A dynamic modeling framework for metro network resilience assessment and improvement under extreme rainstorm disasters

Abstract Extreme rainstorms can severely disrupt metro systems through station inundation, passenger-flow redistribution, secondary overloads, and prolonged service disruption. Despite advances in metro vulnerability and resilience research, hazard exposure, direct disruption, passenger rerouting, and load-induced degradation are often modeled separately, while resilience assessment rarely integrates infrastructure availability, transport efficiency, and satisfied passenger demand. Recovery is also commonly treated as an external schedule rather than a feedback process, although station reopening may redistribute flows and trigger renewed overload. To address these limitations, this study develops a dynamic modeling framework for metro network resilience assessment and improvement across the disruption–recovery process. The metro system is represented as a time-dependent weighted network integrating rainstorm hazard, rainfall-triggered station failures, passenger rerouting, flow-induced cascading degradation, network-state evolution, performance evaluation, and recovery intervention. System performance is measured using a composite indicator of passenger service ratio, network efficiency, and station availability, while resilience is characterized through stage-based capacities and cumulative performance loss. For resilience improvement, resource-constrained recovery decisions are formulated, and a stability-screened performance-gain strategy that prioritizes stations by marginal benefit is compared with a no-intervention reference scenario. An illustrative application to a simplified Shenzhen Metro subnetwork under a normalized rainstorm scenario reproduces the progression from direct failures to cascading disruption, performance deterioration, and recovery. Sensitivity analyses further show consistent and interpretable responses to variations in station-capacity buffers, direct-failure thresholds, and recovery budgets. The framework provides a computationally tractable and transferable basis for dynamic resilience analysis, and practical recovery planning in rainstorm-exposed metro systems.