Machine learning for the definition of landslide alert models: a case study in Campania region, Italy

Abstract

Landslide early warning systems at regional scale are typically based on correlations between landslides that occurred in the past and rainfall monitoring data in order to identify trigger rainfall thresholds. Recently, the availability of large datasets of atmospheric measurements allows including additional variables, increasing the reliability of the models. However, conducting the analyses with traditional techniques can be quite complex and time-consuming. The purpose of this preliminary study is to demonstrate that machine learning techniques can be used to analyze monitoring data in order to select the most relevant variables for the triggering of shallow rainfall-induced landslides at regional scale. The models developed herein were tested in one of the alert zones defined by civil protection for the management of geo-hydrological risk in Campania region, Italy. Two data sources were used in the analysis. The atmospheric variables are derived from the ERA5-Land atmospheric reanalysis. The data on landslide events are retrieved from “FraneItalia”, a georeferenced catalog of landslides occurred in Italy developed by consulting online sources from 2010 onwards. The models developed were calibrated and validated in order to define combinations of rainfall variables and soil water content for the prediction of the occurrence of landslides. Finally, the performance of the models was assessed using statistical indicators derived from contingency matrices.