A novel causal-Bayesian network modelling approach for urban heat vulnerability-mitigation assessment and forecasting

Source abstract

Heat vulnerability refers to the propensity of individuals and communities to be adversely affected by urban heat, shaped by exposure, built environment, and socio-economic conditions. As heatwaves intensify under climate change, understanding not only who is vulnerable but how vulnerability may change under different mitigation strategies has become critical for safeguarding public health, energy performance, and economic productivity. However, most existing studies rely on static, index-based assessments and lack the capacity to robustly forecast how vulnerability changes under interventions while representing the complex interactions across urban systems. To address this gap, this study proposes a causal-Bayesian network modelling approach to assess urban heat vulnerability and forecast its changes under alternative mitigation strategies. The framework integrates structured causal knowledge with probabilistic modelling through three core processes: causality-informed structure construction, hybrid parameter learning, and vulnerability assessment and forecasting through probability propagation. A neighbourhood-scale case study demonstrates the systematic workflow and confirms its practical feasibility using profile-based datasets. By linking real-world data, probabilistic inference, and expert-informed causal reasoning, the proposed modelling approach enables robust forecasting of vulnerability reduction by interventions and uncovers interconnected, cross-sector dynamics. This integrated modelling approach offers a practical and scalable analytical tool for forward-looking, policy-relevant urban planning in the context of urban overheating.