Characterization of Resilience of Networks to Uncertain Disruptions: A Case Study of Houston Road Network during Hurricane Harvey

Existing research that focused on dynamic changes in the networks either studied the phenomenon on theoretical random networks or assumed the disruptions or failures in the real-life networks are completely random, which is hardly the case. Motivated by this observation, this paper has two interwoven objectives: On one hand, the vulnerability of road network under different types of targeted disruptions is characterized; on the other hand, a vulnerability measure based on the expected size of the giant components of the network under an uncertain disruptive event has been proposed. In order to achieve these objectives, first, the road network is represented as planar graphs and targeted disruption percolation patterns were simulated based on the values of the node-significance metrics. Second, the probability distribution of the giant component under uncertain disruptive events is modeled using the co-location index of the road network with flood control infrastructure. Third, percolations at different extents are simulated in order to estimate the giant components of the network under disruptions of different magnitudes, which, together with the probability estimated in the above step, is used to estimate the expected size of the giant connected component of the road network. The proposed method was applied to super neighborhoods in Houston during Hurricane Harvey. Comparisons are made among the vulnerabilities of different geographical locations, which lead to the identification of disaster-prone areas. It is found that the proposed method is capable of identifying areas that would have a lower level of connectivity as a result of a possible flooding event.