and there is a scientific gap in understanding about the combined effects of these groundwater contaminants on human health (Xie and others, 2023; Tesoriero and others, 2024).

DBCP in the Central Valley, California Change Analysis

In this study, DBCP was the only pesticide that was found at high concentrations (above the HHB) in groundwater, and this occurred in one well network in the Central Valley, California. Prior studies have documented that nonpoint, agricultural sources of DBCP were found to contribute to groundwater contamination in the Central Valley, Calif. (Loague and Abrams, 1999). The half-life of DBCP, which has been assessed at 6 years (Burow and others, 1998; Burow and others, 2007), is likely a factor contributing to its persistence in the environment. Our study documented that, even 45 years after being banned, DBCP continued to be present in groundwater from some NWQN-GW wells at concentrations greater than the MCL of 2 µg/L. However, samples collected from the decadal NQWN-GW network of 36 wells between 1993 and 2024 and bimonthly samples collected from two ETN wells from 2014 to 2019 indicated that concentrations of DBCP at the locations and depths represented by these wells are steadily decreasing over time, and most concentrations are now less than the MCL.

The analysis of change in groundwater DBCP concentrations in part of the Central Valley of California provided an opportunity to illustrate a localized example of the effects of a legacy pesticide on the environment. Legacy pesticides are those that have been banned or phased out of use and are detected in either surface or groundwater years to decades after application of active ingredients in the environment has ceased (Rasmussen and others, 2015; McManus and others, 2017). Legacy pesticides in groundwater can become problematic for human health if concentrations are elevated relative to human health benchmarks and the groundwater source is used as a drinking-water source (Gilliom and others, 2006; McKnight and others, 2015; Hakoun and others, 2017; McGinley and others, 2023). Agricultural application of DBCP was banned in the conterminous United States in 1979 (California State Water Resources Control Board, 2002). Forty-five years after being banned, groundwater DBCP concentrations were still greater than the maximum contaminant level of 2 micrograms per liter (µg/L). The long-term monitoring data did show that number of exceedances decreased from 50 percent to 15 percent of the samples between 1993 and 2024, indicating progress toward reducing the human health risks of this pesticide in groundwater.