Improvement approaches and performance of temperature-based remote sensing evapotranspiration models in agricultural fields: A meta analysis

Accurate estimation of cropland evapotranspiration (ETc) is critical for irrigation management and efficient water resource utilization, yet remains challenging due to complex surface heterogeneity and human interventions. Land surface temperature-based remote sensing models (LST-ET) have been widely applied at field and regional scales for their strong physical foundations and computational efficiency. However, their performance in agricultural systems is often constrained by structural limitations, prompting the development of diverse improvement strategies. Despite increasing efforts to enhance LST-ET models, few studies to date have systematically synthesized these strategies or quantitatively assessed their effectiveness. To address this gap, a meta-analysis was conducted using 128 datasets extracted from 90 peer-reviewed studies. A unified classification framework of structural improvement strategies was proposed, and model performance before and after modification was evaluated using three metrics: absolute improvement, relative improvement rate, and residual potential utilization rate. Results showed that all strategies improved ETc estimation to varying degrees. Input mechanism enhancements yielded the highest gains in R2, while error reduction (RMSE%) showed no significant differences across strategies. Model accuracy was more influenced by improvement strategies than by environmental conditions. Additionally, initial model performance influenced outcomes, with R2 showing diminishing returns and RMSE% retaining notable room for optimization. Overall, current LST-ET models have yet to achieve optimal performance in cropland regions. Future improvements should aim to reduce systematic biases while enhancing regional adaptability. This work provides insights into the current landscape of LST-ET model improvements and theoretical guidance for their continued development and application in agricultural systems.