On Friday, I posted a speculation about the causes of the Flt 3407 crash in Buffalo on Thursday evening. I appended an update on Saturday, indicating that the NTSB had reported that the plane's cockpit voice recorder indicated that the pilots were aware they had had a significant icing encounter, though no other particulars were forthcoming.
In that update, I also mentioned that a pilot friend had observed that flying a "coupled approach" (one that uses the autopilot) was not uncommon among commercial pilots, and that if that was the case here, it might mask the effects of icing by re-trimming the airplane automatically, until the effects were quite severe. When an autopilot can no longer deal with the conditions in which it finds itself, it typically sounds an alarm and then shuts off (the pilot's joke is that the machine is effectively saying "OK, hotshot, you got us into this mess, you get us out.").
Since that update, two additional facts have been revealed by the NTSB, both important. First, the crash site investigators have said that contrary to initial reports, all four corners of the aircraft are still recognizable, the nose and both wingtips in addition to the tail. That means that while the concentration of wreckage means the plane hit the ground on a very steep trajectory, it likely didn't crash in a steep nose-down attitude. That doesn't completely rule out the "elevator stall" theory I was espousing, but it does make it much less likely.
The second revelation, that the plane was on autopilot when it crashed, certainly lends credence to my friend's observation about the dangers of such a practice in icing conditions, and indeed, it appears that the airline has a policy of requiring that pilots hand-fly the airplane in potential icing situations. It starts to look more like the plane got a load of ice in a freakishly short period of time, and the wings stalled, possibly at the same time as the elevator. Data recorder analysis will indicate the exact sequence of events, but it's likely that the plane either was too tail heavy (with ice) or too close to the ground to recover, and pancaked into the ground.
Obviously, this too is speculation, and more may well be found wrong, but my pilot's instincts to speculate as an exercise are overwhelming my lack of training as a journalist, which at least should be telling me not to. At least I'm apparently in a lot of journalistic company, but that doesn't really make it right...
49 50 people died last night in a commuter airplane crash just outside Buffalo. An NTSB crash analysis "Go Team" is on the scene, and will no doubt be making their usual thorough study over the next several weeks and months. Fortunately for that analysis, it looks like the only part of the plane which wasn't completely destroyed by the crash itself and the intense fire that followed was the tail, and that's typically where the "black boxes" are mounted, so there's a good chance that the cockpit voice recorder and the telemetry record are preserved. (Update: They have been recovered.)
In advance of that analysis, it's purely speculative (and in the media, pretty irresponsible) to say what made the plane come down, but since I'm pretty much talking to myself here, and irresponsibility is the hallmark of bloggers everywhere, I'm going to engage in a little speculation anyway.
I'm betting that the cause of the crash was an elevator stall caused by icing. Here's why:
1 - There was freezing rain near the flight path, which the plane may well have flown through.
2 - The plane was on final approach to Buffalo's airport (it apparently crashed right near one of the radio beacons marking the instrument approach), which means that it would have normally been gradually losing both airspeed and altitude in preparation for landing. It's last known altitude was about 2000 feet above sea level, or 1500 feet above the ground.
3 - There was no report of trouble from the crew to the approach controller, which means that whatever happened was very quick, no time to report the problem.
4 - The approach controller's last instruction was to turn the plane. The co-pilot read back that instruction, then called approach back about something (the call didn't sound out of the ordinary, but it was just the call sign, so it's hard to be sure) and went off the air. The approach controller tried to raise the plane again to respond to the co-pilot's call, but couldn't.
5 - The plane ended up in a burning pile, taking out only one house,
and producing almost no debris field. This means that it came pretty
much straight down - it wasn't flying anymore, in any real sense of the
6 - The tail survived much better than the rest of the plane, which suggests that the plane may have come down in a steeply nose-down attitude, allowing the destruction of the rest of the plane to absorb some of the forces that would otherwise have broken up the tail.
The plane was a DeHavilland (now Bombardier) Dash-8, which is a high-wing, twin-engine turbo-prop with its elevator at the top of the vertical stabilizer (or tail). I don't know the characteristics of the plane, and it's a very well-regarded airframe, but (and this is really wild-assed speculation)...
a) It looks to me like it used de-icing "boots" (inflatable bladders on the leading edge of the elevator, and possibly on the wing as well), instead of "bleed air" from the engines to warm the surfaces as a jet does, or a "weeping wing" (which dribbles anti-freeze onto the foil surfaces and is a common solution in smaller planes).
b) The Dash-8's elevator is at the top of the tail, which means it sits up there by itself, without any wing in front of it to disturb the flow of air that might be carrying a load of ice to it. I dont know, but it may be that its position allows it to accumulate ice faster. I think it also allows the elevator to be a bit smaller, which may make it a little bit more susceptible to icing effects.)
Here's a possible scenario:
The plane is descending normally for landing, and flies through a patch of freezing rain. This is the worst possible type of precipitation for icing, and can put a great deal of ice on a foil in a very big hurry. It can be particularly bad on descent, because the airframe has been up higher, in colder air, and the airframe can be super-cooled, making ice form on its surfaces that much faster.
The cockpit crew, realizing that they've encountered ice, begins to use the "boots" to pop the ice off the wing and elevator. There are three possibilities here, that the boots failed to inflate for some reason, that they were used too often in rapid succession (one of the failings of boots is that they can become ineffective if used too much too quickly), or that ice accumulated on the wings and elevator surfaces behind the boots in sufficient quantity to bring the airplane down anyway. (There is actually a fourth possibility, that the crew didn't realize that they'd had a serious icing encounter or didn't engage the de-icing system at all for some other reason. That seems unlikely - there'd be ice all over the windscreen - but they didn't report the encounter to approach, so it's possible.)
Neither the wings nor the tail can be seen from the cockpit, so there's no reason for the crew to know that the boots aren't succeeding in shaking the ice off. The plane keeps descending, and losing speed. The ice on the wings and elevator becomes more dangerous as the plane gets slower, because the foils generate less lift at slower speeds.
Then, the controller turns the aircraft toward its approach course. In a turn, some of the lifting force is used to turn the plane, so the plane has to stay in the air on the strength of the lift that's left. It sounds like there wasn't enough left to keep it aloft, and down it went.
Why an "elevator stall", instead of a wing stall? Well, both are possible, but it's useful to know a couple of important details about what a "stall" is in an airplane, and how planes fly.
-- A "stall" is when a plane stops flying and becomes a bunch of heavy equipment in the wrong place (up in the air), and that happens when the angle at which the wind meets the airfoils (the "angle of attack") becomes too great for the foils to generate lift.
-- All airplanes are nose-heavy - they're designed that way, so that in case the plane stalls, the nose immediately drops, the angle of attack of the wings will decrease, they will regain effective airspeed and the plane will start to fly again.
-- That means that while the wings generate lift that is used to keep the plane in the air, the elevator is actually flying "upside down" in normal flight, and its lift is forcing the back of the plane down to counteract the force of gravity on the front of the plane.
When the wing loses lift, it stops flying and stalls, but at 1500 feet, it should have some time to decrease its angle of attack and start flying again (maybe not enough to avoid a crash, but enough to have the plane impact the ground at something other than nearly a straight-down attitude).
When the elevator loses lift, though, it stops holding up the front of the plane, and the plane nose-dives. These sorts of stalls are often unrecoverable, even if there is a lot more than 1500 feet to do it in.
Speculation like this in advance of the evidence really is irresponsible, and the investigation may well reveal some other factor that caused the crash, but from what little we know now, I think this is a plausible explanation.
May they rest in peace...
(Late Update: Apparently, the cockpit voice recorder does indicate that the crew had an icing encounter they were dealing with - I don't know what else it says. Also, a friend points out another possible factor. These pilots often fly what are called "fully coupled" approaches, which means the autopilot is flying until the plane is very close to the ground. If that was the case here, and the plane picked up a load of ice, the autopilot could have automatically re-trimmed the airplane to compensate until it could no longer do so. When it can't cope any more, it would shut off and leave the pilot with a very poorly trimmed airplane to deal with. Alternatively, the pilots could have shut the autopilot off themselves upon receiving the turn instruction from the approach controller, and discovered the same sort of problem in the middle of a turn. Either case could be very dangerous, particularly close to the ground, as they were.)
Yesterday, I wrote something that probably tripled the total number of people in my life that have read something I wrote. Here it is.
Unfortunately, it was a quick note to Josh Marshall, and the first paragraph is badly phrased. I described the US Air ditching in the Hudson on Thursday afternoon as being less "last minute" than many commercial airliner ditchings, but that's not true. What is true is that the US Air plane was under impressive control throughout the ditching, partly due to the early decision to ditch and the planning that went into executing the water landing.
The landing itself was really something - this morning, the mayor's office released a surveillance video that shows a textbook ditch - tail down, as slow as possible. Capt. Sullenburger also appears to have kept the plane from spinning, which is even more impressive once you realize that the plane appears to have lost one engine as it landed, creating uneven drag on one side of the plane.
Sullenburger is a true hero, and saved a lot of lives on Thursday (he'll never buy a drink in a bar with another pilot in it again, I'll bet).
As someone else pointed out on TPM, a lot of credit also should go to the ferry captains who acted so quickly to get people off the plane. This video shows that there was only about four minutes from the time of the impact to the arrival of the first ferry, and there were two there in six minutes and four by nine minutes after the crash. All of those ferries got up to the plane slowly, safely, and ready to take passengers off, and that's not easy in a large vessel either.