An AI approach for multi-risk assessment in the Veneto Region

The increasing frequency and severity of extreme climate events necessitate robust multi-risk assessments. Traditional methods often fail to unravel complex hazard interactions and impacts. Artificial Intelligence provides a powerful tool for analysing environmental data, integrating diverse information sources, and modelling non-linear relationships, crucial for effective risk reduction strategies.

A stepwise AI-based framework to assess the risk posed by extreme climate events was developed for the Veneto region (North-East Italy). The main hazards considered are heatwaves, droughts, storm surges, extreme precipitation, extreme wind, landslide and wildfire. The first step involves identifying single hazard susceptibility maps, using statistical methods for atmospheric hazards and using supervised Machine Learning (XGBoost) for landslide and wildfire. In the second step, the single hazard susceptibility maps are integrated into a multi-hazard map, using a Random Forest model trained and validated on a multi-hazard event dataset in the historical timeframe. The multi hazard event dataset was created considering the spatial and temporal footprints of single hazards from climate data, utilizing statistical methods to detect extreme events, and applying unsupervised machine learning (DBSCAN) for clustering and counts the number of consecutive and compound multi-hazard events. Then, in the third step, the analysis is extended to multi-risk, integrating vulnerability and exposure indicators for multiple socio-economic variables (population, built environment, tourism and agriculture).

This comprehensive approach leverages advanced data-driven and AI techniques to enhance the understanding of the complex dynamics associated with multi-risk events. Applied within the Veneto case study of the Myriad-EU project, this framework has been tested for present and future scenarios considering RCP 4.5 and RCP 8.5, showing an increasing risk from hot and dry events in future multi-risk, especially for the tourism and agriculture sectors.