Contrasting Performance of Empirical Reference Evapotranspiration Models in Different Climatic Regions of China

ABSTRACT Reference evapotranspiration ( ET 0 ) is crucial in agricultural irrigation design and regional water resources allocation. However, understanding the applicability and performance of various empirical ET 0 models across diverse climatic conditions remains challenging. In this study, based on the daily meteorological data during 1994‐2016, including the measured global solar radiation ( R s ), 35 empirical ET 0 models were evaluated against the FAO‐56 Penman–Monteith formula (PM) across four climatic regions of China: temperate monsoon (I), temperate continental (II), mountain plateau (III), and subtropical monsoon (IV) regions. These models were classified into four categories: 13 temperature‐based, 15 radiation‐based, four mass transfer‐based, and three combined models. All models were pre‐corrected using the least square method to ensure comparability. The results showed that the combined models achieved the most accurate ET 0 estimates, followed by radiation‐, temperature‐, and mass transfer‐based models. Regionally, the radiation‐based FAO24 radiation formula exhibited the best performance in region I, with RMSE , MAE , and ENS of 0.245 mm d −1 , 0.210 mm d −1 , 0.975, respectively. In regions II, III, and IV, the combined Pen63 formula performed the best, with RMSE of 0.085, 0.113, and 0.095 mm d −1 , MAE of 0.066, 0.090, and 0.073 mm d −1 , and ENS of 0.998, 0.994, and 0.992, respectively. When u 2 or/and RH was unavailable, the optimal radiation‐based models were suitable for ET 0 estimation (FAO24 radiation formula for regions I and II, Priestley–Taylor formula for regions III and IV, respectively). In the absence of radiation data, the best temperature‐based models served as suitable alternatives to the PM formula (Baier–Roberston formula for regions I and III, McGuinness‐Bordne (2005) formula for region II, and Hargreaves correction formula for region IV, respectively). Therefore, the suitable empirical ET 0 models were recommended based on the data availability, providing flexible choices for ET 0 calculation across diverse climatic regions.