Modeling metro passenger flow dynamics during disruptions considering endogenous crowding in multimodal transportation systems

Urban rail transit (URT) disruptions propagate cascading effects across interdependent multimodal systems, yet existing research inadequately captures the dynamic coupling between risk-sensitive passenger behavior and evolving system status, leading to biased evaluations. To bridge this gap, we propose the Time-Varying Behavior-Status Iterative Coupling (TVBSIC) scheme, which integrates three main components: (1) a capacity-constrained space–time network that integrates metro, bus, ride-hailing, and bikeshare services; (2) incorporating a Cumulative Prospect Theory (CPT)-based behavioral model and schedule-based assignment to formalize risk perceptions and quantify choices across mode, path, and departure-time within system dynamics; and (3) a heuristic iterative optimization to capture the endogenous co-evolution of behavior and flow. Application to Guangzhou Metro power outage disruptions reveals critical adaptation patterns: pre-trip travelers exhibit a 7.5% higher mode-switching rate than mid-journey passengers, and severe platform overcrowding induces cross-modal congestion transfers. A pattern of persistent metro reliance coexists with flexible multimodal integration. Multimodal options dominate short-distance travel, while metro reliance intensifies over longer distances and during service restoration. Multimodal integration reduces peak load rates by 11.53% compared to metro-only scenarios. Comparative experiments demonstrate that TVBSIC reduces total generalized travel cost by up to 11.3% and reproduces minute-level passenger flows with a validation MAPE of 9.4%. By bridging behavior-flow decoupling common in existing models, TVBSIC supports transit agencies in evaluating disruption management strategies given passenger adaptation patterns and network vulnerabilities.