Groundwater is the primary water source for irrigated agriculture in many arid regions. Consequently, groundwater declines in cultivated drylands can threaten the sustainability of irrigation. However, little is known about why groundwater levels decline rapidly in some areas and more slowly in others. Here, we develop an accessible and reproducible workflow that integrates open-source, high-resolution data for groundwater-level records, evapotranspiration, and precipitation data to explore the relationship between agriculture and groundwater decline. We show that groundwater declines tend to be most rapid in areas where evapotranspiration exceeds precipitation. Our finding holds across multiple agricultural regions, spanning from southern California to eastern Arkansas. Our results suggest that regions where evapotranspiration exceeds precipitation are at elevated risk of groundwater depletion, and could be good areas to intensify future monitoring efforts. Altogether, our analyses demonstrate how climate and satellite data can be useful predictors of groundwater decline in cultivated drylands, and may provide a promising proxy for groundwater withdrawals where well metering data is absent.

Column Descriptions:
- year_value: The calendar year of the groundwater level measurement
- site_id: Unique identifier for each monitoring well site
- depth_to_gw_ft: Depth to groundwater surface below land surface, in feet
- depth_to_gw_m: Depth to groundwater surface below land surface, in meters
- latitude: Geographic latitude of the well site (decimal degrees, WGS84)
- longitude: Geographic longitude of the well site (decimal degrees, WGS84)
- data_source: Origin of the groundwater record (e.g., USGS or Jasechko)
- region: Regional area associated with the well site

Background:

Study sites were selected through a combination of data-driven filtering and expert-guided qualitative assessment. Of the 50 total monitoring sites included in this study, 8 were provided directly by professor of Hydrology at University of California, Santa Barbara, Dr. Scott Jasechko, who identified well sites known to be located in cultivated, irrigated areas within arid regions. The remaining 42 sites were drawn from the USGS data publication "Long-term monotonic trends in aquifer and regional groundwater levels in the United States, 1912–2020," which contains Mann-Kendall trend analyses for over 53,000 wells across the conterminous United States.
From this dataset, wells were filtered to retain only those with consistent annual groundwater level records spanning 2000 to 2020, ensuring compatibility with our ET data sources. Dr. Jasechko directed our search toward several regions across the United States known to rely heavily on groundwater for agricultural production, including the Biscayne Aquifer, the Mississippi Embayment in Arkansas, the Columbia River Basin in Washington State, the Harney Basin in Oregon, the Snake River Plain, the High Plains Aquifer in west Texas and eastern New Mexico, and western Kansas south of Garden City.
Within each of these regions, candidate wells from the filtered USGS dataset were examined visually in ArcGIS using satellite imagery. Sites were selected based on the presence of center-pivot irrigation infrastructure in the surrounding landscape, as identified through visual inspection. We targeted approximately 50 sites to provide sufficient data for analysis across the study period. It should be noted that this selection process was qualitative and expert-guided in nature, rather than the product of a fully systematic or reproducible protocol. As such, the final set of sites reflects a practical effort to capture groundwater dynamics in irrigated, arid settings, while acknowledging the inherent subjectivity of the visual selection approach. Region designation is for visualization purposes only.