Title:Extrapolability Improvement of Machine Learning-Based Evapotranspiration Models via Domain-Adversarial Neural Networks

Abstract:Machine learning-based hydrological prediction models, despite their high accuracy, face limitations in extrapolation capabilities when applied globally due to uneven data distribution. This study integrates Domain-Adversarial Neural Networks (DANN) to improve the geographical adaptability of evapotranspiration (ET) models. By employing DANN, we aim to mitigate distributional discrepancies between different sites, significantly enhancing the model's extrapolation capabilities. Our results show that DANN improves ET prediction accuracy with an average increase in the Kling-Gupta Efficiency (KGE) of 0.2 to 0.3 compared to the traditional Leave-One-Out (LOO) method. DANN is particularly effective for isolated sites and transition zones between biomes, reducing data distribution discrepancies and avoiding low-accuracy predictions. By leveraging information from data-rich areas, DANN enhances the reliability of global-scale ET products, especially in ungauged regions. This study highlights the potential of domain adaptation techniques to improve the extrapolation and generalization capabilities of machine learning models in hydrological studies.