Modeling land management effects on the size distribution of eroded sediment

Abstract

Land management that improves soil quality is known to increase the amount of water-stable soil aggregates, but many current soil erosion models do not consider land management effects when estimating the size distribution of eroded sediment. Our objective was to develop a method to improve eroded sediment size estimates in response to alternative land management practices using information available within the RUSLE2 model and its database. We conducted a literature search identifying 274 soil aggregate size distributions where soil clay and soil organic carbon were also reported. The median value and log-normal standard deviation were estimated for each distribution. Median value was found to be linearly related to soil clay fraction and quadratically related to soil carbon content. However, the 10% of samples with observed median aggregate sizes greater than 3 mm were all underestimated by the regression model. All these samples were collected from soils under land use described as grass, pasture, or forest. To then determine whether incorporating management impacts on size distribution could better fit field data than this simple approach based purely on clay and carbon, algorithms termed MAS (Model of Aggregated Sediment) were derived and applied with default values in the RUSLE2 soil erosion model to estimate a coarser sediment size distribution at the point of detachment for land management with increased biomass inputs and reduced tillage disturbances. The MAS estimates of macroaggregated sediment (>0.212 mm) mirrored fragment size observations made over 10 yr when cropland was converted to prairie in the central U.S.A. Thus, MAS estimated dynamic land management effects on aggregation better than regression relationships based solely on soil clay and carbon contents. More aggregate and sediment size datasets that include historical land use information will be needed to fully assess the generality of the MAS algorithms and parameter estimates.