Disentangling Earthquake Damage Correlations to Improve Portfolio Risk and Post-Event Inspections

Earthquakes can damage thousands of buildings and bridges across a region simultaneously. A key question for emergency planners, transportation agencies, and insurers is whether such damage tends to be clustered or more dispersed. The answer depends on how closely the damage to one structure is related to that of nearby similar structures. These relationships, or correlations, strongly influence estimates of the total losses a region may face and the likelihood of unusually severe outcomes. The correlation values used in practice largely rest on expert judgment, because no established method exists to measure them from real earthquake observations. This project develops a way to estimate these damage correlations directly from data collected after past earthquakes and to use them to improve two decisions: estimating the earthquake risk facing a portfolio of buildings or bridges, and sequencing post-earthquake inspections so that the earliest inspections reveal the most about the condition of the larger inventory. More accurate risk estimates support preparation by public agencies, insurers, and communities, advancing national welfare and economic resilience, and smarter inspection sequencing helps return critical infrastructure to service more quickly, reducing disruption and protecting public safety. The project also engages students at multiple levels in engineering research and delivers freely available software and data to the broader community. This research develops statistical methods to estimate structure-to-structure damage correlations from empirical post-earthquake damage data and applies these correlations to improve regional risk analysis and post-earthquake inspection decisions. The approach uses Bayesian inference to jointly estimate ground-motion intensities, fragility parameters, and correlation structure from observed damage, handling the challenge that ground-motion intensity at each structure is itself uncertain and spatially correlated. A central scientific question concerns identifiability: correlation among structural capacities and correlation among ground motions produce distinct signatures in the joint distribution of regional damage, and the estimation framework exploits these signatures to separate the two sources. The work pursues three objectives. The first develops a Bayesian estimation framework for fragility models with correlated capacity structures. The second tests the methodology on damage data from two contrasting events (the 2009 L'Aquila earthquake and the 2023 Kahramanmaraş earthquakes) to quantify the amount of data needed to constrain the estimates and whether they generalize across building types. The third develops an information-theoretic framework for inspection prioritization that applies Bayesian experimental design to post-earthquake response, using the estimated correlation structure to sequence inspections for maximum information gain. Outcomes include calibrated correlation estimates for example portfolios, a correlation-aware inspection prioritization framework, and open-source software released to the research and practitioner communities. 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.