Urban Digital Twin in Daily Fire Risk Prediction: A Group Consensus Model Based on Social Network Hybrid Opinion Dynamics Using Spatial-Temporal Satellite Remote Sensing Data

Urban fire prevention, as a critical global challenge, demands breakthroughs in the daily risk prediction to address the urgent needs of proactive mitigation. However, existing approaches face three bottlenecks: the uncertainty of risk, the real-time coordination of multi-city fire data, and the integration of heterogeneous urban perspectives. To overcome these challenges, which are conceptualized as a group consensus problem, this paper proposes a digital twin (DT)-driven large-scale group decision-making (LSGDM) framework. The framework integrates four interconnected logical loops: First, a DT model for urban fires is developed using spatial-temporal satellite remote sensing data (STSRSD), enabling dynamic fire risk simulations and STSRSD fusion. Second, an improved Leiden algorithm incorporating an social influence index (SII) and trust thresholds is designed to cluster cities by risk characteristics, reducing inter-cluster heterogeneity. Third, decision-making roles (leaders/followers) are assigned based on hesitancy levels and SII metrics, with a bidirectional communication mechanism amplifying the influence of key cities. Fourth, this paper integrates the strengths of the social network DeGroot (SNDG) model and three-dimensional (3D) Deffuant-Weisbuch (DW) dynamics to propose a hybrid opinion dynamics (HOD) with a two-stage confidence and collaboration-based consensus reaching process (CC-CRP), termed HOD-CC-CRP. Finally, the feasibility of the paper is verified through a case study of selected urban fires in China.