image alt text

Never has flying been more frustrating. Air traffic is normalizing after weeks of delays as Congress gives the Federal Aviation Administration (FAA) flexibility to shift spending cuts from operations to infrastructure, avoiding a prolonged furlough of FAA Air Traffic Controllers. Aggrieved fliers (and pundits) continue to argue about whether the delays are genuinely caused by the sequester or are just another example of Washington politics run amok.

In fact, neither answer represents the whole story. The delays are primarily due to the complex and capacity-constrained National Airspace System (NAS).

(First, a disclaimer: I am a pilot myself, and I have tremendous respect for the work of Air Traffic Control (ATC). They are a professional and courteous group who are an integral part of the safety and the efficiency of the NAS. I thoroughly enjoy flying “in the system” and interacting with ATC.)

In the midst of the onset of delays, Bloomberg Government studied FAA staffing levels and concluded in an often-quoted report that the FAA had the option to shift controllers from overstaffed facilities to avoid leaving high-capacity positions, like Newark and LaGuardia airports, understaffed. This makes sense given the topology of air traffic routes: a handful of airports garner most of the arrival and departure traffic.

But even without Bloomberg Government’s nuanced analysis, it’s hard to believe that the FAA could not have staffed these critical positions appropriately. According to the FAA’s 10 year strategy for the Air Traffic Control Workforce, since 2000 the number of controllers has stayed constant while the FAA’s operations budget increased – even though the amount of flying traffic has decreased by approximately 20% (Figure 1.2). If the current level of controllers were sufficient for the peak traffic of the year 2000, could a small reduction of controllers truly lower the capacity of the airspace system today?

image alt text

System Logic conducted a preliminary analysis of data provided by FlightStats. Broken down by day, the correlation between understaffing (as measured by Bloomberg) and excessive delays was around .3, indicating that there may have been some effect. However, as a control, we compared on-time performance in March and April (before the advent of the furloughs) and found a correlation of .15 between increased delays and “understaffed” airports; here we would expect no effect. This suggests that some of the perceived delays are the manifestation of the expected variance in the complex air traffic system. But, it does leave room for an effect from the furlough. We’ll be investigating this effect further and providing updates over the next few weeks.

So how could a very small reduction in air traffic controllers affect our air traffic system so easily? To understand how such a perturbation in the interconnected and overburdened air traffic system could cause delays, we need to understand the nature of the system’s capacity to handle aircraft. Capacity, or the number of planes that can land and/or take-off in a given time, is not a fixed number. It changes constantly even during normal operations. The key determinant is the required separation between aircraft. Different aircraft types, speeds, visibility, and weather conditions change the required separation between aircraft. In addition to dynamically changing required separation, flights need to avoid convective weather (like thunderstorms), further reducing available sky.

Because of the topology of air traffic, high-traffic airports run very close to their ideal capacity, where the number and configuration of runways is what physically limits the number of planes that can takeoff and land. Because of the hub-and-spoke system, certain key airports become even more desirable destinations to serve. This is often beneficial to travelers, as it expands the number of destinations one can conveniently reach from a smaller airport by linking it, through regional carriers and feeder flights, to large hubs.

Additionally, airlines use complicated algorithms to optimize their routes for cost and efficiency. However, these optimizations make their routes vulnerable to cascading delay. Like an intricate house of cards, all the components of the National Airspace System are so tightly coupled that anything other than a minor delay in one small area can easily propagate to larger delays and frustrated travelers across a wide swath of the country. Notably, during the furlough, many departures were “excessively delayed”–that is, beyond 45 minutes late. The rapid change from a small delay to an excessive delay is an example of criticality, which happens when a complex system exceed its ability to absorb delays and provide excess capacity.

It’s hard to know the precise effect of the furlough, but a delay of even a few seconds in communicating a clearance to each aircraft, or even a small increase in required separation, reduces capacity. Coupled with weather, changing winds, and the already stretched air traffic system, consistent delays of a few seconds could have added up and pushed the already overburdened system toward criticality and delays. Indeed, a back-of-the-envelope analysis of departures at Newark Airport reveals that at the start of the furlough, takeoff clearances from controllers at Newark Airport were almost 20% slower than they were the week before.

Unfortunately, the current political controversy detracts from the FAA’s increasingly critical mission to maintain and expand the nation’s air traffic infrastructure. The sequester will reduce funds available to service and maintain critical ATC infrastructure like backup generators and heating and ventilation systems. In addition, it will limit the FAA’s pursuit of its NextGen ATC system, which would increase capacity by allowing for more efficient routing and technologically assisted reduced separation. Given the current state of flight delays–even when all hands are on deck–this infrastructure and capacity development are sorely needed.

Image courtesy of ZHoover123, Wikimedia Commons.