Rating community resilience with a deep learning framework

Gaby Clark

scientific editor

Meet our editorial team
Behind our editorial process

Robert Egan

associate editor

Meet our editorial team
Behind our editorial process
Editors' notes

This article has been reviewed according to Science X's editorial process and policies. Editors have highlighted the following attributes while ensuring the content's credibility:

fact-checked

trusted source

proofread

The GIST
Add as preferred source

An understanding of community resilience and risk analysis is vital when it comes to protecting civilians and infrastructure from natural hazards, such as hurricanes or earthquakes. Artificial intelligence is an efficient way to rate a community's resilience and vulnerability. However, factors that affect community resilience—like power grids and communications systems—are ranked independently by current assessment methods based on the assumed impact of each factor.

Researchers from the Urban Resilience AI Lab at Texas A&M University have developed a new deep learning framework, coined Resili-Net, that examines the interdependence of factors affecting community resilience, providing a more accurate rating of community vulnerability. The findings are published in the journal npj Urban Sustainability.

Resili-Net allows stakeholders to examine the interactions between infrastructure, social conditions and environmental risks. Current assessment methods have a bias on what factors have larger impacts on community resilience, giving high weight to factors like proximity to hospitals and floodplain exposures. Resili-Net uses data to examine the interdependency of these factors, meaning that the same factors can have different impacts across various communities.

"When it comes to community resilience, one solution does not fit all," said Dr. Bo Li, a postdoctoral researcher in the Urban Resilience AI Lab and co-lead investigator of this study. "By acknowledging the interdependence of factors affecting risk, this new model has the potential to change the way cities plan for floods, storms and other natural hazards."

This idea was supported through analysis of simulated scenarios. Researchers found that adding the same improvement—like an additional hospital—to a community does not always lead to an increase in community resilience. This finding highlights the need for targeted, community-specific strategies for improving resilience.

Communities may appear resilient on paper, but function differently under real-world stress. When using Resili-Net to more accurately understand community resilience and risk, researchers discovered communities with high-risk ratings, and low resilience ratings. These communities are considered vulnerable and can use the data provided to create an optimized improvement plan based on the factors most affecting the specific communities.

"Certain areas, some in the Greater Houston region, have an overlap of high flood exposure and low recovery capacity. These communities may face severe impacts, yet lack the ability to adapt or recover quickly," said Kai Yin, a Ph.D. student in the Urban Resilience AI Lab and co-lead investigator. "Urban planners could use Resili-Net to prioritize investment and support in these at-risk areas."

While researchers have not yet partnered directly with agencies, the research offers a blueprint for smarter infrastructure development. The model supports combined resilience and risk analysis, enabling more holistic planning that considers socio-technical factors.

Future plans for Resili-Net include expanding the framework to additional hazards, like wildfires, and integrating ecological data to create an even more well-rounded model of community resilience.

Discover the latest in science, tech, and space with over 100,000 subscribers who rely on Phys.org for daily insights. Sign up for our free newsletter and get updates on breakthroughs, innovations, and research that matter—daily or weekly.

Subscribe

Researchers hope to integrate multi-agent systems into the framework, allowing for automated assessments and updated scenarios in real time. This is a current challenge for researchers due to the lack of data availability.

"Important features affecting risk and resiliency often cannot be included due to lack of access, limiting the depth of analysis," said Dr. Ali Mostafavi, director of the Urban Resilience AI Lab and civil and environmental engineering professor. "Future work would benefit from greater data transparency and sharing."

AI-driven tools are largely beneficial when it comes to improving community resilience against disasters. Improved models like Resili-Net will allow more factors to be considered and data to unbiasedly determine what factors matter most.

More information

Kai Yin et al, Deep learning-driven community resilience rating based on intertwined socio-technical systems features, npj Urban Sustainability (2026). DOI: 10.1038/s42949-026-00364-7

Key concepts

Complex systemsSocial systemsArtificial intelligence

Provided by Texas A&M University