Rainfall simulation experiments were conducted on annual grassland and coastal sage scrub hillslopes to determine the quantities of C and N removed by surface runoff in sediment and solution. Undisturbed coastal sage scrub soils have very high infiltration capacities (> 140 mm h(-1)), preventing the generation of surface runoff. Trampling disturbance to the sage scrub plots dramatically reduced infiltration capacities, increasing the potential for surface runoff and associated nutrient loss. Infiltration capacities in the grassland plots (30-50 mm h(-1)) were lower than in the sage scrub plots. Loss rates of dissolved C and N in surface runoff from grasslands were 0.5 and 0.025 mg m(-2) s(-1) respectively, with organic N accounting for more than 50% of the dissolved N. Total dissolved losses with simulated rainfall were higher than losses in simulations with just surface runoff, demonstrating the importance of raindrop impact in transferring solutes into the flow. Experimental data were incorporated into a numerical model of runoff and sediment transport to estimate hillslope-scale sediment-bound nutrient losses from grasslands. According to the model results, sediment-bound nutrient losses are sensitive to the density of vegetation cover and rainfall intensity. The model estimates annual losses in surface runoff of 0.2 and 0.02 g m(-2) for sediment-hound C and N, respectively. The results of this study suggest that conversion of coastal sage scrub to annual grasslands increases hillslope nutrient losses and may affect stream water quality in the region.