Assessment of calanchi and rill-interrill erosion susceptibility in northern Liguria, Italy: A case study using a probabilistic modelling framework

Soils are a vital part of the natural environment and one of the most important natural resources. The European Mediterranean Regions are particularly sensitive to soil degradation. Especially in Italy many landscapes are prone to water related soil erosion processes such as rill-interrill erosion (sheet erosion), gullies and complex processes forming “calanchi” corresponding to the English term “Badlands”. Calanchi develop preferentially on Plio-Pleistocene deposits, which are highly susceptible to aquatic erosion processes. This study focusses on the assessment of rill-interrill erosion as well as calanchi erosion forms and features in previously none or very little-studied areas of the Langhe-Roero and Monferrato, located in the Liguria and Piedmont Region, Northern Apennines, Italy. In this study we assess the driving factors and the spatial distribution of rill-interrill and calanchi erosion processes using a geostatistical/geostochastic modelling framework. Therefore, a Maximum Entropy Model, a Generalized Linear Model and a Boosted Regression Tree approach were applied. As independent environmental variables we selected DEM based morphometric information as well as pedologic, and geologic input data. Moreover, we explore remote sensing techniques to derive information on the vegetation density and vitality. The stochastic models applied show, that the two very different soil erosion phenomena can be differentiated well with the chosen stochastic techniques. The main drivers controlling the spatial distribution of aquatic soil erosion in the study area are soil type, slope, elevation and topographic wetness index. The results of the MaxEnt and BRT model were confirmed by the GLM application in terms of the significance of predictor variables. Generally, the models show an acceptable to excellent performance with BRT and GLM outperforming MaxEnt. The single independent variable response curves gave valuable insights into the process triggering the erosion forms. We were able to clearly differentiate between the processes looking at the response curves that show reasonable characteristics with all applied models.