Machine Learning-Driven Synthesis of Multi-Hazard Fragility Surfaces for Seismic and Tsunami Resilience

Coastal populations are particularly vulnerable to offshore earthquakes and their cascading effects, such as tsunamis. In response to these risks, this study presents a novel machine learning (ML) model for synthesizing 3D fragility surfaces that capture the combined impacts of earthquake and tsunami hazards. The model leverages independently generated 2D fragility curves for each hazard and integrates physics-based simulations to improve accuracy and reliability. By utilizing data-driven approaches, the model offers computational efficiency, reducing the need for high-performance computing resources typically required for such simulations. An important application of the model lies in retrofitting analysis. By using 2D fragility curves for retrofitted structural systems, the model can generate retrofitted earthquake-tsunami fragility surfaces, allowing for more comprehensive mitigation and resilience planning at both building and community levels. This ability to model structural retrofits provides critical insights into the effectiveness of different mitigation strategies for coastal communities facing multi-hazard scenarios. While the model is primarily demonstrated for earthquake-tsunami hazards, the methodology has the potential to be applied to other hazard combinations, making it a versatile tool for broader multi-hazard resilience assessments. The study concludes that machine learning offers a transformative approach to hazard and vulnerability modelings, enhancing decision-making for urban planning and disaster preparedness.