A Benchmark Dataset of Agricultural Weather Stations over the Contiguous United States for Evapotranspiration Applications

Christian Dunkerly, John Volk, Sayantan Majumdar , Justin Huntington, Richard G. Allen, Christopher Pearson, Yeonuk Kim , Charles Morton, Blake A. Minor, Peter ReVelle, Ayse Kilic, Forrest Melton, Adam J. Purdy, Todd G. Caldwell

Agricultural weather data are fundamental for the accurate estimate of evapotranspiration (ET), irrigation scheduling, and water-use accounting. In particular, reference ET provides a standardized atmospheric demand for water loss from a hypothetical well-watered grass (ETo) or alfalfa (ETr); however, weather stations may not adequately represent such climatic conditions. Weather data commonly contain errors from poor siting, sensor drift, and network management deficiencies. No standardized dataset exists over the contiguous United States (CONUS). Systematic errors affect ETo/ETr calculations and derived products. Notably, satellite-based platforms like OpenET require agricultural weather data to bias correct gridded reference ET to interpolate between satellite overpasses. CONUS-AgWeather is a benchmark dataset of daily agricultural weather data (precipitation, solar radiation, air temperature, humidity, wind speed, ETo, ETr) from 793 stations. This dataset contains 4,191,808 days (11,484 station-years, 1981-2020) and was produced through standardized and systematic quality control procedures and open-source software packages for time series inspection, outlier detection, corrections, and ETo/ETr calculations. CONUS-AgWeather is intended primarily to support OpenET in the Western U.S. but has broader applications.