Bad Breaks? Or a System That's Breaking?
Commercial air travel has been extremely safe. It will take work to keep it that way.
I’ll say it again: Statistically, commercial aviation is the very safest part of modern life.
More than 100 Americans die every day in traffic accidents. More than that die in their own houses, from accidents and falls. On average about 30 Americans per year are killed by lightning strikes.
In comparison, over the past 13 years, when U.S. airlines have conducted more than 10 billion “passenger journeys”—one person on one flight—a total of two people have died in U.S. aviation accidents. On average, once per second a commercial flight takes off somewhere around the world. Virtually all these trips are uneventful. We hear about the ones that aren’t.
But we’re hearing more and more about things that almost went wrong, about close calls and near catastrophes. I have been writing about recent episodes in New York, Hawaii, Texas, and elsewhere.
-Are these sheer coincidence? Results of the law of large numbers, meaning that even a nearly perfect safety record will include some exceptions? Over-attention by the press, making “incidents” into a “trend”?
-Or should they be taken as omens of a system cutting too far into its own safety margins—coasting on its reputation, pushing its luck? The pre-dawn close call last month at Austin’s airport is the most frightening in this regard.
We all hope that these recent cases will eventually be seen as just a temporary rough patch, rather than as “Danger!” signs not given proper heed. A lot of that will depend on how our aviation-safety establishment responds to what these episodes might indicate.
For decades the commercial-aviation culture has been remarkably good at something that is hard for individuals or organizations. That is looking systematically and dispassionately at its own weak spots, learning from mistakes, anticipating problems, and building in extra layers of protection. (The financial system is showing us right now the emergencies it has, and has not, prepared for.) That is what the aviation culture needs to do again.
In China Airborne I wrote about a moment of extreme safety crisis for China’s nascent airline industry. Within the first few months of 1992, five airplanes from its state-owned airlines crashed. To their credit, the Chinese authorities realized that they needed to change everything about the way they operated, from pilot training to aircraft inspections. They worked closely with Boeing, with U.S. advisors from the FAA and the NTSB, with flight trainers from United Airlines, and others. China’s airlines had been as dangerous as those of other developing countries. Within a decade they became among the safest in the world.
Obviously the incidents I am compiling here are nothing like what China was going through. So far in these U.S. episodes no one has been killed. But a lot of people could have been.
1) December 18, 2022, Maui. Plunge toward the ocean on departure.
I wrote about this episode last month but am raising it again because of some larger implications.
Soon after takeoff from Maui, a United 777 dived straight toward the ocean, and came within a few seconds of plunging in. It was less than 800 feet above the water, and descending at some 8000 feet per minute, when pilots reversed the descent. They regained altitude with an emergency 2.7G climb.1
The incident was not publicly known for months. Jon Ostrower of The Air Current broke the story here, in February. (Because no passengers were injured, it didn’t have to be formally reported to the NTSB.)
Here are two reasons I mention it again.
An informative video.
My first purpose is to highlight yet another informative video by the professional pilot Juan Browne. In the YouTube segment below he spends 11 minutes examining possible reasons for the descent.
Might it have been extreme turbulence or downdraft? That same day, a Hawaiian Air flight hit turbulence so severe that several dozen passengers were injured. But that flight was at 38,000 feet; the United flight was all too close to sea level.
Might it have been mis-management of cockpit automation—especially a missed altitude setting on the autopilot? Browne describes several other near-disasters outside the U.S. that support this hypothesis, and explains what the crew might have overlooked.
Like everyone with experience in aviation, Browne suspends judgment until more facts are in. But you can hear his concern.
A loophole that could be closed, tomorrow.
The other reason to mention the incident involves a loophole in accountability. That is the absurdly limited capacity of “cockpit voice recorders.”
Radio transmissions between flight crews and air traffic controllers are all preserved. After an incident, everyone can go back to hear who said what.
But exchanges among the flight crew in the cockpit are captured on cockpit recorders that generally have a two-hour run time. After that, the tape loops back and records over itself. So by the time the Maui plane landed in San Francisco four-plus hours later, there was no way for anyone outside the cockpit to know what the crew members were saying to each other during takeoff and the ensuing crisis. The same was true for the Southwest plane in the close-call in Austin (which then flew on to Cancun), and the American plane that taxied onto the wrong runway at JFK (which then flew on to London).2
In my little propeller plane, I can connect a tiny digital recorder to the intercom, and it will keep running for hours. Surely the airlines and the aircraft companies can solve this problem. Like, tomorrow.
2) January 23, 2023, Honolulu. One plane turns into another’s path.
For this I’ll refer you to local media—Hawaii News Now and Beat of Hawaii — and the Aviation Safety Network, which has a very detailed account.
In brief: the Inouye airport at Honolulu includes two parallel runways, 4 Left and 4 Right. A small turboprop cargo plane was cleared to land on 4L, the shorter runway. That’s shown in green. Meanwhile a big United 777 was cleared to land on 4R, as shown in red. This kind of “parallel runway” operation is routine.
After the United plane landed, it was cleared to turn left off 4R onto Taxiway Kilo. You can see that intersection highlighted by the FAA as “HS 2” — meaning a “Hot Spot” demanding extra caution while taxiing. The United plane crossed into the path of the cargo plane cleared for landing on 4L. The planes were not in real danger of collision but came within 1200 feet of each other.
The who-what-when is complicated, and all laid out in Aviation Safety Network. The point for now is: this was a closer call than the system is supposed to permit.
3) February 4, 2023, Austin. Near collision.
I’m mentioning this one again because the NTSB has come out with its preliminary report. It includes this startling time-cued graphic of how close the planes came to an actual crash. Purple is the FedEx plane that was about to land and then suddenly “went around” as soon as it saw the Southwest plane in front of it. Red is Southwest taking off.
The NTSB report is here; an AVweb article about it is here. As I noted in a previous post, this was “as bad as it gets without body bags.”
4) February 22, 2023, Burbank. One plane taking off while another is preparing to land.
At the Bob Hope Burbank airport outside Los Angeles, controllers cleared a United commuter plane to take off, while an American commuter plane was cleared to land on that same runway. The American plane was 1.3 miles out, or maybe 30 to 40 seconds from touchdown.
That spacing is not shockingly close for busy airports on good-weather days. But to the American crew it looked unsuitable. They aborted the landing and began a “go-around.” Meanwhile the United plane continued its takeoff and climbed toward the American plane’s path. This triggered a “TCAS alert,” the automated collision-avoidance system, in the American cockpit. The controllers eventually vectored the planes away from each other.
Aviation Safety Network again has more details. Another report is on AVweb, with a link to transmissions from the Burbank tower. The relevant part starts at about time 29:00 of this LiveATC clip.
This was not as close a call as the one in Austin, mainly because the weather and visibility were better. But the fundamental issue was the same. The controllers thought the planes had adequate spacing; the inbound pilots did not.
5) February 27, 2023, Boston. Unauthorized takeoff into another plane’s path.
This one got a lot of local and nationwide coverage, for instance here and here.
At face value it seems a simpler case than the others. A Learjet was instructed to “line up and wait” at the end of Runway 9. That phrase has a specific aviation meaning, spelled out here. It means: “taxi onto a runway but do not take off.”
When the tower controller tells the Learjet to “line up and wait,” which you can hear around time 0:33 of this audio, she goes on to explain why. She adds “traffic will land on crossing runway.”
Here is what she means. The Learjet’s path along Runway 9 is shown in red. But a JetBlue plane has already been cleared for landing on Runway 4 Right, as shown in green. The two paths intersect.
Normally a plane told to “line up and wait” would remain in place until it heard “cleared for takeoff.” For whatever reason, rather than waiting, the Learjet crew pushed in the throttle and took off.
The tower controller immediately saw what was happening and told the JetBlue crew to abort the landing and “go around.” You can hear that at about time 0:40. You’ll also notice how unflappable the controllers sound, a trademark of their professional culture.
What is to be done?
What do these episodes have in common? All appear to arise not from mechanical failures or electronic glitches—or even weather, which is the main cause of most aviation problems. (Though weather was a factor in Austin and could prove to be in Maui.)
Instead they are mainly about human judgment, communication3, adherence to procedure. Some of the pilots and controllers involved in these episodes showed themselves at their best. Some did not.
A robust safety system can’t depend on people always being at their best. (Just as a robust financial system can’t depend on people always being honest.) In a column in AVweb, Paul Bertorelli recently made this point about the Austin near-disaster. He said that the acting FAA administrator, Billy Nolen, told a Senate committee that in Austin “the system worked as it is designed to avert… a horrific outcome.”
Bertorelli responded:
Well, yeah, if by “design” he meant the sheer dumb luck of having a [FedEx] crew on approach that had unusually prescient situational awareness and saved the day with a last-second go-around. Somehow, I don’t imagine this was built into “the system”…
The overarching question raised—fairly by the committee, I’ll admit—was whether [all these episodes] are leading indicators of a system unraveling. It’s wholly proper to ask this, in my view, because the answer could just as easily be yes as no.
I don’t know the answer. But this is the right question. And remember the stakes:
Huge aircraft, some weighing hundreds of tons, were headed at high speed toward each other. All were carrying highly flammable jet fuel. Those taking off had brim-full tanks for their journeys. And many had full loads of passengers.
During aviation emergencies, controllers tell pilots, “Say souls on board.” That is the unvarying code for: Tell us how many lives are at stake.
The stakes in understanding what is going wrong, and fixing it, are very high.
For perspective: 2.7G is more force than most people have experienced except on a roller coaster. If you weighed 180 pounds, the force would press you into the seat as if you weighed nearly 500 pounds.
The American crew from JFK had to be subpoenaed before it would answer questions from the NTSB.
From a friend who is a long-time pilot:
”Near misses seem to have fallen off the front pages, or at least below the fold. The aspect of the Boston one that most intrigues me is that I think pulling the whole sequence of communications together requires looking at the tapes of 2 ground, 2 tower and 4 departure frequencies covering the Class B overhead. And maybe there is a tower cab recorder for the conversations off the radio. JFK, ORD and LAX have got to be even worse. How do the controllers coordinate?”
It seems remarkable that in this age of digitization that the cockpit recorders only tape two hours. Particularly since aircraft have become so reliable that they can fly multi leg flights that take nearly all day and night.
Question for James and other knowledgeable people: How much do the systems depend on the relevant people paying full attention for as long as possible?
I ask because, anecdotally, it feels like people get distracted more easily than, say, 20 or 30 years ago. We've all experienced this. Is it possible that a decrease in what's considered paying attention could increase incidents in a system with a now-dated definition of paying attention?
I'm not suggesting anything as egregious as the Lear pilots stealing glances at their phones. Just what would seem the inevitable consequences of how most of us live now.