Multi-variate residential flood loss estimation model for Jakarta: an approach based on a combination of statistical techniques

Jakarta has endured a series of devastating floods in 2002, 2007 and 2013. In the wake of climate change and complex interaction with diverse socio-economic factors, the flood situation in Jakarta will exacerbate from bad to worse. The city requires quantitative evaluation of flood risk to adapt to the changes arising from both existing and future flooding. Consequently, in this study a comprehensive multi-variate residential flood loss estimation model was developed. The model was developed from the outcome of extensive household surveys performed in the aftermath of the 2013 January floods. In addition, a novel procedure for classifying the flood loss into homogenous groups is proposed and implemented on the data from Jakarta. The proposed approach employed principal component analysis supplemented with correlation analysis and mutual information in identifying the factors that can capture the homogeneity in flood damage and loss data. The identified factors are then used as inputs for cluster analysis to for grouping the data into homogenous groups. Depth–damage curves were developed for these groups. The novel approach for demarcating flood loss data produced groups which showed significant reduction in damage variability and improved loss prediction within the surveyed dataset. Further, a scaled flood loss estimation model was developed which quantified the effect of mitigation measures on flood damage and loss.