Influence of Precipitation Sources on Pareto-Optimal Parameter Sets and Hydrological Partitioning in SWAT+

Accurate precipitation inputs are essential for watershed hydrological modeling, yet different precipitation datasets may produce similar streamflow performance through different calibrated parameter combinations and hydrological process representations. This study evaluated the effects of gauge-based and Taiwan Climate Change Projection Information and Adaptation Knowledge Platform (TCCIP) gridded precipitation on SWAT+ simulations for two mountainous watersheds in southern Taiwan using the NSGA-II multi-objective optimization algorithm for calibration. Differences in Pareto-optimal parameter distributions, parameter convergence, and hydrological partitioning were compared between precipitation inputs. Although simulations driven by both precipitation datasets achieved comparable streamflow performance, they produced distinct calibrated parameter sets and, therefore, gauge-based and gridded precipitation resulted in substantially different hydrological partitioning. TCCIP gridded precipitation was approximately 11% and 22% lower than gauge-based precipitation in the ZW and LN watersheds, respectively. Correspondingly, the runoff coefficient decreased from 67.98% to 64.13% in ZW and from 62.13% to 53.18% in LN, while the evapotranspiration-to-precipitation ratio increased from 32.38% to 35.80% and from 37.72% to 45.06%, respectively. The findings indicate that calibrated parameter sets are precipitation-source dependent and should not be transferred directly between different precipitation datasets. Instead, independent model calibration is recommended whenever precipitation inputs are changed. More broadly, this study highlights the importance of complementing conventional streamflow-based evaluation with process-based diagnostics to improve the physical reliability of watershed simulations and water resource assessments.