Platform Prototype for the Prediction of Landslide Susceptibility Through a 4D WebGIS Equipped with Cellular Automata and Neural Networks

Climate change and increasing anthropogenic impact in forested areas (logging and fires) have in recent years intensified the Hydrogeological Instability of many mountain slopes in the Italian territory. This leads to greater difficulty in identifying in the short- and long-term well-defined areas of higher priority for preventive intervention to secure inhabited areas and infrastructure. In fact, extreme weather events limited in space and time, together with possible microseisms, can make the degree of landslide susceptibility attributable to a given slope change abruptly. In this sense, a WebGIS 4D infrastructure capable of producing susceptibility forecast layers through simulators with emergent properties has been implemented, which will cooperate to create a “digital twin” of the territory and context in being, on which simulations are launched to obtain useful forecasts. The digital twin is created from geo-referenced three-dimensional Cellular Automata (C.A.), which are characterized by states and mutual interactions defined by the phenomenology of that context (moisture diffusion, geologic characterization of the terrain, seismic events…). The states then vary over time as simulation iterations proceed and through data assimilation (e.g., rainfall). An AI with SNN investigates the C.A. states looking for specific patterns, from which to generate the susceptibility index and thus the forecast output. The process can be seen as an immediate downstream update of critical events to susceptibility maps distributed by government agencies (e.g., ISPRA). The application of this method was carried out in South Italy.