Among the different transportation modes, air transportation has seen the fastest growth over the last century, and is expected to keep growing. This fact, combined with the difficulty of upgrading the relevant infrastructure, gives rise to socio-economic issues as well as complex dynamical phenomena that are interesting beyond the (however important) scope of revenue maximization. Here, we will focus on the problem of delay propagation: delay can be transferred from one flight to another when they share aircraft, connecting passengers or crew members, or (indirectly) when they are competing for the use of limited airport capacity.
First we will describe an agent-based, data-driven model we developed to simulate the spreading of delays, and briefly discuss the process of validation and the datasets used. Afterwards, we will show one application by studying the interaction between the European and US airport networks through an analysis of the impact of intercontinental flights for several daily schedules. Such impact is defined by setting the initial delay of a single flight to a large amount, letting it propagate through the system, and recording the resulting delay propagation tree. We find that flights going from the US to Europe have higher potential to create disruption that the ones going in the opposite direction, and that tight airline schedules can result in large trees even if the destination of the impacting flight is not a particularly large hub.