Flight delays are the bane of any traveller. They also have an economic impact, an estimated of $40 billion per year in the US alone, according to the 2008 Report of the Congress Joint Economic Committee. So a better understanding of the nature of flight delays is surely of great interest. 

Today, Pablo Fleurquin at the Institute for Cross-Disciplinary Physics and Complex Systems in Spain, and a couple of pals, reveal a unique approach to this problem that shows how flight delays spread across the US.

Fleurquin and co begin by thinking about the air transportation system as a network in which airports are nodes and the flights between them edges. This approach has been hugely successful in simulating and predicting the way that passengers, goods and even diseases spread around the world. This kind of network approach also explains how other phenomena spread, such as wildfires.

So Fleurquin and co examine how delays spread in exactly the same way and use the same kind of simulation to show exactly this happens.

Their simulation compares scheduled and actual departure times for over 6 million flights in 2010 downloaded from the US Bureau of Transport Statistics.  The data includes flights from 18 carriers operating from over 300 airports.  

The model uses this behaviour to track the arrival and departure of each aircraft throughout the year at a resolution of 1 minute, which allows Fleurquin and co to see exactly how delays propagate across the US.  

The results make for interesting reading. The model shows how congestion problems are generally reset each evening when flights stop for the night.  It also shows how local problems such as weather disruptions and labour disputes tend to be confined to a small number of airports.

However, when conditions are ripe, delays can propagate right across America. Fleurquin and co conclude that three main mechanisms lie behind this spread.

The first is through each plane’s flight schedule from one destination to the next–a delay in one leg naturally propagates to the next.

The second comes from passenger and crew connections. Clearly, an aircraft cannot take off if the crew are delayed on another flight.

Finally, there is airport congestion. A given airport can handle only a certain rate of take offs and landings and becomes congested when numbers rise beyond these levels. 

However, Fleurquin and co conclude one of these factors is far more damaging than the others. “Our simulations evidence that passenger and crew connections is the most effective single mechanism to induce network congestion,” they say.

That’s an interesting result that should allow more effective planning. In future, the model could be used to test the robustness of a particular daily schedule to delay propagation and to evaluate different configurations of the network.

If that could prevent just a small fraction of the delays that sometimes spread like wildfire across the air transportation network, the many millions of travellers who benefit would have Fleurquin and co to thank.

Ref:  arxiv.org/abs/1301.1136: Systemic Delay Propagation In The US Airport Network