The Environmental Protection Agency (EPA) requires that sources of contaminants be identified in order for total maximum daily loads (TMDL's) to be developed. One accepted method for doing so is through the use of watershed scale models. In a previous study we found that sediment concentrations from a stream in an agricultural watershed were much higher than those from a reference stream. The Soil Water Assessment Tool (SWAT) model can be used to model the watershed and then aid in determining what best management practices will be most effective for decreasing sediment concentration in the main reach. This study discusses calibrating the SWAT model for sediment concentration out of a stream located in an agricultural watershed. Once the model calibration is completed and it is validated, it can be used to simulate the incorporation of several different best management practice scenarios for the watershed. Different TMDL's can be considered to determine what sediment concentrations should be for the stream to meet its designated use. Flow and total suspended soild (TSS) data collected from the watershed since 2002 were used to calibrate the model for sediment concentration. Fifteen minute precipitation data collected from within the watershed was also used to drive the model. Finally, SSURGO soil data was used in place of STATSGO data to reduce the effect of spatial variability. Statistical tools were used to analyze the differences between observed and simulated flow and TSS leaving the pour-point of the watershed.