RESEARCH Spotlights

This research presents an innovative active learning framework to improve the efficiency and accuracy of post-earthquake reconnaissance planning. By integrating Gaussian process regression with an information-theoretic selection criterion, the method identifies batches of high-value buildings for inspection that are most likely to improve regional damage and loss predictions. An optimized routing strategy then minimizes travel time between inspection sites, allowing reconnaissance teams to cover the most informative locations while conserving limited post-disaster resources. When applied to different earthquake testbeds—including the 2018 Anchorage, Alaska earthquake—the framework achieved up to a 53% reduction in required inspections while improving predictive accuracy compared to conventional approaches, demonstrating its potential for faster, more cost-effective post-disaster assessments.

The NHERI SimCenter’s R2D tool – particularly its backend, rWhale – played a key role in this work by generating testbed data and enabling evaluations with both real-world earthquake records and physics-based simulated ground motions. The approach offers significant societal benefits by accelerating post-earthquake damage assessment, improving agency decision-making, and informing equitable allocation of emergency resources such as shelters. By reducing the time, travel, and costs associated with data collection while preserving or improving accuracy, this framework supports rapid, informed recovery planning and strengthens community resilience in the aftermath of earthquakes.

Researchers are advancing the next generation of seismic risk and resilience assessment through the development of a high-fidelity, high-resolution computational framework capable of analyzing thousands of buildings at the regional scale. Traditional building codes have emphasized life safety, but national initiatives now call for codifying functional recovery as a performance objective—bringing economic loss, downtime, and long-term recovery into focus. This project presents an efficient framework that integrates automation, optimization, and distributed computing to make previously infeasible large-scale risk and resilience evaluations computationally viable. A case study on more than 15,000 residential woodframe buildings in Los Angeles demonstrates how such assessments, once requiring months of computation, can now be completed within days.

The research leveraged NHERI SimCenter tools, including R2D for regional hazard simulation and pelicun for loss estimation, as core components of its workflow. These tools, along with Auto-WoodSDA (which builds on Pelicun), enabled the seamless integration of hazard modeling, probabilistic simulation, and building-specific performance assessments with site-specific ground motion inputs. By validating and advancing computational strategies for regional seismic analysis, the study not only pushes the boundaries of high-resolution resilience assessment but also lays the groundwork for practical policy and planning applications. The resulting framework offers decision-makers the ability to evaluate resilience-based standards, identify vulnerable neighborhoods, and inform mitigation strategies that strengthen community resilience to future earthquakes.

This research presents a machine learning–based optimization framework to support recovery-based design, addressing the challenge of linking building-level engineering decisions to broader community resilience goals. Traditional downtime simulations are computationally intensive, making it difficult to evaluate above-code design strategies across the full range of possible interventions. The framework integrates FEMA P-58 loss assessment, ATC-138 functional recovery modeling, and surrogate modeling to efficiently explore thousands of design alternatives against explicit recovery targets. In a three-story office building case study in Oakland, CA, the study found that targeted nonstructural upgrades—such as improvements to stairs, curtain walls, and HVAC units—were essential to achieve ambitious recovery goals. By contrast, structural-only upgrades sometimes increased recovery times by amplifying acceleration demands, highlighting the complex trade-offs inherent in recovery-based design.

NHERI SimCenter tools played a central role in enabling this research. EE-UQ was used to run nonlinear response history analyses, while pelicun powered the large-scale FEMA P-58 loss assessments necessary to train surrogate models at scale. Together, these tools made it feasible to generate the consistent, high-volume datasets required for optimization. The research team released their workflow and open-source code, providing a reproducible platform that other researchers, engineers, and policymakers can adopt to explore recovery-based design strategies. In the long term, this approach has the potential to inform evolving functional-recovery provisions, guide benefit–cost analyses, and support community resilience planning by offering practical, quantitative tools to balance structural and nonstructural interventions for post-earthquake recovery.

This research advances the understanding of disaster-driven household displacement by developing predictive models that estimate both the likelihood and duration of displacement following natural hazards. Using data from the U.S. Household Pulse Survey, the study introduces three models—ranging in complexity and predictive power—to determine whether a displaced household will return within a month, return after a month, or not return at all. The analysis also examines the relative influence of physical and socioeconomic factors on displacement outcomes. To illustrate practical application, the models were integrated into a hurricane scenario analysis for Atlantic City, NJ, providing valuable insights for community recovery planning.

The NHERI SimCenter’s R2D tool and Atlantic City, NJ Testbed were essential in demonstrating the predictive models within a realistic hazard context. By making all analysis code and fitted models openly available on GitHub, the research provides the broader community with accessible tools for integrating household displacement forecasts into disaster risk assessments. This work supports future efforts to inform public sector decision-making, enhance disaster preparedness, and guide equitable post-disaster resource allocation—ultimately strengthening community resilience in the face of natural hazards.

Researchers addressed the challenge of prioritizing and focussing retrofit efforts on critical bridges to maintain transportation network resilience after seismic events. Conventional approaches rely on computationally expensive traffic simulations to assess network performance and identify critical infrastructure. To accelerate this process, researchers developed a “simulation-free” surrogate model combining Markov random walk and random forest methods, along with a novel bridge ranking framework integrating One-at-a-time, Sobol index, and Gini importance measures. Applied to the San Francisco Bay Area highway network, the surrogate model achieved a 98% reduction in computation time—approximately 400 times faster than traditional agent-based traffic flow models—while maintaining high predictive accuracy (R² = 0.95, MAPE = 1.62%). This efficiency significantly reduces the need for extensive training datasets, enabling more rapid assessment of network vulnerabilities across numerous earthquake scenarios.

Leveraging NHERI SimCenter’s R2D tool to generate seismic scenarios in the Bay Area region, the research team successfully demonstrated their model’s ability to identify critical bridges whose retrofit would most improve network performance after earthquake events. By streamlining the process of seismic risk evaluation for large-scale transportation systems, the study provides decision-makers with a fast and practical method for prioritizing infrastructure investments. Beyond technical advancements, the approach offers tangible societal benefits: enabling more effective use of limited resources, improving community resilience through targeted retrofitting, and supporting faster recovery of mobility in earthquake-prone regions.