Abstract
Study region: Seneca River Basin South Carolina, USA.
Study focus: This study focuses on evaluating different precipitation products on streamflow and water quality modeling. We used 13 years (2001-2014) of daily data from multiple precipitation products. The precipitation products are classified into three categories: (1) products based on observed precipitation data (rain gauges located within the study area and PRISM data), (2) five satellite-based precipitation products (3B42, 3B42RT, PERSIANN-CDR, PERSIANN, and GPCP-1DD); and (3) one Reanalysis product (CFSR). The Soil and Water Assessment Tool (SWAT) was calibrated and validated using these precipitation products against monthly observed streamflow, total nitrogen (TN) loads, and total phosphorus (TP) loads within the Seneca River Basin situated in northwest of South Carolina, USA. For each precipitation product, the best performing model parameters was determined using the Sequential Uncertainty Fitting Procedure (SUFP2).
New hydrological insights for the region: Results showed that the spatio-temporal pattern of precipitation products varies and further influenced the streamflow and water quality simulations. Most of the precipitation products achieved at least satisfactory performance (R-2 > 0.50, NS > 0.50, and PBIAS >= +/- 15) for streamflow and water quality. PRISM product showed the best overall performance, followed by 3B42 and PERSIANN-CDR, whereas CFSR and PERSIANN products comparatively performed poorly for streamflow and water quality simulations. The uncertainty in the precipitation products was reflected in the streamflow that further affected the water quality simulations (TN and TP loads). In addition, this study showed that the parameter uncertainty significantly varies among precipitation products, leading to a considerable uncertainty in simulation results, especially for TN and TP. Our results suggest that using multiple precipitation products can increase the level of confidence in hydrologic model simulation, particularly in water quality modeling.