Extreme weather events (EWEs) like typhoons increasingly disrupt metro systems, threatening urban mobility with widespread travel delays. However, existing studies often oversimplify delay mechanisms and lack context-specific insights. To disentangle the complex impacts of EWEs, we propose and validate a novel analytical framework that quantitatively links macro-scale passenger delays to micro-scale facility performance. The framework integrates two core diagnostic modules: a data-driven attribution method to identify critical facility-level delay sources, and a causally-informed deep learning model to robustly pinpoint the root causes of delay formation. This approach enables the detection of actionable bottlenecks, reveals opportunities for proactive demand management, and infers the dominant impact pathways through which EWEs degrade various facilities. Its strong performance and computational efficiency support both post-hoc diagnosis of past EWEs and near-real-time operational decision-making for ongoing EWEs. Validated on the Shenzhen Metro, this study provides a robust foundation for enhancing system resilience through targeted interventions.
Li et al. (2026) studied this question.