Comparison and application of remote sensing evapotranspiration retrieval models for the Inner Mongolian grassland region

Grasslands provide essential ecosystem functions, but their water and carbon cycles are highly sensitive to climate change and environmental disturbances. Quantifying grassland water consumption and water use efficiency ( WUE ) is critical for sustainable water management and conservation. This study, conducted in arid and semi-arid Inner Mongolia, northern China, used 20 site-years of in situ evapotranspiration ( ET ) observations to optimize ET parameters and evaluate remote-sensing ET models. The Shuttleworth-Wallace-Hu (SWH), Penman-Monteith (PM), and Surface Energy Balance Algorithm for Land (SEBAL) models were evaluated using the Bayesian analysis. The best-performing model was used to estimate water use in Inner Mongolian Grassland region in 2003–2018. The dynamics of ET , evaporation( E ), transpiration( T ), transpiration fraction( T/ET ), and WUE were also analyzed. Results suggested that SWH outperformed PM and SEBAL in accuracy and uncertainty, and was therefore recommended as the preferred model. For 2003–2018, spatial patterns of ET , T , T/ET , and WUE exhibited high values in eastern than western regions, reflecting regional gradients in vegetation conditions and water availability. While Xilingol grassland had the highest ET and T , western desert grassland had higher E and lower T , T/ET , and WUE . Unimodal seasonal cycles were observed for ET and its related factors, with the most water consumption occurring during the growing season. For 2003–2018, increases in ET , E , and T was weak and insignificant. WUE increased significantly, mainly associated with enhanced gross primary productivity and transpiration ratio rather than changes in T/ET. These findings provide a basis for understanding grassland responses to climate change and supporting sustainable water management and ecological conservation.