In recent years, the rate of soil erosion exceeds the rate of soil formation over wide areas of Korea, resulting in the depletion of soil resources and productive potential. This disparity between erosion and soil-formation rates usually is the result of human activities. Especially, as haphazard land development and unauthorized farming increase in mountainous regions of Korea, soil depletion proceeds at faster rates and over wider areas. According to the National Soil Erosion Research report in Korea (Ministry of Environment, 2012), more than 30% of the land in Korea lead to an annual soil loss of 33 ton/ha, which corresponds to the ‘severe’ level of OECD soil erosion standards. In addition, about 20% of the land suffers from a serious soil loss problem exceeding 50 ton/hec·year. Computer models are good alternatives for soil erosion simulation at catchment scale; they take into account many of the complex interactions of soil erosion mechanisms in time and space. Commonly used soil erosion models developed in the last decades tend to shift in their methodology from empirical or conceptual approaches such as USLE family models in the 1970s to physics-based approaches in the present. In this study, we developed the process-oriented soil erosion model which consists of two basic modules: rainfall-runoff module and erosion-sediment yield module and then applied the model for soil erosion simulation due to typhoon events. Moreover, we assessed model uncertainties by coupling the model with particle-filtering method.