We model how extreme events ripple through infrastructure and communities, and design decisions that make both more resilient.
Natural disasters, cyberattacks and pandemics keep revealing how vulnerable our communities and built environment are. We combine advanced AI, statistical learning and robust optimization to examine the impacts of chronic and acute shocks on sociotechnical systems, develop risk-informed decision models, and identify cost-effective adaptation measures.
A Human-centered, Equity-focused, Risk-informed Decision-making (HERD) framework for critical infrastructure in the wildland-urban interface. Study regions: Mora and Taos Counties, NM, and Larimer County, CO. Workshop photos and publications.
How grid and transportation failures affected socially vulnerable Puerto Rican communities, and how recovery was prioritized. Led with Dr. Diana Ramirez-Rios. Photos, video and publications.
A synthesis of the challenges and opportunities of integrating renewable energy into the grid, with policy recommendations. OASIS UB was one of eight winning teams. Poster and outputs.
Resilient Engineering through Knowledge Integration for NetworkeD Lifelines and Ecosystems. UB leads this collaboration with UC Berkeley, Northeastern University and Pacific Northwest National Laboratory. Project details.
Wildfires can set off debris flows and flash floods that cut roads and interrupt power and water for communities already struggling to recover. REKINDLE (Resilient Engineering through Knowledge Integration for NetworkeD Lifelines and Ecosystems) will develop AI-powered tools that connect these hazards, the infrastructure they disrupt and the people affected, so decision-makers can plan for them together.
GeoAI and physics-based models of wildfire and the debris flows and flash floods that follow, with uncertainty bounds.
Graph machine learning and network science for cascading failures across power, water and transportation.
Household surveys, focus groups and serious games to measure the social burden of lifeline disruptions.
Multi-level optimization, scaled with AI, for pre-fire hardening and post-fire recovery.
A Human-centered, Equity-focused, Risk-informed Decision-making (HERD) framework that helps emergency managers and infrastructure operators plan wildfire preparedness and recovery around the needs of socially vulnerable communities in the wildland-urban interface.
Cascading failure of interdependent power, gas, water and road networks under incomplete information.
A Capabilities Approach metric for how service disruptions affect households, from surveys and focus groups.
Bilevel optimization to coordinate recovery and allocate funds across independent stakeholders.
Decision models for collaboration among agencies, utilities and communities.
How did failures of the electric grid and the road network affect socially vulnerable communities after Hurricane Fiona, and how were the recovery and reconstruction of these interdependent systems prioritized?
PI: Dr. Diana G. Ramirez-Rios. Co-PI: Dr. Sayanti Mukherjee.
Researchers: Poulomee Roy, Prasangsha Ganguly, Angelo Soto-Vergel, Judith Brennan, Huize Hou and Esneyder Gonzalez Ponzon.
Collaborators in Puerto Rico: Arq. Astrid Díaz and Carmen Villanueva (PRODEV).
In-depth interviews with emergency management, utility and municipal stakeholders, and focus groups with residents, analyzed with a hybrid qualitative approach that pairs thematic analysis with AI and natural language processing.
The Science Synthesis Prize, launched in May 2024 as part of DOE’s Grid Resource Integration Technology (GRIT) prize series, asked teams to map the state of renewable energy integration and identify the key barriers and investment opportunities. Team OASIS UB was one of eight winners, each awarded $10,000.
A systematic literature review of solar, wind, geothermal and hydrogen integration: current U.S. integration practices, the main barriers (intermittency, integration cost, storage, forecasting and aging grid resilience), the opportunities that address them, and policy recommendations grounded in those opportunities.
The poster extends this to decentralized grids, matching AI methods such as deep forecasting, reinforcement learning, anomaly detection and spatiotemporal models to each challenge, with real case studies like the Texas 2021 freeze and California microgrids.
Lifelines such as power, gas, water and transportation depend on each other, yet topology and interdependency data are often missing. We use simulation, Bayesian networks, generative AI and predictive analytics to assess vulnerability under incomplete information.
Extreme heat, cold, storms and drought threaten the energy system. We use statistical learning, GeoAI, remote sensing and deep learning to quantify risks and support planning and policy.
Wildfires, heat waves, storms and flooding affect health, infrastructure and communities. We use GeoAI, remote sensing, interpretable ML and econometrics to design interventions.
Data-centric models for mortality and morbidity risk, disease spread and mental health, informing health policy and resource allocation.
Interpretable AI and optimization to understand the geographic and socioeconomic context of crime and support fair allocation of public safety resources.
GPS, phone, social media and transit data reveal how people move before, during and after disasters, and who loses access to critical services.