Narita MD-11 crash: the disaster sequence examined

The prevailing situation:

The FedEx MD-11 was approaching runway 34L at Tokyo Narita, with fairly high gusting winds forecasted. Gusting winds always raise the spectre of potential windshear, and Narita is renowned for it.

The forecast wind (320deg at 26kt gusting to 40kt) would have provided a crosswind from the left that was some 20deg off the runway heading, although that may not have been what actually prevailed on landing.

So although it was not going to be particularly easy to land any aircraft type in an elegant way that day, the conditions were far from extreme and the visibility was excellent.

Now watch the video below, be ready to pause it from time to time to examine the very rapid transition from a relatively normal landing to a disastrous one, and then check the text below for my interpretation of what you are seeing at each point:

 

The landing sequence shown in the video tells us the following:

1. On the last part of short final approach the aircraft appears to be stable, if slightly low, with wings level and a normal pitch attitude for the circumstances (given that we don’t know what the airspeed is);

2. the touchdown is very firm, but under gusty circumstances the pilots would naturally aim for a firm touchdown;

3. the nosewheel was lowered onto the runway at a high rate. Although the crew would want to put the nosewheel on the runway quickly to stop the aeroplane flying, the rate at which it was lowered might have threatened damage to the nose-gear – but it looks as if it survived the impact anyway.

Note: up until this point the aircraft’s landing performance and behaviour has been well within the normal range. But then:

4. immediately following nosewheel touchdown the aircraft pitched up dramatically and the aircraft ballooned into the air again.

Note: the rate at which the nosewheel was lowered may have been a part of the cause of  the pitch-up following nose oleo compression, and that pitch-up might also have been exacerbated by the automatic extension of the spoilers which, in this type, are renowned for producing a pitch-up moment;

5. Now the aircraft is airborne again. This ballooning following first touchdown might have been made worse by a sudden gust of wind, momentarily raising the airspeed. But if that were true, the spoilers would have been simultaneously destroying a lot of the lift, and producing considerable drag; so, as the gust died (if it did) the aircraft might have been at or below stalling speed;

6. then – and this is what leads to disaster – the nose drops and stays low until the nosewheel’s impact with the runway. This happens either because of lack of elevator authority, or because the pilot flying was tempted into a classic pilot induced oscillation.

Note: there are no circumstances under which a pilot of any type should deliberately select a nose-down attitude at that point – if, indeed, pilot selection of the nose-down attitude is what actually happened. During ballooning following touchdown the nose MUST be held up (if the elevator authority allows it) and appropriate power applied, whether the crew are trying for a successful second touchdown or for a go-around.

7. finally the nosewheel hits the ground extremely hard and the nose instantly rebounds upward, the main gear touching the surface momentarily a fraction of a second later. Almost simultaneously, the aircraft begins its fatal bank to the left, from which recovery was impossible once the left wingtip had hit the ground.

Note: banking to the left is not what the forecast crosswind would have been expected to produce. Normally, especially in a swept-wing aircraft like this one, the upwind wing has a tendency to lift, but in this case it didn’t. So the crosswind does not appear to be the critical factor here, although windshear is very likely to be one of the causal factors.

For all those with different interpretations (or even to agree), please feel free to file your comments,

38 Responses to Narita MD-11 crash: the disaster sequence examined

  1. peter haller 24 March, 2009 at 3:15 pm #

    Swissair MD-11 captain told me once,that in last stage of landing,depending on wind you sometimes had practically no rudder or elevator authority…

  2. Uwe 24 March, 2009 at 5:11 pm #

    When does the port wing break?

    At second touchdown or during the roll/backflip?

    My impression is that the swash of fire starts
    before the backflip ( and near the fuselage ).

    uwe

  3. michael 24 March, 2009 at 5:23 pm #

    it can be also a common crewcoordination mishappening occuring within seconds between co-pilot flying initiating a go around after first touch and the captain overriding to land within half second later, so with a delay of a second all inputs are coming a bit time shifted causing oscillation and a general problem four hands are generating at the columm especially in this conditions, but of course it can be also a ripped off freight in a half full freighter or a combination of everything which has been written so far, as also unknown reasons.

  4. michael 24 March, 2009 at 5:40 pm #

    the bank starts same time the left mainwheel touches down, letting me think it collapsed simultaniuosly while touching down, initiating the the bank, with the wind coming from the left, left aileron was helping that roll moment.

  5. Uwe 24 March, 2009 at 6:11 pm #

    Well, some white mist appears behind
    the craft from second touchdown onwards.
    ( when the nose is just under the tower+boxontop )

    The fire seems to start from under the tail/portside
    elevator when the starbord wingtip just goes past
    the right side end of the box_on_the_tower.
    The flip accellerates here and the flames
    race (flash?) forward up to the wing root.

    The port engine seems to scrape the ground
    for most of the time.

    uwe

  6. Russell Southgate 24 March, 2009 at 7:38 pm #

    I was very interested in the sequence of events during the touchdown of the MD11. I certainly had missed many of the aspects that David Learmount notes. However the most fascinating aspect, probably just a coincidence, is that this is the third MD11/DC10 this company has lost in a landing accident in recent years. I presume there is no connection.
    Photos of these aircraft types laying in a crumpled heap and inverted seem somehow familiar.

  7. Jay Pierce 24 March, 2009 at 7:49 pm #

    In my estimation, the left rear spar fails on the second impact, ripping the wing box open. The dense fuel mist cloud can be seen behind the plane very briefly, maybe 100-200 feet past where the left main hits the ground, just before the fire flashes. The bounce of the nose gear causes an abrupt increase in left wing AoA, which rapidly accelerates the left roll rate already started by the lack of support on the left side due to the left wing fracturing. Note the contact of the left engine with the ground, something that could not have happened had the LH gear been intact…

    BTW, the failure mode I describe above is nearly identical to that experienced by another FedEx MD-11 back on July 31, 1999. Reading the NTSB report is very enlightening, but you have to search for it on the web I used the aircraft’s N-Number) because the NTSB site does not contain the complete report.

  8. Kevin 24 March, 2009 at 7:57 pm #

    For the most part, I agree with David’s analysis of the accident. From my personal flight experience from small aircraft and commercial flight simulators, this is what I see:

    During the ballooning, it appears that the aircraft held a nose up attitude for a full second before violently pitching down roughly 40 to 50 ft above the runway. At this point, I see three factors that could have contributed to a sudden loss of airspeed, reducing elevator effectiveness to cause the sudden pitch down:

    1. Spoilers deployed – If the spoilers had been deployed automatically, the aircraft would have lost considerable airspeed during the ballooning regardless of the gust.

    2. The calming of the gust – if the ballooning was indeed caused by a gust, the sudden calming of that gust would have immediately removed an equal amount of airspeed, further stalling the aircraft.

    3. Reverse thrust delay (guess only) – I have little direct experience with large commercial aircraft, but I would like to ask if this is possible: Before the ballooning began, one of the pilots instinctively commanded reverse thrust while the main gear was still in contact with the runway. In this case, the engines would have spooled up in the air right before the aircraft pitched forward, further contributing to the loss of airspeed. A forward pitching moment would also be induced since the engines are below the aircraft’s center of gravity.

    I believe that the sudden pitching forward was unexpected and the pilot did not immediately produce a large enough corrective action. It appears to me that the aircraft stopped pitching down a half second before impact, which leads me to believe that the pilot did eventually apply full aft column, although nearly a second too late.

    On impact, the plane was banked to the left into the crosswind. This makes sense if the pilot instinctively was slipping the aircraft to the left to track the runway centerline. With full up elevator coupled with the impact of the nose wheel, we see the aircraft pitch up suddenly. At this point, I also will guess that the left main gear collapsed, leading to the impact of the left engine and wing on the runway, causing the fuel to spill onto the runway. Sparks from the left horizontal stabilizer ignited the fuel.

  9. Chris 24 March, 2009 at 8:16 pm #

    It would appear that the failure mode is somewhat similar to the HKG China Airlines/Mandarin Airlines MD-11 – after the failure of the wing at or near to the root, the rollover is inevitable.
    What caused the pitching related to the aircraft becoming airborne again and the failure to be able to keep the aircraft in the normal flared attitude which would be normal when trying to recover a “bounced landing” (which this was not since the nose gear was lowered to the ground after the first ground contact) is going to be the most important part of the investigation.
    Unlike the HKG accident, there was a significant impact from an airborne height of well over 10ft when inverted which would have tested the structural integrity of the flight deck section.

    Importantly, we must remember that two professional pilots died in this accident – RIP

  10. J3 24 March, 2009 at 8:30 pm #

    My sense is that when the plane came down after the first big bounce, the left main landing gear broke or collapsed, causing the left wing to hit the ground and roll the plane over. I have seen some post crash photos. They show the right main gear appears to be normally extended, while the left is mangled. What a tragedy!

  11. Romulo Sobral 24 March, 2009 at 8:33 pm #

    Although we can not see the airspeed, it looks like they had excess airspeed during the first touchdown, maybe bringing a couple of more knots guarding against windshear. That could explain the first “bouncing”. What really look strange to me is the abrupt nose drop in the second touchdown. I don’t know if the MD-11 beheaves that way when stalled, but I don’t believe. Even if they didn’t have any elevator authority, is hard to believe that the nose could drop that fast.
    If the spoilers deployed in the first touchdown, they didin’t remain in that position during the ballooning, as long as there were no weight on wheels.
    One should consider any inadvertent displacement of the cargo (CG) during touchdown, altering the balance and making the nose beahaves that way.
    The fact is that we are watching one major accident per week, witch is preety coincident with the projections, taking into the account the increase in the air traffic and the difficulties in further reducing the accident rates.
    Now, we are just sitting and waiting when and where the security cameras will capture the next espectacular, but terrible air accident.

  12. Anonymous 24 March, 2009 at 10:19 pm #

    1) The aircraft appears to leave a vapour trail behind it after the first bounce. This could be smoke from the tyres or liquid from a ruptured tank or pipe.

    2) It’s hard to tell, and could be a visual edge-sharpening artefact from the low-res video, but the starboard MLG appears to extend lower than the port MLG during the second bounce, which could indicate that the port MLG had already collapsed. OTOH:

    3) The aircraft also appears to bank to starboard during the 2nd bounce, rolling to port as the nose drops. That would also account for the starboard MLG appearing to extend lower than the port MLG during the second bounce. The FDR will confirm that.

  13. mike 25 March, 2009 at 2:46 am #

    They bounced the first landing, and might have attempted to force the nosewheel down. The spoilers likely automatically deployed and they bounced high then stalled it and ran out of elevator.

  14. Chagrin 25 March, 2009 at 3:24 am #

    It seems to me that the plane came in too fast and hit the ground heavy damaging the Landing gear after which the Pilots tried briefly to “powerup” to go around but there may have been a load/cargo shift due to the initial first attempt which helped to destabilize the Aircraft. Could it be? The USAir Pilot “Schully” would have put this thing down gently I believe. Excess airspeed the culprit?

  15. jorge pedreira 25 March, 2009 at 8:36 am #

    I heard that the MD11 had some control and pitch stability problems that made it dificult to handle in the final app and flare,
    Pilots found its long fuselage tough to handle in crosswinds and were troubled by a tendency of the nose to pitch up at the critical moment of touchdown.
    Pilots also tell of pulling with all their might and finding the plane hardly responded. That can be a problem during landing. In several instances where pilots brought down their MD-11s too rapidly and tried to compensate at the last minute, they smacked the aircraft’s tail on the runway or caused other damage.

    For its difficulty to be controlled during landing it was called it “the scud” after the Iraq´s unpredictable Gulf War missile.
    Jorge Pedreira

  16. Brian Duxbury 25 March, 2009 at 1:29 pm #

    Two items of data are needed here:

    1 – weight of the aircraft – I would expect a lightly loaded aircraft to ‘bounce’ higher after a firm touchdown required in gusty conditions. Was this the case?

    2 – Did the pilot, realising the aircraft had ballooned to a higher than normal height and realising he was rapidly loosing runway length, over react and force the nose down with a greater than survivable input? The FDR will have this answer.

  17. Speed 25 March, 2009 at 2:44 pm #

    Readers may find the NTSB final report on the FedEx MD-11 crash at Newark in 1997 interesting.

    http://www.ntsb.gov/publictn/2000/AAR0002.pdf

    The analysis beginning on page 51 describes an accident sequence quite similar to this one.

  18. Paul 25 March, 2009 at 4:09 pm #

    Sequence 1 : Very heavy landing, left wing, left gear and engine number two reverser control damaged. Thrust reverser activated seconds before touch down.
    Sequence 2 : Rebound, and a sudden gust of wind, the aircraft is in a stalled condition, engine number 2 producing thrust (downward force), engines one and three producing braking force below the center of gravity, the resulting down forces produced the fast
    nose down movement. At this point the elevator had
    not enough authority to either produce or correct such a rapid change.
    God bless them…

  19. David Learmount 25 March, 2009 at 4:24 pm #

    Paul

    Do you have some reliable inside information here? Because if not you’ve indulged in some interesting speculation about the condition of No 2 engine’s reverser plus suggesting the Nos 1 and 3 were in reverse.

    34L at Narita is very long, and this sequence of events happened so soon after the first touchdown I doubt if the crew had had time to select reverse. They certainly wouldn’t have needed to hurry selection of it.

    Any comment?

    David Learmount

  20. J3 25 March, 2009 at 5:01 pm #

    Does anyone have any information as to whether or not the USAF has had problems such as these with its KC-10s. I believe the AF is today the largest fleet of DC-10/KC-30s operating and perhaps the most flight time on the planes of any user.

    Christopher Dye

  21. Uwe 25 March, 2009 at 6:59 pm #

    To 3 on March 25, 2009 5:01 PM

    Tankers do get “senile” before their cycles are used up.
    Imho a major reason why the Airforce wanted to go for the
    Airbus Tanker cum Transport.

    So, my guess is civil airliners ( passenger or freight )
    will get lots more hours and landing cycles per year
    than most military planes.

    uwe

  22. Jean-François Didisheim 26 March, 2009 at 9:44 am #

    I am of the opinion that David Learmount’s analysis is an explanation of what the likely initial pilot’s actions were. The central question though relates to the cause of the roll. What could have caused such a dissymetry that an aircraft this size could develop such a high roll rate at such a low speed? Dissymetric spoiler deflection comes to my mind, possibly as a result of dissymetric retraction during an attempted go-around, caused by a structural failure during the previous heavy bouces.

  23. fabius 26 March, 2009 at 12:51 pm #

    May be, a structural damaged occurred to the first section of the fuselage at the second heavy nlg bounce, leading an uncommanded unnatural movement of the flight control surfaces cables. Besides, the spoilers deploy after a time delay if ground condition occurs at the nose landing gear or detected by FCC and spin-up of the aft wheels. This delay could be support the fatal pitch down of the MD11.

  24. Little Plane 26 March, 2009 at 2:03 pm #

    Checking the stats on the MD-11, there is some interesting data. Out of 5 total losses, 3 ended with the aircraft on inverted on the runway. Is it a coincidence? I do not recall any other design with the same behavior.

  25. Bill Eddy 26 March, 2009 at 9:42 pm #

    I flew the MD-11 for a couple of years. I agree with your analysis. I would add that roll moment was caused by the failure of the left wing/landing gear. The MD-11 has very sensitive controls and is easy to over control especially in windy and gusty conditions. Personally, I liked it. However, pilot over control was the cause of the Newark accident with the same results. Looking at the flight data recorder will tell for sure. Looking at the video, I believe if the nose had not come back up after the first touchdown the aircraft would have parked at the gate, the pilots would have gone to their layover and at worst case had to write up a hard landing. It just goes to show how unforgiving aviation can be.

  26. Rick 27 March, 2009 at 12:27 am #

    No one has mentioned use of the autopilot. I was on the team validating and verifying the avionics for the FEDEX MD-11. These types of oscillations are possible in gusty conditions when the AP is in control (and did not disconnect on impact). Any MD11 drivers can tell me if FEDEX would have been using AP here?

  27. Cxflyer 27 March, 2009 at 1:20 am #

    This crash reminds me one of my first flights in the simulator. When I try to get the plane to noise up little bit more at 30 feet (because i was trying to get a smooth landing,) the airplane’s nose doesn’t respond until a few seconds later… the plane hits the ground then bouce back up. That pitch up action causes my speed to decrease even more which also causing the plane to pitch nose down.
    When that happens, you actually have no control of your aircraft because it is going too slow, and theres not enough lift to get the plane back up or even powers to nose up. A little unbalance on the main wheel during touch down can cause the aircraft to over turn. You guys can actually try that in the FLX.

    It is pretty easy for me to redo the accident again :P

  28. Paul 27 March, 2009 at 5:56 pm #

    Hello David
    Early engine reverser activation is fairly
    common even with long runways when landing conditions are tough. Maximum release of kinetic energy is of prime importance while landing in less than ideal conditions.
    Look at the video closely and you will notice the abnormal downward movement of the aft fuselage section during the landing impact. Control wires and hydraulics are under the floor (as you know) and such a movement might have been damaged.
    Sorry for the bad English written…i am french !
    Let’ s wait and see the enquiry.
    Have a good day David

  29. Thomas 29 March, 2009 at 11:31 am #

    Hi, the MD11 has a strong pitch up tendency when the ground spoilers are deployed. The GND spoilers move to a intermediate ext position with wheel spin up, then at nose wheel touch down to full deployed position. This is different to the design of the system on the DC-10, and was changed during the flight tests in the late eighties. rgds thomas – swissair -

  30. jack pratt 31 March, 2009 at 5:04 pm #

    the md-11 has a serious design flaw that is being argued in a l.a. courtroom (china airlines) to do with auto-pilot stabilization for landings under these circumstances—maybe fedex should look into this?? jack

  31. Paul 2 April, 2009 at 6:38 pm #

    This is a botched landing gone bad. The first touchdown was very hard, the plane bounced. No big deal. A go-around should have been (was being?) initiated at this point. Then, someone, probably the pilot NOT flying, the Captain, stuffed the nose down to get it back on the ground. I really can’t see any way to get that kind of pitch down without shoving the yoke fwd. It looked very deliberate. With those kind of winds the jet would have had plenty of “smash”, extra speed, for a GA. Possibly they damaged it on the first touchdown, which contributed to the breakup on the second extremely hard touchdown.

  32. Otto 15 April, 2009 at 8:07 pm #

    Prior to providing any comment is must be stressed that any comments made are purely speculative. Cause and circumstance are things that will be be made clear in the final investigative report.

    Therefore my best guess:

    The important aspect of all large jet landing is energy management. A high indicated airspeed will give any aircraft greater maneuvre capability. If the MD11 had a forward centre of gravity the landing flare would have to be initiated at an earlier moment in time to establish the correct landing attitude. If this was not done or the flare was simply executed late with a forward centre of gravity the landing would be rough. If the aircraft was late in responding to an up-elevator command the nose would still be coming down after landing. However if then the up-elevator was still maintained and the spoilers extended the pilots might be surprised that the aircraft pitched up and became airborne again. This is however totally plausible in regards to the well known pitch-up tendency of the MD11 after groundspoiler extension in combination with a high energy ( high indicated airspeed ) situation. It is realistic that the final approach speed was higher due to the high windspeeds at the time of the accident. Now the best option at this stage is to perform a goaround especially when becoming – unexpectedly – airborne again after the nosewheel has touched down. The strong pitch down wich occurs when the aircraft is about 30-50 feet high is inexplainable from an aerodynamic point of view. The high rate of change in pitch angle must be due to pilot input. The subsequent second touchdown is very violent but amazingly doest not break the nose landing gear. The left main oleo strut however is fully compressed during touchdown. This causes the strut to bottom and break the above lying main wing spar wich then fails. The lift of the righthand wing then causes the aircraft to flip over onto its back. This accident sequence has been shown to occur during landings with high rate of descent and touchdown on a single main landing gear. This was the case in the MD11 accidents of FedEx in Newark and China Airlines in Hong-Kong as well as the Martinair DC10 accident in Faro.

  33. Peter Bore 18 April, 2009 at 2:06 am #

    D.P. Davies is quoted as saying that, as a certification pilot, he had to fly the aeroplane as badly as a just adequate airline pilot having a bad day, and still survive. Would he have agreed to UK certification of the MD 11?

  34. Gabriel 29 April, 2009 at 4:08 am #

    You can not know if the conditions were far from extreme using a forcast. You never make a statement with any supposed data. You should first know exactly where did the actual wind come from at that specific minute (moment) and conclude if weather was far extreme or not. Until then, all your analysis has as much value as any other based on any supossition.

  35. Richard Reed 26 June, 2009 at 2:28 pm #

    This is a bit of a late post, but what the heck, I just found this piece and thought I’d add a theory.

    Regarding the second pitchdown that dumped the aircraft onto the ground at high speed, David mentions that no professional pilot would make such a manoeuvre close to the ground (with good reason, reducing the AoA when ballooning is a great way of getting rid of what little lift your aircraft still has).

    Is it possible therefore that the manoeuvre was unintended? Consider that during the first heavy nose-gear touchdown the rear of the aircraft almost seemed to be bouncing/rising. Could this have been strong enough to physically pitch the crew forwards in their seats? Could that have been sufficient to have shoved the control columns forward just as the aircraft was getting airborne again. I know previous posters have commented about a lack of control authority, but if the yokes were shoved right forwards to the stops by the crewmen being pushed forwards onto them, and the aircraft goes nose-high in that same moment, if there was enough control authority left (and they seemed fast over the numbers) that could have been enough to make them involuntarily do what they would not have consciously done, and shoved the nose hard downwards.

    I think so far the speculation has been about the state of the aircraft and not the unfortunate people being thrown around inside it. The question then is how much “give” there is in their harnesses and whether a sudden downward movement would be enough to cause the restraints to “lock” and stop the crew being pitched forwards.

  36. Richard Reed 26 June, 2009 at 2:51 pm #

    Having just read the accont of the Newark accident that was broadly similar, it would also be interesting to know what the aileron effectiveness of the MD11 is at landing speeds (and high AoA), since in both instances the aircraft rolled quite quickly into wind. This suggests to me that at the moment of touchdown the crew input some into-wind aileron that became dangerous once the aircraft had bounced and become airborne again at low speed, without the gear to keep the aircraft level, but with the rapid sequence of dramatic events unfolding around them, they simply omitted to centralise the controls for the bounce.

    I can quite understand them not doing that, after all, the Hudson A320 didn’t have the “ditch switch” (the one that closes all below-waterline orifices prior to ditching) actuated in the extreme pressure that the crew were under, and the Hudson crew had a little longer to work things out.

  37. Professional Airline Pilot 13 June, 2010 at 11:29 pm #

    1. Bounced landing
    2. Failed attempt to Go Around (due to spoiler deployment)

  38. Mick 9 August, 2010 at 11:16 am #

    This probably isn’t news to anyone, but you can see a different perspective of the crash here: http://www.flightglobal.com/articles/2010/04/17/340740/pictures-investigators-detail-narita-md-11-roll-over.html
    The spoilers are clearly deployed after the initial touchdown but (quality of the photos notwithstanding) appear to be retracted by the time of the second bounce. Could this have happened? Also the touchdown speed was 166kts. That seems more like a take-off speed to me. Anyone got some info?