This study is to assess the reduction of non-point source pollution loads for rice straw mulching of upland crop cultivation at a watershed scale. For Byulmi-cheon watershed (1.21 $km^2$) located in the upstream of Gyeongan-cheon, the HSPF (Hydrological Simulation Program-Fortran) and SWAT (Soil and Water Assesment Tool), physically based distributed hydrological models were applied. Before evaluation, the model was calibrated and validated using 9 rainfall events. The Nash-Sutcliffe model efficiency (NSE) for streamflow using the HSPF was 0.62~0.76 and the determination coefficient ($R^2$) for water quality (sediment, total nitrogen T-N, and total phosphorus T-P) were 0.72, 0.62, and 0.63 respectively. The NSE for streamflow using the SWAT were 0.43~0.81 and the $R^2$ for water quality (sediment, T-N, and T-P) were 0.54, 0.87, and 0.64 respectively. From the field experiment of 16 rainfall events, the rice straw cover condition reduced surface runoff average 10.0 % compared to normal surface condition. By handling infiltration capacity (INFILT) in HSPF model, the value of 16.0 mm/hr was found to reduce about 10.0 % reduction of surface runoff. For this condition, the reduction effect of sediment, T-N, and T-P loads were 87.2, 28.5, and 85.1 % respectively. By handling soil hydraulic conductivity (SOL_K) in SWAT model, the value of 111.2 mm/hr was found to reduce about 10.0 point reduction of surface runoff. For this condition, the reduction effect of sediment, T-N, and T-P loads were 80.0, 83.2, and 78.7 % respectively. The rice straw surface covering was effective for removing surface runoff dependent loads such as sediment and T-P.