Estimating flood damage in Italy: empirical vs expert-based modelling approach

Abstract. Flood risk management generally relies on economic assessments performed using flood loss models of different complexity, ranging from simple univariable to more complex multivariable models. These latter accounts for a large number of hazard, exposure and vulnerability factors, being potentially more robust when extensive input information is available. In this paper we collected a comprehensive dataset related to three recent major flood events in Northern Italy (Adda 2002, Bacchiglione 2010 and Secchia 2014), including flood hazard features (depth, velocity and duration), buildings characteristics (size, type, quality, economic value) as well as reported losses. The objective of this study is to compare the performances of expert-based and empirical (both uni- and multivariable) damage models for estimating the potential economic costs of flood events to residential buildings. The performance of four literature flood damage models of different nature and complexity are compared with the performance of univariable, bivariable and multivariable models empirically developed for Italy and tested at the micro scale based upon observed records. The uni- and bivariable models are produced testing linear, logarithmic and square root regression while multivariable models are based on two machine learning techniques, namely Random Forest and Artificial Neural Networks. Results provide important insights about the choice of the damage modelling approach for operational disaster risk management.