Several weeks ago there were massive flight cancellations on Southwest Airlines and then a while later American Airlines experienced similar widespread cancellations. Immediately people began to speculate if this was a reaction to corporate decisions for mandatory vaccinations – a “blue flu” epidemic as people began to call in sick. Cancellations and very long delays were the norm of two different holiday weekend. And why were only two airlines so largely affected? It is not as though this was the first weather event these two airlines had experienced.

Let’s return to one topic that always fascinates me: chaos.  As mentioned in many other posts, mathematical chaos is not the randomness of a butterfly in China who flutters all the time, but on one occasion, the flutters give way to a tornado in Kansas – how random one might think. Such an event depends on the initial conditions in which the butterfly flutters (…. and why aren’t they called flutterbys?). Given the same initial conditions, the same tornado will predictably appear in Kansas.

The initial conditions have changed for Southwest and American. The workforce for Southwest shrank about 10% from 60,000 to 54,00; American’s workforce fell from 134,000 to 103,000. The job losses were primarily through furloughs and early retirements across pilots, flight attendants and ground crews. The baggage handling ground crews have been especially difficult to bring back and travel increased. Add to this that each airline, in the face of the pandemic impact, increased the number of flights and destinations. Southwest’s expansion was the most aggressive in its 50-year history. As business travel stayed lack luster, each company added non-stop service to vacation friendly destinations in the U.S. and Mexico – including stops far flung from their normal operations area, e.g., Yucatan Peninsula

The initial conditions were dramatically changing. Both airlines had a larger network with fewer pilots. In September the pilot’s union complained that flight operations were not matching pilots, planes, and destinations causing pilots and flight crews to be out of position even when the weather was good.

And then the weather wasn’t.

Two-thirds of cancellations were because there were not adequate flight crews available. Almost all the rest were associated with a lack of available pilots. People were not able to get to work. That might give you pause, but if you have ever chatted with airline folks lots of them “commute” to work. A friend of mine lived outside Tampa, but his airline hub was Atlanta.

Southwest’s model is a city-to-city network – but 50% of all Southwest flights touch down in Florida on any given day.  Thunderstorms in Orlando and Tampa will start a chain of delays. But when it happened in early October, the initial conditions were very different. Delays became cancellations.

American uses a hub-and-spoke model. In late October with high winds in Dallas-Ft. Worth, the impact was similar for the same basic reasons: flight attendants and pilots could not get to where they needed to be to start operations.

Flight and crew scheduling is a complex operation in an area of mathematics. If you remember from high school math that if you have three unknowns and you want to solve the problem for an exact answer, you need at least three equations. But what if you only have two equations? There are no exact, perfect answers, but…. depending on initial conditions there can be localized optimal answers. Sounding familiar? Once flights are scheduled, crews are assigned (and there is a seniority and bid system behind this), and reserve crews designated. The assignment models are working on historically thin margins with less room for “bumps.”

Weather, initial conditions, crew assignments… it is complex and operating closer to the edge. Now, we are entering the holiday season. Buckle up.