A NEW SPECTRAL-SPATIAL SUBSPACE CLUSTERING ALGORITHM FOR HYPERSPECTRAL IMAGE ANALYSIS

Ho Chi Minh City (HCMC) faces severe flood risks from climatic and socio-economic changes, requiring effective adaptation solutions. Flood loss estimation is crucial, but advanced probabilistic models addressing key drivers and uncertainty are lacking. This study presents a probabilistic flood loss model with a feature selection paradigm for HCMC’s residential sector. Experiments using new survey data from flood-affected households demonstrate the model's superior performance.

Many households in Vietnam depend on revenues from microbusinesses (shop-houses). However, losses caused by regular flooding to the microbusinesses are not modelled. Business turnover, building age and water depth are found to be the main drivers of flood losses to microbusinesses. We built and validated probabilistic models (Non-parametric Bayesian Networks) that estimate flood losses to microbusinesses. The results help in flood risk management and adaption decision making for microbusinesses.