Statistical modeling of agricultural chemical occurrence in midwestern rivers

W.A. Battaglin; D. A. Goolsby

doi:10.1016/s0022-1694(97)00011-5

Home/Archive/USGS

USGS · 70019606

Statistical modeling of agricultural chemical occurrence in midwestern rivers

W.A. Battaglin·D. A. Goolsby

View original source ↗Export metadata ↓

Abstract

Agricultural chemicals in surface water may constitute a human health risk or have adverse effects on aquatic life. Recent research on unregulated rivers in the midwestern USA documents that elevated concentrations of herbicides occur for 1-4 months following application in late spring and early summer. In contrast, nitrate concentrations in unregulated rivers are elevated during fall, winter, and spring months. Natural and anthropogenic variables of fiver drainage basins, such as soil permeability, amount of agricultural chemicals applied, or percentage of land planted in corn, affect agricultural chemical concentration and mass transport in rivers. Presented is an analysis of selected data on agricultural chemicals collected for three regional studies conducted by the US Geological Survey. Statistical techniques such as multiple linear and logistic regression were used to identify natural and anthropogenic variables of drainage basins that have strong relations to agricultural chemical concentrations and mass transport measured in rivers. A geographic information system (GIS) was used to manage and analyze spatial data. Statistical models were developed that estimated the concentration, annual transport, and annual mean concentration of selected agricultural chemicals in midwestern rivers. Multiple linear regression models were not very successful (R2 from 0.162 to 0.517) in explaining the variance in observed agricultural chemical concentrations during post-planting runoff. Logistic regression models were somewhat more successful, correctly matching the observed concentration category in 61-80% of observations. Linear and multiple linear regression models were moderately successful (R2 from 0.522 to 0.995) in explaining the variance in observed annual transport and annual mean concentration of agricultural chemicals. Explanatory variables that were commonly significant in the regression models include estimates of agricultural chemical use, crop acreage, soil characteristics, and basin topography.

Explore related subjects

Journal of Hydrology Article

Keep this discovery

Explore connections, maps & timelines

BibTeX RIS

W.A. Battaglin, D. A. Goolsby. 1997. Statistical modeling of agricultural chemical occurrence in midwestern rivers. https://doi.org/10.1016/s0022-1694(97)00011-5

Cite the original work for its findings. Save a collection to share your selection of sources.

KEEP EXPLORING

Related USGS reports

USGS2026

Reservoir releases and land cover interact to drive event-scale nitrate export in a large agricultural basin

Understanding the drivers of nitrate export in rivers is critical for developing effective nutrient management strategies. However, few studies have explored event-scale drivers of export in large river basins with human modifications like reservoirs. Here, we analyzed nitrate concentration-discharge (C-Q) relationships from 215 events at the outlet of the Kansas River Basin, USA (155,690 km 2 ) from 2014 to 2022 to (1) characterize event-scale nitrate export behaviors in a large agricultural basin, and (2) determine how different event characteristics are linked to these behaviors. We found that C-Q behaviors varied greatly, with 60% of events producing nitrate enrichment (n = 130) and 40% of events producing nitrate dilution (n = 85). These behaviors were correlated with complex spatial interlinkages between climate and land cover: across most of the basin, nitrate enrichment was correlated with drier antecedent conditions, but in wetter areas with higher proportions of urban and forested land cover, enrichment was more strongly correlated with precipitation magnitude/intensity. This difference in hydroclimatic controls on nitrate export might be related to differential distributions of nitrate sources within these land covers. Upstream of major reservoirs, however, neither variable was strongly correlated with C-Q behavior, suggesting that nitrate attenuation within reservoirs decouples event-scale concentration signals in upstream waters from those downstream. Reservoir outflows had variable impacts on C-Q behavior, reflecting reservoir-specific variations in nitrate attenuation efficiency. Together, these results identify specific complex interactions between hydroclimate, land cover, reservoir positioning, and individual reservoir properties that control event-scale nitrate export from large basins.

Colorado, Kansas, Nebraska↗

USGS2026

Deep learning error post-processing improves stochastic watershed modeling

Hydrologic extremes, including floods and droughts, pose substantial societal risks that are expected to intensify with climate change. Deterministic watershed models (DWMs) remain a mainstay for modeling these extremes, but lack explicit representation of uncertainty, limiting their utility for risk-informed planning. Stochastic watershed models (SWMs) address this limitation by generating ensembles of streamflow via models of observed DWM residuals. However, most SWMs struggle with the complex dependence between DWM residuals and the underlying hydrologic state, which can complicate stochastic simulations under nonstationary climates. Deep learning (DL) models, whether used as standalone models or post-processors for process-based DWMs, offer a pathway to address this challenge by reducing conditional dependence. In this study, we evaluate SWMs applied to seven models: three process-based models (PRMS, Hymod, and HBV), their hybrid process-DL counterparts, and a pure DL DWM, focusing on daily simulations and extremes under both historical conditions and synthetic climate change scenarios. Results for a case study watershed in Massachusetts show that SWMs applied to hybrid or pure DL DWMs consistently outperform those applied to process-based DWMs. However, an SWM applied to the pure DL model exhibits weaknesses at low flows for this study basin, underscoring the value of hybrid approaches. Extending the analysis across 73 additional basins demonstrates that these improvements are robust and generalizable statewide. This work highlights the potential of a DL-enhanced stochastic watershed modeling framework to advance hydrologic risk prediction under changing climate conditions, offering a scalable methodology for integrating uncertainty into watershed modeling for long-term planning.

Journal of Hydrology↗

USGS2026

Groundwater drought in the United States: Spatial and temporal variability

Many communities and ecosystems in the United States that are dependent on groundwater are potentially adversely affected by groundwater drought. We computed yearly groundwater-drought metrics and mean groundwater levels at well locations across the conterminous United States (CONUS), using data from wells and remotely sensed and modeled Gravity Recovery and Climate Experiment Drought Monitor Data Assimilation (GRACE-DADM). We also modeled the probability of low or high human impact at each well location. The spatial distribution of groundwater-drought duration and severity from 2001 to 2020 for 1,510 wells shows longer maximum duration and higher maximum severity events in drier regions like the Southwest than in wetter regions like the Northeast. Based on 613 wells in CONUS from 1981 to 2020, there are many significant decreases in drought duration and severity in the Northeast and many significant increases in annual-mean groundwater levels. In contrast, there are many significant increases in drought metrics and decreases in mean water levels in parts of the Southeast. There are major differences in trends from 2001 to 2020 between well-based and GRACE-DADM-based groundwater metrics in some CONUS regions and a very low correlation between trends at individual locations across CONUS. A potential reason for this disparity is the low GRACE-DADM resolution (∼12 km) and the potential for a large amount of groundwater variation at the local scale. Also, GRACE-DADM represents shallow, unconfined aquifers which may not match the screened interval of the monitoring wells we evaluated. Large spatial gaps in long-term, high frequency, and quality-assured groundwater-well monitoring data present a challenge for understanding groundwater-drought variability across CONUS. Remote sensing tools such as GRACE can help but cannot fully replace well monitoring, as highlighted by our study results. Substantially more long-term monitoring wells would more accurately represent groundwater-drought trends and spatial variability across CONUS, particularly in western regions.

conterminous United States↗