Estimating Potential and Reference Evapotranspiration in the Central European Region: The Challenge of Model Selection

Abstract An accurate assessment of evapotranspiration (ET) is crucial for determining water balance, particularly in agricultural areas. However, misconceptions surrounding the definition and concept of ET contribute to significant uncertainty in its measurement and estimation. This study focuses on comparing the performance of 25 common models, including mass transfer, radiation-based, temperature-based, and combined methods. The results were compared with observed pan evaporation data from 18 stations across various climatic conditions in the temperate zone of the Czech Republic. The best results were achieved with the combined methods, which performed well in terms of both absolute values and daily patterns. The mean RMSEs between modeled and observed PET/RET were 1.2/18.6/33.3 mm day⁻¹/month⁻¹/season⁻¹, with correlation coefficients of 0.75, 0.94, and 0.75 for daily, monthly, and seasonal analyses, respectively. Individually, the radiation-based Makkink method (J Inst Water Eng 11:277–288, 1957) and the temperature-based approach by Oudin et al. (J Hydrol 303(1-4):290–306, 2005) stood out, both providing reliable results in terms of absolute PET values. The daily course was most accurately simulated by the Penman method (1963) and the WMO method (1966). Overall, model performance indicators such as RMSE and BIAS were more method-specific, whereas the correlation between modeled and observed data varied by station.