Influence of Data Preprocessing and Optimization in Multivariate Statistical Analysis of Landslide Susceptibility

As part of the PNRR – Extended Partnership RETURN, the research group is developing a procedure for producing landslide susceptibility maps in a GIS (Geographic Information System) environment. The use of the statistical technique of Logistic Regression has proved suitable for the purpose, but the reliability of the resulting models depends greatly on the input data and their pre-processing. To make the proposed procedure effectively usable by land managers, transferable to different areas, and allowing for the comparison of different scenarios, it is necessary to define a minimum standard for the input data, as well as some specifications for their preprocessing. The present work, based on the case study of the Province of Savona (IT), identifies the basic data requirements, and provides instructions for the preparation of spatial datasets (relative to the Italian territory) and for the processing flow for the effective use of Logistic Regression for statistical landslide prediction.