Comparative Analysis of Exposure Databases and Portfolio Diversification in Catastrophe Modeling for (Re)insurance Applications in Italy

Modern risk management relies heavily on catastrophe modeling, which offers a structured methodology to estimate potential losses from natural and anthropogenic hazards. By combining hazard, exposure, and vulnerability components, these models quantify the expected impact of catastrophic events on assets. This thesis discusses how the core risk components are incorporated within catastrophe models, examines the primary outputs generated by catastrophe modeling analyses, and identifies the main sources of uncertainty involved. A particular emphasis is also given on the subjects of spatial correlation and portfolio diversification. A central part of the work presents a comparative analysis of publicly available exposure databases used in catastrophe risk assessment, with a particular focus on applications for reinsurance broker services in Italy. Six sources were evaluated; the European Seismic Risk Model, PERILS, the Global Human Settlement Layer, CORINE Land Cover, OpenStreetMap, and Ambiente Italia 3D. The comparison assessed structural detail, financial attributes, and overall suitability for portfolio applications. While ESRM provides the most comprehensive information, each model contributes uniquely, serving complementary roles rather than acting as direct substitutes. To demonstrate practical implications, the thesis incorporates a portfolio diversification study that explores how different exposure resolutions affect the assessment of asset correlation and the quantification of associated losses. By testing alternative representations of the same portfolio, the study highlights the sensitivity of diversification benefits to the level of spatial detail available in exposure data. This analysis emphasizes that the level of spatial detail directly impacts the loss distribution. Overall, higher-resolution and more disaggregated exposure inputs provide more realistic and consistent insights. In addition, the thesis presents average annual loss ratio maps for Greece, focusing on a representative industrial building type. These maps were developed through the integration of two of the evaluated exposure databases, providing an applied example of how different data sources can be combined to generate risk insights. The results illustrate regional variations in expected losses, while also demonstrating the added value of harmonizing datasets to support more reliable catastrophe risk assessments.