Damage susceptibility in Italy: a case study for Tuscany

With our work, we’re proposing an approach to damage modeling oriented to the potential damage evaluation, with the aim to use it to address the prevention strategies in a more efficient way.

The methodology will be presented on a case study developed in the Italian region of Tuscany. For this application, our data collection on perceived damage, made up of claims compiled by citizens in the aftermath of relevant flood events, has been enriched with new data to cover the whole set of national state of emergency for Tuscany in the period 2013/2023.

Claims have been geolocalized and extracted on the plane areas of the region. We came up with more than 10800 points, providing a picture of where damage occurred, the declared economic losses and the areas affected by more than one event.

This dataset has been later adopted to train a machine learning model which combines the occurred damages, the characteristics of the territory (obtained from digital terrain model and other open data) and the communities’ social variables primarly derived from national census. Instead of superpose hazard and exposure, we have been working combining data sources which are different for origine, scale and semantic area in a big database to be provided to the algorithms.

We are here presenting the results of our work as well as the lesson learned in the modeling procedure.