Non-linear transitions in air transport delays: models and data

The appearance and propagation of delays in air transport are phenomena intuitively connected to the amount of traffic in the system. High traffic volumes imply that airports and airspaces can become saturated, and, as a result, any small perturbation can get amplified and generate a snowball effect. On the contrary, if only a few aircraft were flying over Europe, these would not (could not) interact, and hence delays would not propagate. This is what can intuitively be observed by any passenger and has further been confirmed through the analysis of historical data. Yet, those analyses are mostly based of data representing the normal dynamics of the system, i.e., for relatively homogeneous levels of traffic. Notably, in the last few years, an event has shocked air transport in an unprecedented way: the COVID-19 pandemics has supposed a reduction of traffic never experienced before, but also the possibility of observing the dynamics of the system far from its normal condition. The aim of the present project is to test the previous theory, using a large data set of flights and associated delays spanning from 2015 to the present days. We will analyse those data from a statistical point of view, obtaining models relating the observed average delays with operational variable as traffic volume.