Capturing Regional Differences in Flood Vulnerability Improves Flood Loss Estimation

Flood vulnerability is quantified by loss models which are developed using either empirical or synthetic approaches. In reality, processes influencing flood risk stochastic and predictions bear significant uncertainty, especially due to differences in across exposed objects regions. However, many state-of-the-art deterministic, i.e., they do not account for data model uncertainty. The Bayesian Data-Driven Synthetic (BDDS) was one of the first approaches that used reduce prediction errors at object-level enhance reliability models. BDDS does regional may result over-/under-estimation losses some order overcome this limitation, study introduces a hierarchical parameterization enhances quantifying vulnerability. makes optimal use process information contained overall set various applications, so it particularly suitable cases only small amount available. implementation performance parametrization demonstrated with Multi-Colored Manual (MCM) functions damage dataset from UK consisting residential buildings regions Appleby, Carlisle, Kendal Cockermouth suffered during 2015 event. improves accuracy compared MCM reducing absolute error bias least 23 90%, respectively. terms hit rate (i.e., probability observed value lies 90% high density interval predictions) 88% same calibration 73% new approach practical relevance all where limited amounts