Characterization of compound earthquake damage empowered by AI remote sensing

Rapid post-earthquake response remains a significant challenge for humanity. Emergency response to earthquakes requires accurate and timely information about the geographic locations of secondary hazards and the likely compound effects, such as landslides, liquefaction, and building damage. Current methods rely on data-driven approaches, and also start to consider the complex causal dependencies associated with earthquake-induced disasters. However, the accuracy of existing pipeline is limited due to factors like atmospheric noise contaminating satellite imagery.

To improve the accuracy of predicting multiple hazards and impacts, we introduce the principles of time-series Interferometric Synthetic Aperture Radar (InSAR) analysis to generate high-quality Damage Proxy Maps (DPM). Subsequently, we adopt a rapid seismic multi-hazard and impact estimation system leveraging advanced statistical causal inference and remote sensing techniques. This approach, by modeling causal dependencies from satellite images, infers multiple hazard scenarios on a regional scale at high accuracy and resolution.

Data we using include landslides, liquefaction, and building damage. We also created DPMs using SAR images from the Sentinel-1 satellite. Beides the accuracy, our approach’s results also reveal quantitative causal mechanisms among earthquake-triggered multi-hazard and impact events. Our system provides a new approach to InSAR data processing and offers a novel avenue for understanding the complex interactions of multiple hazards and impacts in seismic geological processes.