Abstract
In the context of a rapidly changing climate, understanding and addressing the unequal distribution of flood risk is critical for effective disaster risk reduction (DRR) and equitable policy design. This research presents a novel, data-driven approach to mapping social vulnerability to floods across Canada using Bayesian machine learning. Leveraging national census data from 2016 and 2021 at the dissemination area level, this study develops Canada’s first Bayesian Neural Network (BNN)-based Social Vulnerability Index model. The JBA 2020 Canada flood map was used to assess flood risk. By applying probabilistic modeling, the approach captures uncertainty in vulnerability assessments and offers improved accuracy in identifying neighborhood-level risk profiles. Monte Carlo simulations and mutual information measures were used to identify and quantify the influence of socioeconomic, demographic, and housing-related indicators on vulnerability. The model results reveal spatial patterns of heightened vulnerability across urban and rural areas, where recent immigrants, low-income populations, racially marginalized groups, and seniors are disproportionately exposed to flood risks. This nuanced understanding of vulnerability enables policymakers and emergency professionals to prioritize interventions and allocate resources more effectively. The findings highlight the potential of integrating machine learning with national-scale demographic data to inform inclusive climate adaptation strategies. By advancing a probabilistic, neighborhood-level understanding of social vulnerability, this research supports the development of targeted policy tools for DRR, emergency management, and climate resilience planning in Canada. It also demonstrates how a social justice lens, grounded in environmental equity, can be operationalized in the design and implementation of flood risk reduction programs.
Presenters
Liton ChakrabortyAdjunct Professor, Disaster and Emergency Management, York University, Ontario, Canada Dan Henstra
Hirsa Taherimashhadi
Researcher, University of Tehran
Details
Presentation Type
Paper Presentation in a Themed Session
Theme
2026 Special Focus—Unseen Unsustainability: Addressing Hidden Risks to Long-Term Wellbeing for All
KEYWORDS
Social Vulnerability, Bayesian Neural Networks, Flood Risk Assessment, Environmental Justice
