Machine learning-based estimation of daily ETo under limited meteorological data

Accurate estimation of reference crop evapotranspiration (ETo) is essential for sustainable irrigation management, particularly in regions facing water scarcity challenges. This study evaluates the performance of three machine learning (ML) models: Artificial Neural Networks (ANN), Light Gradient Boosting Machines (LGBM), and Random Forest Regression (RFR) for estimating daily ETo in Alanya, Turkey, under varying scenarios of limited meteorological input availability. Ten input scenarios comprising different combinations of temperature, sunshine duration, relative humidity, and wind speed were analyzed using data spanning from 1975 to 2023. Results indicate significant variability in model performance, with ANN and LGBM consistently outperforming RFR across most scenarios. Among single-variable scenarios, temperature-based predictions were the most reliable (R2=0.66). Two variable scenarios combining temperature and sunshine duration notably enhanced prediction accuracy (R2=0.85). The highest predictive accuracy was achieved with a three-variable combination of temperature, sunshine duration, and wind speed (R2=0.89). This research underscores the potential of ML models, particularly ANN and LGBM, in accurately estimating ETo with limited meteorological data, contributing significantly to sustainable water management practices in Mediterranean climates.