Residential building flood damage: Insights on processes and implications for risk assessments

Abstract

Flood damage assessments provide critical insights on processes controlling building damage and loss. Here, we present a novel damage assessment approach to develop an empirical residential building damage database from five flood events in New Zealand. Object-level damage data was collected for flood hazard and building characteristics, along with relative building component and sub-components damage ratios. A Random Forest Model and Spearman's Rank correlation test were applied to analyse damage data variable importance and monotonic relationships. Model and test results reveal flood inundation depth above first finished floor level is highly important and strongly correlated with total building damage ratios while flow velocity is important for structure component damage. Internal finishes components contribute highly to total building damage ratios as higher value sub-component materials are susceptible to direct damage from water contact and indirect damage during repair. The empirical damage data has several implications for damage model development due to the limited heterogeneity of flood hazard intensities and building attributes observed. Extending empirical damage data with synthetic damage data in future would support development of more representative object-specific damage models to evaluate direct tangible damages for local contexts.