Abstract:
The preventive assessment of flood risk is key to understand the costs of hazard scenarios in a changing climate so that adequate adaptation policies can be timely implemented. Italy is a flood-prone country that suffers among the highest economic impacts in the EU. Nevertheless, there is no established approach for estimating the economic impact of flood events. This is critical considering that such impacts are expected to increase by 2050 in Europe due to the effect of increased climate variability.
This thesis improves the customary flood risk framework commonly applied to translate the changes in flood hazard probability and magnitude into variation of Expected Annual Damage. It does so by focusing on the key components of the analysis, particularly the representation of exposed value and the characterisation of vulnerability.
In Paper #1 I test the performance of existing flood damage models for estimating direct impact to different land use categories.
In Paper #2 I calibrate a new damage curve for residential units using empirical damage records and a statistical calibration procedure adapted from an Australian study.
In Paper #3, I collect heterogeneous country-wide exposure data and combine them using a dasymetric approach in order to draw a new homogeneous dataset including asset, population, GDP, and social vulnerability.
Paper #4 takes a further step in the elaboration of a tool that can be practically employed for country-wide risk assessment. Damage observations from three recent flood events are employed to evaluate the relative importance of risk predictive variables and to test the accuracy of different damage models. Two machine learning algorithms are applied to assess the predictive ability of a multivariable setup compared to a customary univariable models.
