Bayesian Network Modeling of Flood Cascade Dynamics and Climate-induced Risks in the Pearl River Delta Urban Agglomeration

Climate change and rapid urbanization are intensifying flood vulnerability in highly urbanized delta regions. This study develops a Bayesian network model to assess the vulnerability and cascading failure dynamics of flood control infrastructure in the Pearl River Delta (PRD) urban agglomeration under future climatic scenarios. By integrating multi-source hydrological data, network topology, and conditional probability inference, the model quantifies systemic flood susceptibility and identifies critical failure nodes within the regional flood network. To evaluate future flood hazards, downscaled climate projections from global climate models (GCMs) and a stochastic weather generator (WeaGETS) were employed to simulate extreme precipitation patterns. Results indicate that in certain areas of the PRD, 100-year return period design rainstorm values may increase by up to 100% by the 2050s, particularly under high-emission scenarios. Spatial analysis reveals that the central and southeastern PRD (including Guangzhou, Shenzhen, and Dongguan) are most susceptible to cascading flood failures due to dense hydrological interconnectivity and topographic constraints. These findings underscore the urgent need for climate-adaptive infrastructure planning, enhanced early-warning systems, and integrated watershed management. By bridging hydrological modeling, climate impact assessment, and network-based vulnerability analysis, this study offers a systematic, data-driven framework to support resilient urban flood governance in deltaic megacities facing compounding environmental risks.