Trees of Reactionary Delay: Addressing the Dynamical Robustness of the US Air Transportation Network

Recent works in the area of Complex Systems have addressed the robustness of networks such as power grids, social groups and the Internet. The robustness is evaluated against an external perturbation that can be different in nature depending on the particular network. For instance, the failure of power transmission lines that can trigger a nationwide blackout or a general shutdown of routers and the consequent connectivity loss. In this work, we introduce metrics inspired by Complexity Science to explore the robustness of the air transportation system in the US with respect to delay propagation. We use an agent-based model recently developed to simulate delay propagation and assess the effect of disruptions in the network. These disruptions are introduced as initial conditions and can affect single flights or full airports. The model is then run with and without disruptions and the outcome is compared to quantify the system robustness. Our results indicate that large hubs (in the sense of number of offered destinations) are more vulnerable to flight delays than small or medium sized airports. However, the impact in the whole network of delays initiated in an airport does not depend on whether it is a hub or not. We also detect a set of high impact flights and explore the drivers that generate these long tail extreme events.