Landslide inventory update and susceptibility mapping: integration of field survey, remote sensing, and UAVs

The evaluation of landslide susceptibility, particularly when conducted through semi-automatic methods, is significantly influenced by the quality of the landslide inventory. The last landslide inventory update in Sardinia dates back to 2005, leaving many uncovered areas. Updating the inventory is crucial to assess the spatial distribution of landslides and to identify their source areas, which are both the major concerns in determining geological and geomorphological susceptibility factors. Mapping landslides is very time-consuming, especially with classical field survey, considering that many wild zones on the island are difficult to reach. The dominant landslide type are rockfalls, often occurring on very steep to vertical slopes. For these reasons, in this study field surveys were supported by recent technologies, such as UAVs, remote sensing data and GIS tools analysis. In the framework of two PNRR Projects, Geosciences IR and e.INS, we focused on two test areas poorly covered by the IFFI inventory, both showing diffuse instability: a mountainous area in North-East Sardinia, characterized by limestones, and a coastal area in South-West Sardinia with cliffs made of highly fractured rhyolites and exposed to heavy swells. At first, we used existing maps, multi-temporal analysis on orthophotos and satellite images to identify areas with evidence of instability and diffuse rockfalls. Consequently, we performed UAV surveys, precisely delineating several rockfalls that were sometimes not visible in existing images due to shadows or low resolution. We investigated two test landslides in detail, evaluating their morphometric parameters and geo-mechanical characteristics by post processing the UAV data. In the mountainous test area, rockfalls were predominant, mostly distributed in steep cliffs. The resulting rockfall deposits often overlay a different lithology than the source one with less steep terrain. This poses the problem of how mapping susceptibility based on semi-automatic procedures can be misleading with such context, very diffuse in Sardinia. In the coastal test area, rockfalls are regulated by swells and storms, with most deposits under water. Mapping and analyzing them could only be conducted with remote UAV techniques. In both areas the 3D terrain models, obtained by UAV and post processed with semi-automatic fractures detection algorithms facilitated the identification of failure surfaces and allowed a geo-mechanical analysis. The integration of these recent technologies enables the identification of landslides, speeds up their analysis, and meets the need to update inventories on a large scale in a relatively short time and to improve the accuracy of susceptibility mapping based on semi-automatic probabilistic methods.