A long-term aerobic incubation and leaching technique was used to measure N mineralization of surface and subsurface soils (sandy loam) from agricultural,forest, and fallow sites in a Virginia Coastal Plain watershed. N mineralization potential was measured to refine models used to describe this process in a watershed-scale nutrient export assessment. Potentially mineralizable N (N-0) and reaction rate constants (k) were estimated using a first-order model and a nonlinear regression procedure. Large variations in cumulative N mineralized, N-0, and k, were found for the surface soils from agricultural areas. Forest soils had much higher potentially mineralizable N than agricultural soils. For subsurface soils, the differences among land uses were less variable than those observed for the surface soils. The first order model (single-pool approach) was adequate for predicting N mineralization in surface soils from agricultural and fallow areas, but less suitable for forest surface soils. Consideration of a double exponential (two-pool approach) model did not improve the performance of N mineralization prediction for forested or agricultural soils. Large variations occurred in the field-predicted values of mineralized N due to temperature and moisture ranges commonly occurring throughout the season. Variability in the N mineralization potential of soils in the watershed suggests that individual k and N-0 should be derived for soils with similar properties to obtain better predictions of N mineralization and thus N movement to groundwater.