Latest Edition: 18 March 1998
|18 March:||Flight Profile data supplemented by information from the Aviation Week report of 16.3.98|
|10 March:||Data from Taiwanese CAA incorporated from Flight International report of 11.3.98|
|9 March:||Flight Profile graph and legend from Taiwanese CAA added|
|4 March:||Information from Aviation Week 23.2.98 and Flight International 25.2-3.3.98 added|
The publically-available facts concerning the accident to CI676 are given, and it is suggested where they point - and where they don't point. ([03.11] Most recent data released show a manual loss of control after an approach that was far too high. There is no apparent involvement of the automation.) Specifically, they cannot show a repeat of the Nagoya accident, because of mandatory A300 system modifications since that accident. This article will be continually modified as information is gathered.
Note: The information in each section is preceded by a date in square parentheses, e.g., [02.19] for February 19, 1998. The information in the section below this date was written on or before that date. New information will be added to the front of (at the top of) each section, with date of entry, as it arrives. Thus this article will serve as a record of how information was gathered, and its timing, for those who are interested in the sociology of accident reporting.
There is an exception to this scheme. Information from the International Herald Tribune/New York Times Service that appeared on 17-18 February, along with aircraft and fleet information and some deductions from this information, has been included in-paragraph on 21 February, between lines consisting of `[02.21:' and `]' which act as text delimiters. Just like this paragraph.
On Monday 16 February, 1998, at approximately 20.09 local time
China Airlines Flight 676
from Bali, an Airbus A300, crashed into a residential area while
performing an ILS/DME Rwy 5L approach to Chiang Kai Shek International
Airport near Tapei, Taiwan.
I infer from the comment in the International Herald Tribune that "Taiwan civil aviation authorities ordered China Airlines to ground its nine remaining Airbus A300-600R jets" (IHT.98.02.18) that the aircraft was an A300-600R variant, of which China Air had 10, now 9 (see below - it also has 6 earlier A300B4 variants). Edward A. Gargan of the New York Times Service reported the China News Agency as stating the time of the crash as 20.05 local time (IHT.98.02.17). The International Herald Tribune reported a "China Airlines spokesman" as saying that "The Airbus underwent a maintenance check in February. We found no problems." (IHT.98.02.18).
There was some suggestion, supported by the title of CNN's original piece (Taiwan crash raises questions about Airbus A-300), and apparently taken up by the Japanese and Taiwanese presses immediately following the accident, that there were computer-related problems. From the evidence that I have available to me at the moment, I see no reason yet to suspect a computer-related problem.
More recent reports corroborate this view.
` Loss of control appears to have been the cause [...]. Preliminary flight data recorder readings [...] indicate that the [aircraft's] approach to runway 05L was far too high for a safe landing, and that the crew lost control during a manually flown go-around in which extreme pitch attitudes and speeds were allowed to develop. The [Taiwanese CAA] released details from 1m 6s of the tape. '
The aircraft, registration number B-1814, manufacturer's serial number 578, was an Airbus A300B4-622R, equipped with two Pratt and Whitney 4158 engines (Swa). It was delivered to China Airlines on December 14, 1990, and had accumulated 20,070 hours and 8,800 cycles (AW.98.02.23).
Captain Lin Kang Long was 49 years old, joined China Airlines in 1990,
and had 7,210 hours total flight time. First Officer Jiang Der-Sheng was 44
years old, joined China Airlines in 1996, and had 3,530 hours total flight
time. Both pilots were formerly with the Taiwanese Air Force
The Flight Profile
Back to Contents
Details of the flight profile have been released by the Taiwanese CAA. These show the aircraft very much too high on the glideslope on approach At some point near the runway threshold, where the aircraft is still at well over 1000ft altitude ASL, it climbs steeply, evidently stalls, and augurs in some 2 statute miles past the runway threshold. The horizontal legend is distance in statute miles from the runway threshold and the vertical legend is altitude in feet MSL. The Outer Marker/Final Approach Fix is at approximately 3.9nm (nautical miles, = 4.5 statute miles) from the runway threshold.
The Outer Marker is at 4.1DME from the the localiser ITIA (the localiser signals the inbound ILS course centerline, and has collocated Distance Measuring Equipment, DME, which responds to query signals from the aircraft and measures distance in nautical miles). This means that the DME is located (with the localiser) about 0.2nm beyond the runway threshold. At 140kts ground speed, time FAF to Missed Approach Point (MAP) is 1min 40sec (AERAD-Taipei).
Approach altitude is 3000ft until established on the ILS. ILS glideslope is 3°, inbound course 053° magnetic (magnetic variation is 3° West). Runway threshold elevation is 73ft MSL. Crossing altitude at the Outer Marker is 1400ft MSL (i.e., QNH; 1330 QFE); at 4DME, 1370ft QNH (1300 QFE); at 3DME 1050 QNH (980 QFE); at 2DME 730 QNH (660 QFE) (AERAD-Taipei).
A translation of the plan legend follows the plan. The plan appeared in The Liberty Times on 4 March 1998. The plan appears to denote the Outer Marker position (the second point shown on the profile) as 3.9 statute miles. This seems to be a mistake.
It may be important to observe that all of the altitudes at which control problems occurred up to and including the stall were altitudes at which sufficient pilot control-column pressure takes precedence over autopilot control, according to SB A300-22-6021, whose basic requirements have been a Taiwanese AD for some time. Thus the Nagoya situation, in which the autopilot worked against the pilot flying, because go-around mode had inadvertently been commanded, should not have happened in this case. In fact, the incident started at well above the altitude at which the original design which SB A300-22-6021 modifies allows pilot take-over with control-column force (it is also the case that the pilot has two other means of disconnecting the autopilot - the `big red button' on the control column, and the ON/OFF switch on the Flight Control Unit (FCU), on the glareshield on the right-hand side).
Data for the last two seconds are lost due to impact. Legend: 1 nautical mile = 1.82 kilometer 1 statute mile = 1.60 kilometer Actual approach of China Air Flight 676 Normal Glide Path Drawing/ Reporter [Names not Translated] Computer Graphics/Editing [Name not Translated] Scale: 6 statute miles, 5 miles, 4 miles, 3 miles, 2 miles, 1 mile, RUNWAY 5L, 1 mile, 2 miles (all statute miles)
The following parameters were released by the Taiwanese CAA, and the chart is based on this information. Numbers should be easily recognisable for non-Chinese readers. Two parameters have been corrected at 05m14s (left bank was 40° at this point), and 05m11s (pitch was +40.5°) and one parameter added at 05m16s (bank 9° right), from information in (FI.98.03.11), confirmed against (AW.98.03.16). Altitude at 20:05:04 added, and altitude at 20:05:11 corrected, from (AW.98.03.16). Altitude is reported twice in (AW.98.03.16) as higher than reported by the Liberty Times, namely at 20h05m29s, as 2,627ft ASL, 100ft higher; at 20h05m24s, as 1,799ft ASL, 2ft higher. I include the Aviation Week figures in parentheses. The time of the accident aircraft's final approach was 12:03-12:06 hours GMT; 20:03-06 local time. The time in the table is given as local time. `IAS' is Indicated Air Speed. The interpolating of the AC/ATC communication and the threshold distances and altitude above glideslope (where applicable) are from (AW.98.03.16).
|20h03m13s||187||3187||Flaps 15°. Engines idle.|
|20h03m46s||185||2483||Flaps 20°. At the Outer Marker.|
|20h03m55s||CAL676: "Tower CAL676, 3 mi. on final".|
|20h04m00s||TWR: "CAL676 cleared to land, wind 360 at 3".|
|20h04m30s||decreasing.||1515||1,000 ft too high on glideslope, 1.2nm to touchdown. Elevator angle decreases and fluctuates.|
|20h04m43s||167||1339||Flaps 40°. 0.5nm to threshold.|
|20h04m46s||159||1323||increasing.||Minimum altitude of the flight.|
|20h04m50s||149||1375||Autopilot disconnected. 0.20nm to threshold.|
|20h04m59s||1475||Go-around thrust. Engine rpm increases. At threshold.|
|20h05m09s||134||1723||35.51°.||Flaps commanded up to 20°. Landing gear raised.|
|20h05m12s||CAL676: (Two loud rings are audible but no communication from crew)|
|20h05m13s||CAL676: "Tower CAL?".|
|20h05m14s||104||2151||40 ° left|
|20h05m16s||2327||42.72°.||9° right.||Flaps 20°. 1nm down runway.|
|20h05m21s||TWR: "CAL676 confirm go-around" (No response from CAL676).|
|20h05m34s||173||1779||-44.65°.||20° right.||1.5nm down runway.|
|No further data|
The pilot was performing an ILS/DME 5L approach to Chiang Kai Shek International (Lad97-approaches) . The threshhold elevation is 75 ft above Mean Sea Level (MSL). The Decision Height is 200 ft QFE (QFE = elevation above threshold), that is, Decision Altitude is 275 ft QNH (QNH = with altimeter setting for MSL); the Runway Visual Range (RVR) requirement is 600 meters (AERAD-Taipei). The Category II ILS ceiling minimum would be 100ft QFE. The airport is on flattish land close to the ocean, which lies a couple of nautical miles (nm) away to the north-west. All three runways are north-east/south-west oriented, between 50 and 60 degrees North, and lie approximately parallel to the shoreline. All manoeuvering for approaches and missed approaches takes place over the ocean. Higher terrain, rising to 2000 feet, lies 5-8 miles east, and 5-12 miles south. There is terrain to 2,400 feet about 11nm east, and terrain to 4,000 feet some 13-15 nm south-south-east. Runway 05L lies at 053 degrees North (magnetic), and the magnetic variation is 3 degrees West (AERAD-Taipei-ILS5L) .
Weather conditions were reported by CNN:
Heavy fog was reported around the airport throughout the afternoon and evening, and a light rain was falling at the time of the crashCNN also quoted `airline spokesman' Liu Kuo-chien as saying that "Visibility was extremely bad. The pilot said he was having trouble seeing the runway as he made his approach and asked to come around for another try" (CNN.98.02.17).
Gargan reported Hsu Lu, general manager of the Voice of Taipei radio station, as saying that her staff had reported that the control tower had asked the pilot not to land (IHT.98.02.17). I note that procedurally it is the pilot that determines if the conditions are met for landing, not the control tower. Had the control tower required a go-around, that would likely have been for reasons of traffic conflict. If a pilot lands during a period in which ATC believes that the minimal conditions do not pertain, they might conduct an inquiry or discipline the pilot if (s)he is found to be in violation. But this is after-the-fact. Requiring a go-around is not an ATC prerogative in marginal conditions. It's the pilot's call. No news organisation has reported a potential traffic conflict.
The Airbus A300 is built with Category III ILS reception equipment
as standard. Category III ILS allows an approach under autopilot control
until main gear touches the runway. It requires particular ground-based
equipment and maintenance standards, specially-qualified crew following
a recurrent training regimen, and specific recurrent maintenance
and calibration standards for the aircraft-based equipment.
It has not been reported whether this aircraft and crew were Category III
A reliable source told me that the visibility was reported by the
Taiwanese authorities to be between Category I and Category II minima,
although this information is unconfirmed
That is, ceiling between 100 and 200 feet and RVR of 600 meters.
The source also said that the `light rain' was reported to have been
very, very light.
Rain and fog normally don't go together -- rain disperses fog. So
one would expect that a little light rain would mean that the fog
is not pea-soup thick.
I noted from the BBC film of the crash site
that street lights that appeared to be some hundreds of meters
distant were clearly and distinctly visible.
The unconfirmed indirect report of between Category I and Category II
seems consistent with other evidence, and with what one might expect
the weather to have been.
Conditions for the Approach
Back to Contents
Flight International reports "Visibility [...] at about 1,000m [was reported, and the CAA said] the aircraft was on a Category 1 [...] ILS approach (FI.98.02.25), for which the landing minima are 600m Runway Visual Range (RVR) (AERAD-Taipei-ILS5L). Aviation Week reported the ATIS broadcast as giving 2,400ft visibility, with RVR 05L at 3,900ft, RVR 06 at 4,500ft, ceiling 300ft broken, 3,000ft overcast (AW.98.02.23).
Such weather conditions are easily within the capabilities of airline pilots to fly, and they do so routinely in Europe in winter. In fact, most aircraft are Category III capable in Europe and it is not unusual to fly Category III approaches. Category I and Category II approaches are routine, and it would be reasonable to expect any airline pilot to fly Cateogy II approaches routinely. Private pilots such as myself fly Category I approaches routinely. The difference is between a Decision Height of 200 feet above threshold (Category I) or 100 feet above threshold (Category II). The profile flown, and the way it is flown, is identical in each case. The A300 autopilot is easily capable of flying a Category II approach with the required precision -- as mentioned above, it is in fact Category III capable, although it may not be in fact maintained to a Category III schedule if the normal use is not into Category III capable airports, and if the crews are not Category III qualified.
The normal procedure is that at or before decision height, some part of the runway, its boundary, or its lighting, must be clearly visible to the pilots for the landing to be continued. If there is no such visual contact, a `go-around' is required, whereby the aircraft flies up and away under take-off performance, and either returns for another try or proceeds to the alternate airport. The aircraft is required to have on board enough fuel for a flight to the alternate airport plus reserves of 45 minutes or 30 minutes, depending where in the world one is (different countries have different regulations). We have noted the CNN report of the airline spokesman that the pilot didn't have visual contact and declared a go-around -- "[...] The pilot said he was having trouble seeing the runway as he made his approach and asked to come around for another try" (CNN.98.02.17). One point to note is that it is up to the pilot to declare a go-around if he does not see the runway, it's environment or lighting. He then executes the missed approach procedure (in this case, to climb staight ahead to 800 ft MSL, then to turn left and continued climbing to a specific fix over the ocean and hold over that fix at 3000 ft MSL. That fix is called `SD', and is at 11DME (11 nm on the Distance Measuring Equiment collocated with the localiser ITYA, the ILS/DME 05L localiser), and on the 294 degree (magnetic) radial from the TIA VOR (Very-high-frequency Omni Range) beacon located near the departure end of Runways 05. An unconfirmed report corroborated the suggestion that the pilot had declared a go-around (Source).
According to Aviation Week and Space Technology, the aircraft veered to the left on the ILS, "impacting on a road and striking buildings that run parallel to Runway 05/23L (Category 2), approximately a mile past the threshold. CAA officials said the [aircraft] possibly suffered a tail strike to the left of the runway while a go-around was attempted." (AW.98.02.23). Flight International said that "The impact point was 2,400-2,700m (8-9,000ft) from the threshold of the adjacent 3,660m-long runway. After impact, the aircraft crossed a road, went through a wall and hit housing. Wreckage was spread over an area some 400m in length along a projected track of 017-020°, some 30° left of the 053° runway centreline alignment. [...] Preliminary investigation of the crash site shows a V-shaped gouge at the start of the debris trail, believed to have been made by the A300's tailplane and indicating that the aircraft had a nose-up pitch attitude, thought possibly to be consistent with a stall during a go-around attempt. (FI.98.02.25). Flight's figure of impact point at 2,400-2,700m+ beyond the threshold is not compatible with Aviation Week's figure of 1 mile (roughly 1,700m).
The crash site was in a residential area. Although it is trivially
possible for any of the many television and newspaper journalists
on the spot to describe the specific location of the crash, I
have been unable to obtain this information.
CNN reported witnesses as saying that the airplane hit
"hundreds of yards short of the runway"
Edward A. Gargan of the New York Times Service reported "several hundred yards short of the runway" (IHT.98.02.17),
and the BBC said "hundreds of meters before the runway" (BBC.98.02.16).
The International Herald Tribune reported Chang Kuo-Cheng, Deputy Director of the Civil Aeronautics Administration in Taiwan, as saying "that the plane hit a utility pole and a traffic median in a road about 70 meters (200 feet) from the runway"; according to this report it "then skidded into several houses and exploded" (IHT.98.02.18).
Other reports have varied between 3km before the runway to 1.5km after it (Source). An unconfirmed report has indicated, however, that the site is 300m-500m left of the extended centerline of the runway (Source). If true, the aircraft was seriously off course. This would be significant for the accident investigation.
However, it should be clear that there are very different but superficially `reliable' descriptions here. To take Kuo-Cheng's comment, an aircraft impacting initially 70m short, with forward momentum, could be expected to `skid' quite a few meters before finally stopping. See the report of the Guam crash of flight KE801 in August 1997 (Lad97-Guam), in which the aircraft struck on a hill at 650 feet MSL and came to rest at 560 feet MSL. Even a very steep slope of 30 degrees would make that some 170 feet of slide, and in fact the map at (McK-18.8) indicates it to be about a quarter of a nautical mile (over a quarter of a statute mile), some 400 or so meters. This is consistent with the photograph in (McK-11.8). The presence of houses in the Taipei accident would stop such a slide abruptly, of course. But some of those amongst which the wrechage lies are three storeys high, as seen in photographs (IHT.98.02.18). Is it plausible to believe that there are three-storey residences well within 70 meters of the approach end of the runway at a major international airport?
One should not forget that the A300-600/600R is 54.08 meters long
(the A300B4-200 is 53.75 meters long)
That's comparable with the 70 meters distance from the runway of the
first impact attributed to Kuo-Cheng.
Did anyone report wreckage as lying
adjacent to the approach end of the runway, as the A300 length figures
above when put together with the 70m figure of first impact, along
with momentum in the direction of the runway, would suggest?
The communications transcript between CI 676 and ATC gives no indication of difficulties:
|20:02:18||CAL676||Taipei Tower, Good evening. CAL 676, 9 miles on final ILS Runway 05 approach.|
|20:02:26||Tower||CAL 676, Taipei Tower, Runway 05L, wind 350 at 5, QNH 1015 cleared to land [grinding sound on tape]|
|20:02:33||CAL676||1015, cleared to land|
|20:03:55||CAL676||Tower CAL 676, 3 miles on final|
|20:04:00||Tower||CAL 676 cleared to land, wind 360 at 3|
|20:05:12||CAL676||[Two loud rings audible, described by CAA as "warning bells" without elaboration. Loud grinding sound is heard again]|
|20:05:21||Tower||CAL 676 confirm go-around|
Both Flight International and Aviation Week reported that the gear appeared to be up, and Flight reported that "the slats and flaps were not set fully down", although the crew had acknowledged clearance to land (FI.98.02.23) (FI.98.02.25). If the gear was up, and the pilot had noticed that he had not put the gear down, that would explain a go-around attempt. The engines "were reported to have disintegrated, indicating they were at high thrust" (FI.98.02.25), which is consistent with a go-around attempt but not with a landing.
Chiang Kai Shek International has both primary radar (two-dimensional location on the surface) and secondary radar (Source). Secondary radar interrogates a transponder -- transmitter-responder -- in the aircraft, which responds with an aircraft ID and altitude (pressure altitude -- the altitude at which the airplane would be at the current static air pressure on the ICAO standard altitude. The standard altitude has a 0 MSL pressure of 1013 hectoPascal (millibars), or 29.92 inches Hg, and a standard rate at which temperature varies with altitude (the `lapse rate'). This pressure altitude would appear on ATC tapes. Its reading from the tapes can be corrected for non-standard MSL pressure (which is measured regularly, because it must be set in the `Kollsman window' of the altimeter of every approaching or departing aircraft so that the altimeter reads QNH), but it's not apparent to me whether it could be corrected for non-standard lapse rate. In any case, within 1,000 feet of the ground, a difference in measured altitude caused by non-standard lapse rate would be trivial, so we can probably ignore it.
According to unconfirmed current information, the airport does not have a Minimum Safe Altitude Warning (MSAW) syste m Such a system consists of software that compares an aircraft's reported altitude (via transponder and secondary radar) with the minimum safe altitudes for obstacle clearance known to and used by controllers. MSAW is prevalent amongst major airports in the USA, and in about half of major Japanese airports (Mae96). An MSAW software malfunction in Guam caused the system to fail to give warning of the fatally low profile of KE801 in August 1997 (Lad97-Guam).
According to Airbus Industrie, reported by CNN, the aircraft had been
delivered to China Airlines in December 1990, and had at the end of
January 1998 "accumulated approximately 20,070 flight hours in 8,800
China Airlines and its Safety Record
Back to Contents
China Airlines is the 34th largest airline in the world, ranked by
sales, and the 10th largest in the Asia-Pacific region
Unlike most other airlines in the region, it is not a member of
the International Air Transport Association (IATA)
A questionnaire circulated by Flight International, updated by reference to information supplied by the insurance company Airclaims, and valid up to and including 1 January 1997, list the following information for China Airlines.
It flies 41 machines, comprising 6 Airbus A300-600R, 6 Airbus A300B4, 3 Boeing B737-200, 3 Boeing B747-200, 6 Boeing B747-200F, 6 Boeing B747-400B, 4 Boeing B747SP, 2 Boeing B747-400, and 4 MD-11 machines. It has 6 Boeing B737-800 and 8 Boeing B747-400 on order, with options on 9 further B737-800, 4 more B747-400, and 4 Boeing B777-200 machines. China Airlines has simulators for Airbus A300B4, Boeing B747-200, B747-400 and MD-11 machines; and maintainance facilities (Airframe and engine) with 450+ airframe engineers and just under 100 powerplant engineers, with just under 30 avionics engineers. It carries maintenance approvals from the FAA (US), DGAC (France) and CAA ROC (Taiwan), and services principally A300B4-220, A300B4-622R, B727, B737, B747-200, B747-400 and MD-11 machines (FI.97.04.02).
As of August 1997, the Airclaims CASE database for passenger aircraft -- so no mention of the B747-200F -- shows China Airlines as operating 10 A300-600R machines with two on firm order (instead of the 6 mentioned above), as well as the 6 A300B4 machines; no B737-200 (instead of the 3 mentioned above), 3 B747-200, 5 B747-400 with 5 on firm order (rather than 2 with 8 on order, and no mention of the 400B variant) (FI.97.10.15). The comment in the International Herald Tribune that "Taiwan civil aviation authorities ordered China Airlines to ground its nine remaining Airbus A300-600R jets" substantiates the figure of 10 A300-600R machines from (FI.97.10.15). If this data is accurate, the airline acquired 4 A300-600s between January and August 1997, and placed firm orders for 2 more. But those wishing to assess the accuracy of the rest of the data will wonder 2+6 B747-400/400B machines became 5 B747-400 with 5 on firm order. Puzzling also is that Flight International's directory of Commercial Aircraft of the World includes no B747-400B variant (FI.97.10.15). One wonders if the "-400B" from (FI.97.04.02) is a misprint.
China Airlines Accident Record
Back to Contents
Todd Curtis lists four fatal accidents to China Airlines in the last 15 years (that is, since 1983), including Taipei and Nagoya. The other two are:
16 February 1986; China Airlines 737-200; Pescadores Islands, Taiwan: The aircraft touched down on the runway but crashed during an attempted go around. 26 October 1989; China Airlines 737-200; near Hualien, Taiwan: The crew was using an incorrect departure procedure and hit cloud shrouded high ground at 7000 feet (2130 meters). [...]
In 1985, a non-fatal accident over the Pacific, 300nm north-west of
San Francisco caused a lot of attention. The accident started when
an engine failure at 41,000 caused the autopilot to put
in control inputs under which the aircraft could not sustain straight
and level flight. It departed, and had rolled nearly 160 degrees
inverted before the autopilot was switched off. The ensuing dive
registered over 5g before the pilots recovered to straight and level
flight at about 9,000 ft.
More details, including the final accident report
NTSB/AAR-86/03 from the U.S. National Transportation Safety Board,
may be found in
This accident was non-fatal.
Back to Contents
Peter Neumann reported in the RISKS Digest a UPI story from the
San Francisco Chronicle of 31 July 1987. A monkey escaped from
a cage on a China Airlines cargo aircraft before landing at JFK
airport in New York. It forced the crew from the cockpit and played
with the controls before being caught
Back to Contents
On 4 November 1993, a Boeing B747-400 overran the runway and ended up
with its nose wheel in the Bay at Kai Tak airport in Hong Kong,
having landed in heavy rain and squalls associated with a nearby tropical
storm. This accident was non-fatal.
Back to Contents
On 26 April 1994, an Airbus A300-600 crashed tail-first while trying to land at the airport in Nagoya, Japan. The co-pilot, who was flying, had inadvertently triggered the Take-off/Go-around (TOGA) switch on the engine power levers. This caused the Flight Director (an advisory device on the Attitude Indicator) to switch to GO-AROUND mode, and increased thrust on the engines. The autopilots were subsequently engaged, with GO-AROUND mode still engaged. The co-pilot applied heavy nose-down forces to the control column under the Captain's instructions, and continued to do so. The autopilot on this A300 did not disengage under these forces, specifically because of its design. Thus it attempted to counteract the nose-down attitude of the aircraft by putting in nose-up trim on the Horizontal Stabiliser (the moveable, horizontal part of the tailplane, the empennage), causing an abnormal out-of-trim situation. The copilot and the autopilot were thus fighting each other. The captain took over control, judged that landing would be difficult, and opted for a go-around. The abnormal nose-up trim caused the aircraft to pitch up drastically, losing airspeed very quickly, stall, and impact the ground tail-first. The final report from the Japanese authorities singled out 12 causal factors. I quote in full from the report. My comments are in parentheses with the form "[..... PBL]"
While the aircraft was making an ILS approach to Runway 34 of Nagoya Airport, under manual control by the F/O [First Officer PBL], the F/O inadvertently activated the GO [GO-AROUND PBL] lever [positioned on the thrust levers PBL], which changed the FD (Flight Director) to GO AROUND mode and caused a thrust increase. This made the aircraft deviate above its normal glide path.
The APs [Autopilots PBL] were subsequently engaged, with GO AROUND mode still engaged. Under these conditions the F/O continued pushing the control wheel in accordance with the CAP's [Captain's PBL] instructions. As a result of this, the THS (Horizontal Stabilizer) [the horizontal part of the tailplane, the empennage PBL] moved to its full nose-up position and caused an abnormal out-of-trim situation.
The crew continued approach, unaware of the abnormal situation. The AOA increased, the Alpha Floor function [a particular automatic control function intended for situations in which a high angle-of-attack (AOA), the incidence of the wings to the air, is very high. A high AOA leads to a stall PBL] was activated and the pitch angle increased.
It is considered that, at this time, the CAP (who had now taken the controls), judged that landing would be difficult and opted for go-around. The aircraft began to climb steeply with a high pitch angle attitude. The CAP and the F/O did not carry out an effective recovery operation, and the aircraft stalled and crashed.
The AAIC [Japanese Air Accident Investigation Commission PBL] determined that the following factors, as a chain or a combination thereof, caused the accident:
- The F/O inadvertently triggered the Go lever It is considered that the design of the GO lever contributed to it: normal operation of the thrust lever allows the possibility of an inadvertent triggering of the GO lever.
- The crew engaged the APs while GO AROUND mode was still engaged, and continued approach.
- The F/O continued pushing the control wheel in accordance with the CAP's instructions, despite its strong resistive force, in order to continue the approach.
- The movement of the THS conflicted with that of the elevators, causing an abnormal out-of-trim situation.
- There was no warning and recognition function to alert the crew directly and actively to the onset of the abnormal out-of-trim condition.
- The CAP and F/O did not sufficiently understand the FD mode change and the AP override function. It is considered that unclear descriptions of the AFS (Automatic Flight System) in the FCOM (Flight Crew Operating Manual) prepared by the aircraft manufacturer contributed to this.
- The CAP's judgment of the flight situation while continuing approach was inadequate, control take-over was delayed, and appropriate actions were not taken.
- The Alpha-Floor function was activated; this was incompatible with the abnormal out-of-trim situation, and generated a large pitch-up moment. This narrowed the range of selection for recovery operations and reduced the time allowance for such operations.
- The CAP's and F/O's awareness of the flight conditions, after the PlC took over the controls and during their recovery operation, was inadequate respectively.
- Crew coordination between the CAP and the F/O was inadequate.
- The modification prescribed in Service Bulletin SB A300-22-6021 had not been incorporated into the aircraft.
- The aircraft manufacturer did not categorise the SB A300-22-6021 as "Mandatory", which would have given it the highest priority. The airworthiness authority of the nation of design and manufacture did not issue promptly an airworthiness directive pertaining to implementation of the above SB.
It should be fairly clear that most of these causal factors pertain to the crew's actions and understanding of the aircraft systems. However, contributing were unclear manual descriptions (also a factor in the Lufthansa Warsaw A320 crash in 1993 and the Birgenair Puerto Plata B757 crash in 1996), and the placement and design of the GO-AROUND lever on the thrust levers. Also that China Airlines had not performed a recommended modification to the aircraft.
The modification has to do with pilot inputs to the control column overriding the autopilot. The A300 was originally designed so that this could not happen below 1,500 feet. The intent was to prevent inadvertent control column input from disturbing an autopilot-controlled landing. The Airbus A300 is delivered with Category III capable autoland already installed. Category III autolands require the autopilot and may be performed in circumstances in which a manual landing by the pilots is disallowed. The control is very sensitive, and Airbus wished to eliminate the possibility that inadvertent inputs from the pilot could disturb an autoland.
However, most aircraft allow heavy pilot inputs to disengage the autopilot, and there was some discussion about this. The Service Bulletin SB A300-22-6021 referred to above changed the altitude below which pilots could not override the autopilot to at or below 400 feet ASL. This Bulletin was made mandatory, and the altitude was subsequently changed in March 1997 to 0 feet ASL, i.e., pilots could, with sufficient force (greater than or equal to 45 pounds push, greater than or equal to 100 pounds pull) always disengage the autopilot (AW.97.07.29). In September 1997, the FAA issued an airworthiness directive largely conforming to the conditions of the new SB, and requiring testing of the disconnect every eighteen months (AW.97.09.15).
It should be noted that the autopilot may be manually disconnected by the (standard) red disconnect button on the control wheel, and by an ON/OFF switch on the cockpit forward control panel (AW.97.07.29).
On May 3 1994, the Taiwanese authorities ordered China Airline to complete the modification specified in SB A300-22-6021 promptly. On May 7, they ordered supplementary training and a reevaluation of proficiency to be given to A300-600R pilots. On September 5, they issued an Airworthiness Directive (AD) requiring flight manual revision and Flight Control Computer modification within 24 months. An Airworthiness Directive is a compulsory order, without compliance with which an aircraft is unairworthy and may not fly. China Airlines had completed the modifications according to the SB by September 7. They rechecked the proficiency of all pilots under supervision of the Taiwanese authorities, and they carried out unscheduled inspections of engines, flight control systems, and autopilot systems on all aircraft, those on the A300-600R being completed by May 31.
The French Bureau Enquetes Accidents (BEA) recommended to the Direction General de l'Aviation Civile (DGAC) that a study be performed concerning pilot override of autopilot while in Land and Go-Around modes, and that modifications recommended by this study be made mandatory. On August 17, DGAC responded by issuing an AD similar to that subsequently issued by the Taiwanese authorities.
Airbus Industrie notified all operators of A300, A310 and A300-600 aircraft on May 5 of the hazards of overriding APs by means of the elevators when the APs are engaged in Land or Go-Around mode. On December 13 they reclassified the SB to mandatory.
The Japanese and US authorities issued similar ADs, making the Airbus SB compulsory, within similar time periods. The NTSB recommended to the FAA that an alert system be provided in the A300/310 autopilots to warn the pilots when the THS is in motion. The FAA had not taken action on this by the date of publication of the Nagoya final report on July 16 1996.
The required modification in the Airbus SB, which became an airworthiness directive in Taiwan, France, the US and Japan, ensured that the Nagoya accident could not be repeated in these aircraft. The pilot's control input at low altitudes would cause disengagement of the autopilot.
The information above which does not contain an explicit citation is taken from the final report, a digitised version of which is available from (ChinaAir-Nagoya).
It is very important to note that, because of the ADs,
the circumstances causing the Nagoya accident cannot be repeated.
In particular, they cannot have repeated in the February 16 1998
Back to Contents
On 11 December 1995, a taxiing B747-200B at Manila in the Phillipines
struck a stationary vehicle close to the taxiway with the Number 4
engine, damaging the engine. On 12 December, the crew ferrying the
aircraft on three engines to a repair station lost directional
control on takeoff, veered right off the runway, and the nose gear
Back to Contents
The accident on February 16 is the first fatal accident for
China Airlines in the five preceding years.
The FAA's International Safety Assessment Program Rating
Back to Contents
Countries are assessed by the US FAA according to their aviation safety
records and oversight procedures. Taiwan was rated Category 2, in
which flights to the USA had been subjected to extra FAA surveillance.
After the Nagoya accident, the FAA worked closely with China Airlines
and the Taiwanese authorities on their safety and training programs.
The Taiwanese safety-oversight system was upgraded to Category 1,
signifying full compliance with International Civil Aviation Organisation
(ICAO) standards in April 1997
Back to Contents
CNN has questioned the safety of the A300 (CNN98.02.16). They cite as evidence the four A300 crashes:
CNN also quotes Professor John Hansman of MIT as saying
There have been issues with the design of the Airbus autopilots having to do with the automation philosophy Airbus uses. Again, we don't know whether that's related to this accident.(He is of course right. We don't have any idea at the moment whether anything to do with any of the airplane systems is related to this accident.)
The only A300 accident which has `raised issues' about the Airbus automation philosophy was Nagoya, concerning manual override/disengage of the autopilot close to the ground, which was discussed above. Most of the questions concern the Airbus fly-by-wire aircraft, the A319/320/321/330/340 series.
Two of these four accidents were CFIT. All CFIT accidents have a large component of human error. Pilots are legally responsible for knowing where the aircraft is at all times. If an aircraft is flown into terrain or ocean inadvertently, as in a CFIT accident, then it is necessarily true that the aircraft was flying outside safe or cleared airspace. The minimum safe altitudes are clearly marked on approach charts for each sector, and on en-route charts for cruise. Looking at it strictly legally, therefore, the pilot in a CFIT accident contributed to the crash by not fulfilling his responsibility to know where (s)he was. Computer issues have entered into just one reknowned CFIT accident, that of a B757 in Cali, Colombia in December 1995 (AA-Cali). However, the probably causes had mostly to do with pilot actions.
In short, CNN's suggestion that there are `questions' about
the A300's safety can be seen to rest on the Nagoya accident alone.
Hansman's quote gives no support, and the Nepal and Indonesian accidents
Conclusions and Questions
Back to Contents
Assuming that the weather was between Category I and II, with no unusual winds, and that the pilot had declared no emergency and there was no signal of anything unusual in ATC communications, this accident has the hallmarks of CFIT (Lad97-CFIT) (KhRo96), (Wie77). CFIT was the biggest killer by far in 1997 (FI.98.01.21). If this accident is indeed a CFIT accident, there must be a large component of human error.
There are some unconfirmed reports that the Japanese and Taiwanese
press are following CNN's lead and questioning the computer systems
on the A300. If the accident is indeed CFIT, the computer systems will
not be the major causal factor in the accident -- some sort of human
error will be.
In any case, even if the accident turns out not to be CFIT,
since an A300 is not a fly-by-wire aircraft, it is not
possible for a `computer error' alone to cause an A300 accident.
(I described a similar argument in the case of the Aeroperu B757
crash in 1996, in response to similar premature claims from various
Peruvian official sources
Back to Contents
As with any accident, various questions need to be answered before any serious conclusions can be drawn about this accident. These answers can come fast, or slowly.
I thank David Learmount, Operations and Safety Editor of Flight International, for some informative discussions on many of these topics.
(AW.98.02.23): Geoffrey Thomas and Pierre Sparaco, Retracted Landing Gear Cited in China Airlines Crash, Aviation Week and Space Technology, February 23, 1998, p45, available at http://www.aviationweek.com/safety/news/nz030298.htm#Retracted Landing Gear Cited In China Airlines Crash (Back)
(CNN98.02.16): CNN, Taiwan Crash raises questions about Airbus A-300, at http://www.cnn.com/WORLD/9802/16/airbus.300.safety/ Author, February 16, 1998. (Back)
(CNN98.02.16.2): CNN, Airliner crashes in Taiwan, at http://www.cnn.com/WORLD/9802/16/taiwan.crash.update/ Author, February 16, 1998.
(CNN98.02.16.3): CNN, Nearly 200 feared dead in Taiwanese plane crash, at http://www.cnn.com/WORLD/9802/16/taiwan.crash.update2/ Author, February 16, 1998.
(CNN98.02.16.4): CNN, More than 200 feared dead in Taiwanese plane crash, at http://www.cnn.com/WORLD/9802/16/taiwan.crash.update3/ Author, February 16, 1998. (Back)
(Cur): Todd Curtis, Airsafe.com : Airline Safety Information for the Airline Passenger, http://www.airsafe.com --> Fatal Airline Events by Airline Since 1970 --> Asia/Pacific Region --> China Airlines (Back)
(FI.98.03.11): David Learmount, Loss of control is key to China Airlines accident, Flight International, 11-17 march 1998, p9.
(IHT.98.02.17): Edward A. Gargan, China Airlines' Jet Crashes in Taiper, Killing 197 Aboard, article from the New York Times Service, International Herald Tribune, 17 February 1998, p1.
(IHT.98.02.18): Airbuses in Taiwan Grounded After Crash, story `compiled from dispatches', International Herald Tribune, 18 February 1998, p4.
(KhRo96): R. Khatwa and A. L. C. Roelen, An Analysis of Controlled-flight-into-terrain Accidents of Commercial Operators, 1988 Through 1994, Flight Safety Digest 15(4/5), April-May 1996. Available at http://www.flightsafety.org/FSD1996.html. (Back)
(LadComp): Peter B. Ladkin, ed., Computer-Related Incidents with Commercial Aircraft, Compendium RVS-Comp-01, Networks and Distributed Systems Research Group, Faculty of Technology, University of Bielefeld. Available at http://www.rvs.uni-bielefeld.de/~ladkin/FBW.html.
(Lad96-Aeroperu): Peter B. Ladkin News and Comment on the Aeroperu B757 Accident, Research Report RVS-RR-96-16, Networks and Distributed Systems Research Group, Faculty of Technology, University of Bielefeld, 8 November 1996. Available at http://www.rvs.uni-bielefeld.de/publications/abstracts.html#AeroPeru. (Back)
(Lad97-approaches): Peter B. Ladkin Flying an ILS or Localiser Approach - An Example, Report RVS-J-97-07, Networks and Distributed Systems Research Group, Faculty of Technology, University of Bielefeld, 25 August 1997. Available at http://www.rvs.uni-bielefeld.de/publications/abstracts.html#approaches. (Back)
(Lad97-CFIT): Peter B. Ladkin Controlled Flight Into Terrain: What Is Being Done?, Report RVS-J-97-08, Networks and Distributed Systems Research Group, Faculty of Technology, University of Bielefeld, 21 August 1997. Available at http://www.rvs.uni-bielefeld.de/publications/abstracts.html#CFIT. (Back)
(Lad97-Guam): Peter B. Ladkin, The Crash of Flight KE801, a Boeing B747-300, Guam, Wednesday 6 August, 1997: What We Know So Far, Report RVS-J-97-06, Networks and Distributed Systems Research Group, Faculty of Technology, University of Bielefeld 11 September 1997. Available at http://www.rvs.uni-bielefeld.de/publications/abstracts.html#guam. (Back)
(Mae96): Hirokazu (Mike) Maekawa, An Effort to Prevent Controlled Flight Into Terrain (CFIT), International Society of Air Safety Investigators Regional Seminar, Asia Pacific Region, May 29, 1996. (Back)