Residual herbicide concentrations in on‐farm water storage–tailwater recovery systems: Preliminary assessment

On‐farm water storage–tailwater recovery systems reduce groundwater usage and intercept agrochemical loads, but pesticide residue dynamics in these systems are not well understood. This study monitored concentrations of seven herbicides in seven northeast Arkansas tailwater recovery systems (April 2017–March 2018). Clomazone, glyphosate, metolachlor, and quinclorac were frequently detected, with minimal detections of 2,4‐D, dicamba, and propanil. Concentrations peaked during the growing season (1 Apr.–15 Sept.), reflecting an interaction of application and precipitation. Clomazone, glyphosate, and quinclorac concentrations were greater in ditches (<0.80–67, <0.50–6.2, and <0.40–62 μg L−1, respectively) than in the associated reservoir (<0.80–6.0, <0.50–4.1, and <0.40–6.0 μg L−1, respectively), but metolachlor concentrations were not different between structure types (maximum 22–32 μg L−1). Off‐season concentrations were mostly below detection, except for quinclorac. Cycling recovered tailwater through the system and irrigating from reservoirs may minimize risk of cross‐crop contaminations with residual herbicides. Managed groundwater recharge should use reservoir water during winter to protect groundwater quality.


INTRODUCTION
Producers in the U.S. mid-South facing groundwater decline have incorporated on-farm water storage-tailwater recovery (OFWS-TWR) systems into their irrigation practices (Fugitt et al., 2011;Yaeger, Massey, Reba, & Adviento-Borbe, 2018;. These networks of ditches paired with a storage reservoir replace 25-50% of a production system's groundwater irrigation on average (Sullivan & Delp, 2012). Reservoirs also have the potential to supply managed groundwater recharge (MAR) . Other benefits of OFWS-TWR systems include reduction of solids and agrochemical loads that contribute to impaired water quality downstream (USDA-NRCS, 2011;USEPA, 2009). Studies of spatial and temporal dynamics of solids and nutrients suggest that OFWS-TWR systems can reduce loads and watershed yields substantially (Omer & Baker, 2019, Omer, Dyer, Czarnecki, Kröger, & Allen, 2018a) by retaining flow during periods of high precipitation (Czarnecki, Omer, & Dyer, 2017;Omer et al., 2018b).
Pesticide residue concentrations and loads may also be reduced by OFWS-TWR systems. Some pesticides may degrade within the extended hydraulic residence time in an OFWS-TWR system (Lewis, Tzilivakis, Warner, & Green, 2016). For others, removal may be promoted by increased spatial and temporal interaction with sediment or aquatic vegetation. For these reasons, models identify ponds as a best management practice (BMP) to reduce chlorpyrifos and diazinon export from agricultural lands (Luo & Zhang, 2009). Longitudinal studies of pesticide dissipation in drainage ditches suggest rapid sorption and/or uptake (Bennett et al., 2005;Moore et al., 2001). But pesticide monitoring in OFWS-TWR systems is rare. The most extensive study to date detected a number of rice (Oryza sativa L.) production herbicides throughout the OFWS-TWR flow system during the growing season (Dewell & Lavy, 1996). In addition, pesticide residue accumulation was observed in sediments of eight Mississippi drainage ditches (Kröger, Moore, & Brandt, 2012), suggesting some "removed" pesticides are stored long-term, with the potential to build up and reenter the water column through sediment desorption or resuspension.
As sources for irrigation and MAR, OFWS-TWR systems pose potential agronomic and environmental challenges due to pesticide persistence and recirculation. Agronomically, OFWS-TWR may result in low-level application of herbicides to nontarget crops, potentially reducing yields. Environmentally, OFWS-TWR systems may be sources of pesticide contamination via discharge to streams or as a water supply for MAR. Thus, integrated monitoring across seasons and system locations is needed for thorough assessment of the effectiveness of OFWS-TWR as a BMP for reducing or removing pesticide residues. The objectives of this study were (a) to monitor target herbicide concentrations in seven northeast Arkansas OFWS-TWR systems (Supplemental Figures S1-S2) over a full year (Apr. 2017-Mar. 2018) and (b) to analyze concentrations for differences between seasons and structure types (ditch or reservoir), with the goal of facilitating management of OFWS-TWR systems to reduce risk of cross-crop and groundwater contamination.

Core Ideas
• Herbicides were frequently detected in on-farm water storage-tailwater recovery systems. • Concentrations were greater in ditches than associated reservoirs during the growing season. • Irrigation from reservoirs may minimize crosscrop contamination risk. • Targeting managed aquifer recharge during the off-season should best protect groundwater quality.
ethyl-6-methylphenyl)-N-(2-methoxy-1-methylethyl) acetamide (metolachlor), N-(3,4-dichlorophenyl)propenamide (propanil), and 3,7-dichloroquinoline-8-carboxylic acid (quinclorac). Ditches and reservoirs were sampled at accessible locations in proximity to the pumps that transport water between structures. Grab samples were collected weekly (Apr.-Aug. 2017) to monthly (Sept. 2017-Mar. 2018) in high density polyethylene bottles from ∼0.5 m depth using a pole sampler and were shipped overnight on ice to the University of Arkansas Residue Laboratory. Samples were stored at 4 • C until filtration through a 0.45-μm nylon membrane within 48 h of receipt. Filtered samples were preserved by freezing until analysis by enzyme-linked immunosorbent assay with photometric detection (glyphosate) or by high performance liquid chromatography with photodiode array detection following solid phase extraction (SPE). During SPE, acidified samples (0.5% v/v phosphoric acid), were concentrated from 200 to 8 ml 50:50 acetonitrile/methanol using Strata-X reverse-phase polymer columns. Columns were conditioned with 100% methanol, equilibrated with 0.5% v/v phosphoric acid in ultrapure water, and rinsed with 20% v/v methanol and 0.5% v/v phosphoric acid in ultrapure water. Eluates were analyzed using a mobile phase gradient (34-64% v/v acetonitrile in 0.1% v/v phosphoric acid over 20 min). For all analytes, SPE recovery was 100% ± 5%, confirmed with fortified samples for each new Strata-X lot. Bulk water sample concentrations were calculated by multiplying the measured eluate concentration by the ratio of eluate and sample volumes. Nondetections or concentrations below reporting limits (i.e., 10 times the detection limit; 2,4-D, propanil, and quinclorac = 0.40 μg L −1 ; dicamba and clomazone = 0.80 μg L −1 ; glyphosate = 0.50 μg L −1 ; metolachlor = 2.0 μg L −1 ) were censored in subsequent analysis. were used when censoring frequency was <50% or ≥50-80%, respectively (Helsel, 2012). Summary statistics were not calculated when censoring frequency was >80%. Herbicide concentrations were compared for differences between seasons using generalized Wilcoxon tests and between adjacent ditches and reservoirs using paired Prentice-Wilcoxon tests. Summary statistic calculations and generalized Wilcoxon tests were conducted in R, version 3.1.6 (R Core Team, 2019) using the NADA (Lee, 2017) and interval (Fay & Shaw, 2010) packages. Paired Prentice-Wilcoxon tests were conducted in Minitab 19 (Minitab, 1998) using the PPW macro (Helsel, 2012).
Seasonal dynamics in OFWS-TWR residual herbicide concentrations are congruent with previous studies of agricultural watersheds, especially the trend of a "spring flush" For censoring frequency >80%, median is known only to be below the reporting limit.
c For censoring frequency >80%, mean and standard deviation cannot be estimated. F I G U R E 1 Frequency of (a) clomazone, (b) glyphosate, (c) metolachlor, and (d) quinclorac detections greater than the reporting limit, >1.0 μg L −1 (except metolachlor, which had a reporting limit of 2.0 μg L −1 ), and >10 μg L −1 by month during the study period April 2017-March 2018. (e) Craighead County, AR, monthly precipitation totals and 1981-2010 climate normals (Arguez, Durre, Applequist, Squires, & Vose, 2010) by month. Shaded areas (a-d) approximate common application timing for the detected herbicides in the region (Barber et al., 2019) (Thurman, Goolsby, Meyer, & Kolpin, 1991). For 62 lake and river sites in four Arkansas counties, spring and summer samples comprised 73% of pesticide detections (Senseman, Lavy, Mattice, Gbur, & Skulman, 1997). However, unex-plained spikes in concentrations occurred, including elevated clomazone (2.9 μg L −1 ) and quinclorac (20 μg L −1 ) concentrations in a November ditch sample. Recurrence of springapplied herbicides could indicate re-entrainment of sediments T A B L E 2 Summary statistics of the four herbicides frequently detected in the on-farm water storage-tailwater recovery systems by structure type, calculated across seasons and for the growing season (1 Apr.-15 Sept. 2017) andoff-season (16 Sept. 2017-15 Mar.  Note. GS, growing season; OS, off-season; Rsvr, reservoir. a A positive or negative median of differences between paired observations in Prentice-Wilcoxon tests indicates higher or lower concentrations, respectively, in ditches when p < .05. b For censoring frequency >80%, median is known only to be below the reporting limit. c For censoring frequency >80%, mean and standard deviation cannot be estimated. and/or desorption from sediment. Further inquiry, including mass spectrometric identification of compounds, is required to interpret such events.

Differences between OFWS-TWR structure types
During the growing season, clomazone (median ditch, 0.29 μg L −1 ; median reservoir, <0.80 μg L −1 ), glyphosate (median ditch, 0.71 μg L −1 ; median reservoir, 0.28 μg L −1 ), and quinclorac (median ditch, 2.0 μg L −1 ; median reservoir, 0.90 μg L −1 ) were greater and more frequently detected in ditches than in the adjacent reservoir (Table 2; p < .001). These findings differ from comparisons of nutrients and solids between system structures (Moore, Pierce, & Farris, 2015) but are congruent with the concept that residues are diluted along the flow path by mixing with increasingly large water volumes at lower concentrations, in tandem with sedimentation and degradation over time. Mattice, Skulman, Norman, and Gbur (2010) observed this pattern in four regional river networks, with pesticide concentrations decreasing with increasing discharge. Further, 62% of pesticide detections occurred in rivers and streams relative to lakes and reservoirs in the survey by Senseman et al. (1997). In contrast, growing season metolachlor concentrations (median 0.48-0.78 μg L −1 ) were not different between ditches and reservoirs (p = .83). Elevated metolachlor concentrations in reservoirs at the end of the flow system, compared with the other detected herbicides, may reflect low meto-lachlor sorption to soil/sediment and transport mainly in the dissolved phase (Lerch, Lin, Goyne, Kremer, & Anderson, 2017), as well as slow degradation in environmental waters (Liu, Maguire, & Pacepavicius, 1995). Indeed, metolachlor and its metabolites are frequently detected in surface and groundwaters of agricultural regions (Battaglin, Furlong, Burkhardt, & Peter, 2000;Rebich, Coupe, & Thurman, 2004;Tagert, Massey, & Shaw, 2014).
During the off-season, reservoir quinclorac concentrations (median 0.70 μg L −1 ) exceeded ditch concentrations (median 0.40 μg L −1 ; p < .001), but differences were small in magnitude compared with the growing season. Quinclorac has an estimated degradation half-life of 168-913 d (Lewis et al., 2016) and was persistent within the growing season in the OFWS-TWR systems monitored by Dewell and Lavy (1996) and rivers monitored by Mattice et al. (2010). Therefore, it is not surprising to find quinclorac in OFWS-TWR systems during the off-season. Reservoirs may be a location for quinclorac residue build-up that persists into winter months.

Preliminary examination of OFWS-TWR herbicide residues for potential risks
Maximum detected herbicide concentrations did not exceed national drinking water standards (glyphosate <0.7 mg L −1 ; USEPA, 2018) or human health advisories (metolachlor <0.7-2 mg L −1 ) and benchmarks (clomazone <5.4-30 mg L-1; quinclorac <2.4-60 mg L −1 ; USEPA, 2014b). Observed clomazone, glyphosate, and quinclorac concentrations are not expected to be lethal to fish, invertebrates, or nonvascular plants after acute or chronic exposures (USEPA, 2014a). This is congruent with a previous ecotoxicological assessment of OFWS-TWR system water and sediment (Moore et al., 2015). Maximum metolachlor concentrations, however, were within ranges known to be toxic to fish, invertebrates, and nonvascular plants. Assessments of crop sensitivities to residual herbicides in irrigation are limited. Soybean is sensitive to dicamba doses in irrigation exceeding ∼30-160 g ha −1 equivalent to 0.05-0.14 mg L −1 in 3 ac-in of irrigation (Grantz, Lee, Willett, & Norsworthy, 2020;Willett, Grantz, Lee, Thompson, & Norsworthy, 2019). This study detected 2,4-D, quinclorac, clomazone, glyphosate, and propanil concentration maxima of similar magnitude to the dicamba concentration range reported to injure soybean. However, the detected herbicides in irrigation may be less likely to damage crops, as soybean and dicamba are a known high-sensitivity pairing (Egan, Barlow, & Mortensen, 2014).

CONCLUSIONS
Herbicides applied to fields adjacent to OFWS-TWR systems were frequently detected in ditches and reservoirs, with peak concentrations during the growing season in ditches. While it is not known if residual herbicides at detected concentrations can lead to cross-crop injury, risk may be minimized by cycling tailwater through the reservoir for treatment and for sourcing irrigation. This strategy may be effective for clomazone, glyphosate, and quinclorac, but managing metolachlor residues may be more complex due to its persistence in reservoirs during the growing season. Residual herbicide concentrations were lowest in the OFWS-TWR systems in winter, although evidence of off-season quinclorac persistence in reservoirs was observed. Targeting winter months for MAR using OFWS-TWR reservoirs should best protect groundwater quality.

ACKNOWLEDGMENTS
We thank our producer cooperators: Lindy Alexander, Roger Bradley, Jerry Don Clark, Bryan Huber, Mickey Seeman, Gary Sitzer, and Tom Wimpy. Funding was provided by the Arkansas Water Resources Center through the U.S. Geological Survey 104b National Institute of Water program and the Arkansas Soybean Promotion Board. The research was supported by the University of Arkansas System Division of Agriculture and the U.S. Department of Agriculture, Agricultural Research Service (USDA-ARS). The authors acknowledge the technical assistance of Ian Godwin and Klarissa Kahill of the USDA-ARS Delta Water Management Research Unit. Mention of trade names or commercial products in this publication is solely for the purpose of providing specific information and does not imply recommendation or endorsement by the U.S. Department of Agriculture, which is an equal opportunity provider and employer.