Daniel's been staring at the Singapore Airlines footage from May twenty twenty-four — flight three twenty-one, six thousand feet of altitude gone in under a minute, one passenger dead, dozens injured. The aircraft landed safely and was back in service within days. And his question is basically three parts. Is this kind of thing becoming the new normal — is climate change actually driving up turbulence? When passengers are white-knuckling the armrests, how often is the plane in danger? And how wide is the gap between what we feel back there and what the airframe actually experiences?
To answer any of that we have to start with what turbulence actually is — and what it isn't. Because most people picture the plane getting batted around by angry clouds, and that's maybe a third of the story.
There are types.
Three main ones. Convective turbulence — that's your thunderstorms, the stuff weather radar can see. Mechanical turbulence — air flowing over mountains or buildings, predictable if you know the terrain. And then clear-air turbulence. CAT. The invisible kind. No clouds, no radar return, hits at cruise altitude without warning. That's the one that sent the Singapore Airlines triple seven into that dive, and it's the one linked to climate change.
And the mechanism there — warmer atmosphere, something happens to the jet stream?
The jet stream is driven by the temperature difference between the equator and the poles. As the troposphere warms and the stratosphere actually cools — both measured trends — that gradient sharpens. Sharper gradient means stronger wind shear. Stronger wind shear means the air layers start tumbling over each other like rapids in a river. The Nature paper from twenty twenty-three quantified it: severe CAT over the North Atlantic is up fifty-five percent since nineteen seventy-nine. Moderate CAT up thirty-seven percent. Same flight corridors, same altitudes, four decades of reanalysis data.
Fifty-five percent sounds alarming. But I want to know what "severe" means in that sentence — is that severe to the airframe or severe to the drinks cart?
That's the right question. In that study, severe CAT is defined by the vertical acceleration threshold — it's the kind that would throw an unbelted person into the ceiling. So it's severe to the passenger. Structurally, it's still well within what the aircraft is designed to handle.
So let's look at the data properly. Is it really getting worse, or does it just feel that way?
It is getting worse, but the absolute numbers matter. The fifty-five percent increase is real, but we're talking about a baseline that was already very low. The BBC ran the numbers after that Nature paper — even with the increase, a given flight's chance of encountering severe CAT is still something like one in many thousands. The trend line is up, the per-flight risk is still tiny.
And that's the North Atlantic. What about other routes?
The North Atlantic has the best reanalysis data because it's the most heavily trafficked long-haul corridor. The physics should apply anywhere with a strong jet stream — the North Pacific, the routes over East Asia. But the measurements are spottier. What we can say is the mechanism is global, even if the best data is from one corridor.
Walk me through what's actually happening to the air when CAT forms. You said rapids in a river.
Imagine the jet stream as a fast-moving river of air at thirty-five thousand feet. The air above it and below it is moving at different speeds. When the shear gets strong enough, the boundary between those layers goes unstable — it starts to roll up into eddies. Some of those eddies are hundreds of feet across. The plane flies through them and the wing sees a sudden change in angle of attack. Lift changes instantly. That's the bump.
And there's no visual signature at all.
None. No clouds, no precipitation. The air looks identical. The only warning pilots get is a pilot report from another aircraft that just flew through the same patch. And those reports are only as good as the last plane that was there — if the wind field is evolving, the report might be stale in twenty minutes.
So you're flying blind into invisible rapids. That's... comforting.
Well, this is where the engineering comes in, and this is the part most coverage misses. The plane is built for this. Certification standards require an airliner to withstand one point five times the maximum load it's ever expected to encounter in service. For a typical narrow-body, that's three point eight g positive and one point five g negative.
Let's put numbers on that for someone who's never thought about g-loads. What does a passenger actually feel during what they'd call "severe" turbulence?
When you're sitting in your seat and the plane drops, you feel a fraction of a g negative — your stomach lifts, your body interprets it as falling. Most turbulence that passengers describe as terrifying peaks at about zero point five g. Maybe zero point seven in a really rough patch. The airframe is certified to one point five negative. That's more than double.
So the plane's not even at half its limit.
Not even close. The Singapore Airlines flight — six thousand feet of altitude change in under a minute, which sounds catastrophic, but that's a controlled descent after the initial upset. The actual g-loads during the event were enough to throw unbelted people around, but the aircraft was never structurally threatened. It landed, got inspected, and was back carrying passengers within days.
Compare that to the one case where CAT actually did break a plane — the BOAC flight nine eleven over Mount Fuji in nineteen sixty-six.
Right. A seven-oh-seven encountered extreme CAT on the lee side of Mount Fuji. The vertical stabilizer separated. The aircraft broke up in flight. Everyone aboard died. But that was before modern turbulence detection, before revised certification standards that came directly out of that accident, and before we understood mountain wave phenomena the way we do now. That specific failure mode — the fin separating from extreme oscillatory loads — essentially doesn't happen in the modern fleet.
So what's the actual danger when a plane hits rough air at cruise?
Unsecured objects and unsecured people. The Singapore Airlines fatality was a passenger not wearing a seatbelt. The NTSB reports roughly thirty turbulence-related injuries per year on US carriers. Almost every single one is an unbelted passenger or a flight attendant who was up and about. Zero structural failures. The last fatal turbulence accident on a US commercial jet was nineteen ninety-seven — United seven forty-seven over the Pacific. Severe CAT broke the horizontal stabilizer, but the crew still landed safely.
Wait — the stabilizer broke and they landed?
They landed. That event actually drove changes to turbulence penetration speed procedures. Before that, the guidance was looser. Now pilots are trained to slow to a specific speed — typically around two hundred eighty knots indicated — when they hit significant turbulence. That reduces the structural loads. The plane is flown differently when it's bumpy.
So the pilots have a whole set of responses that passengers never see.
They do. Weather radar for convective turbulence — that's the stuff they can see and route around. For CAT, they rely on pilot reports and satellite-based turbulence forecasting products that have gotten much better in the last decade. When they know it's coming, they'll adjust altitude, change course, or at minimum get the seatbelt sign on and slow to penetration speed. The turbulence might still be uncomfortable, but the aircraft is being operated inside a safety envelope that accounts for it.
And that forecasting — is it keeping up with the increase?
That's an open question and it's one the FAA and EASA are actively reviewing. The current generation of turbulence forecasts works reasonably well for large-scale patterns, but CAT can form and dissipate on scales smaller than the models resolve. As the frequency increases, the forecasting tools need to get finer-grained. There's work on using machine learning to predict CAT from satellite data and pilot reports in near real time.
But here's where it gets interesting — because what the plane feels and what you feel are two very different things.
The perception gap is enormous. And it's not just about g-loads — it's about how the human body interprets sudden motion.
Explain the sensory mismatch.
Your vestibular system evolved to keep you upright on the ground. When a plane drops ten feet in half a second, your inner ear registers a loss of support. There's no visual reference — you're in a tube, the horizon might not be visible, and if it's clear-air turbulence there's not even a cloud to suggest why you're moving. Your brain gets one signal: falling. It triggers a primal fear response. Heart rate spikes, muscles tense, you grab the armrest. The plane dropped ten feet. It has another thirty-five thousand to go.
So the body is screaming "we're crashing" and the plane is basically fine.
And the gap goes both ways. Psychological research on this — there have been studies where they instrument passengers and compare their ratings to the flight data recorder. Passengers consistently rate turbulence as "severe" at g-loads pilots classify as "light" or "moderate." The gap is widest during CAT because there's no visual cue. You can't see the bump coming, so you can't brace for it, so it feels worse.
Does the pilot even notice the same event the same way?
They're strapped in, they have instruments, they can see the horizon or at least the attitude indicator. They know what the plane can take. Their frame of reference is completely different. A pilot might log "moderate chop" for something that has half the cabin convinced they're about to meet their maker.
Which creates its own problem — if the pilots are under-rating it, the next aircraft doesn't get an accurate pilot report.
That's a real operational issue. The system depends on honest reporting. And there's some evidence — anecdotally, from safety surveys — that experienced pilots can get desensitized. They've been through so much turbulence that "moderate" to them might be "severe" to a junior first officer and "terrifying" to the passengers. The data recorders don't lie, but the voice reports are subjective.
So we've got passengers overestimating, pilots potentially underestimating, and a black box logging the truth somewhere in between.
And the black box usually shows the plane barely noticed. I've seen data from a turbulence event on a triple seven test flight — but we'll get to that.
Let's talk about what the industry is actually doing about the trend. You mentioned the FAA and EASA reviewing forecasting tools. What else?
Fuel planning is one knock-on effect. If CAT is more frequent on certain routes, airlines may need to carry more contingency fuel for diversions. That's weight, that's cost, that's carbon — there's a feedback loop there. Seatbelt sign policies are another. Some carriers are getting more conservative — keeping the sign on longer, even in light chop, because the liability calculus is changing. After Singapore Airlines, every airline's risk department took a fresh look at their turbulence procedures.
And the cabin crew — they're the ones actually at risk.
Flight attendants are disproportionately represented in turbulence injury statistics because they're standing. The FAA has been pushing for better real-time turbulence data delivery to the cabin so crews get more warning. Some airlines now have procedures where the cockpit calls back with a specific time estimate — "we expect rough air in three minutes" — so the crew can secure the galley and strap in.
Three minutes sounds tight if you're halfway through a beverage service.
It is. And with CAT, you might get zero minutes. That's the nature of the beast.
Let's put a finer point on the structural question, because I think this is where the public imagination goes straight to disaster movie territory. Can turbulence tear the wings off?
No. The certification margin is enormous. The wing is tested to one hundred fifty percent of the maximum load it would ever see in the most extreme turbulence event in the operational envelope. That's a static test — they literally bend the wing until it breaks, and it has to break at no less than one point five times the limit load. In practice, most wings break well above that. The safety factor is baked into the aluminum and the carbon fiber.
So what would it actually take?
You'd need a combination of extreme turbulence and pilot error — flying well above turbulence penetration speed, or making abrupt control inputs that compound the loads. That's what happened in a few historical cases where aircraft broke up in turbulence. It wasn't the turbulence alone. It was the turbulence plus flying too fast, or turbulence plus a structural flaw that should have been caught in maintenance.
The Mount Fuji case again.
And even that led to changes. After BOAC nine eleven, the FAA mandated that the seven-oh-seven's vertical stabilizer be reinforced. The entire certification philosophy shifted. Modern aircraft are designed with damage tolerance — the structure can sustain a certain amount of cracking or damage between inspections without failing. That philosophy covers turbulence loads too.
So the engineering story is basically: the plane is overbuilt for this, and it's been getting more overbuilt for decades.
And the operational story is: pilots have tools to avoid the worst of it, and procedures to minimize stress when they can't avoid it. The combination means the structural risk is vanishingly small.
Which brings us back to the passenger experience. Because the structural risk is small, but the fear is real, and the fear is what people remember.
And the fear has consequences. There's a whole literature on flight anxiety and turbulence. For some people, one bad turbulence event is enough to make them avoid flying for years. The perception gap isn't just an academic curiosity — it has real effects on behavior.
I'm thinking about the sensory mismatch you described. Is there anything passengers can actually do to recalibrate?
The standard advice is to look at the flight attendants. If they're still walking around, it's not severe. But honestly, that's not great advice during CAT — the attendants might be strapped in too. The better heuristic is to understand the physics. Your drink just left the tray table because the plane dropped ten feet. It has thirty-five thousand more feet of air underneath it. The plane is not falling out of the sky — it's riding a wave.
A wave you can't see.
A wave you can't see. And the plane is designed to surf it.
There's something almost perverse about CAT specifically — the fact that the most frightening turbulence is the kind with zero visual warning. Your brain has nothing to anchor to.
That's exactly what the research shows. In convective turbulence, you see the clouds, you might see lightning, there's a narrative — we're flying through a storm, this makes sense. In CAT, there's no narrative. It's just sudden violence in clear air. The lack of a cause amplifies the fear.
So the same g-load in a storm might be rated "moderate" by passengers, and in clear air it's rated "severe."
There's data suggesting exactly that. Context shapes perception. And the context for CAT is "I have no idea why this is happening."
Which is worse.
Much worse.
Let's talk about the climate piece a bit more. The fifty-five percent increase — is that a linear trend, or is it accelerating?
The Nature paper found a statistically significant linear trend over the satellite era. The question of acceleration is harder to answer because you need longer data. But the underlying physics suggests it should accelerate — as the temperature gradient continues to sharpen, the jet stream shear should increase nonlinearly in some regions. The models project it, but the observations aren't long enough yet to confirm acceleration versus a steady climb.
This is all at cruise altitude. What about turbulence lower down?
Convective turbulence is also expected to increase — warmer air holds more moisture, thunderstorms get more energetic. But that's easier to avoid with radar. The CAT trend is the one that keeps safety people up at night because it's the one you can't see.
The one that hits at the altitude where everyone's unbuckled and walking around.
Cruise is when the seatbelt sign is off, people are queuing for the lavatory, the cabin crew is serving meals. That's when CAT does its damage.
The practical advice from all of this is just: keep your seatbelt on.
Loosely fastened, even when the sign is off. It's the single most effective thing you can do. The Singapore Airlines fatality was an unbelted passenger. The thirty NTSB injuries per year — almost all unbelted. The physics is simple and the solution is simpler.
The boring advice is always the right advice.
It really is.
On that note, Hilbert has been waiting to tell us about a flight test he worked on.
Hilbert: The seven-seven-seven's turbulence load monitoring system. Nineteen ninety-seven. I was installing accelerometers in the cargo hold.
Go on.
Hilbert: We put them at the wing roots, the tail cone, a few points along the fuselage. The system logged every g-load the airframe experienced, sampled at something like twenty hertz. After every test flight, we'd download the data and compare it to the pilot reports.
What did you find?
Hilbert: The pilots almost always under-called it. A flight they logged as "light chop" would show zero point six g peaks in the data. Not dangerous — but not light. The one that stuck with me was a flight over the Rockies where the data showed two point one g. That's getting into real territory. The pilots wrote "moderate chop" in the log. I asked the flight test engineer about it and he shrugged. Said they'd been through worse.
Two point one g and they called it moderate.
Hilbert: The plane was fine. But you start getting above two point five g at the wrong angle, and you're bending things that aren't supposed to bend. The safety margin is huge but it's not infinite.
How often did you see numbers above two g?
Hilbert: In the test program? Twice, across maybe four hundred flight hours. And test programs deliberately go looking for rough air — they fly through weather a revenue flight would route around. So twice in four hundred hours of hunting for it.
That's... reassuring, actually.
Hilbert: The plane is tougher than you think. But it's not tougher than a two point five g hit at the wrong angle. And those are getting more common. Not common — but more common.
Did the monitoring system go into production aircraft?
Hilbert: A version of it. The airline can download the data if there's a report of severe turbulence. Most of the time they don't bother because the pilots already said it was moderate and the plane landed fine. But the data's there if they want it.
When they do pull the data after a passenger-reported "terrifying" flight?
Hilbert: Usually zero point three g peak. The plane barely noticed. I had a colleague who used to say the passengers were riding a different aircraft than the one we built.
That's the whole episode in one sentence.
Hilbert: The snout on those accelerometers was a pain to calibrate, by the way. You had to get the angle exactly right or the readings drifted. We spent three days on one sensor because the mounting bracket didn't account for the curvature of the fuselage. Turned out we needed a shim — a little wedge of aluminum. I think I still have one in a box somewhere.
Did the shim fix it?
Hilbert: Mostly. We got it within half a percent. Good enough for the FAA.
Where does that leave us? The data says CAT is increasing, the engineering says the plane can handle it, and the passengers are experiencing a completely different flight than the airframe is. Daniel's three questions — yes, no, and enormous.
Yes, clear-air turbulence is becoming more common — fifty-five percent more severe CAT over the North Atlantic since nineteen seventy-nine, driven by a warming climate sharpening the jet stream. No, it almost never threatens the structural integrity of a modern airliner — the certification margins are huge, and the last fatal turbulence accident on a US carrier was nineteen ninety-seven. And the perception gap is vast — passengers feel zero point five g as a crisis, the airframe is certified to three point eight positive and one point five negative, and pilots sometimes call two g "moderate."
The misconception I keep coming back to is the idea that turbulence can rip the wings off. It can't — not in a modern aircraft flown within its envelope. The wing bends, a lot, by design. If you're sitting over the wing and you look out the window during rough air, you can see it flexing. That's not failure. That's the wing doing exactly what it was engineered to do.
The flex is the safety. A rigid wing would snap. A flexible wing absorbs the energy and returns to shape. The Boeing seven-eight-seven's wing can flex something like ten feet at the tip before you're anywhere near the limit load.
The open question I'm left with is whether the trend line eventually erodes the margin. If CAT keeps increasing at this rate for another forty years, do we reach a point where the probability of a two point five g hit becomes high enough that certification standards need to shift? Or does forecasting get good enough that we just route around it?
I think it'll be forecasting. The physics of CAT formation is better understood every year, and the machine learning approaches are promising. We might get to a point where the cockpit gets a real-time turbulence map as detailed as the weather radar, but for clear air. That would change the game entirely.
Until then, the advice is boring and true. Keep your seatbelt on. The plane is fine. You might not be.
The next time your drink jumps off the tray table, remember: the plane dropped ten feet. It's got another thirty-five thousand to go. It knows what it's doing.
This has been My Weird Prompts. Thanks to our producer Hilbert Flumingtop, who apparently still has a shim from a triple seven test program in a box somewhere.
If you enjoyed this episode, leave us a review wherever you get your podcasts — it helps. Find us at my weird prompts dot com.
We'll be back soon.
