Evapotranspiration Estimation Using Multimodel Ensemble

Evapotranspiration (ET) plays a crucial role in the Earth's hydrological cycle, influencing water resource management, agricultural productivity, and ecosystem dynamics. Accurate estimation of ET is essential for sustainable water resource planning and management. Multiple models based on different physical principles were developed. Past studies have reported varying accuracies, with performance of the models differing across sites, climate, biomes etc. and a single model could not be identified best under all conditions. This also limits the use of ET products in various applications. In recent years, ensemble modeling techniques have emerged as a promising approach to enhance the accuracy of ET predictions by combining the strengths of multiple individual models. Studies indicate that the ET estimates are improved even when basic methods of mean or weighted average techniques are used for the ensemble. Integrating machine learning techniques into ensemble framework utilizes the physical nature of model as well as the predictive power of machine learning algorithms. Against this backdrop, this study aims to develop an ensemble ET over Indian region. The primary objective of the study is to improve the accuracy and reliability of ET predictions by utilizing the strengths of multiple modeling approaches. Three popular ET models for creating ensemble models namely: Priestley Taylor – Jet Propulsion Lab (PT-JPL), Soil Plant Atmosphere and Remote Sensing Evapotranspiration (SPARSE – Layer and Patch), and Surface Temperature Initiated Closure (STIC) belonging to different categories of ET models are used to create the ensemble. The study tests the ensemble based on mean, Bayesian Model Averaging, and k-Nearest Neighbor. The preliminary results show improvement in ET estimates using the ensemble models compared to that of individual models and thus the research holds the potential to contribute significantly to more accurate ET assessments.

Key words: Evapotranspiration, ensemble, PT-JPL, SPARSE, STIC