Estimating cropland evapotranspiration based on remote sensing models: A global meta-analysis

In recent decades, numerous studies have utilized remote sensing (RS)-based models to estimate actual evapotranspiration of crops and cropland (ETc), considering various influencing factors. However, these findings have not been systematically integrated, and no quantitative assessment has been conducted on the factors influencing model accuracy, nor has there been a thorough evaluation of the specific applicability of each model. Therefore, this study conducted a global meta-analysis based on an analysis of 690 published articles, focusing on exploring the differences in ETc estimation performance among various RS models. The results showed that (1) the use of RS models to estimate ETc has increased rapidly in recent years; nevertheless, the accuracy of many models did not exhibit a simple upward trend over time; (2) most studies (62 %) reported an increasing trend in ETc over time due to changes in precipitation and temperature; (3) environmental factors significantly affected the performance of single-source and two-source models. PM and PT models exhibited high accuracy under homogeneous crop cover, while empirical and semi-empirical models demonstrated the greatest stability though they remained data-dependent; (4) soil moisture input had a significant positive impact on model accuracy, especially in irrigated semi-arid areas. Future research directions in ETc were also highlighted and discussed. This study will help to understand the current states of ETc and provide important guidance for the selection of RS-based ETc models and water resource allocation, which can better serve efficient water use and precision agriculture.