Spatial uncertainty of soil erodibility using joint stochastic simulation: application to a basin of south Italy

Soil erosion varies greally over space and neglecting information about estimation uncertainty may lead to inappropriate decision-making. One geostatistical approach to spatial analysis is joint stochastic simulation. Differences between the realilizations provide a measure of spatial uncertainty and allow to carry out an error propagation analysis. The objective of this study was lo assess the model output error of soil erodibility resulting from the uncertainties in the Input parameters. The study area (about 1.4 km2) was a forest watershed (South Italy). From superficial soil horizon 110 soil samples were collected and analyzed for soil texture and organic matter content. A Monle Carlo analysis was performed, which consisted of drawing a large number of identical1y distributed inputparameters from the multivariable joint probability distribution function. Spatial cross-correlation information was incorporated through joint turning bands simulation. A linear coregionalization model was fitted to all simple and cross- variograms of the input variables including a nugget effect and spherical structure with a range of 129 m. The erodibility model was estimated for each set of the 500 joint realizations of the input variables and the ensemb1e of the model outputs was used to infer the erodibility probability distribution function. thus providing a confidence interval at each simulated point. This approach has also allowed to delineate the areas characterized by greater uncertainty and then to suggest efficient strategies for funther improving the accuracy of soil erodibility value predictions, in the contest to identify potential erosion risk area with accuracy and precision.