Spatio‐temporal calibration of Hargreaves‐Samani model to estimate reference evapotranspiration across U.S. High Plains

Abstract Temperature‐based grass‐reference evapotranspiration (ET o ) estimation methods (e.g., Hargreaves−Samani [HS] model) present advantages over combination‐based methods that require full‐suite weather data. The U.S. High Plains region has scarce and short‐term full‐suite weather sites. This data scarcity presents challenges for combination‐based ET o estimation. The performance of HS model against the American Society of Civil Engineers (ASCE) standardized Penman−Monteith (PM) model was assessed using long‐term data at 124 full‐suite weather sites across nine states in the U.S. High Plains. The HS model underestimated ET o at arid (mean bias error [MBE] = −1.68 mm d −1 ), semi‐arid (MBE = −0.34 mm d −1 ), and dry subhumid sites (MBE = −0.16 mm d −1 ) and overestimated ET o at humid sites (MBE = 0.14 mm d −1 ). There was a significant relationship ( p  < .01) between HS model performance and aridity index. The HS model performed better (27% lower root mean squared difference [RMSD]) in summer months than the rest of the year at semi‐arid and dry subhumid sites. The model performance was non‐ideal during the summer months in subhumid climates. Spatio‐temporal annual zonal (climate division), monthly zonal, annual site‐specific, and monthly site‐specific calibration resulted in 12, 16, 20, and 26% reduction in RMSD and 11, 16, 17, and 23% reduction in relative error, respectively. Monthly site‐specific calibration performed the best and was used to quantify annual and growing season ET o across the region. The research characterized performance patterns of the HS model over an important agroecosystem‐dominated region. Practical data‐driven strategies were proposed to better estimate PM ET o using limited weather data at any given site (with similar aridity) and time of the year.