GNGTS 2024 - Atti del 42° Convegno Nazionale

Session 2.2 GNGTS 2024 Machine Learning-based modelling for Near Real-Time predicHon of liquefacHon C. Varone 1 , F. Mori 1 , A. Mendicelli 1 , G. Ciotoli 1 , G. Acunzo 2 , G. Naso 3 , M. Moscatelli 1 1 CNR Italian Na9onal Research Council, Ins9tute of Environmental Geology and Geoengineering (IGAG), Montelibreo, Italy 2 Theta Group Srls, Rome, Italy 3 Presidenza del Consiglio dei Ministri, Dipar9mento della Protezione Civile (DPC), Rome, Italy The occurrence of an earthquake may involve extensive consequences, highlighCng the criCcal requirement from prompt and reliable informaCon to minimize the effects of the disaster. The rapid execuCon of emergency response strategies also depends on the accurate predicCon of seismically induced effects. This study aims to develop an approach based on machine learning to predict seismically induced liquefacCon phenomena in near real-Cme. This method combines factors such as ground shaking scenarios (Magnitude and Peak Ground AcceleraCon PGA), shear-waves velocity to a depth of 30 meters (Vs30), and groundwater table depth to predict the likelihood of liquefacCon phenomena. The implementaCon is planned within the SEARCH (Seismic Emergency Assessment and Response CompuCng Hub) sopware, which is an innovaCve soluCon able to generate impedance maps for landslides, liquefacCon, and building collapses. These maps provide valuable assistance for emergency management. Corresponding author: chiara.varone@cnr.it