Abstract: Predicting storm peak flow in small watersheds is critical to model soil erosion in the LoessPlateau. In this study, we simulated the relationships between storm peak discharge and the parameters which represent precipitation, watershed geometry, and runoff so as to sort out the dominant factorsaffecting storm peak flow in the Loess Plateau. We collected precipitation and runoff data from 65 small watersheds in the Loess Plateau and built a database which included 2 441 storms. The results showed that peak flow was positively correlated with the product of precipitation volume and precipitation intensity, watershed area, watershed length, watershed width, runoff volume, runoff depth, and runoff coefficient, but negatively correlated with the slope of the main channel. The impacts of these parameters of precipitation, watershed geometry and runoff on peak flow were divergent at different subzones in this plateau as well as at different watershed scales. Our results are helpful for the modeling of the storm peak flow with empirical equations in the Loess Plateau.