Mosimann, Markus; Frossard, Linda; Keiler, Margreth; Weingartner, Rolf; Zischg, Andreas Paul (2018). A Robust and Transferable Model for the Prediction of Flood Losses on Household Contents. Water, 10(11), p. 1596. MDPI 10.3390/w10111596
|
Text
water-10-01596.pdf - Published Version Available under License Creative Commons: Attribution (CC-BY). Download (447kB) | Preview |
Beside the flood hazard analysis, a comprehensive flood risk assessment requires the analysis of the exposure of values at risk and their vulnerability. Currently, the main focus of such analysis is on losses on building structure. However, loss on household contents accounts for up to 30% of the total losses on buildings due to floods. Based on insurance claim records, we developed and (cross-)validated two functions. The models based on linear regressions estimate the monetary loss and the degree of loss of household contents by the monetary and degree of loss for building structure, respectively. The main focus herein is to develop functions which provide robustness in prediction and transferability to other regions. Both models generate appropriate results with a comparative advantage of the relative over the absolute loss model. Our results indicate that the ratio of household content to building structure loss is decreasing relatively in regions with comparatively high losses or degrees of loss. A detailed examination of the model residuals, shows that the Box-Cox transformation works well to accurately fit a standard regression model to general right-skewed loss data as the transformed data meet the assumptions of a regression model.
Actions (login required)
 |
Edit item |