Predicting postfire sediment yields at the hillslope scale: Testing RUSLE and Disturbed WEPP

High-severity wildfires can increase hillslope-scale sediment yields by several orders of magnitude. Accurate predictions of postfire sediment yields are needed to guide management decisions and assess the potential impact of soil loss on site productivity and downstream aquatic resources. The Revised Universal Soil Loss Equation (RUSLE) and Disturbed WEPP are the most commonly used models to predict postfire sediment yields at the hillslope scale, but neither model has been extensively tested against field data. The objectives of this paper are to ( 1) compare predicted sediment yields from RUSLE and Disturbed WEPP against 252 plot years of data from nine fires in the Colorado Front Range; and ( 2) suggest how each model might be improved. Predicted and measured sediment yields were poorly correlated for RUSLE (R-2 = 0.16) and only slightly better correlated for Disturbed WEPP (R-2 = 0.25). Both models tended to over-predict sediment yields when the measured values were less than 1 Mg ha(-1) yr(-1) and to under-predict higher sediment yields. Model accuracy was not improved by increasing the soil erodibility ( K) factor in RUSLE and was only slightly improved by slowing the vegetative recovery sequence in Disturbed WEPP. Both models much more accurately predicted the mean sediment yields for hillslopes grouped by fire and severity (R-2 = 0.54 to 0.66) than for individual plots. The performance of RUSLE could be improved by incorporating an erosivity threshold and a nonlinear relationship between rainfall erosivity and sediment yields. The performance of WEPP could be improved by reducing the effective hydraulic conductivity in sites that have recently burned at high severity. The results suggest that neither model can fully capture the complexity of the different controlling factors and the resultant plot-scale variability in sediment yields.