Multivariate Marked Point Process Fusion and Statistical Emulation for Uncertainty Quantification in Geophysical Extremes

Rare geophysical hazards, including hurricane storm surges, earthquake-generated tsunamis, and volcanic eruptions, threaten lives, infrastructure, and ecosystems worldwide. Reliable assessment of these rare events is essential for disaster preparedness and resilient infrastructure design, but historical observations alone are often insufficient to accurately characterize their risks. Advances in physics-based simulation can generate large collections of realistic synthetic hazard events, creating new opportunities to improve hazard assessment beyond the historical record. This project aims to develop statistical methods that integrate limited observations with synthetic event datasets to produce more reliable estimates of rare geophysical hazards. The resulting methodology will improve hazard assessment and risk-informed decision making, provide interdisciplinary research opportunities for students, and strengthen collaborations among statisticians and geoscientists. The project will develop a unified statistical framework for integrating sparse observations with large synthetic event datasets generated by physics-based simulations within the realm of machine learning and artificial intelligence. The framework models rare geophysical events as multivariate spatio-temporal point processes and employs optimal-transport-based data fusion to improve statistical inference under severe data scarcity. It also incorporates statistical emulators for computationally intensive simulators together with a comprehensive uncertainty quantification strategy that accounts for limited observations, simulation bias, and emulator approximation. Applications to hurricane storm surges and volcanic eruptions will demonstrate the framework and establish broadly applicable statistical methodology for quantifying rare geophysical hazards. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.