From Environmental Simulators to Building-Scale Risk Maps: A Digital Twin Framework for Multi-Hazard Risk Assessment

Abstract

Risk maps are traditionally constructed from historical observations of hazards and impacts. Yet such records are often sparse, incomplete or poorly representative of future conditions, limiting their usefulness for assessing rare or emerging events. This paper proposes a Digital-Twin-based methodological framework for multi-hazard risk analysis that generates synthetic, forward-looking hazard scenarios and translates them into building-scale risk indicators. The framework combines physics-based simulators for atmospheric pollution, wildfire spread and surface flooding with scenario generation informed by environmental drivers such as wind regimes, ignition locations, rainfall intensities and soil parameters. These scenarios are propagated through a probabilistic workflow to derive empirical distributions, exceedance probabilities, tail-oriented indicators and spatial risk maps. Three case studies demonstrate how heterogeneous hazard models can be organised within a common pipeline. Pollution simulations combine deterministic advection with wind-rose frequencies to produce weighted concentration indicators; wildfire simulations use a stochastic cellular-automaton model to estimate burn probabilities; and flood simulations use a rainfall--soil design of experiments to evaluate maximum water-depth exceedance. The results show that synthetic yet physically grounded scenarios can reveal low-probability, high-impact situations that may be absent from historical records. The framework provides a transparent basis for decision support, urban planning and longer-term risk-informed assessment.