Residential flood loss estimated from Bayesian multilevel models

Abstract. Models for the predictions of monetary losses from floods mainly blend data deemed to represent a single flood type and region. Moreover, these approaches largely ignore indicators preparedness how predictors may vary between regions events, challenging transferability loss models. We use database 1812 German flood-affected households explore Bayesian multilevel models can estimate normalised damage stratified by event, region, or process type. Multilevel acknowledge natural groups in allow each group learn others. obtain posterior estimates that differ types, with credibly varying influences water depth, contamination, duration, implementation property-level precautionary measures, insurance, previous experience; overlap across most events regions, however. infer underlying damaging processes distinct types deserve further attention. Each reported affected region involved mixed likely explaining uncertainty coefficients. Our results emphasise need consider as an important step towards applying elsewhere. argue failing do so unduly generalise model systematically bias estimations empirical data.