Considering parameter seasonal variation to enhance process-based ecosystem model performance, evidence from the SWH model

• Empirical parameters in the SWH models effectively characterize evapotranspiration dynamics. • Introducing temporal dynamics into parameter calibration scheme enhanced model predictive accuracy. • Improved Monte Carlo-based calibration scheme significantly enhanced model performance. Modeling and partitioning ecosystem evapotranspiration (ET) are critical for predicting how ecosystem water cycles respond to global climate change. In this study, we used the widely applied SWH model, an ET partitioning model, as an example. We demonstrated that certain empirical calibrated parameters (e.g., scaling coefficients used in the calculation of surface soil resistance and canopy stomatal resistance) can effectively characterize ET states, as evidenced by global FLUXNET data. These parameters showed distinct discrepancy across ecosystem types, and exhibited strong correlations with climatic factors such as precipitation, soil moisture and temperature. However, existing studies often assume these parameters to be temporally constant, which limits the model’s ability to capture seasonal ET dynamics. To address this, we developed a series of parameter calibration schemes by adjusting parameters within adaptive time windows (L), which allow parameters to vary temporally. Results showed that for ecosystems with pronounced seasonal dynamics, such as savannas and temperate forests, the length of L aligns well with the specific seasonal characteristics. The improved Monte Carlo-based calibration scheme achieved a 95% success rate, with R 2 significantly higher than traditional strategy (10–30% improvement). At 20 long-term observation sites, its performance approached that of the Extended Kalman Filtering (EKF), which yielded R 2 values between 0.83 and 0.98.​ This calibration scheme is sound in theory and easy to apply in practice, and would enhance the capability of similar models in simulating ET dynamics. Future studies should pay more attention to the temporal variability of parameters to more accurately capture specific ecosystem indicators and enhance predictive capacity.