Riparian zone, as the interlaced zone between land and water, plays an important role in society, landscape and environmental quality. Riparian ecosystems have critical impacts on controlling the non-point source pollution (NPSP) and maintaining the health of aquatic ecosystems, especially on nitrogen (N) removal. The processes that affect N removal in riparian ecological system mainly include soil nitrous gas emission, plant uptake and sediment retention, of which nitrous gas release by soil denitrification is one of the most important functions for riparian system. Therefore, it's critical to build an N removal model including soil denitrification, nitrification and ammonium volatilization to evaluate the riparian ecological function and management practice. In this study, the Soil and Water Assessment Tool (SWAT) was extended with algorithms from a simple soil denitrification model and remote sensing data to enhance the model performance with regard to predicting soil N removal in the Guanting reservoir riparian catchment. The N removal model is based on chemical and physical relationships that govern soil heat, moisture and nitrogen movement. Processes considered include denitrification, nitrification and ammonia (NH3) volatilization. SPOT-5 and Landsat5-TM satellite data were used to interpret the spatial land surface information and derive model parameters. Results of laboratory-scale anaerobic incubation experiment were applied to estimate the soil denitrification model parameters for the different soil types. In an in situ field-scale experiment conducted to calibrate and validate models and an indirect method was used to test simulated N removal load in the Guanting reservoir riparian catchment. Results showed that the process-based model performed well and produced sound simulation results for the riparian reservoir catchment, with the coefficient of determination (R2) between the simulated and observed values being 0.71.