Abstract: Evapotranspiration plays a crucial role in regulating the global water and energy cycles. How to accurately simulate evapotranspiration in terrestrial ecosystems is of great significance to the study of global ecological and hydrological processes. In this study, 1 774 watershed hydrological sample data from around the world were used to evaluate the adaptability of the typical Budyko framework evapotranspiration estimation model namely Fu′s model, across different vegetation coverage areas and different climate zones. The correlation between vegetation index and characteristic parameters were used to improve the Fu′s model. The results show that: ① The evapotranspiration calculated by Fu′s model using the empirical values of the basin characteristic parameters at most stations is lower than the actual evapotranspiration calculated by the water balance component. The model tends to underestimate the regional evapotranspiration. ② Differences in climate and vegetation types significantly affect the simulation results of Fu′s model. The empirical values of the model parameters considerably overestimate evapotranspiration in non-humid areas (P/E₀ < 1), while considerably underestimate it in humid areas. ③ By introducing the relationship between vegetation index and watershed characteristic parameters to optimize the watershed characteristic parameters, the simulation accuracy of watershed evapotranspiration can be significantly improved. For the optimized model, the accuracy of grassland and farmland in temperate zones increased by 21.68 mm; the accuracy of shrubland in temperate zones and arid zones increased by 67.50 mm and 2.57 mm respectively; the accuracy of forestland in arid zones and boreal high latitude zones increased by 10.54 mm and 10.42 mm respectively. These findings contribute to improving the accuracy of global hydrological simulations and provide valuable support for addressing climate change and enhancing water resource management.