A probabilistic performance-based framework for heat vulnerability and risk assessment of buildings

Abstract

As climate-induced hazards increase rapidly, the built environment's limited preparedness highlights the urgent need to enhance resilience. Although risk assessment frameworks inform resilient design decisions for many hazards, heat stress at the building scale are often addressed using simplified code-based approaches. Emerging heat fragility models provide a basis for risk quantification; however, a unified heat risk framework that estimates multi-domain consequences (social, economic, environmental) accounting for multi-factor uncertainties (climate, building, occupant) is still lacking. To address this gap, this paper proposes a probabilistic method grounded in performance-based engineering principles. The approach integrates hazard analysis, building performance evaluation, fragility modeling and loss estimation, and employs a Monte Carlo framework to propagate uncertainty across all stages. Its step-by-step implementation is demonstrated on a multi-story building under varying design conditions and climate scenarios, proving the framework's ability to quantify probable maximum losses and incorporate them into risk matrices. The case study results show energy demand and carbon costs increasing by around 13% under high-warming scenarios, with heat-related mortality nearly tripling in naturally ventilated conditions, enabling building-level comparison across performance thresholds and hazard severities. Probabilistic loss functions translate these impacts into expected annual losses, further highlighting the importance of passive survivability, as cost and carbon metrics are projected to increase by 50% while mortality risk rises sharply for the analyzed building. These annualized losses can inform building-level design decisions and multi-hazard resilience planning, as the proposed approach aligns with probabilistic models and risk metrics used in catastrophe modeling to compare natural and climate hazards.