PEMIP: Post-fire erosion model inter-comparison project.

Land managers often need to predict watershed-scale erosion rates after disturbance or other land cover changes. This study compared commonly used hillslope erosion models to simulate post-fire sediment yields (SY) at both hillslope and watershed scales within the High Park Fire, Colorado, U.S.A. At hillslope scale, simulated SY from four models— RUSLE, AGWA/KINEROS2, WEPP, and a site-specific regression model—were compared to observed SY at 29 hillslopes. At the watershed scale, RUSLE, AGWA/KINEROS2, and WEPP were applied to simulate spatial patterns of SY for two 14–16 km2 watersheds using different scales (0.5–25 ha) of hillslope discretization. Simulated spatial patterns were compared between models and to densities of channel heads across the watersheds. Three additional erosion algorithms were implemented within a land surface model to evaluate effects of parameter uncertainty. At the hillslope scale, SY was only significantly correlated to observed SY for the empirical model, but at the watershed scale, sediment loads were significantly correlated to observed channel head densities for all models. Watershed sediment load increased with the size of the hillslope sub-units due to the nonlinear effects of hillslope length on simulated erosion. SY's were closest in magnitude to expected watershed-scale SY when models were divided into the smallest hillslopes. These findings demonstrate that current erosion models are fairly consistent at identifying areas with low and high erosion potential, but the wide range of predicted SY and poor fit to observed SY highlight the need for better field observations and model calibration to obtain more accurate simulations. • Evaluated performance of erosion models at hillslope and watershed scales. • Simulated hillslope sediment yields did not correlate with measured values. • Simulated watershed sediment yields were correlated with observed rill densities. • Longer hillslopes within watershed simulations led to greater sediment loads. • Parameter uncertainty caused >2 orders of magnitude variability in sediment yields. [ABSTRACT FROM AUTHOR]