Sediment runoff from areas of disturbed soil, such as those common on construction sites, is an issue of great concern both from an environmental impact standpoint and as a regulatory compliance issue for construction site mangers. In order to address these concerns and to help develop appropriate mitigating strategies, it is first necessary to be able to quantify the expected sediment runoff for a given site. To this end, this study will build on work performed by the Soil Erosion Research Laboratory (SERL) of San Diego State University in order to assess the effectiveness of using a physically based model of sediment transport to accurately predict sediment discharge due to erosion. While there are empirical models such as the Revised Universal Soil Loss Equation (RUSLE) that are widely used, such approaches remain limited in that they draw strictly from observed phenomenon. While extensive research has been conducted, the charts and tables associated with RUSLE are not all inclusive and cannot, as the name would imply, be “universally” applied. By using a physically based model it is hoped that as long as the physical characteristics of given soil/site are known, a single model would be sufficient to make accurate predictions of sediment runoff. Using the KINEROS2 software developed by the U.S. Department of Agriculture, it has been possible to run numerous iterations of a simulated rain event on a small watershed based on the kinematic wave theory of sediment transport. Simulations were conducted on an 8 meter by 2 meter (L x W) plane at three separate grades and were intended to mimic the conditions of experiments performed by SERL. By comparing the sediment discharge of the KINEROS2 simulations to observed values, an optimal conceptual geometry and appropriate erosion coefficients were determined for conditions that are likely to be found on many construction sites. The results from the simulations show that the kinematic wave model is indeed capable of reproducing sediment discharge rates within the observed ranges. Further, it was determined that the simplest configuration, that of a single plane with no channels, produced results most consistent with the average values obtained by SERL.