Ground delay programs

Why does air traffic control hold an aircraft at its gate for an hour when the sky ahead of it is empty?

The aircraft is loaded, the crew are ready, the weather outside is fine, and the departure airport has runway to spare. The captain then announces a wait of an hour, imposed by a flow control centre, for a sky that is visibly empty.

The natural reading is that somebody has made a scheduling mistake. But notice what the announcement never claims: it does not say the airspace ahead is full. It says the aircraft has been given a time. And a time to arrive somewhere is a very different object from permission to be somewhere now. What could make an authority care about when this aircraft lands, three hours from now, rather than whether it can safely leave?

Begin with the constraint, which is almost never the departure end. An airport's ability to accept arrivals is a rate, and that rate is weather-dependent. In good visibility, controllers can run approaches close together because pilots can see and follow the aircraft ahead. Drop the cloud base and every approach must be separated by radar to a wider standard, a runway may lose its parallel partner, and the acceptance rate falls.

Do the arithmetic with round illustrative figures, because the shape matters more than the values. Suppose an airport normally accepts sixty arrivals an hour and low cloud cuts that to forty, while sixty aircraft an hour are still scheduled to arrive. Twenty an hour more are coming than can land. After three hours, sixty aircraft — an entire hour's worth of traffic — are surplus. They exist. They are somewhere. The only question the system gets to answer is where they wait.

So ask that question directly. Waiting in the air means burning fuel, and fuel aboard is finite: an aircraft holding is running down the reserve that also covers a diversion, a missed approach, and the unexpected. Hold long enough and it becomes a fuel declaration and then a diversion, which moves the aircraft to a different airport with a crew running out of duty hours. Waiting on the ground burns almost nothing and consumes no reserve. The delay is the same length either way; only its cost and its risk differ — and the aircraft cannot take it on the ground unless somebody tells it to before it leaves.

That is the mechanism, and it is worth stating baldly: a ground delay program does not create the delay. The capacity shortfall creates the delay. The program relocates it to the cheapest and safest place to hold it, which is the only place where holding costs nothing — stationary, on the ground, at the origin.

Now the part that explains why this needs a central authority rather than good sense. Suppose the airlines were left to decide individually. Every operator knows the destination will be congested. What is the best move for one of them? Launch immediately, because arrival sequencing is broadly first-come-first-served, and an aircraft already airborne and inbound is ahead of one still at a gate. Each operator reasons this way, all launch, and the surplus arrives overhead as a scramble of aircraft competing for the same approaches with fuel clocks running. Individually sensible, collectively the worst available outcome. This is a coordination failure, not a knowledge failure — everyone can see it coming and nobody can act on it alone.

The fix is to make the queue positional rather than physical. The flow manager sets a program, allots arrival slots, and assigns each flight a controlled departure time computed backwards from its slot. Slots are rationed by original schedule, so a carrier's place in the queue is the place its timetable already gave it, and no advantage is available from launching early. Once your position is guaranteed, the incentive to scramble disappears — and the queue that was going to form somewhere forms in departure lounges instead of stacks.

One feature makes the scheme tolerable to airlines: they may trade. Cancel a flight and its slot is released, with later flights compressing forward into it; a carrier that would rather protect a long-haul departure than a shuttle may swap its own two slots. That collaborative layer turned flow metering from something imposed on airlines into something they help run.

Take that last point seriously, because it is the honest cost of the whole design and it is routinely glossed over. The program is built on a predicted capacity. If the forecast is pessimistic and the cloud lifts early, the airport spends the afternoon running below its true rate while dozens of aircraft sit at gates that could have been airborne. Delay has then been created rather than relocated. Revision and compression exist to recover some of it, but not all: an aircraft you have already told to wait cannot be un-waited instantly, because its crew have stood down and its slot in the departure sequence has gone. The trade is real — less airborne holding and fewer diversions, at the price of over-delaying whenever the weather beats the forecast.

One refinement is worth adding. The visible emptiness that started the question is not evidence against the program. The metering is aimed at a time and a place several hours downstream. An empty departure sky is exactly what you should expect to see when the scheme is working, and it would be a strange system that could only justify itself by producing congestion you could point at.

• Why the same logic is harder to apply to en-route airspace capacity than to a single congested airport.

Every term the collection defines is gathered in the glossary.