Artificial Intelligence for climate change multi-risk assessment: a Myriad-EU case study in the Veneto Region

The escalating frequency and intensity of extreme climate events underscore the need for robust multi-risk assessment methodologies. Conventional approaches often struggle unravelling the intricate interplays among diverse hazards and their impacts on vulnerability and exposure factors. Understanding the complex impact chains and the consequences of extreme climate events on socio-economic and natural systems is crucial for formulating effective risk reduction and preparedness strategies. Artificial Intelligence (AI) has emerged as a powerful tool for analysing intricate environmental data, fusing information from different heterogeneous sources, and modeling non-linear relationships.

A stepwise AI-based framework has been developed to assess the risk induced by extreme climate events—specifically, heatwaves, droughts, storm surges, extreme precipitation, and extreme wind events—in the Veneto Region (North-East Italy). The first step consists in the identification of single hazard spatial and temporal footprints from climate data, using statistical methods (quantiles and percentiles) for identifying anomalies and extreme events and unsupervised machine learning (DBSCAN) for clustering. The second step aims at building multi-hazard event sets, by combining the dynamic single hazard clusters extracted in the first step with static footprints of other hazards, such as wildfires and landslides. In particular, different time lags and spatial overlaps are applied to identify compound or consecutive events. Finally, the third step employs supervised ML algorithms, such as Random Forest, Support Vector Machine (SVM), and Convolutional Neural Networks (CNN), to model multi-hazard susceptibility over different multi-hazard combinations. Footprints of past single and multi-hazard events are used as assessment endpoints to train the ML model and identify the most important vulnerability and exposure factors and multi-risk hotspots within the Veneto region.

This comprehensive approach integrates advanced data driven and AI techniques to enhance the understanding of the complex dynamics associated with multi-risk events. This framework has been applied and tested within the Myriad-EU project, in the Veneto Region case study, demonstrating its efficacy in assessing and predicting the impacts of multi-risk events under different climate change scenarios.