Air travel congestion can quickly spread from a few cities to a whole network.
Researchers studying air travel congestion have typically focused on a handful of hubs, those problem airports that seem to routinely struggle getting flights on and off the ground on time (we’re looking at you, Newark and JFK). Air travel congestion, however, is really more a phenomenon built on networks of airports than any individual one. In fact, it may be most useful to think of flight delays spreading across a region of the United States in the same way that disease travels through our connections to each other.
"Our approach was more from epidemiology, from epidemics spreading," says Pablo Fleurquin, a researcher with the Spanish National Research Council. He and several colleagues recently published a study in the journal Scientific Reports analyzing air travel congestion in more than 8 million domestic flights in the United States in 2010.
That year, according to Airline On-Time Performance Data from the Bureau of Transportation Statistics, 37.5 percent of all flights were delayed. They generally followed a similar pattern: a large number of the delays were quite small, while a small number of them lasted for hours. These three charts illustrate that pattern in average delays (A), in summer and winter (B) and at the three airports of Atlanta, JFK and Honolulu (C):
The average delay in 2010 was 29 minutes. The researchers then took that figure to assess individual airports for their overall congestion. On any given day, an airport with an average delay among all of its flights of more than 29 minutes was considered "congested" in the study. And the picture of congestion varied wildly from day to day. Here are three maps of the U.S. airport network from April 4th, March 9th and March 12th.
The map at the right shows a massive cluster of congestion, illustrated by all of airports in red dots connected through direct flights (the middle map shows a day with moderate congestion, with a few airports in orange that have not amassed into a kind of cluster). "As can be seen," the researchers write, "the scenario dramatically changes from day to day: in some days a large cluster surges covering 1/3 of all airports, while in others only one or two airports cluster together."
March 12th was not, in fact, a day with some crazy weather phenomenon. So what was going on then? "When we analyze the cluster," Fleurquin says, "what we saw is that it’s not that it starts in only one airport."
Rather, the really problematic congestion more likely results from delays that occur in multiple airports connected to each other. Those delays then merge into a kind of mega-congestion as passengers and crew members fail to make connections through the system. This means that there are primary delays and reactionary delays.
"The way we see this problem, primary delays could be anything: a broken aircraft or some meteorological stuff, strikes in some airports," Fleurquin says. "These primary delays are propagated throughout the network and are sometimes magnified by the connectivity."
Atlanta, for instance, is a highly connected airport. It has flights to and from dozens of cities, and it has dozens of flights a day connecting just to New York. If you have a plane malfunction in Atlanta and a snowstorm in New York, those problems proliferate from there to Washington and Philadelphia.
"The thing here that we realized – and this is related to other modeling approaches for disease spreading – is that what is important here in this case in flight delays is the connectivity," Fleurquin says.
Connectivity, however, is also the very thing that gives us a robust air travel network in the first place. Connectivity creates redundancies (so you can still get to Chicago from San Francisco, even if your connecting airport in Denver is shut down in a blizzard). Connectivity gives us options (so you can get from St. Louis to Kansas City, Washington or Atlanta at any time of day). But the downside of all of this connectivity is that problems travel throughout the network just as easily as passengers do. And this is a pattern, Fleurquin adds, that applies equally to our local transit systems and our international supply chains.