Enhancing predictive accuracy of daily reference evapotranspiration (ETO) in diverse climates: a comparative analysis of simplified algorithms and model performance

Abstract The simplification of algorithms for predicting natural events, particularly in the context of environmental and agricultural applications, has gained significant attention due to the need for reliable, efficient, and adaptable models. This study aims to assess the performance of various simplified algorithms for predicting daily reference evapotranspiration (ET O ) across diverse climatic conditions. A comparative analysis of multiple models, including combination-based, mass transfer-based, radiation-based, and temperature-based approaches, was conducted to evaluate their precision and adaptability in different environmental settings. Among the combination-based models, PME and ResNet₄ exhibited outstanding performance, with standardized index (SI) values consistently below 0.1 and generalized predictive index (GPI) values under 5% across all climatic conditions, making them ideal for applications in regions with diverse environmental characteristics. Other models such as LSSVR₄ and ANF-PSO₄ demonstrated moderate effectiveness in arid climates but struggled in more humid conditions, highlighting the need for further model refinement in extreme environments. The mass transfer-based models, including A-LSTM 3 and ResNet 3 , showed strong performance in very dry climates, although their precision decreased in humid regions, indicating the sensitivity of these models to changes in moisture availability. Radiation-based models such as A-LSTM 2 and ResNet 2 performed well in humid and semidry conditions, while LSSVR 2 and ANF-PSO 2 were most effective in dry climates. Temperature-based models, particularly LSSVR 1 and ANF-PSO 1 , demonstrated remarkable stability across all climates, with low GPI values, making them well-suited for temperature-sensitive environments. Overall, PME emerged as the most reliable model, offering consistent high performance across all climates. The findings of this study emphasize the importance of selecting and calibrating models based on climatic variability, ensuring accurate predictions for sustainable environmental management and agricultural planning.