Flood risk management generally relies on economic assessments performed by using flood loss models of different complexity, ranging from simple univariable models to more complex multivariable models. The latter account for a large number of hazard, exposure and vulnerability factors, being potentially more robust when extensive input information is available. We collected a comprehensive data set 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), building characteristics (size, type, quality, economic value) and 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 performances of four literature flood damage models of different natures and complexities are compared with those of univariable, bivariable and multivariable models trained and tested by using empirical records from Italy. The uni- and bivariable models are developed by using linear, logarithmic and square root regression, whereas multivariable models are based on two machine-learning techniques: random forest and artificial neural networks. Results provide important insights about the choice of the damage modelling approach for operational disaster risk management. Our findings suggest that multivariable models have better potential for producing reliable damage estimates when extensive ancillary data for flood event characterisation are available, while univariable models can be adequate if data are scarce. The analysis also highlights that expert-based synthetic models are likely better suited for transferability to other areas compared to empirically based flood damage models.

Flood loss models are one important source of uncertainty in flood risk assessments. Many countries experience sparseness or absence of comprehensive high-quality flood loss data, which is often rooted in a lack of protocols and reference procedures for compiling loss datasets after flood events. Such data are an important reference for developing and validating flood loss models. We consider the Secchia River flood event of January 2014, when a sudden levee breach caused the inundation of nearly 52 km² in northern Italy. After this event local authorities collected a comprehensive flood loss dataset of affected private households including building footprints and structures and damages to buildings and contents. The dataset was enriched with further information compiled by us, including economic building values, maximum water depths, velocities and flood durations for each building. By analyzing this dataset we tackle the problem of flood damage estimation in Emilia-Romagna (Italy) by identifying empirical uni- and multivariable loss models for residential buildings and contents. The accuracy of the proposed models is compared with that of several flood damage models reported in the literature, providing additional insights into the transferability of the models among different contexts. Our results show that (1) even simple univariable damage models based on local data are significantly more accurate than literature models derived for different contexts; (2) multivariable models that consider several explanatory variables outperform univariable models, which use only water depth. However, multivariable models can only be effectively developed and applied if sufficient and detailed information is available.

Effective flood risk management requires a realistic estimation of flood losses. However, available flood damage estimates are still characterized by significant levels of uncertainty, questioning the capacity of flood damage models to depict real damages. With a joint effort of eight international research groups, the objective of this study was to compare, in a blind-validation test, the performances of different models for the assessment of the direct flood damage to the residential sector at the building level (i.e. microscale). The test consisted of a common flood case study characterized by high availability of hazard and building data but with undisclosed information on observed losses in the implementation stage of the models. The nine selected models were chosen in order to guarantee a good mastery of the models by the research teams, variety of the modelling approaches, and heterogeneity of the original calibration context in relation to both hazard and vulnerability features. By avoiding possible biases in model implementation, this blind comparison provided more objective insights on the transferability of the models and on the reliability of their estimations, especially regarding the potentials of local and multivariable models. From another perspective, the exercise allowed us to increase awareness of strengths and limits of flood damage modelling, which are summarized in the paper in the form of take-home messages from a modeller's perspective.

We investigate possible changes in flood hazard over a 77-km² area around the city of Ravenna. The subsidence rate in the area, naturally a few mm year⁻¹, increased dramatically after World War II because of groundwater and natural gas extraction, exceeding 110 mm year⁻¹ and resulting in cumulative drops larger than 1.5 m. The Montone–Ronco river system flows in the southern portion of the area, which is protected against frequent flooding by levees. We performed two-dimensional simulations of inundation events associated with levee breaching by considering four different terrain configurations: current topography and a reconstruction of ground elevations before anthropogenic land subsidence, both neglecting and representing the main linear infrastructures (e.g. roads, artificial channels). Results show that flood-hazard changes due to anthropogenic land subsidence (e.g. significant changes in computed water depth and velocity) are observed over less than 10% of the study area and are definitely less important than those resulting from construction of the linear infrastructures.

Effective flood risk management requires a realistic estimation of flood losses. However, available flood damage estimates are still characterised by significant levels of uncertainty, questioning the capacity of flood damage models to depict real damages. With a joint effort of eight international research groups, the objective of this study was to compare the performances of different damage models for the estimation of the direct flood damage to the residential sector at the building level (i.e. micro scale) in a blind validation test. The test consisted in a common flood case study characterised by high availability of hazard and building data, but with undisclosed information on observed losses in the implementation stage of the models. The selected nine models were chosen in order to guarantee a good mastery of the models by the research teams, variety of the modelling approaches and heterogeneity of the original calibration context, in relation to both hazard and vulnerability features. By avoiding possible biases in model implementation, this blind comparison provided more objective insights on the transferability of the models and on the reliability of their estimations, especially regarding the potentials of local and multi-variable models. From another perspective, the exercise allowed to increase authors’ awareness on strengths and limits of flood damage modelling, which are summarised in the paper in the form of take-home messages from a modeller's perspective.