Local and Network-Wide Time Scales of Delay Propagation in Air Transport: A Granger Causality Approach

Complex network theory, in conjunction with metrics able to detect causality relationships from time series, has recently emerged as an effective and intuitive way of studying delay propagation in air transport. One important step in such analysis is converting the discrete set of landing events into a time series representing the average delay evolution. Most works have hitherto focused on fixed-size windows, whose size is defined based on a priori considerations. Here, we show that an optimal airport-dependent window size, which allows maximising the number of detected causality relationships, can be calculated. We further show how the macro-scale but not the micro-scale structure is modified by such a choice and how airport centrality, and hence its importance in the propagation process, is strongly affected. We finally discuss the implications of these results in terms of detecting the characteristic time scales of delay propagation.

Automated summary

Complex network theory, along with metrics for detecting causality relationships from time series, has been increasingly used to study delay propagation in air transport. To convert landing events into time series, most previous works have focused on fixed-size windows. However, we demonstrate that an optimal airport-specific window size can be calculated to maximize the number of detected causality relationships. We also show that this choice affects the macro-scale structure and airport centrality, highlighting its importance in delay propagation.