Delay propagation patterns in Japan's domestic air transport network

TL;DR

This study analyzes recurrent delay propagation patterns in Japan's domestic air transport network in 2019, revealing structured groups, significant motifs, and seasonal influences that could improve delay prediction.

Contribution

It introduces a novel integration of delay causality networks with temporal analysis to classify delay propagation patterns and identify their structural characteristics.

Findings

Delay patterns form distinct groups with different delay impacts.

Certain network motifs are significantly over- or under-represented in delay networks.

Delay propagation varies seasonally and geographically in Japan.

Abstract

We experience air traffic delays every day, but are there any recurrent patterns in these delays? In this study, we investigate the recurrence of delay propagation patterns in Japan's domestic air transport network in 2019 by integrating delay causality networks and temporal network analysis. Additionally, we examine characteristics unique to delay propagation by comparing delay causality networks with corresponding randomized networks generated by a directed configuration model. As a result, we found that the structure of the delay propagation patterns can be classified into several groups. The identified groups exhibit statistically significant differences in total delay time and average out-degree, with different airports playing central roles in spreading delays. The results also suggest that some delay propagation patterns are particularly prominent during specific times of the year, which could be influenced by Japan's seasonal and geographical factors. Moreover, we discovered that specific network motifs appear significantly more (or less) frequently in delay causality networks than their corresponding randomized counterparts. This characteristic is particularly pronounced in groups with more significant delays. These results suggest that delays propagate following specific directional patterns, which could significantly contribute to predicting air traffic delays. We expect the present study to trigger further research on recurrent and non-recurrent natures of air traffic delay propagation.