Somewhere over the middle of the North Atlantic, a few hundred airliners are flying through airspace where nobody can see them. There is no radar out there. For most of the crossing there is no VHF radio either. The nearest controller is in Prestwick or Gander, a thousand miles away, working a screen full of predictions rather than returns.

And yet the North Atlantic is the busiest stretch of oceanic airspace on the planet, and aircraft are routinely separated from each other by distances a European controller would consider snug. That works because oceanic flight was never really about knowing where you are. It is about everyone agreeing, in advance and in writing, where everyone else is going to be.

In March 2026 that system took another step away from the way it has worked for half a century. The clearance itself is going away.

Two rivers of aeroplanes, flowing opposite ways

The North Atlantic has a daily rhythm, and it is set not by air traffic control but by what passengers want and what airports will tolerate. People want to leave Europe in the morning and arrive in North America in time for the afternoon. They want to leave North America in the evening and arrive in Europe for breakfast. Night curfews at European airports enforce the second half of that.

The result, in the words of NAT Doc 007 itself, is that “much of the North Atlantic (NAT) air traffic contributes to two major alternating flows: a westbound flow departing Europe in the morning, and an eastbound flow departing North America in the evening”. The manual then puts numbers on it: peak westbound traffic crosses 30 degrees West between 1130 and 1900 UTC, and peak eastbound traffic crosses the same meridian between 0100 and 0800 UTC.

Those are also the validity windows of the two track systems, and they are not a coincidence. Rather than fight the surge, the system is built around it. For eight hours the ocean is a one-way street westbound. For seven hours overnight it is a one-way street eastbound. In between are the changeover periods, which are the awkward hours for anyone flying against the grain.

The tracks move every day, and the airlines help draw them

The tracks are redrawn from scratch twice a day. The reason is the jet stream, which wanders by hundreds of miles between one day and the next, and which is strong enough that riding it or avoiding it is worth real money. A track system that stayed put would be a track system that ignored the single largest variable in the crossing.

Construction starts from minimum-time tracks, then gets bent by reality. The appropriate oceanic area control centre builds the system “with due consideration of airlines' preferred routes and taking into account airspace restrictions such as danger areas and military airspace reservations”, coordinating with the neighbouring centres in New York, Reykjavik, Bodo and Santa Maria so the result is actually flyable. Gander produces the night-time system; Shanwick, at Prestwick, produces the day-time one.

What is less widely known is that the airlines get a vote. The process begins with the Preferred Route Message system, and operators are urged to tell the oceanic centres what routing they would like for the coming peak, no later than 1900 UTC for the next day-time system and 1000 UTC for the next night-time one. The proposed tracks are then published on a website, and there is a formal one-hour window in which anyone interested can comment before the design is frozen. The highway is drawn collaboratively, every single day, in public, by the people who are about to drive on it.

And then nobody has to use it. This is the detail that surprises people most, and the manual says it in one line.

The catch is in the next sentence, which warns that while controllers will try to clear random traffic across the system at the levels requested, “re-routes or significant changes in flight level from those planned are likely to be necessary during most of the OTS traffic periods”. You may cross the motorway. You should not expect to do it at your chosen altitude during rush hour.

The clearance you no longer need

For decades the ritual before coast-out was fixed: an hour or so from the oceanic boundary, the crew requested an oceanic clearance, and waited for a controller to issue one. Getting it wrong, or getting it late, meant holding at the boundary.

That is being dismantled. Under the current manual the crew sends a Request for Clearance, or RCL, which carries the oceanic entry point, the estimated time over it, the Mach number and the requested flight level, plus the highest level the aircraft could accept there. The timing differs by centre: 90 to 60 minutes before the entry point for Gander, 90 to 30 for Shanwick, at least 40 minutes for Santa Maria, at least 20 for Bodo. Reykjavik and New York Oceanic East do not want one at all, and from 19 March 2026 a crew that sends Reykjavik an RCL anyway will simply be told it was not needed.

And then, crucially, the manual states the thing that would have been unthinkable a few years ago: “Enroute aircraft shall enter the oceanic airspace in accordance with their existing ATC clearance. No oceanic clearance is required.” The RCL is not a request for permission. It is a declaration of intent that lets the ground system model the crossing, and the manual is blunt that “The RCL, or RCL RECEIVED, is not a clearance.”

The rollout is uneven, and that unevenness is itself a live operational fact rather than a footnote. Shanwick has delayed its implementation of oceanic clearance removal, so a flight entering Shanwick's airspace still receives a conventional oceanic clearance after its RCL, and will do until further notice. The result is an ocean where the procedure on one side of a boundary is not the procedure on the other, which is exactly the kind of asymmetry that gets written into airline briefing notes in bold type.

NATS, which runs the Shanwick oceanic centre from Prestwick, made its own film about the crossing. It is worth watching for the control-room footage alone: the ocean as a planning problem, worked by people looking at predicted positions rather than live ones.

Nobody is watching, so you have to tell them

Since there is no radar, the aircraft has to be its own sensor. That is now a hard equipment requirement. The NAT Data Link Mandate, in the manual's words, “requires aircraft to be equipped with, and operating, FANS 1/A CPDLC and ADS-C over Inmarsat or Iridium SATCOM in the NAT region”, and currently covers FL290 to FL410 inclusive. CPDLC is text messaging between controller and cockpit. ADS-C is the aircraft reporting its own position, automatically, under a contract the ground system sets up with it.

There are exclusions, and they are logical rather than generous: airspace north of 80 degrees North, the New York Oceanic East flight information region, and areas where a surveillance service genuinely exists, such as the Iceland to Greenland corridor, where ADS-B is required of all instrument flights instead. An aircraft whose data link fails before departure is expected to refile clear of the mandated airspace. One that fails after departure gets handled tactically, and must tell air traffic control before entering.

Around that sit the older habits, still in force. Position reports at the waypoints in the flight plan, with controllers aiming to have positional information at roughly hourly intervals. A revised estimate passed as soon as possible if the next waypoint time slips by three minutes or more, for anyone not reporting by ADS-C. The transponder set to 2000 ten minutes after passing the oceanic entry point. And, when a crew sends one, an operations normal report consisting of nothing more than the callsign and the words “operations normal”.

Separation is a promise about time

This is where the design pays off. Because the system knows what every aircraft intends to do, and because the aircraft keep confirming it, separations can be cut to something close to radar standards without radar. Performance-based separation minima “as small as 15 NM lateral and 14 NM longitudinal” have been implemented in the region, predicated on performance-based communication and surveillance rather than on anybody watching a screen.

Inside the track system the standard is 23 nautical miles laterally, achieved by spacing the tracks through whole and half degrees of latitude between FL340 and FL400. Earning that is not automatic: an aircraft must hold RNP 4 approval and formal state authorisation to file RCP 240 and RSP 180, which are numerical promises about how quickly a message gets through and how quickly a position report arrives. Fly without them and you get the wider spacing. Vertically, reduced vertical separation gives 1,000 feet between FL290 and FL410.

And because mathematically perfect tracks would put every aircraft on exactly the same line, the system deliberately adds noise. Strategic lateral offset procedures should be standard operating procedure across the region, distributing traffic between the centreline and two miles right of it. The randomness is the point; it is what stops a vertical error from also being a lateral coincidence. Left offsets are prohibited, which is the sort of rule that only needs stating once you understand what the right-hand bias is protecting against.

When something breaks in the middle

All of this depends on the crew being able to ask permission. The interesting case is when they cannot: a depressurisation, a medical diversion across the flow, a navigation system failure, weather that has to be avoided now. The manual's answer starts with an instruction to try anyway.

If it cannot be obtained, the crew acts on its own and the procedure is a small masterpiece of pessimism. Turn at least 30 degrees left or right, chosen on the basis of where the track system is, what is flying on the neighbouring tracks, where the diversion airfield lies, which lateral offset you were already flying, and the terrain. Establish a parallel track five nautical miles off the cleared one. Hold the assigned level until you are established on it, and if you cannot, descend as slowly as is operationally feasible.

Then put yourself somewhere nobody is supposed to be. Either descend below FL290 and sit 500 feet off the levels normally used, or stay where you are and take the same 500-foot offset, 1,000 feet if you are above FL410. The manual notes drily that with a 500-foot offset applied, collision avoidance alerts may occur, and that altimetry error may mean you have less than the nominal 500 feet in reality. This is a procedure that assumes it is not the only thing that has gone wrong.

Meanwhile: all exterior lights on, collision avoidance left in resolution advisory mode, transponder squawking 7700 when able, and the distress or urgency call spoken three times by whatever means works, voice or data link. Tell somebody as soon as practicable. The aircraft has just stopped being a predictable object in a system built entirely on predictability, and the only remaining defences are being visible, being loud and being somewhere unusual.

The new problem is not the ocean

The 2026 edition of the manual treats satellite navigation jamming and spoofing as a routine hazard rather than an exotic one, and it makes a point that is easy to miss. A crew whose position accuracy recovers has not necessarily recovered everything else. Time, the automatic position reports, the data link and the longitudinal separation that depends on accurate estimates over waypoints can all still be affected. In an airspace where separation is a promise about arrival times, a corrupted clock is a safety issue in its own right.

Which is a strange place for the North Atlantic to have arrived at. The crossing stopped being about navigation a long time ago; modern inertial and satellite systems made knowing your position the easy part. What replaced it was a vast exercise in collective bookkeeping, and the threats to it now are threats to the bookkeeping, not to the airmanship. The jet stream still moves the tracks every day. The geometry of a great-circle route still bends them north. But what keeps two aeroplanes apart over 30 degrees West is a set of agreements made before either of them left the ground.

Sources: ICAO NAT Doc 007, North Atlantic Operations and Airspace Manual, edition V.2026-1, applicable 19 March 2026, paragraphs 1.7, 1.9, 2.1, 2.2, 5.3, 6.3, 10.2 and 10.3, read directly from the ICAO EUR/NAT publication; OPSGROUP's summary of the March 2026 NAT Doc 007 changes for the removal of the flight level allocation scheme and the Reykjavik RCL change; NATS, North Atlantic Skies.