Digital soil erodibility mapping by soilscape trending and kriging

Spatial representation of soil erodibility (Universal Soil Loss Equation's [USLE] K factor) is critical for soil conservation and erosion modeling. K factor is directly linked to the soil properties, which have a spatially continuous and soilscape related variability. The objective of this study was to test a methodology to map the spatial distribution of soil erodibility in a 1,200 ha sub‐basin making use of available spatial covariates and field data. The analysis was run for the Posses sub‐basin, in southeast Brazil. The topsoil erodibility was calculated at 85 sampled locations. The spatial prediction of soil erodibility was performed using the scorpan approach, in which the trend term for kriging with external drift (KED) was modeled by soilscape covariates selected by multiple linear regression analysis. The results confirmed that relief data could produce feasible results for digital soil erodibility mapping, especially when combined with geostatistical procedures. A comparison with ordinary kriging showed better error statistics and decreased variance of the estimates for the KED model. This could affect significantly the uncertainty of further USLE applications. The best agreement between KED erodibility values and direct measurements of the K factor was observed for the Red–Yellow Argisol (Red–Yellow Ultisol), which is the dominant soil class in the sub‐basin.