The objectives of this paper were to estimate cell based pollutant loadings for total maximum daily load (TMDL) programs and to evaluate the applicability of the agricultural nonpoint source (AGNPS) model for an intensive agricultural watershed in Korea. The model was calibrated and validated at a watershed of 384.8 ha of drainage area using the observed data from 1996 through 2000 in terms of runoff, suspended solid, total nitrogen, and total phosphorus on a hourly basis. Analysis of spatial variations of pollutant loadings for rainfall frequencies of various intensities and durations were conducted. In addition, the validated model was applied to estimated the TMDL removal efficiency for best management practices (BMPs) scenarios which were selected by taking into account the pollutant characteristics of the study watershed. The model can help to understand the problems and to find solutions through landuse changes and BMPs. Thus, the method used for this study was able to identify TMDL quantitatively as well as qualitatively for various sources pollution that are spatially dispersed. Also it provides an assessment of the impact of BMPs on the water bodies studied, allowing the TMDL programs to be complemented more effectively.