The short answer
Ice keeps bringing down planes because so many fixes were written for the ice that had just killed people: a warning light for the speed probes, a table of minutes for the wait on the ground, bigger boots for one wing. Ice kept finding the gaps those fixes left. It has brought down planes for more than fifty years, on the ground, in the clouds and very likely in air too cold for liquid droplets.
That chain is our reading of the record. No investigator has written it down as one story, and some fixes came before a fatal crash, such as the ATR orders that followed non-fatal upsets. NTSB report AAR-96/01, page 80
The question is back because of the Palm Beach Learjet. On November 1, 2026, its crew told controllers “they were encountering light, clear ice, which was manageable,” and minutes later the jet started to fall. All nine people on board were killed. NTSB briefing, November 2 Nobody knows yet whether ice played any part in that crash. The NTSB says weather is a major focus of the investigation, and that it won’t speculate about the cause. NTSB briefing
Watch the chain
The episode runs the news footage, NASA flight film and report pages behind each crash, with diagrams of how ice fooled the instruments and got past the boots.
1974: The dial that lied
On December 1, 1974, Northwest Flight 6231, a Boeing 727 with three crew and no passengers, left New York for Buffalo. Small tubes on the outside of a jet, called pitot tubes, measure its speed from the air rushing in. They have heaters to keep ice out, and that night the heater switches had not been turned on. NTSB report AAR-75-13, pages 1 and 20
At about 16,000 feet, ice blocked the tubes and trapped that height’s air pressure inside. A speed dial reads the difference between the pressure in the tube and the pressure outside. As the jet climbed and the outside air thinned, that difference grew. The higher the jet went, the faster the dial said it was going, even as it was really slowing down. AAR-75-13, pages 13 and 20
The crew raised the nose to slow down, which made the dial read faster still, and the overspeed horn and the stall warning went off together. When the stall warning first fired, the recorder showed 412 knots, about 474 miles an hour. A working dial would have read about 165 knots, or 190 miles an hour. The jet fell from 24,800 feet in 83 seconds. The NTSB said the crew “failed to recognize and correct” the stall. Near the end, the force of pulling out failed part of the tail, and all three died. AAR-75-13, pages 7, 10, 20 and 21
The fix looked simple: a light that tells pilots the heat is off. The NTSB asked for one in 1975. The FAA wrote back, “We do not concur.” AAR-75-13, PDF pages 29 and 31 In 1978 it required one anyway. 43 FR 10338 The rule doesn’t name the flight, but a 1979 industry petition that the FAA printed says it came from the NTSB’s recommendation after the Northwest crash. 46 FR 78
1982 and 1992: Ice on the ground
Air Florida Flight 90 sat in a Washington snowstorm for about 49 minutes after it was deiced on January 13, 1982. With the engine heat left off, a probe on each engine iced over, so the gauges showed the target reading of 2.04 while the engines were really at 1.70. By our arithmetic from the report’s figures, that’s about a quarter less thrust than the crew thought they had. There was ice on the wings too. The jet hit the 14th Street Bridge and went into the frozen Potomac, and 78 people died, four of them on the bridge. The NTSB wrote that “either condition alone should not have prevented continued flight.” NTSB report AAR-82-8, pages 5, 27, 29, 80 and 81
Taking off with ice on the wings had been against the rules since 1950, NTSB letter A-82-6 to -15, page 2 and this crew knew there was snow on theirs. AAR-82-8, pages 80 and 114 So how long does deicing fluid protect a wing? That December, the FAA put its answer in writing: “the time that may be considered a safe interval between ground deicing and takeoff cannot be estimated.” FAA AC 20-117, page 3
Ten years later, on March 22, 1992, USAir Flight 405 was stuck in line on a snowy night at LaGuardia. This crew was careful: they had the jet deiced twice, and the first officer checked the wing with a light again and again. “Looks pretty good to me from what I can see,” he said, about five minutes before takeoff clearance. But the jet had sat in the snow for 35 minutes, on fluid the NTSB calculated was good for about 11. At night, the NTSB found, crews can’t see minute amounts of ice from the cockpit. A layer as rough as medium or coarse sandpaper can cut a wing’s lift by as much as 30 percent. NTSB report AAR-93/02, pages 3, 36, 74, 75 and 100
The jet stayed in the air about 11 seconds, and 27 of the 51 people on board died. This time the NTSB named “the failure of the airline industry and the Federal Aviation Administration to provide flightcrews with procedures, requirements, and criteria compatible with departure delays in conditions conducive to airframe icing.” Its probable cause also named the crew’s decision to take off without being sure the wings were clean. AAR-93/02, pages 6, 50 and 77
From November 1, 1992, U.S. airlines needed deicing programs with tables of how long the fluid lasts. Once the time ran out, a crew needed a fresh check within five minutes of takeoff, an approved alternate procedure, or another deicing. The tables are for planning, and they aren’t hard limits. 57 FR 44924, 67 FR 54946
Since then, our search of NTSB records finds no fatal crash of a large U.S. airliner blamed on ice picked up on the ground. That’s our own reading of 55 fatal Part 121 records, and the last published statement we found dates from 1996. DOT Inspector General, page 7 It doesn’t cover smaller scheduled flights. PenAir Flight 350 in 2001 and Hageland Flight 500 in 1997 were both lost to ice or frost on the ground.
1994: Outside the rulebook
In January 1990, a NASA film said, “Icing is no longer a major problem for many of today’s aircraft.” NASA Lewis Research Center film
On October 31, 1994, American Eagle Flight 4184, an ATR 72 with 68 people on board, was holding over Indiana, waiting to land at Chicago. The crew had the deicing boots working, rubber strips on the wing’s front edge that inflate to crack ice off. But this was freezing drizzle. Some drops were estimated at over 100 microns and some as large as 2,000, against a certification limit of 50. Big drops run back past the boots and freeze where nothing can reach them. Investigators concluded a ridge of ice grew there, ahead of the ailerons, the hinged panels near the wingtips that roll a plane. Nobody saw it on this airplane; tanker and wind tunnel tests reproduced it. NTSB report AAR-96/01, pages 29, 30, 95, 158, 159 and 203
When the flaps came up, the autopilot lifted the nose and the air over that ridge broke away. In about half a second the ailerons went almost to their stops, faster than a pilot or the autopilot could move them. No stall warning sounded. About 24 seconds after the ailerons first moved, the recording ended. All 68 people on board died. AAR-96/01, pages 8, 63 and 91
About two hours later and some 50 miles away, by their own account, another ATR 72 crew felt their plane shake. They added power and it stopped. Then a light showed a ridge just behind the boot, about half an inch high, and they changed nothing else until the ice began to melt. They landed normally. ALPA comments, July 2001, PDF pages 13 and 14 Later tests found that half an inch was enough. AAR-96/01, page 96
The NTSB had flagged the gap 13 years earlier. In 1981 it asked the FAA to “expand the certification envelope to include freezing rain.” The FAA called that “excessively penalizing and economically prohibitive.” On September 16, 1994, the NTSB received the FAA’s latest reply: the icing criteria were “adequate.” That was 45 days before Indiana. The NTSB later wrote, “If the FAA had acted more positively upon the Safety Board’s aircraft icing recommendations issued in 1981, this accident may not have occurred.” AAR-96/01, pages 130 to 132 and 204
The first fix came fast. Every U.S.-registered ATR had bigger boots before June 1, 1995, about seven months later. AAR-96/01, pages 39 and 40 The rule took 20 years. The FAA’s own 1997 schedule set it for March 2001. 62 FR 64622 In February 2010 it still wasn’t done, and a congressman asked the NTSB’s chair, Deborah Hersman, whether her agency was satisfied with the FAA’s progress. “No, Congressman Coble,” she said. House hearing, February 24, 2010
It came on November 4, 2014, for new designs and for changes that touch the affected areas. The FAA wrote, “we do not agree that the part 25 and part 33 rule changes discussed in this amendment should apply to existing airplanes.” For those, it pointed to orders on when to switch on ice protection and when to get out of severe icing. 79 FR 65507, at 65508 and 65514
In 2024, an ATR 72 in Brazil, Voepass Flight 2283, lost speed in severe icing and fell in a flat spin. All 62 people on board died. CENIPA interim statement That was a different failure from Indiana. Press accounts of the final report say its deicing fault had shown up on earlier flights and was never formally recorded. Reuters, via Tuoi Tre News
Where rain is impossible
Around 1960, a Navy training film told pilots, “The danger to the flying aircraft is not from the ice crystals, but from the water droplets which freeze upon contact.” U.S. Navy, Ice Formation on Aircraft That’s still true for a cold wing. The film says nothing about engines or probes.
For years, jet crews near big storms reported heavy rain on the windshield, at heights and temperatures where rain is not possible. Sometimes an engine then lost power. By about 2008 those reports had been tied to ice crystals, perhaps as small as flour, melting on the heated glass. AeroSafety World, June 2008 A cold wing shrugs them off. On anything warm, like a heated probe or the inside of an engine, they can melt, stick and freeze again. Boeing AERO, 2007
So the lying instruments came back, sometimes with the heat on. In 2009, crystals likely blocked the speed probes on Air France 447 over the Atlantic. The autopilot quit, and that started the chain that killed all 228 people on board. BEA final report In 2014, Air Algérie 5017 crashed in Mali. Investigators later said crystals probably blocked a sensor on each engine, the same kind of probe that fooled Air Florida’s crew, with the engine heat off. The engines were left short of power, the autopilot raised the nose until the jet stalled, the crew reacted too late, and all 116 people on board died. The report couldn’t say whether the heat was left off by oversight or by choice. BEA report on Air Algérie 5017, pages 125 and 126
NASA and the FAA went looking. In 2015 they started flying a research DC-8 into these storms, first out of Fort Lauderdale. With the heaters working, on probes built to an older standard, crystals still plugged the tubes. Of 71 anomalies logged, 54 made the reading drop. The other 17, about one in four, trapped the pressure, and NASA wrote down what happens then: “ascending will cause the indicated airspeed to increase.” NASA doesn’t mention the 727 from 1974, and the crews’ situations were different, but it’s the same effect. That kind usually happened colder than minus 40 Celsius. Ratvasky and others, NASA, 2024, PDF pages 1, 10 and 22
Back to Palm Beach
Clear ice forms from water droplets that are still liquid below freezing, and the FAA puts the limit for those droplets at minus 40 Celsius. FAA Aviation Weather Handbook, pages 20-2 and 20-3 In August 1999, scientists flew a Lear jet into growing storm clouds in Texas and measured liquid water down to minus 37.5. A little colder, it was gone. Rosenfeld and Woodley, Nature, 2000 But the Palm Beach crew reported clear ice at minus 47, and then said, “We are picking up precipitation.” NTSB briefing, November 2
The FAA’s pilot guide has one line that might explain it: crystals “may melt on a heated surface, run back, and refreeze.” The same guide says crystals “typically do not accrete on external airframe surfaces.” FAA AC 91-74B, pages 3 and 11 Nothing on record says crystals were there that day, or that this jet’s speed readings were wrong. Neither NTSB briefing mentions crystals, pitot probes or instruments.
A crew talking about ice doesn’t mean ice did it. In 2009, Colgan Air Flight 3407 crashed near Buffalo and 50 people died. The crew had talked about “lots of ice,” and ice was one of the first guesses on the news. A year later the NTSB found “minimal aircraft performance degradation” from the ice. The cause was the captain’s response to a stall warning. NTSB report AAR-10/01, pages x, 3 and 151
The Learjet 60 was approved on January 15, 1993, including for flight in icing, under the older standard. The 2014 rule, the one that added ice crystals, doesn’t apply to airplanes already flying, so it never applied to that model. FAA type certificate data sheet A10CE, pages 22 and 30, and 79 FR 65507, at 65514. Nobody has said that has anything to do with this crash.
As of the morning of November 4, investigators were working to get into the flooded crater. NTSB briefing, November 3 Maintenance records show a cockpit voice recorder was installed, but the NTSB didn’t yet know whether it was still on board. NTSB briefing, November 2 The agency’s first report usually comes out about a month after a crash. When it does, we’ll check it against everything here.
We remember the nine people on that jet, everyone else in this story who didn’t get home, and the people who still miss them.
Primary sources
- NTSB case page, ERA27MA026, with the November 2 and November 3 briefings: the Palm Beach Learjet.
- NTSB report AAR-75-13: Northwest Flight 6231, in the FAA’s copy.
- NTSB report AAR-82-8: Air Florida Flight 90.
- NTSB report AAR-93/02: USAir Flight 405.
- NTSB report AAR-96/01: American Eagle Flight 4184.
- NTSB report AAR-10/01: Colgan Air Flight 3407.
- BEA final report: Air France 447.
- Final report, English: Air Algérie 5017.
- CENIPA record: Voepass Flight 2283.
- FAA AC 20-117 and AC 91-74B: ground deicing and flight in icing.
- FAA Aviation Weather Handbook: chapter 20, icing.
- Federal Register: the 1978 warning light rule, the 1992 ground deicing rule, the 1997 tasking and the 2014 icing rule.
- NASA, Air Data Probe Anomalies in High Ice Water Content: the DC-8 flights.
- Reporting: Reuters, via Tuoi Tre News on the Voepass final report.
Sources reflect the public record as of the morning of November 4, 2026. Quotes from the NTSB’s briefings and the NASA and Navy films come from machine captions. CENIPA’s final report on Voepass was not read for this article; the line on its deicing fault comes from press accounts. Miles an hour and the thrust figure are our conversions. The search behind the line on ground ice is our own reading of the NTSB’s public CAROL records, not an NTSB statement.
