Putting regional landslide susceptibility maps to work: bridging statistical assessment to territorial planning and civil protection (the SUFRA project in Sicily, Italy)

Applied geomorphology has to furnish society effective and reasonable tools for governing earth surface processes related risk issues. This is a challenge which takes a critical perspective in the case of landslides, which certainly are among the most hazardous morphodynamic phenomena, both in terms of magnitude and recurrence sensitivity to climate change scenarios. In light of its geomorphological setting and meteorological/seismic/anthropogenic recurrent stressing, the regional territory of Sicily is strongly affected by landslide phenomena. In the framework of the SUFRA project, systematic regional susceptibility maps have been recently completed and delivered to the use of the Basin Authority of the Hydrographic District of Sicily (BAHD_Sicily), which is the regional administrative agency responsible for landslide risk compliant land use planning policies and rules. In particular, each of the SUFRA project activities have been carried out with a strict view toward the needs of (BAHD_Sicily). This perspective actually conforms to all the steps which structure reliable landslide susceptibility modelling and mapping: landslide classification, mapping unit and diagnostic landform selection (pre-processing); model building and validation strategies (processing); score classification and map designing (post-processing). Differently from a research prospective, optimal solutions had to be targeted to find the best trade-off between the quality of models and derived maps and their functioning. Landslide susceptibility stochastic assessment is based on the concept that past and future landslides share common underlying causes (geo-environmental factors), which can be explored by evaluating the distribution between past events in the hyperspace of these geo-environmental predictors. A number of blind-testing methods can be then applied to estimate the quality of the models resolution/reliability/robustness/adequacy. However, the final maps should cope with the need of matching the local scale, which is the one typically involved in engineering geology. Objective, unambiguous susceptibility classes have to be set and associated with each specific site which has to be analysed by taking into account the geomorphological connectivity both for causative geomorphological factors and gravitational outcome. The regional landslide susceptibility maps of Sicily were calibrated by exploiting inventories made of near 50000 flows, 4000 slides, for the twenty-one catchments or catchment groups in which the near 22.000 km2 was split to preserve the required density of data. Independent landslide susceptibility models and maps were produced for: flows, rotational/translational slides, rebounds/falls, debris flows. To solve for the issues above recalled, slope units partition was used for slope phenomena (flows, slides), whilst pixel based models were prepared for debris flows and falls. Besides, criteria and laws and tools were established for defining the spatial connectivity of a specific site and its surrounding susceptibility scenario. Multiple cross-validation strategies were adopted to check the models for adequacy, accuracy and reliability. Four classes of susceptibility were objectively derived by a recursive nested double application of the Youden-index criteria on the AUC plots. The results attested as the SUFRA procedure can be taken as an example of effective application of geomorphology, as attested by the ongoing implementation of new rules by BAHD_Sicily including the susceptibility maps which have been produced. This study was carried out within the RETURN Extended Partnership and received funding from the European Union Next-GenerationEU (National Recovery and Resilience Plan – NRRP, Mission 4, Component 2, Investment 1.3 – D.D. 1243 2/8/2022, PE0000005).