How does a reduction in airport capacity create delays?

We'd left off at having explained what airport capacity is and how it varies. In our disruption outlooks, we've also referenced the relationship between flight schedules (i.e. demand for slots) and airport capacity (i.e supply of slots); more aptly, when demand exceeds supply, queues are essentially created and delays result.

Let’s create a demand overage.

So what does that process look like? Let's consider a hypothetical airport with one runway that is continuously used for arrivals. We're not going to worry about departures for the time being - maybe there's an endless supply of gates, maybe there's a The Good Place-like archway in the woods that aircraft proceed through when they exit the runway. Regardless, given blue skies and favorable winds, 40 aircraft can land on the runway each hour (approximately 1 every 90 seconds). There's only one airline that serves this airport and they've built their schedule to fully utilize those 40 "good day" landing slots.

Eventually, however, the weather will stop cooperating. Let's imagine a day when drizzle starts at noon and lasts for the rest of the day. The drizzle lowers visibility to about 1/8 mile (they're really socked in): in these conditions, pilots are no longer able to maintain visual separation during their approach (i.e. they're unable to rely on sight while operating around other aircraft). In response to the loss of visual approaches, the FAA is going to increase the spacing between arrivals - in this case, aircraft will land approximately every 140 seconds. The extra 50 seconds between arrivals means 14 fewer aircraft will land each hour and the FAA has set the arrival rate to 26.

Let’s create delay.

Alright, we've created our demand overage - 40 arrivals are scheduled for a runway that will only be able to accommodate 26. You can see below how the longer-than-scheduled spacing between arrivals creates delays, as a flight's scheduled arrival time no longer matches their slot time. Though we'll focus on just the first 10 minutes of the hour to keep things visually (and conceptually) accessibly, you may also note that delays increase. If things continue at the same rates, delays will have grown to more than 64 minutes by the end of the second hour of scheduled activity; and the last scheduled arrival of the 8th hour would be looking at a delay of more than 4 hours.

What we've reproduced mostly closely resembles a ground delay program (GDP), which is one type of traffic management initiative in the FAA's toolbox. When a GDP is active for an airport, arriving flights will be assigned an expected departure clearance time (EDCT; sometimes referred to as a wheels up time) from their origin station, where they're held on the ground. Let's take flight 81 above as an example, which we'll say was scheduled to pushback from the gate at its origin at 12:25 p.m. and scheduled to takeoff at 12:38 p.m. In our hypothetical GDP, flight 81 will be assigned a 1:42 p.m. wheels up time. Where will this delay be absorbed? Though the FAA enforces the delay, the airline largely gets to decide where it will incur the delay. It will typically be most environmentally- and customer-friendly, as well as cost-effective, to postpone boarding and delay the pushback from the gate. Absorbing the delay on the gate reduces fuel burn and may save crew costs, depending on contracts. Occasionally, however, flight 81's gate is needed for the next flight (or the airline myopically pursues on-time departure metrics) and flight 81 will pushback from the gate as scheduled then absorb the delay on the tarmac before taking off. There's also an oft-employed compromise wherein the delay is split between the gate and tarmac, which positions the flight to accept an earlier wheels up if it becomes available.

Let’s redistribute delay.

Airlines can also exert control over delays via slot swaps. While the FAA creates the slots - and sets the initial order based on scheduled arrival time - it's ultimately up to the airline to distribute slots within their allotted portfolio. Let's say flight 4 is a lightly-booked small aircraft (e.g. 50 seats) while flight 6 is a fully-booked large aircraft (e.g. 350 seats). The airline may deem it more optimal to give flight 6 the slot that was originally assigned to flight 4, thus allowing flight 6 to land on-time. Let's also say that flight 8 is a medium-sized aircraft (e.g. 150 seats). If the airline sacrificed flight 4 once, then they might be inclined to do so a second time: flight 8 is given flight 6's original slot (thus minimizing its delay), leaving slot 8's original 12:16 slot (and a 12 minute delay) for poor flight 4. If our hypothetical airport had departures, the airline might also consider the number of connecting passengers (and where they're connecting to), crew duty time (i.e. legalities), aircraft turnaround buffers and curfews when swapping slots.

While our three-way swap makes a difference of less than 10 minutes, swaps are increasingly consequential as the GDP drags on and delays inflate. If our hypothetical GDP runs for 8 hours at the same rates, the airline will need to distribute more than 41,000 minutes of delays (and the average delay settles out at 129 minutes). It's also important to note that these swaps did not change total delays (or average delay), only shifted delays around. The only way to reduce delays in the aggregate? Cancellations.

Let's continue to pick on flight 4, but rather than kicking the can down the road, we're just going to cancel it. This serves to remove 12 minutes of total delay and reduces the average delay by more than a minute. Perhaps you're unimpressed with how much delay the cancellation saves, so let's reframe it in percentage terms: a schedule reduction of less than 15% shrinks delays by nearly 40%.

At this point, we imagine many readers might be saying, "I'd prefer a longer delay if it meant no risk of the airline cancelling my flight - why can't the airline absorb all of the delays?" The short answer is staffing, especially of the pilot variety. To ensure pilots are able to safely operate their aircraft - without the influence of fatigue - the length of a pilot's workday is federally regulated. While the allowable duty period varies by number of pilots, start time and number of flights, it begins when the pilot reports for duty at the start of their day and ends when their last flight is parked; 13 hours is a serviceable rule of thumb. Airlines build some buffer between a pilot's scheduled end of day and allowable end of day, however delays consume that buffer.

We think this is an appropriately sized explainer, so we'll hit pause. GDPs are just one of the options in the FAA's traffic management toolbox.. we'll pick it up there next time!