Technical Abstract: Natural gas exploration and production, with nearly 30,000 new wells drilled each year in the US, requires land disturbing activities that can accelerate soil loss. Erosion modeling has been successfully used for decades to predict soil loss and conservation effects on agricultural fields, rangeland, and forests, but little research has been conducted on applying erosion models to construction sites. The objective of this research was to evaluate Water Erosion Prediction Project (WEPP) runoff and sediment yield predictions relative to measured data. Model parameters were adjusted from WEPP defaults based on available literature and model observations. A low effective hydraulic conductivity value resulted in successful runoff predictions. Similarly, increasing default rill and interrill erodibility and decreasing critical shear stress resulted in successful sediment yield prediction. WEPP performance was evaluated with the Nash-Sutcliffe efficiency (NSE), root mean square error (RMSE)-observation standard deviation ratio (RSR), and percent bias (PBIAS). For one site, NSE and RSR indicated “good” performance for runoff and “satisfactory” performance for sediment yield. For the second site, NSE and RSE values were “very good” for runoff but “unsatisfactory” for sediment yield. Use of modified NSE and RSR to consider measurement uncertainty improved model performance to “very good” for all instances. PBIAS values indicated “very good” prediction of runoff and sediment yield from both sites. Overall, these results indicate that WEPP can effectively model runoff and sediment yields from natural gas well sites, which makes it useful for evaluating management alternatives to minimize sediment loss from such sites.