Research
Our research examines how rare, uncertain, and cascading hazards affect the infrastructure systems that cities depend on. We integrate engineering mechanics, probabilistic modeling, network analysis, decision analysis, and scalable computation to understand how infrastructure failures propagate and to develop strategies that protect lives, maintain access to critical services, and accelerate recovery.
Healthcare Infrastructure & Earthquake Emergency Response
Earthquake emergencies are systems problems: injuries, search and rescue, transportation, hospital operations, and recovery are tightly coupled. We develop models that connect seismic damage to patient movement and healthcare capacity to identify where communities may lose timely access to care and to evaluate strategies that strengthen emergency response.
Our recent work includes the Earthquake Survival Chain and regional-scale modeling of healthcare and transportation disruptions across the San Francisco Bay Area.
Figure description: The map shows the ratio of post- to pre-earthquake travel times to the nearest functional acute-care hospital, with darker areas indicating larger losses in healthcare accessibility. Ceferino et al., Nature Communications (2025).
Transportation Infrastructure & Mobility
We study how infrastructure, information, behavior, and inequality shape mobility during disasters. Our work combines probabilistic risk analysis, network models, urban data, and human behavior to understand whether people can move safely through disrupted cities.
Current research includes hyperlocal flood-sensor prioritization, emergency routing, virtual-reality experiments on flood decision-making, evacuation modeling, and inequality-aware models of commuting and migration.
Figure description: The map shows the relative priority of NYC census tracts for hyperlocal flood-sensor deployment, integrating flood exposure, infrastructure needs, and social vulnerability. Darker areas indicate higher-priority locations. Negri et al., Natural Hazards Review (2025).
Energy Infrastructure
We examine how earthquakes, hurricanes, and compound threats disrupt power systems and how distributed energy resources can strengthen community resilience. Our research spans solar-panel vulnerability, hurricane-driven reductions in solar generation, outage prediction and satellite-based monitoring, distributed energy sharing, cyber-physical risk, and power-system recovery.
We increasingly connect probabilistic risk models with remote sensing, digital twins, and data-driven methods for rapid infrastructure assessment and recovery.
Figure description: Satellite observations are used to infer the spatial distribution of power outages following Hurricane Fiona in Puerto Rico, providing a scalable approach for rapid post-disaster power-system assessment.
Arora, Ceferino & Dobler, npj Natural Hazards (2026).
Housing Infrastructure & Community Resilience
We study housing as a dynamic component of disaster resilience. Our research examines how buildings and communities change over time, how those changes affect vulnerability and life safety, and how households adapt to persistent and emerging hazards.
Current work links incremental construction, seismic vulnerability, collapse and survivability, tsunami and flood hazards, relocation, housing abandonment, and long-term community resilience.
Figure description: Communities with similar physical flood risk can face very different long-term housing outcomes when market conditions are considered.
Negri et al., Earth’s Future (2026).
Computational Methods, Scientific AI & Digital Twins
Across these applications, we develop computational methods that make rare-event infrastructure risk analysis feasible at city and regional scales. Our work spans adaptive importance sampling, uncertainty propagation, scalable regional risk modeling, stochastic optimization, surrogate modeling, Bayesian online learning, and digital twins.
These methods form the computational core of our research program, allowing us to move beyond application-specific simulations toward general approaches for rare-event estimation, uncertainty quantification, and resilience decision-making.
Figure description: Adaptive importance sampling progressively concentrates computation on the scenarios that contribute most to rare-event seismic hazard.
Houng & Ceferino, Bulletin of the Seismological Society of America (2025).
Research Vision: Physics-Guided AI for Engineering Resilience
The next phase of our research moves beyond risk assessment toward methods that directly improve physical infrastructure before and after disasters. We are integrating physics-guided AI, digital twins, engineering mechanics, and decision models to connect infrastructure assessment, repair, retrofit, and reconstruction.
Our goal is to bring together structural engineering, systems engineering, and construction to enable faster, safer, and more cost-effective infrastructure resilience.
Selected Research Support
Our research is supported by federal agencies, industry partners, and UC Berkeley programs advancing infrastructure resilience, computational methods, and disaster risk reduction.
NSF CAREER
Rare-event resilience planning for interdependent infrastructure systems.
NSF MATH-DT
Scalable Bayesian online learning and digital twins for power-network resilience under sequential extreme events.
NSF CLIMA
Climate adaptation of housing infrastructure for resilient coastal communities.
NSF Earthquake Casualty Research
Interdisciplinary modeling of severe earthquake injuries, human-building interactions, and life-safety interventions.
Pacific Gas & Electric Company (PG&E)
Power-system resilience, outage modeling, and post-disaster recovery.
Hellman Fellows Program, UC Berkeley
Support for emerging research directions in disaster resilience and computational modeling.