Longitudinal assessment of changes in household flood resilience in Ho Chi Minh City, Vietnam

Flood resilience reflects the capacity to anticipate, withstand, and recover from flood impacts through a combination of available resources and adaptive responses. Despite its prominence in flood risk research, flood resilience is rarely measured empirically in urban environments, where exposure and vulnerabilities evolve dynamically over time. This study examines changes in household-level flood resilience in Ho Chi Minh City (HCMC) between 2020 and 2023 using two longitudinal survey waves (1,000 and 750 households, respectively, including a panel of 560 households that participated in both surveys). Our goal is to identify trends and dynamics of different resilience dimensions over time, along with the drivers of persistent vulnerability. We develop a multi-stage data-driven approach that combines indicator screening, dimension construction, and statistical modelling. A comprehensive set of survey-based indicators capturing flood characteristics, socioeconomic conditions, behavioural responses, and flood damage are first formulated to represent human, social, physical, financial, and natural capitals (5C). Tree-based models are then applied to identify the feature importance associated to the factors most strongly related with changes in flood outcomes. Based on this screening, selected indicators are then aggregated into latent resilience dimensions corresponding to the 5R framework (robustness, redundancy, resourcefulness, rapidity, and recovery). These are combined, producing individual 5R scores and an overall resilience score. The longitudinal design enables comparison of resilience profiles over time and supports the analysis of variation in resilience within Ho Chi Minh. By linking observed household-level capacities to resilience processes, this study supports the empirical measurement of systemic resilience and provides actionable insights for flood risk reduction and adaptation planning in rapidly urbanising flood-prone contexts.