Radar has been the backbone of air traffic control since the Second World War, and it is genuinely impressive technology. But it has a fundamental limitation that every controller knows and quietly works around: it tells you where an aircraft was, not quite where it is, and it tells you almost nothing about what the airspace around it is actually doing. Wind shear, wake turbulence, microscale weather — radar sees through most of it entirely. For decades, controllers have filled those gaps with experience, intuition, and conservative separation standards that sometimes feel more like educated guessing than precision science. AI is now starting to close that gap in ways that feel, to this avgeek at least, genuinely extraordinary.
The specific problem worth understanding is wake turbulence. Every aircraft above a certain weight leaves behind a pair of counter-rotating vortices — invisible, intensely energetic, and capable of rolling a following aircraft completely inverted if the spacing is wrong. Controllers manage this with time-based and distance-based separation minima that were established largely through empirical observation in earlier decades. They work. But they are also conservative by design, because the alternative to conservative is dangerous. At busy airports, that conservatism costs capacity. More separation means fewer arrivals per hour, which is part of why places like Heathrow or JFK are essentially running at their limits during peak periods even on a calm, clear day.
What AI systems are now beginning to do is model wake turbulence dynamically, in real time, by pulling together data streams that no human controller could synthesise fast enough to be useful. Wind speed and direction at multiple altitudes, aircraft weight and configuration, temperature gradients, the precise flight paths of preceding aircraft — all of it feeds into models that can predict where a vortex will be, how long it will persist, and when it will have dissipated enough to be irrelevant. The result, in trials at a number of major European and North American airports, is that separation minima can be safely tightened when conditions allow, and widened automatically when conditions are more turbulent. Dynamic rather than fixed. The runway handles more aircraft on a good day and stays appropriately cautious on a bad one.
There is a subtler benefit that is easy to overlook. When controllers are working a busy arrival stream, cognitive load is the real constraint. Every judgment call about spacing is mental work, and mental work compounds. Systems that handle the wake turbulence calculation in the background and surface a simple, trustworthy recommendation free the controller to focus on the things that genuinely require human judgment — unusual situations, communication, the kind of pattern recognition that comes from years of watching aircraft behave unexpectedly. AI handles the arithmetic; the human handles the edge cases. That is a good division of labour.
What strikes me most about this particular application is how invisible it is from the passenger seat. You will never feel the vortex that was predicted and avoided. You will never know that the aircraft ahead of you landed with slightly less separation than it would have received a decade ago, because a model confirmed the vortex had already drifted clear of the glidepath. Aviation safety has always worked this way — the best outcome is nothing happening — but there is something quietly wonderful about knowing that the airspace above a busy airport is now understood in more granular detail than at any point in the history of flight.
Radar is not going anywhere. It remains the bedrock. But the AI layer being built on top of it is starting to show us what the atmosphere is actually doing inside those returns, and that is a genuinely new kind of seeing. For anyone who has spent time thinking about how complex managed airspace really is, watching that capability mature is about as exciting as aviation gets.