Northeastern Joins $2M NSF-Funded Effort to Build Resilience Against Wildfire and Its Cascading Aftershocks
Multi-institution team will develop new models linking wildfire risk, infrastructure failure, and community impact in the wildland-urban interface
A team of researchers spanning four institutions has been awarded an approximately $2 million grant from the National Science Foundation’s Fire Science Innovations through Research and Education (FIRE) program to develop new tools for understanding and preparing for the cascading disasters that wildfires can unleash on communities and infrastructure.
The three-year project, titled “Collaborative Research: FIRE-WUI: REKINDLE: Resilient Engineering through Knowledge Integration for NetworkeD Lifelines and Ecosystems,” brings together investigators from the University at Buffalo (lead institution), the University of California, Berkeley, Northeastern University, and the U.S. Department of Energy’s Pacific Northwest National Laboratory (PNNL). The award runs from September 2026 through August 2029.
Northeastern’s share of the funding, led by Professor Auroop R. Ganguly, is approximately $419,000.
Addressing a Gap in Wildfire Resilience Planning
Wildfires in the wildland-urban interface, or WUI, increasingly trigger secondary disasters—debris flows, flash floods, and other compound extremes—that cascade through power grids, transportation networks, water systems, and the communities that depend on them. Existing models tend to treat these hazards in isolation, missing the cross-scale interactions that drive real-world disaster impacts.
The REKINDLE project aims to close that gap by building a new class of models that formally couple wildfire and compound-hazard risk, interdependent lifeline infrastructure, and human adaptive behavior within a single, decision-oriented framework. The research team describes the approach as a way to move wildfire resilience planning “from reactive, fragmented, sequential analysis into a proactive, unified, and decision-oriented system.”
The work is organized around four interconnected modules. The first is a physics-informed Geo-AI model that combines multiscale Earth observation data with a Wildfire Knowledge Graph to generate probabilistic wildfire and compound-hazard risk maps. The second embeds that hazard uncertainty into a multilayer graph machine-learning model that simulates how failures cascade across interdependent infrastructure systems—power, transportation, and water—during compound hazard events. The third introduces a social burden metric, informed by household surveys, focus groups, and simulation games, that captures how disruptions to lifelines constrain households’ livelihoods and capacity to recover in ways standard infrastructure metrics do not. The fourth ties the previous three together through a bi-level optimization model, supported by an LLM-based heuristic pruning framework, to identify optimal strategies for hardening infrastructure before a fire and guiding recovery afterward, while accounting for budget, time, and social costs.
The research will draw on data-rich case studies from the Montecito and Los Angeles regions of California, developed in collaboration with government agencies, utilities, and community stakeholders.
A Multi-Institution, Multidisciplinary Team
The project is led by Professor Sayanti Mukherjee at the University at Buffalo, working alongside Professor Susan Clark, also at Buffalo. At UC Berkeley, the team includes Professor Marta Gonzalez—who has longstanding research ties to Northeastern—and Dr. Anna Serra-Llobet. PNNL’s contributors are Dr. Andre Coleman and Dr. Sam Chatterjee, who also holds a joint faculty appointment at Northeastern’s Department of Civil and Environmental Engineering.
At Northeastern, the project is led by Professor Auroop R. Ganguly, who directs the Sustainability and Data Sciences Laboratory (SDS Lab) in the Department of Civil and Environmental Engineering and leads the AI for Climate and Sustainability (AI4CaS) focus area within the university’s Institute for Experiential AI (IEAI). Ganguly’s current and former postdoctoral researchers, Dr. Somnath Mondal and Dr. Rachindra Mawalagedara, contributed to development of the proposal—experience that also gave them hands-on exposure to building mid-sized, multi-institutional research collaborations.
Broader Impact
Beyond its technical contributions, the project is expected to feed into the Wildfire Risk & Resilience Program at PNNL and inform how emergency managers, urban planners, and infrastructure managers design community-specific wildfire preparedness and recovery strategies. The team also plans to recruit graduate and undergraduate researchers, run summer research camps for middle and high school students, and develop open, publicly available educational materials on compound hazard risk.
ABSTRACT SOURCE: NSF
Wildfires and their cascading impacts, such as debris flows and flash floods, pose growing threats to communities in the Wildland-Urban Interface (WUI), disrupting critical lifeline infrastructures and amplifying risks to public safety and well-being. This project addresses a national priority by developing a unified, decision-oriented framework that integrates wildfire and compound hazard modeling, infrastructure interdependency analysis, and human adaptive behavior. This approach enables emergency managers, planners, and infrastructure operators to anticipate, prepare for, and recover from complex wildfire-driven disasters, ultimately reducing societal costs and enhancing community resilience. This project has an effective workforce training strategy of providing educational opportunities to learners at several stages: middle school students, high school students, undergraduate students, and graduate students. Internships will be provided with partners in industry and national labs. The project also develops multidisciplinary education modules and tutorials on compound hazard risks.
This project advances the science of wildfire resilience by coupling environmental, technological, and social systems within a single, uncertainty-aware model. The project introduces a physics-informed geospatial artificial intelligence (GeoAI) module that fuses Earth observation data with a wildfire knowledge graph to generate probabilistic risk maps for wildfires and related hazards. A multilayer graph machine learning network simulates cascading failures across interdependent lifeline systems, such as power, water, and transportation, under compound hazard scenarios. The project also captures how service disruptions affect community capabilities related to wildland fire response. A bi-level optimization model identifies robust strategies for infrastructure hardening and post-fire recovery, balancing technical performance with economic needs. The research is grounded in robust case studies and co-developed with government, utility, and community stakeholders through a structured engagement process. By integrating artificial intelligence, machine learning, and stakeholder input, this project transforms wildfire resilience planning from fragmented, reactive analysis into a proactive, unified system that supports national efforts to protect lives, infrastructure, and communities.
Related NSF FIRE stories: Liu to Lead New $1.2M Collaborative NSF Grant on Wildfire Emissions and Interdisciplinary Team Receives NSF FIRE Grant to Study FireScreens for Wildfire Safety