China Airlines Disputes Official Finding of Pilot Error

Carrier Says Evidence Points to Wind Shear in MD-11 Crash

China Airlines (CAL) is refusing to accept the official government conclusion that pilot error caused the crash of an MD-11 B-150 at Hong Kong’s new Chek Lap Kok Airport (CLC) on Aug. 22, 1999. Instead, the carrier claims there is new factual information proving that wind shear or a “microburst” at the last moment caused the aircraft to land hard, flip over and kill three on board.

The Hong Kong Civil Aviation Department (HKCAD) is stubbornly sticking to its report’s “pilot error” bottom line. The accident was caused by “the commander’s inability to arrest the high rate of descent,” the HKCAD report says. It was almost guaranteed that the pilot, Italian expatriate Gerrardo Lettich, would have to take the fall for an error of judgment for placing his aircraft in jeopardy, but there were significant other factors in play. One of these — monsoonal wind conditions — relates to the overall safety of Chek Lap Kok in severe weather conditions, according to the Taiwan Aviation Safety Council (ASC). More on this below.

On the day of the accident, there was a severe tropical storm in the vicinity and strong winds and severe turbulence was forecast. China Airlines Flight 642’s highly experienced captain “lost it,” according to the report, in the last 50 feet of his stable approach and landed hard enough to break the right wing’s landing gear, lose the right wing, roll inverted and somersault through 180�. The aircraft came to a final stop upside-down alongside the runway and facing in the direction of the approach. The accident killed three people and seriously injured 50 of the 300 passengers and crew on board. The quick reaction of airport firefighters to extinguish the blaze before it engulfed those trapped in the wreckage saved hundreds of lives. It took nearly three hours for rescue workers to extract all the survivors.

Aircraft serviceability was not deemed to be a factor so the accident had to be attributed to the pilot’s decision-making, handling and use of systems in the severe weather conditions. Of 26 approaches flown in the period of three and three-quarter hours up to the accident, 10 resulted in go-rounds as a result of the high winds. The only other MD- 11/DC-10 to arrive in the period of deteriorating weather landed 3.5 hours earlier than Flight 642.

Prevailing visibility and braking action weren’t factors, and the commander’s crosswind technique wasn’t faulted. The aircraft was not fitted with the optional rain repellent system, however, the commander’s assessment of visibility through his windshield on final approach was “moderate.” Despite that assessment, it is possible that sunset, overcast conditions, and rainwater on the windshields outside the sweep of the windshield wipers and on the unswept side-windows, may have affected his peripheral vision; this may have resulted in him not appreciating the aircraft’s developing high rate of descent as it passed the normal flare height.

The right side landing gear broke because of a very late-developing high descent rate and a slight 4� right wing down touchdown. The investigation therefore concentrated on airspeeds and wind profiling. The MD-11’s Quick Access Recorder (QAR) buffers data and writes it every 35 seconds. If power is lost, as it was here, that final critical accident data is not written and is lost. The flight data recorder (FDR) similarly had a defective longitudinal accelerometer input so the wind profiling had to be “reconstructed” from the four touchdown zone anemometers. Boeing [BA] did two studies of the wind affecting the final approach, one in 2000 and another in 2003 (after the earlier one was found to have been mathematically flawed). The National Transportation Safety Board (NTSB) agreed with the method but never checked either sets of figures.

The Boeing conclusion was that there had been a 20-knot loss in airspeed in the last 50 feet of the approach. Part of that could be attributable to wind-shear or loss of headwind, Boeing said, and part to the scheduled closure of the throttles by the auto-throttle (A/T) system at 50-feet radio altitude (RA). Simply put, the pilot had cancelled the autopilot-flown instrument landing system (ILS) upon becoming visual with the runway lights at 700 feet but in accordance with company SOP (standard operating procedure), he’d left the auto-throttles active. According to other MD-11 pilots, that might be SOP but it was also his fatal error. The Flight 642 pilots had determined the landing reference speed (Vref) as 152 knots, and selected a target approach speed of 170 knots in view of the expected strong wind and turbulence on final approach.

The HKCAD report states:

“The auto-throttle controlled the speed adequately within a four or five knot tolerance either side of a mean speed of 165 kt until just below 300 ft RA [radio altitude] when the indicated airspeed fell to 157 kt. The co-pilot called ‘Speed’ and claimed to have moved the thrust levers forward when there was no apparent response from the commander; however, in a later statement the commander claimed that he had moved the thrust levers forward. The thrust then increased significantly from a previous average of 1.05 EPR [computed engine pressure ratio] to almost 1.3 EPR, with a consequential increase in speed to 175 kt. The A/T [auto-throttle] response to this excessive speed was to put the thrust levers at the fully closed position by about 70 ft RA, and the thrust decayed to an average of 1.0 EPR by 50 ft RA (the altitude at which the auto-throttle would’ve otherwise normally commenced a thrust lever retard), and to idle thrust by 35 ft AGL.”

So here we have a gust-induced loss of speed, human intervention (perhaps unwittingly by both pilots), a resulting slight speed excess that destabilized the approach, and an auto-throttle system’s closure (to idle) to compensate. But before the auto-throttle could reestablish thrust to support speed and control rate of descent, the RA said: “no you don’t, we’re passing 50ft RA, you must stay at idle.” So from 70 ft onward, the throttles were at idle and any thrust intervention had to be thereafter manually injected by the pilot. Unfortunately at just that stage, the bottom fell out of the pilot’s expat world as he hit the other side of the “gust” that had caused that speed instability. Appendix 21 of the HKCAD report shows that the gust affecting Flight 642 was an “about the same speed” increment but an order of magnitude greater in direction-change than the 777’s that had landed a few minutes prior, and occurred closer to the ground. Or to put it more clearly, it was “a much nastier gust and right before touchdown.”

The Taiwan ASC also commented on the report, saying that the severity of these late finals gusts can also be attributed to the Runway 25L flare zone being in the lee of the airport’s large Passenger Terminal Building during monsoonal wind conditions. In its detailed wind analysis, the ASC claims that it is apparent that Flight 642 entered an intensifying downdraft at 200 feet above ground level (6 seconds to go) and that it peaked at 2 seconds to go.

To explain, it must be realized that winds near the ground flow both laterally and vertically. In the case of transiting through a microburst or mechanical eddy, you’d have to expect an instant airspeed increase (the outflow toward you), a higher sink rate (inside) and then the further deterioration of a speed loss (due the outflow behind you). Auto- throttles (A/Ts) cannot compensate as instinctively as an adrenalized pilot can. Throttle-wise, he would react with all the instincts of a trapped mongoose. But up until this point (50 feet radio altitude), the A/Ts had been ceded responsibility for speed maintenance (i.e., his throttles were out of hand, out of sight and out of mind). Suddenly the bottom drops out and the pilot instinctively pulls back (flares) only and tries to arrest that alarming rate of descent. Just prior to that, per SOPs, he has correctly removed the crab due to crosswind and is pointing down the center-line, side-slipping immaculately with the right wing down into the crosswind (that is right on his wet runway limits).

The effect of a late sudden flare that compensates for airspeed loss is threefold. It may cause a tail-strike; it will pitch-rotate that one wing-held-low gearset harder into the ground (which is what happened here) and at landing reference speed (Vref) minus 20 knots it will not affect the high sink-rate at all — certainly not at the high inertia of Max Landing Weight (MLW). Note (below) that passing 21 feet radio altitude the elevator went from 8.5� down to almost 16� UP — immediately before touchdown — i.e., attitudinally rotating the right-side landing gear into the hard deck. The aircraft touched down slightly right wing low (3.5 degrees-4 degrees) on its right main landing gear at a rate of descent calculated at approximately 18 feet per second, well beyond the design structural limit of 12 feet per second.

Therefore, it is correct to say that the aircraft’s destructive 18ft/sec final impact was not all “sink-rate” and that the destructive dividend was achieved by the pilot’s final desperate back-stick input (that rotated the gear into the deck). You may say it’s still pilot error, and he “should have cobbed the throttles or gone around.” Well, other MD-11 pilots assert that such a decision is not so easy given the inherent handling characteristics of the MD-11, aka the “Mad Dog”.

A Repeat Performance by This Breed of Trijet

The overall impression one gets is that the pilot may have allowed this accident to happen, but that it was more likely to happen in this breed of Trijet than in any other airplane. The MD-11 was an outgrowth derivative of the DC-10. If further proof (of inbuilt risk) is required, just look at the compendium of similar MD approach and landing accidents at http://www.iasa.com.au/md.htm, entitled “the Tritanic (MD Trijet series) and the Trishaw” (Lockheed’s L-1011 TriStar). A misnomer maybe, because you won’t find any similar Lockheed Tristar accidents there — there weren’t any.

In fact this crash is almost identical to the first crash shown on that Web page (the Dec. 21, 1992, Martinair Flight 495 DC-10 at Faro, Portugal, that killed 54). But then again, so were the Newark, N.J. (FedEx MD-11F, July 31, 1997) and Memphis, Tenn. (FedEx MD-10F, Dec. 18, 2003) crashes. There are some subtle differences (only) in the various failure modes in all these crashes. In Memphis (MEM) the RH gear catastrophically failed — the unit broke into six pieces and didn’t take out the wing-spar. In Newark, the wing spar failed (from loads transmitted through the gear). Touchdown of the RH main (MEM) was estimated to have occurred in the vicinity of 14.5 ft/sec (870 ft/min), in a crab (sideways drift). Although in excess of certification limits, the design limit (when it breaks) should be 1.5 times the certified limit — 900 ft/min (15 ft/sec).

The China Airlines RH forward gear trunnion and rear spar web failed at CLC, initiating the RH wing-fold sequence. The right wing rear spar web fracture of ship 518 (B-150) was found to be identical to that of Ship 553’s (FedEx at Newark).

Meanwhile, there are very few MD-11’s that haven’t been converted to freighters (or aren’t planned to be) — and the main reason for that lies in its mottled and motley history.

Pilots’ Over-Reliance on Automated Systems

The auto-throttle system in the MD-11 is a ‘full-time’ system capable of automatically controlling a variety of parameters of the flight’s progress from the initiation of the takeoff roll until 50 feet radio altitude on final approach, after which it remains armed but normally inactive unless the “go-around” switch is pressed to discontinue an approach. The pilot may disconnect the system by simply pressing a button on the outside of the No. 1 or No. 3 thrust lever, or by selecting reverse thrust after landing. He may also intervene and adjust the thrust temporarily in flight by manually moving the thrust levers.

While the operations manuals are not explicit regarding use of the auto-throttle system, full time use of the system is known to have been encouraged by the manufacturer in operation of MD-11 aircraft, and also in that of its predecessor, the DC 10. As in other areas of automation on the flight deck, this may encourage over-reliance on the automated system, to the point where the pilot may no longer be aware of the need to intervene when the system is either not coping with the operating conditions affecting the aircraft, or the operational situation is outside the system’s design parameters.

At least one of China Airlines Flight 642 pilots did intervene by advancing the thrust levers when the speed fell to 157 knots just below 250 feet. However, more critically, the commander did not react to override the early retardation of the thrust levers and apply thrust to counteract the increasing rate of descent in the flare, as the commander of the previously landing aircraft necessarily did. It was therefore recommended that China Airlines should review its MD-11 training syllabuses to ensure that the crew monitor the automated systems on the flight deck, so as to be ready to intervene, or override manually, whenever necessary.

The final accident report, on p. 100, says blithely: “It was therefore recommended to China Airlines that, in association with the Boeing Company, they amend the recommended landing procedures in the MD-11 SOP to include procedures for approaches and landings in more demanding weather conditions.” Ralph Nader in his hey-day would have called foul, written another “Unsafe at any Speed,” and demanded a product recall. Nevertheless, and despite having said that, it is believed to be a very profitable freighter. The freight fraternity is well aware that going around with the mighty “Scud” is the best way to emerge unscathed. But going ten rounds with it might eventually remind one too vividly of that old Humphrey Bogart movie about punch-drunk pugilists: “The Harder they Fall.” Flight 642’s captain, Gerrardo Lettich, would probably agree. He has probably kissed the canvas for the last time, and in the view of many of his brethren, quite unfairly.

The Finality of It All

The MD-11 is not fly-by-wire. It is, however, fly by CONSTANT pilot input. MD decided to make the empennage 40 percent smaller than the DC-10’s to save on both parasitic drag and induced drag by keeping the center of gravity near the aft limit during high-speed cruise. This airplane doesn’t really have a “slot” when you are on final; it doesn’t seem to really stay at a trimmed angle of attack/deck angle at a specific power setting/airspeed. As such, the pilot is constantly making little corrections, like flying a dynamically unstable fly-by-wire fighter with the computer out.

As at least one pilot has said, part of the problem is the Longitudinal Stability Augmentation System (LSAS), which is a computer that constantly trims the stab to make up for the shortcomings of the tail size. The landing is also “software dyslexic,” according to one pilot. “As soon as the plane touches down I have to push on the yoke to counteract a severe pitch-up from the spoilers coming to 2/3 extension. Less than a second later, the auto-brakes kick in, so you have to pull back on the yoke to gently lower the nose to the runway.”

Judge for yourself. There is a nine megabyte film clip of the actual fiery CLC accident at: http://streaming.scmp.com/aircrash/CAL_Plane_crash1f.avi

Michael Crichton’s 1996 novel “Airframe” is transparently based on the checkered history of the MD-11 and, in particular, on the deficient designs of its slats and pitch stability systems.

The HKCAD’s report and the review findings can be accessed at http://www.info.gov.hk/cad/english/news.html; the MD-11’s accident history is available at http://www.airlinesafety.com/faq/faq9.htm

Loading In Poor Weather Conditions

On Aug. 22, 1999, the day of the China Airlines Flight 642 accident, with the heavy wind observed as backing to northwesterly, air traffic control at Chek Lap Kok Airport changed to Runway 25L. Two further go-rounds followed, but the successful landing rate then improved so that in the period between 0947 hours and the accident at 1043 hours, six aircraft landed and only one had to go-round, the latter occurring at 1034 hours.

Effect of ‘Microburst’ or Wind Shear on Aircraft

China Airlines (CAL) on Feb. 4 said it could not accept the investigation results that ascribed the fatal jet crash in August 1999 at Chek Lap Kok Airport to pilot error. The carrier says that Boeing [BA], the maker of the MD-11, found new evidence during the investigation that a wind gust suddenly changed direction 1.5 seconds before the plane touched down. Evidence collected from the plane’s flight data recorder (FDR) showed that swift wind changes, triggered by a tropical storm sweeping through Hong Kong at that time, made the pilot unable to control the aircraft’s sudden descent. CAL has forwarded the evidence to the Hong Kong Civil Aviation Department and asked it to attach the information to its investigation report.

One Pilot’s Jaundiced View

“The report strikes me as rather too focused on blaming the captain. After all, he wasn’t the only pilot in the vicinity at the time. If the (none too subtle) inference of the report is that he shouldn’t have tried to land in the first place (because of the hostile ambient conditions) — then what does that say about all the other aircraft that did the same thing (tried to land) but were fortunate enough to get away with it? The point that he did not arrest the sink rate overlooks the well known fact that wind shear sometimes overwhelms aerodynamics. That he ‘did not’ does not mean the same thing as ‘he did not try.’ And thus, if this new airport is subject to inescapable wind shear — why the hell is it still operating (full steam ahead) during typhoons? My first officer (F/O) summed it up for me yesterday: “He (Capt. Gerrardo Lettich) tried too hard to get in.” My view is that this airport is dangerous in windy conditions. If the HKCAD [Hong Kong Civil Aviation Department] is blithely going to ‘blame the pilot,’ then the solution is simple. I most certainly am not going to try at all to get in there in poor conditions in the future. I will make a token effort and then shamelessly divert.”

Source: “HIALS,” an anonymous Dragonair A330 Captain, Hong Kong

Other Pilot’s Experience

The commander of a B777 aircraft that landed on Runway 25L some four minutes before China Airlines Flight 642 later stated that he had become fully visual by 400 feet, although in driving rain. Between 200 feet and 100 feet, the aircraft encountered some violent gusts that resulted in speed fluctuations of 10-15 knots, and “a large speed reduction” on entering the flare, which was successfully countered by a rapid, manual, application of power.

In the Lee of the Passenger Terminal Building

Appendix A of the Hong Kong Civil Aviation Department’s accident report on China Airlines Flight 642 shows a variation in TDZ (touchdown zone) wind direction of between 314� and 326� with speeds from 39 knots to 43 knots (Runway 25R) in comparison to a variation in TDZ wind direction of between 283� and 339� at 14 knots to 28 knots for Runway 25L in the lee of the Passenger Terminal Building. This kind of late sudden wind change will greatly affect the sink-rate in the flare.

Source: Taiwan Aviation Safety Council