Adequate and reliable erosion prediction tools are needed for sound forest resources management. Numerous watershed models have been developed during the past. These models, however, are often limited in their applications largely due to their inappropriate representations of the hydrological processes involved. The Water Erosion Prediction Project (WEPP) model has demonstrated its usefulness in certain forest applications, such as modeling erosion from a segment of in-sloped or out-sloped road, and harvested or burned units. Nevertheless, when used for modeling water flow and sediment discharge from a forest watershed of complex topography and channel systems, WEPP consistently underestimates these quantities, in particular, the water flow at the watershed outlet. The goal of this study was to improve the WEPP model such that it can be applied to adequately simulate forest watershed hydrology and erosion. Specific objectives were to: (1) identify and correct WEPP algorithms that inappropriately represent forest hydrologic processes; and (2) verify the modified model. Substantial changes were made in the approach to, and algorithms for modeling percolation of soil water and subsurface lateral flow in WEPP. The modified codes were subsequently applied to Hermada watershed, a small watershed located in the Boise National Forest in northern Idaho. The modeling results were compared with those obtained by using the original WEPP and the field-observed runoff and erosion data. Conclusions of this study include: (1) compared to the original model, the modified WEPP more realistically and properly represents the hydrologic processes in a forest setting; and (2) application of the modified model produced satisfactory results, demonstrating the adequacy of the model modifications.