The vector physics of wind-driven rain (WDR) differs from that of wind-free rain, and the interrill soil detachment equations in the Water Erosion Prediction Project (WEPP) model were not originally developed to deal with this phenomenon. This article provides an evaluation of the performance of the interrill component of the WEPP model for WDR events. The interrill delivery rates were measured in the wind tunnel facility of the International Center for Eremology (ICE), Ghent University, Belgium with an experimental setup to study different raindrop impact velocity vectors. Synchronized wind and rain simulations with wind velocities of 6, 10 and 14 m s–1 were applied to a test surface placed on windward and leeward slopes of 7, 15 and 20%. Since both rainfall intensity and raindrop impact velocity varied greatly depending on differences in the horizontal wind velocity under WDRs, the resultant kinetic energy flux (KEr, in J m–2 s–1) was initially used in place of the WEPP model intensity term in order to incorporate the effect of wind on impact velocity and frequency of raindrops. However, our results showed only minor improvement in the model predictions. For all research data, the model Coefficients of Determination (r2) were 0·63 and 0·71, when using the WEPP and the KEr approaches, respectively. Alternately, integrating the angle of rain incidence into the model by vectorally partitioning normal kinetic energy flux (KErn, in J m–2 s–1) from the KEr greatly improved the model's ability to estimate the interrill sediment delivery rates (r2 = 0·91). This finding suggested that along with the fall trajectory of wind-driven raindrops with a given frequency, raindrop velocity and direction at the point of impact onto the soil surface provided sufficient physical information to improve WEPP sediment delivery rate predictions under WDR. Copyright © 2012 John Wiley & Sons, Ltd.