Web-GIS 4D for prediction of landslide susceptibility and hazard with emerging properties such as 3D cellular automata, neural networks and SPH fluids

Rapid rain-induced landslides involve different types of terrain and often cause loss of life and major socio-economic disasters. This is related to the limited presence of warning signs in the pre-break-up phase (especially in the absence of efficient monitoring and warning systems), the high velocities reached in the propagation phase, and the increased volume of debris involved, due to soil erosion along the way. The research aims to implement a 3D WebGIS system integrating innovative models and tools to generate zoning layers based on landslide susceptibility. Such models will cooperate to create a 'digital twin' of a territory to be used as the test field of the simulations to be carried out to obtain useful forecasts. The modelling techniques include: (a) Smoothed-particle hydrodynamics (SPH) simulations to downscale meteorological surveys, or the outputs of forecasting models; (b) Cellular Automata ('CA') to represent the digital twin of mountain slopes and track the evolution of parts of the terrain downstream of critical events; (c) Neural networks, applied to the state of each cellular automata, generating indexes which indicate landslide susceptibility at a simulation time.