Landslide Susceptibility Analysis with Artificial Neural Networks Used in a GIS Environment

The identification of areas potentially subject to hydrogeological risk is an important issue, especially for Institutions that spend substantial sums each year on remediation. In 2018, north-eastern Italy was hit by an extreme event, Storm Vaia, which produced extensive damage, such as the collapse of trees and the reactivation of numerous landslide phenomena. In particular, the study area of this research, located in the municipality of Feltre, saw the reactivation of 9 landslide phenomena, which prompted the assessment of landslide susceptibility for the area in question. The objective of this research is to identify areas of high potential landslide hazard, using GIS-type software to prepare the spatial analysis and generating an algorithm inspired by the weight of evidence technique. For the calculation of susceptibility, eight environmental variables were considered, which can substantially influence the activation of landslide phenomena: permeability of the lithotype, type of vegetation cover, exposure, slope, ancient and stable deposits, recent slope deposits, bedrock, and existing landslides. The first level contains all the areas in which is present at least one of the conditions under our phenomena, the second at least two and so on. The algorithm works with three main processing tools, extraction by position, clip and merge, clipping the layers according to a number of simple combinations, without repetition, increasing for each susceptibility level: 8, 28, 56, and 70. The final map contains four levels of susceptibility in ascending order of hazard, and as further proof of the validity of the model it overlapped very well with already well-known and surveyed phenomena. The method used, unlike the weight of evidence, assigns the same weight to each underlying condition, so that the map can be more conservative and not underestimate risks. Potentially, the analysis could be further extended in the future by adding additional factors influencing slope stability, such as the amount of precipitation.