Improving SWAT model performance and characterization of water cycle components through dual-objective spatially distributed calibration of streamflow and evapotranspiration

Study region Shaxi River Basin of China. Study focus Accurate simulation of water cycle components is critical for understanding basin hydrological processes and supporting water resources management. This study evaluates the performance of the SWAT model under two calibration strategies: dual-objective lumped calibration of streamflow and evapotranspiration (ET) (M1) and dual-objective spatially distributed calibration (M2). Model calibration and validation were conducted using observed outlet streamflow and spatially distributed ET data, and terrestrial water storage change (TWSC) was estimated based on water balance analysis. New hydrological insights for the region Both calibration strategies reproduce the temporal dynamics of streamflow and ET, while M2 demonstrates consistently superior performance. For streamflow, M2 improves NSE from 0.73 to 0.94 during calibration and from 0.51 to 0.89 during validation, with PBIAS reduced from 15.04–17.45% to 3.79–5.10%. For ET, PBIAS decreases from −3.31% to 0.69% during calibration and from −7.98% to −2.79% during validation. Spatially, M2 enhances sub-basin ET simulation, with NSE generally increasing and systematic biases reduced across most sub-basins. ET-related parameters exhibit physically consistent relationships with soil types and vegetation cover, indicating improved representation of underlying surface heterogeneity. Long-term simulations reveal pronounced spatial gradients, with precipitation and runoff decreasing from northwest to southeast, ET increasing from west to east, and TWSC showing relatively limited spatial variability. Overall, spatially distributed dual-objective calibration improves both the accuracy and spatial consistency of water cycle simulations, providing a robust framework for hydrological modeling in heterogeneous basins.