GNGTS 2024 - Atti del 42° Convegno Nazionale

Session 2.2 GNGTS 2024 Towards an earthquake-induced landslide triggering map for Italy S. Azhideh, S. Barani, G. FerreU, D. Scafidi, G. Pepe 1 Università degli Studi di Genova, Genova, Italy Landslides are one of the most frequent geohazards that have caused devasCng damage throughout history. Landslides open occur as a consequence of other natural hazards among which earthquakes can be considered as one of the main triggering factors. When an earthquake occurs, the effects of the induced ground shaking are open sufficient to cause failure of slopes that were marginally to moderately stable before the earthquake. In the present study, we present a first abempt to define a screening map for all of Italy that classifies sites in terms of their potenCality of triggering earthquake-induced landslides based on seismic hazard. To this end, we analyze seismic hazard disaggregaCon results on a naConal scale (Barani et al., 2009) and compare magnitude-distance ( M - R ) scenarios with the upper- bound M - R curves defined by Keefer (1984) for different types of landslides: disrupted slides and falls, coherent slides, and lateral spreads and flows. For a given magnitude value, these curves define the criCcal distance below which earthquake-induced failures may occur and, as a consequence, the possibility of triggering a landslide can not be discounted. First, for all computaCon nodes considered in the hazard assessment of Italy (MPS Working Group, 2004; Stucchi et al., 2011), joint distribuCons (i.e., probability mass funcCons, PMFs) of M and R are analyzed to idenCfy all modal scenarios (i.e., local maxima). To this end, we treat each PMF as an image and apply morphological image processing techniques to find local maxima. Specifically, we apply the maximum (dilaCon) filter operaCon (e.g., Gonzales and Woods, 2018). Each M - R scenario in the PMF matrix is treated as a pixel. Each pixel is updated based on comparing it against the surrounding pixels in a running window process (a 3-by-3 square window around the target pixel is used). Specifically, the maximum filter replaces the value of the PMF associated with the central pixel with the greatest one in the running window. Finally, local maxima in the distribuCon are obtained by checking for element-wise equality within the original and filtered matrices, creaCng an array of Boolean values (Boolean matrix) within which True values indicate the modes. Figure 1 shows an example M - R distribuCon with indicaCon of the modal scenarios resulCng from maximum filtering and Boolean mask.