Continuous (1-hour maximum interval) streamflow was measured at the Highway 50 and Sedgwick sites using standard USGS methods (Sauer and Turnipseed, 2010; Turnipseed and Sauer, 2010). Physicochemical properties affecting water quality were continuously measured (1-hour maximum interval) at the Highway 50 and Sedgwick sites. Physicochemical properties that were measured included specific conductance and turbidity. Specific conductance and turbidity have been used as surrogates for pesticide concentrations at the Highway 50 and Sedgwick sites (Christensen and others, 2003; Rasmussen and others, 2016; Stone and Klager, 2022, 2023). Water-quality monitors were installed near the centroid of the stream cross section. The Highway 50 and Sedgwick sites were equipped with a YSI Incorporated 6600 Extended Deployment System water-quality monitor (YSI Incorporated, 2012a) in May and April 1998, respectively, to continuously (60-minute interval) measure specific conductance. Turbidity sensors were added to the water-quality monitors in October and September 1998, respectively. Water-quality monitors were maintained following standard USGS procedures (Wagner and others, 2006).

Some equipment was upgraded during the project. The YSI 6026 turbidity sensors installed at the Highway 50 and Sedgwick sites in October 1998 were replaced with YSI 6136 turbidity sensors in January 2007 and July 2004, respectively. A YSI EXO2 water-quality monitor (YSI Incorporated, 2012b) equipped with specific conductance and YSI EXO turbidity sensors was installed in January 2017 at the Highway 50 site and in September 2014 at the Sedgwick site. The YSI 6026 and YSI 6136 turbidity measurements were converted to YSI EXO2 turbidity measurement values using conversion factors developed from side-by-side sensor measurements documented in figure 1.1 and table 2.1 in Stone and Klager (2023).

Discrete surface-water-quality samples were collected at the two study sites during a range of streamflow conditions and seasons. Samples were primarily collected using depth- and width-integrated sample collection techniques (U.S. Geological Survey, 2006) and are coded in WDFN as either equal-width increment “EWI” or “Multiple verticals” under “SampleCollectionMethod_IdentifierContext.” Discrete water-quality samples were analyzed for pesticides at both study sites. Specific pesticides were included in this report because they had a documented history of frequently being detected in Little Arkansas River water samples at study sites (table 1.5 in Stone and others, 2019, and table 4 in Stone and Klager, 2023). Discrete samples collected at the Sedgwick site were additionally analyzed for glyphosate and AMPA. Pesticide collection and analyses followed methods described by Ziegler and Combs (1997), Ziegler and others (1999, 2010), Stone and others (2012, 2016, 2019), Tappa and others (2015), and Stone and Klager (2022, 2023). Water-quality samples were analyzed for atrazine by the USGS National Water Quality Laboratory (NWQL) using a gas chromatography/mass spectrometry (GCMS) method using NWQL schedule 2003 (for samples collected from 1995 through September 2017) or 2033 (for samples collected from November 2017 through 2023) and the Method identifier “GCM35” (“ResultAnalyticalMethod_Identifier” in WDFN). Atrazine was isolated from filtered water samples by solid-phase extraction and analyzed using capillary-column GCMS with selected-ion monitoring (Zaugg and others, 1995). Atrazine samples analyzed using NWQL schedule 2003 had detection and reporting limits of 0.004 µg/L and 0.007 µg/L, respectively, for samples collected from 1995 through September 2010 and detection and reporting limits of 0.004 µg/L and 0.008 µg/L, respectively, for samples collected from October 2010 through September 2017. Atrazine samples analyzed using NWQL schedule 2033 had detection and reporting limits of 0.004 µg/L and 0.008 µg/L, respectively, for samples collected from October 2017 through December 2019 and detection and reporting limits of 0.001 µg/L and 0.008 µg/L, respectively, for samples collected during January 2020 through December 2023. Water-quality samples were analyzed for glyphosate and AMPA by the USGS Organic Geochemistry Research Laboratory using a liquid chromatography/tandem mass spectrometry method (WDFN “ResultAnalyticalMethod_Identifier” of either “LCM41” or “LCM57”) as described in Meyer and others (2009). Glyphosate and AMPA had detection and reporting limits of 0.01 µg/L and 0.02 µg/L, respectively.

Continuous streamflow, specific conductance, and turbidity values were obtained using the dataRetrieval package (De Cicco and others, 2025) in the R (version 4.3.0) programming language (R Core Team, 2023) to get unit (hourly) values. Duration curves were used to summarize these data. Duration curves are cumulative distribution functions and were constructed using hourly values to evaluate and compare frequency and magnitude characteristics at the Highway 50 and Sedgwick sites during the study period. Duration curves indicate the percentage of time that specified conditions were equaled or exceeded, or the frequency of exceedance (Searcy, 1959). The Weibull formula (Weibull, 1939; Helsel and others, 2020) was used to plot position. Statistical summaries of discrete pesticide data were computed with the Regression on Order Statistics method using the censtats function from the NADA package (Lee, 2020) in the R (version 4.3.0) programming language (R Core Team, 2023).

Agricultural pesticide-use data were summed for 1992 through 2019 using data from Baker and Stone (2013), Stone (2013), and Wieben (2019, 2021) for selected pesticides. Estimation methods to get data for county-level pesticide use are described in Thelin and Stone (2013). These data include two annual estimates called Epest-low and Epest-high that are variations on estimates that treat missing data differently (Thelin and Stone, 2013). Epest-low considers missing-use reports to be zero use, and Epest-high estimates consider missing-use reports based on surrounding area use (Thelin and Stone, 2013). The 2018 and 2019 use estimates are currently (2026) considered preliminary (Wieben, 2021). Epest-high estimates were used in this report for simplification.

Annual agricultural pesticide use in the Highway 50 and Sedgwick site drainage basins was calculated from 2001 to 2019 by proportioning Epest-high county-level pesticide use estimates to the cropland in each county for counties having any land area in the study drainage basins. A geographic information system was used to estimate cultivated cropland in square miles for Kansas, study drainage basin counties (Ellsworth, Rice, McPherson, Marion, Reno, Harvey, and Sedgwick), study drainage basin areas, and parts of counties within the study drainage basin areas. Cultivated cropland data were downloaded from the USGS National Land Cover Database (Dewitz and U.S. Geological Survey, 2021) for the years 2001, 2004, 2006, 2008, 2011, 2013, 2016, 2019, and 2021. Cultivated cropland area for years in between (2002, 2003, 2005, 2007, 2009, 2010, 2012, 2014, 2015, 2017, 2018, and 2020) were estimated using linear interpolation. Estimated drainage basin pesticide use in study drainage basins was computed as the proportion of cropland in each county that was contained in the drainage basin.

Trends for atrazine (herbicide) were analyzed at the Highway 50 and Sedgwick sites. Trends for glyphosate (herbicide) and AMPA (a degradation product of glyphosate) were analyzed at the Sedgwick site. Pesticide trends were analyzed using SEAWAVE-Q, a parametric regression model that uses a seasonal wave (seawave), adjusts for streamflow (Q), and uses other ancillary variables (Ryberg and Vecchia, 2013; Ryberg and York, 2020). Although discrete pesticide samples collected within 1 to 7 days of each other are typically screened to prevent autocorrelation using SEAWAVE-Q, these occasional samples were retained in the datasets for analysis in this report. The SEAWAVE-Q model is specifically designed for analyzing seasonal- and flow-related variability and trends in pesticide concentration. SEAWAVE-Q is capable of robustly fitting seasonal patterns of pesticides. In addition to including seasonality and streamflow as variables, the SEAWAVE-Q model allows for other continuously measured physicochemical parameters, such as daily mean specific conductance and turbidity, to be included as surrogate variables; however, if daily mean surrogate variable data were missing on a sample day, then those sample data were not used in the model. For SEAWAVE-Q analysis, continuous daily streamflow values were obtained using the importDVs function from the waterData package (Ryberg and Vecchia, 2012) in the R (version 4.3.0) programming language (R Core Team, 2023). Flow anomaly variables were obtained using the compAnom function from the waterData package (Ryberg and Vecchia, 2012) in the R (version 4.3.0) programming language (R Core Team, 2023). Specific conductance and turbidity values were obtained using the dataRetrieval package (De Cicco and others, 2025) in the R (version 4.3.0) programming language (R Core Team, 2023) to pull unit (hourly) values and then compute daily means from the unit values.

SEAWAVE-Q models that included (1) only seasonality and streamflow variables and (2) seasonality and streamflow variables, daily mean specific conductance, and daily mean turbidity variables were developed for comparison. Three dimensionless streamflow anomalies were included in the SEAWAVE-Q models to account for flow-related pesticide concentration variability: short-term (day-to-day) flow anomaly (STFA), mid-term (30- to 365-day) flow anomaly (MTFA), and long-term (greater than 365 days) flow anomaly (LTFA). Option 1 from the compAnom function of the waterData package (Ryberg and Vecchia, 2012) was selected, which calculated the 1-day, 30-day, and 1-year anomalies for STFA, MTFA, and LTFA, respectively. The Restricted Cubic Spline model option (for long trend periods) and the default 4 knots were selected to model pesticide trends. The SEAWAVE-Q models were fitted to the pesticide data using the seawaveQ (Ryberg and York, 2020) package in the R (version 4.3.0) programming language (R Core Team, 2023). Models were selected that minimized Aikake’s Information Criterion (AIC; Akaike, 1974) and maximized the coefficient of determination (R²). AIC is the model deviance plus two times the number of fitted coefficients, and R² is the fraction of the variance explained by regression (Helsel and others, 2020). Pesticide loads were computed using daily SEAWAVE-Q model-computed concentration estimates that were corrected for retransformation bias. Percentage change in starting and ending pesticide concentrations is the trend expressed as a percentage in net change during the period of record (Ryberg and York, 2020). Descriptive statements of trend likelihood were used to assign trend likelihood categories following the methods of Hirsch and others (2015) for water-quality constituent trend analysis using a WRTDS model bootstrap test using the SEAWAVE-Q trend likelihoods and thresholds defined in Hirsch and others (2015).

Atrazine is a chlorinated triazine synthetic herbicide used to selectively control annual broadleaf weeds and grasses before and after they emerge (Pohanish, 2015). Atrazine is also commonly used on corn and sorghum, which are crops commonly grown in the study area. Atrazine can cause cardiovascular and reproductive problems in humans (U.S. Environmental Protection Agency, 2009). The National Pollutant Discharge Elimination System ASR phase II treatment facility (Kansas permit number: I–LA24–PO01; Federal permit number: KS0099694) required monthly atrazine monitoring during facility operation for the periods 2010–14 and 2015–19. Atrazine affects aquatic plant communities and there is a potential for chronic risk to aquatic invertebrates, amphibians, and fish in areas where atrazine use is substantial (U.S. Environmental Protection Agency, 2016).

Atrazine was detected in all discrete samples collected at the study sites (table 3, table 2.1). The Federal MCL for atrazine is 3 µg/L (U.S. Environmental Protection Agency, 2009) and was exceeded in discrete samples from the Highway 50 and Sedgwick sites (table 3), primarily in the months of May and June. Mean atrazine was 4.05 µg/L and 3.36 µg/L at the Highway 50 and Sedgwick sites, respectively, and concentrations did not exceed previously reported ranges from 1995 through 2021 (Tappa and others, 2015; Stone and others, 2016, 2019; Stone and Klager, 2022, 2023) at either site during the study period (table 3).

Glyphosate is a broad-spectrum herbicide used to control annual and perennial plants (Pohanish, 2015). Glyphosate can cause kidney problems and reproductive difficulties in humans (U.S. Environmental Protection Agency, 2009). AMPA is a chemical that results from the breakdown of glyphosate. The largest glyphosate and AMPA concentrations were during May, June, and July of the study. The EPA MCL for glyphosate is 700 µg/L (U.S. Environmental Protection Agency, 2009). Glyphosate and AMPA were detected in all the samples collected during the study but were analyzed only for the Sedgwick site. During the study period, glyphosate concentrations ranged from 0.050 to 7.80 µg/L, and AMPA concentrations ranged from 0.290 to 4.30 µg/L (table 3). Mean concentrations of glyphosate and AMPA were 0.934 and 1.45 µg/L, respectively (table 3); these concentrations are substantially larger than the median annual averages (0.09 µg/L and 0.35 µg/L, respectively) for Midwest streams during 2015 through 2017 (Medalie and others, 2020). Glyphosate and AMPA concentration ranges reported in this study did not exceed the ranges previously reported (Stone and Klager, 2023; table 3).

Atrazine is consistently one of the most frequently detected pesticides in the Nation’s rivers and streams (Gilliom, 2007; Stone and others, 2014; Stackpoole and others, 2021; Shoda and others, 2026). The SEAWAVE-Q models using streamflow and additional timeseries data selected for atrazine included explanatory variables streamflow, turbidity, STFA, MTFA, and LTFA at the Highway 50 site and streamflow, specific conductance, turbidity, STFA, MTFA, and LTFA at the Sedgwick site (table 4). The amount of variance explained by the SEAWAVE-Q atrazine models was 64 to 66 percent (table 4). Surrogate regression models for atrazine at the Highway 50 and Sedgwick sites published in Stone and Klager (2023) included turbidity and seasonal components as explanatory variables, and the amount of variance explained by these models was 58 to 69 percent at the Highway 50 and Sedgwick sites, respectively.

SEAWAVE-Q computed atrazine concentrations ranged from 0.006 to 20.6 µg/L with a mean of 0.954 µg/L from October 1998 through December 2023 at the Highway 50 site and from 0.051 to 21.1 µg/L with a mean of 1.15 µg/L from September 1998 through December 2023 at the Sedgwick site (table 5). SEAWAVE-Q computed atrazine concentrations exceeded the Federal MCL of 3.0 µg/L about 8 percent of the time at both study sites from 1998 to 2023 (fig. 3), and this typically happened at both sites during May and June. SEAWAVE-Q computed atrazine concentrations for this study had relatively similar maxima (within 5 percent) and relatively smaller means (21 to 36 percent smaller) and MCL exceedances (25 percent smaller) than those computed previously using surrogate regression models (Stone and Klager, 2023) during the same periods. SEAWAVE-Q computed atrazine concentrations at the Highway 50 site ranged from 0.041 to 12.0 µg/L (mean=0.699 µg/L) and exceeded the Federal MCL about 3 percent of the time in comparison with previously published surrogate regression-computed atrazine concentrations that ranged from 0.034 to 12.0 µg/L (mean=1.10 µg/L) and exceeded the Federal MCL about 10 percent of the time from 2017 through 2021 (table 5; Stone and Klager, 2023). SEAWAVE-Q computed atrazine concentrations at the Sedgwick site ranged from 0.088 to 12.1 µg/L (mean=0.883 µg/L) and exceeded the Federal MCL about 4 percent of the time in comparison with previously published surrogate regression-computed atrazine concentrations that ranged from 0.030 to 11.6 µg/L (mean=1.12 µg/L) and exceeded the Federal MCL about 10 percent of the time from 2014 through 2021 (table 5; Stone and Klager, 2023).

Atrazine trends were inconsistent between study sites. Atrazine had an increasing trend at the Highway 50 site that was highly likely to occur and a decreasing trend at the Sedgwick site that was likely to occur during the study period; neither of these trends were statistically significant (table 6, fig. 4A, B). However, the atrazine trend at the Sedgwick site did not appear to follow atrazine use, which did not change substantially in the study drainage basin from 2001 to 2019 (figs. 7.1A, 8.2). Atrazine concentrations increased by 98 percent at the Highway 50 site and decreased by 31 percent at the Sedgwick site from 1998 through 2023 (table 6). Atrazine concentrations showed a strong seasonal pattern (fig. 4A, B), with larger estimated concentrations happening in late spring and early summer. Decreasing atrazine trend results at the Sedgwick site during this study are consistent with the decreasing atrazine trend results at most of the sites studied for national pesticide trend assessments (Ryberg and others, 2014; Ryberg and Gilliom, 2015; Oelsner and others, 2017; Shoda and others, 2026). Differences in long-term water-quality constituent trend directions between the Highway 50 and Sedgwick sites are not uncommon (Stone and Klager, 2022). Sulfate, nitrate plus nitrite, and total phosphorus had different concentration trend directions at the study sites from 1996 to 2021. Sulfate had a decreasing trend at the Highway 50 site and an increasing trend at the Sedgwick site; nitrate plus nitrite and total phosphorus had increasing trends at the Highway 50 site and decreasing trends at the Sedgwick site (Stone and Klager, 2022).

Kansas has been underrepresented in efforts that have quantified pesticide trends in rivers and streams across the Nation and regionally (Sullivan and others, 2009; Ryberg and others, 2010, 2014; Oelsner and others, 2017) until recently (Shoda and others, 2026). Atrazine trend results in this study generally follow those found across the Nation (Ryberg and others, 2014; Ryberg and Gilliom, 2015; Oelsner and others, 2017; Shoda and others, 2026). Average atrazine concentrations at the Little Arkansas River study sites were commonly substantially greater (tables 3 and 5) in comparison with those reported by Ryberg and others (2017), and both measured and model-computed atrazine concentrations in this study exceeded the MCL more frequently. National atrazine trends were assessed in Ryberg and others (2014), Ryberg and Gilliom (2015) and Oelsner and others (2017) using SEAWAVE-Q. Pesticides that were used predominately in agricultural settings, such as atrazine, had similar riverine concentration and use trends and represent pesticide application in agricultural settings (Ryberg and others, 2014). Ryberg and others (2014) found that instream atrazine concentration trends were generally consistent with use trends, which declined from 2003 to 2010. Ryberg and Gilliom (2015) found decreasing atrazine trends at most sites in the Mississippi River region from 1992 to 2010. Oelsner and others (2017) found increasing atrazine trends at 21 sites and decreasing atrazine trends at 53 sites across the United States from 1992 to 2012. Breitmeyer and others (2026) described atrazine trends that were decreasing at 60 percent and increasing at 40 percent of 45 sites across the conterminous United States during 2013–22; 2 of the 45 study sites were in Kansas, included the Sedgwick site, and had nonsignificant increasing atrazine trends during 2013–22 (Breitmeyer and others, 2026).

Although both study sites had strong seasonal patterns and interannual variability in their atrazine concentrations, atrazine concentrations regularly (about 8 percent of the time) exceeded 3 µg/L over the study period (fig. 3). Despite the lack of significant changes in atrazine concentrations over time at the Little Arkansas River study sites, trend analysis during targeted flow conditions could help identify the drivers of long-term water quality, including changes in pesticide concentrations. Given the challenges associated with quantifying BMP effectiveness, future study efforts in the Little Arkansas River drainage basin could be focused on smaller drainage basins in the study area and include consistent and long-term discrete pesticide sampling and accompanying in situ water-quality monitoring to better quantify BMP effectiveness.

Glyphosate is one of the most widely used herbicides in the United States and worldwide, and its use has increased substantially since its introduction in 1974, particularly after the introduction of glyphosate-resistant crops (Richmond, 2018; Duke, 2020). Glyphosate use is expected to continue to increase in the United States (Richmond, 2018). The selected SEAWAVE-Q glyphosate model in this study used streamflow, specific conductance, turbidity, STFA, MTFA, and LTFA as explanatory variables, and the amount of variance explained by the SEAWAVE-Q model was 72 percent (generalized R² of 0.72; table 4). A previously developed glyphosate surrogate regression model used for the Sedgwick site used turbidity and seasonal components as explanatory variables and had an adjusted R² value of 0.59 (Stone and Klager, 2023). SEAWAVE-Q computed glyphosate concentrations for this study had relatively smaller minima (32 percent smaller), maxima (30 percent smaller), and mean concentrations (20 percent smaller) than those computed previously with surrogate regression models (Stone and Klager, 2023) during similar periods (March 2015 through December 2021 for this study [table 5] compared with September 2014 through December 2021 for Stone and Klager [2023]). Previously published surrogate regression-computed glyphosate concentrations ranged from 0.108 to 5.81 µg/L (mean=0.662 µg/L) from September 2014 through December 2021 at the Sedgwick site (Stone and Klager, 2023). Computed glyphosate concentrations for this study ranged from 0.073 to 4.04 µg/L with a mean value of 0.528 µg/L from 2015 to 2023 (table 5). Larger computed glyphosate concentrations occurred during June and July. Computed glyphosate concentrations at the Sedgwick site never exceeded the Federal MCL of 700 µg/L during the study (table 5; fig. 5A).

The AMPA model in this study included streamflow and specific conductance as explanatory variables, and the amount of variance in AMPA concentration explained by the model was 78 percent (table 4). Stone and Klager (2023) developed an AMPA surrogate relation for the Sedgwick site using turbidity and seasonal variables as explanatory variables; the amount of variance in AMPA concentrations explained by this (Stone and Klager, 2023) previously developed AMPA model was 76 percent. SEAWAVE-Q model-computed AMPA concentrations for this study ranged from 0.337 to 3.45 µg/L and had a mean of 1.18 µg/L during March 2014 through December 2023 (table 5). SEAWAVE-Q computed AMPA concentrations for this study had relatively smaller minimum values (20 percent smaller), larger maximum values (22 percent larger), and similar average values (<1 percent) than those computed previously using surrogate regression models for the same period (September 2014 through December 2021; Stone and Klager, 2023). Previously published surrogate regression-computed AMPA concentrations ranged from 0.425 to 2.82 µg/L (mean=1.19 µg/L) during September 2014 through December 2021 at the Sedgwick site (Stone and Klager, 2023).

Trends in glyphosate and AMPA concentrations were not detected; decreasing trends were about as likely as not to occur and were not significant (table 6). No substantial glyphosate use changes were detected during the coinciding years of available glyphosate use data and trend analysis (2015–19); however, from 1992 through 2019, the Sedgwick drainage basin counties had substantial glyphosate use increases (app. 7 and 8, figs. 7.1B, 8.3) and so did Kansas as a whole (app. 4 and 5, fig. 4.1 and tables 5.1 and 5.2). The measured and computed glyphosate concentrations showed a strong seasonal pattern, with larger estimated concentrations in early summer months of June and July (fig. 6A). AMPA concentrations decreased by 4 percent from 2014 to 2023 (table 6, fig. 6B).