Evaluating combinations of rainfall datasets and optimization techniques for improved hydrological predictions using the SWAT+ model

Study Region: This study focuses on the Cape Fear and Tar-Pamlico watersheds in North Carolina, which are characterized by diverse hydrological conditions, varied land use, soil types, and hydrological characteristics. Study Focus: The primary goal of this study is to examine the combined effects of three satellite precipitation products (SPPs) — ERA-5, gridMET, and GPM IMERG — along with three autocalibration techniques — DDS, GLUE, and LHS — on SWAT+ river flow predictions. Flow accuracy was assessed using three evaluation metrics: NSE, KGE, and R². New Hydrological Insights for the Region: Key findings revealed that five SWAT+ parameters (cn2, revap_co, flo_min, revap_min, and awc) were consistently sensitive across all SPPs and watersheds, with rainfall products exerting a greater influence on simulated river flow than optimization techniques. Among the SPPs, GPM IMERG performed the best, followed by ERA-5 and gridMET, while NSE was more responsive to changes in SPPs and calibration methods than KGE and R². For the Cape Fear and Tar-Pamlico watersheds, the study highlighted SWAT+ 's challenges in predicting base flow for groundwater-driven systems and demonstrated the potential of optimization techniques to improve flow simulations despite poor satellite-gauge rainfall correlation. The combination of the GPM IMERG dataset and the GLUE method proved most effective, offering valuable guidance for selecting optimal datasets and methods to enhance prediction accuracy in complex watersheds.