Linking Disaster-Driven Destruction and Reconstruction to Population Health Outcomes over Two Decades

Rising sea levels and shifting weather patterns linked to climate change have increased the frequency and severity of exposures to extreme events, particularly flooding, across the globe. A paucity of high-quality longitudinal data has limited scientific understanding of the implications of these exposures and their aftermath for population health and well-being over the long term. Focusing on the 2004 Indian Ocean Tsunami, we measure small-scale geographic area exposures to initial destruction and gradual reconstruction of built and natural environments using convolutional neural network methods applied to high-resolution satellite imagery. We illustrate the value of combining these measures with individual-level data on health and well-being that we collected over two decades starting before the tsunami as part of a population-representative longitudinal household survey, the Study of the Tsunami Aftermath and Recovery.