Validation of 2D flood models with insurance claims

Abstract

Flood impact modelling requires reliable models for the simulation of flood processes. In recent years, flood inundation models have been remarkably improved and widely used for flood hazard simulation, flood exposure and loss analyses. In this study, we validate a 2D inundation model for the purpose of flood exposure analysis at the river reach scale. We validate the BASEMENT simulation model with insurance claims using conventional validation metrics. The flood model is established on the basis of available topographic data in a high spatial resolution for four test cases. The validation metrics were calculated with two different datasets; a dataset of event documentations reporting flooded areas and a dataset of insurance claims. The model fit relating to insurance claims is in three out of four test cases slightly lower than the model fit computed on the basis of the observed inundation areas. This comparison between two independent validation data sets suggests that validation metrics using insurance claims can be compared to conventional validation data, such as the flooded area. However, a validation on the basis of insurance claims might be more conservative in cases where model errors are more pronounced in areas with a high density of values at risk.

The relationship between precipitation and insurance data for floods in a Mediterranean region (northeast Spain)

“…Also our findings align with the results of previous studies (Spekkers et al, 2013;Zhou et al, 2013;Wobus et al, 2014;Torgersen et al, 2015) and further indicate that insurance databases are a promising source for flood damage assessment at local (Garrote et al, 2016;Le Bihan et al, 2017;Zischg et al, 2018;Zhou et al, 2013) and at regional scale (Barredo et al, 2012;Kim et al, 2012;Wobus et al, 2014;Zhou et al, 2017).…”

Characterizing precipitation events leading to surface water flood damage over large regions of complex terrain

“…A review of using damage data and insurance data in the analysis of SWFs is given by Gradeci et al [29]. Insurance claim data are also used for validating flood models [30][31][32]. Hurford et al [33] evaluated the return periods of rainfall thresholds that are used for warnings by linking rainfall intensities with observed SWFs using data from fire and rescue services.…”

A Robust and Transferable Model for the Prediction of Flood Losses on Household Contents

“…The availability of spatially and temporally high resolved flood models and the possibility to develop flood scenarios lead to new perspectives in detecting regions or even single buildings with a high loss potential. Especially, flood losses are increasingly estimated at the scale of single buildings [11][12][13][14][15][16]. Because inundations rarely lead to a total destruction of buildings, e.g., Papathoma-Köhle et al [17] and Fuchs et al [18] use the term "(physical) vulnerability" to describe the ratio of the monetary loss to the value of a building and thus, this term corresponds to the relative loss occurring on a building (loss divided by insurance sum).…”