Modeling urban rail transit system resilience under natural disasters: A two-layer network framework based on link flow

Service disruption of urban rail transit systems in natural disasters induces direct and simultaneous impacts (e.g., delay of travel) that may propagate to indirect and long-term impacts (e.g., economic losses). Models that can identify the critical components of infrastructures, if with proper granularity, will help to form actionable strategies for infrastructure protection and resilience. To better understand the system response under service disruption and followed recovery processes, a purely data-driven approach is proposed to model the system's absorptive and adaptive resilience. To do so, the urban rail transit system is modeled as a customized two-layer network to distinguish its infrastructure layer and service layer. A new localized measure, i.e., link flow, is suggested to construct the system functionality in attack-repair scenarios. It is estimated based on recorded smart card data. Applied to the Hangzhou Metro system, service restoration relying on the new metric is 2 similar to 3 times faster and more robust (i.e., smaller normalized standard deviation) than that relying on the common metric (i.e., recovery cost) under a portfolio of attacks tested with varying severity. Stably high correlations (i.e., 0.65 to 0.8) are observed between the two set of rank indices, i.e., of infrastructure topology-based importance and of flow based importance. Furthermore, the system response to attacks with the estimated link flow is consistent with simulated ones with known and exact link flow.