Riparian buffer zones in agriculture dominated watersheds play important roles in reducing nonpoint source pollution into aquatic ecosystems and are widely used as a Best Management Practice. Assessment of the effectiveness of riparian buffer zones by modeling method is widely used for watershed management as field measurement-based assessment is difficult and expensive.The integration of Riparian Ecosystem Management Model (REMM) and Soil and Water Assessment Tool (SWAT) has been developed to simulate the effect of nonpoint source pollution reduction by riparian buffer zones at subbasin scale. However, there are problems in using the integrated model at subbasin scale, as the size of subbasin partition could affect the pollutant reduction rate by riparian buffers. In this study, we partitioned a large watershed with size of 1331 ha into sub-watersheds with sizes of 666, 333, 166, 83, 51, and 29 ha, and then compared the different simulation results. We found that the modeling could yield more convergent results when the sub-watersheds were partitioned into suitable size. In the studied area, the suitable sub-watershed size was less than about 166 ha for runoff and nitrogen and 83 ha for sediment and phosphorus. Among the eight sub-watersheds (partitioned based on the size of 166 ha), results showed that the effects of riparian buffers on runoff and nutrient loading varied drastically. The reduction rate varied from 0.26% to 30.13% for runoff, 29.4% to 74.07% for sediment, 9.61% to 57.85% for nitrogen, and 18.61% to 68.12% for phosphorus, respectively.