The Crash of Flight KE801, a Boeing B747-300, Guam, Wednesday 6 August, 1997: What We Know So Far

Peter B. Ladkin

Article RVS-J-97-09

The publically-available facts concerning the accident to KE801 are given, along with references to sources, and the features of the accident that arise from these facts are discussed and analysed, along with various pieces of reasoning and commentary from the press.

On Wednesday 6 August, just before 02.00am (16.00 GMT Tuesday) when it was still dark, Korean Airlines Flight 801 from Seoul to Guam crashed into terrain on approach to Won Pat International airport in Agana, Guam. The impact site lies at 658ft on Nimitz Hill (Lew) ( `about 650ft. msl.' according to (McK-18.8) and a `few hundred yards' from the VORTAC antenna (McK-11.8)) and came to rest at about 560ft MSL (McK-18.8) just to the north of and downslope of the Nimitz VOR. The VOR is located at 680ft altitude according to (Airnav), but the Jeppesen chart (JeppGuam) (also reproduced in (McK-18.8)) shows an obstacle of 724ft at the VOR; (McK-18.8) shows the UNZ VORTAC at 723ft and the elevation of Nimitz Hill as 709ft. This yields a height of 14ft for the VORTAC antenna, which is described as being 6.7m high (22ft high) in (Lew). The wreckage lay within a few hundred meters of the VOR antenna (pictures in (McK-11.8), (McK-18.8), also (NTSB)). The VORTAC is on the approach path, and is a `step-down fix' (Lad97-approaches) for the localiser-only ILS 6L approach, which is a non-precision approach.

The B747-300, registration HL7468, was delivered December 12, 1984, and had accumulated 49,526 flight hours as of June 30, 1997, and 8,340 cycles (takeoffs and landings) (CNN-05b) (McK-11.8).

KE801 had been cleared for the ILS 6L approach at 1:39:44 and was advised the glide slope (GS) was out of service (OTS) (McK-18.8). The pilot acknowledged receiving the ATIS information, which also included the GS OTS NOTAM, and `a copy of the NOTAM, with the glide slope information highlighted, was found in the cockpit' (McK-18.8), indicating that this information had been briefed before departure. `An initial analysis of the cockpit voice recorder indicated that the crew also discussed the unusability of the glideslope' (McK-18.8). It was also reported that KA 801 had been cleared to land ((Rea), in (IHT-06)). The last recorded `hit' by the Agana Cerap radar was at 1:42:26, and based on that data and timing off the CVR, investigators calculated the accident to have occurred at 1:42:30 (McK-18.8).

The control tower at Aguna is operated privately by the Johnson-Barton company (Airnav).

KE801 was nearly 800ft below its legal minimum altitude when it impacted the hillside. The overseer of the on-site investigating team, NTSB member George Black, said the crash "has all the earmarks of controlled flight into terrain" (McK-11.8) (Jor-08). Controlled flight into terrain (CFIT) accidents caused the deaths of 2,200 people, over half the air carrier fatalities, in the period 1988-1995 and is of great concern to aviation safety organisations (Lad97-CFIT) . Let us start with a look at some CFIT questions, and compare the recommendations of the Flight Safety Foundation CFIT Task Force in 1994 directly with the situation at Agana.

Controlled Flight into Terrain Questions

The 1994 Flight Safety Foundation Task Force anti-CFIT recommendations (Lea94) were

The following situation obtains at Agana with respect to these recommendations:

One `classic' definition of a CFIT accident is

A CFIT accident is one in which an otherwise-serviceable aircraft, under control of the crew, is flown (unintentionally) into terrain, obstacles or water, with no prior awareness on the part of the crew of the impending collision.
(Wie77), quoted in (KhRo96).
and another, used in the recent Dutch National Aerospace Lab study, is
A CFIT accident is one in which an aircraft, under control of the crew, is flown (unintentionally) into terrain, obstacles or water with no prior awareness on the part of the crew of the impending collision.
(KhRo96).
The difference lies in omission of the phrase `otherwise-serviceable'.

Corresponding to these definitions are criteria. Thus the major reasons why the NTSB will be considering CFIT closely are that

There are five major pieces of evidence so far that suggest that the aircraft was under control when it impacted:

The first two items convey that the wreckage pattern and the track of the aircraft over the ground are judged to be consistent with a relatively shallow impact angle. The fifth item confirms this rather precisely, if indeed it turns out to be correct. One would expect an aircraft whose control has been lost to have impacted at a greater angle, and presumably to have broken up in a different pattern. to aircraft who impact more shallowly. The first two and fifth pieces of evidence would therefore be quite telling to investigators looking at them that the aircraft impacted under control. In associate with the third and fourth pieces of indirect evidence, they would be quite compelling. This suffices to explain why the NTSB has confirmed that CFIT is a strong candidate, but there may be other reasons that we do not yet know about.

However, it is always possible that one or more of the conditions for CFIT (see the definitions above) may be found not to pertain. It is possible that a subtle mechanical failure will be discovered; it is also possible that after careful consideration of the cockpit voice recorder (CVR), the crew will be found to have been considering and dealing with an as-yet-unrecognised problem causing them to be dangerously low on the approach. The intra-cockpit talk is in Korean, and expert translators are needed for accurate transcription. It is also possible that the 950fpm descent rate turns out not to be accurate. Let us call these `counter-states' Because it has not yet been determined that such counter-states did not occur, no definite conclusion can be drawn at present. However, the evidence known at the moment contrasts very strongly with the way that the evidence would have been expected to look if one of these counter-states occurred. Thus the `no-counter-state' assumption appears at this point to be justified, and it follows from the evidence plus the no-counter-state assumption that this was a CFIT accident. The no-counter-state assumption shows that this presumption is non-monotonic reasoning (and will remain so). To conclude CFIT, the non-counter-state assumption must therefore be extensively justified, and we may presume that the ongoing investigation will attempt to justify the no-counter-state assumption or find a counter-state.

GPWS False-Positives from Nimitz Hill

Apparently Nimitz Hill is one of those `specific locations' that cause `nuisance' GPWS warnings - false positives.

`Airline crews flying to that runway routinely get warnings from their aircraft's ground proximity wasnint system as they pass the UNZ VORTAC. That beacon, which is used for the final fix for the approach, sits atop Nimitz Hill
    Some airlines instruct their crews to expect the ground-proximity warning ahd to ignore it, double-checking their clearance of terrain with their aircraft's radio altimeters. The Korean Air 747 had a recent-model ground proximity warning system that provided altitude callouts to the crew before the accident.
'
(McK-11.8)
The crew obtained a GPWS warning clearly discernible on the CVR tape: "The conversation on the voice recorder was minimal, but on the the data recorder a ground proximity alarm did sound before impact", [NTSB member George] Black said' (CNN-07), also reported in (Phi-09). Had the GPWS not been late-model, it might not have given a warning at all: concerning enhanced GPWS (EGPWS) that " unlike conventional GPWS, sensing is not automatically disabled when the aircraft is in the landing configuration [..]" (Pro97.2). This implies that `conventional' GPWS is disabled when the aircraft is in the landing configuration, as KE801 was when it impacted. In fact, as noted above, the GPWS gave three altitude callouts that do not appear to be consistent with the kinds of false-positive warnings that one obtained from earlier-model GPWSs. It therefore remains a serious question as to why the crew did not appear to react to the altitude callouts, and we may expect the investigation to address this issue.

Radar Service and MSAW

Flight International reported that the airport primary surveillance radar was NOTAM'd OTS (out of service) (Lew) at the time of KE801's flight. This appears to be incorrect. Radar data was available from Cerap (McK-18.8), and preliminary investigation showed the surveillance radar was generally in working order (FAA).

However, the tower controller's radar at Agana is also equipped with a Minimum Safe Altitude Warning (MSAW) system, which is integrated into the system software. The MSAW antenna is located at Andersen AFB, roughly some 10nm (11.5 statute miles) beyond the departure end of Runway 6L at Agana. This system normally has a 55-mile-radius service range from the antenna (IHT-11.8) (CNN-10) but at this time the effective coverage was restricted to a mile-wide circumferential strip only (CNN-10). This reduction in coverage was discovered after inquiries by NTSB investigators Wentworth and Pereira as to why the MSAW alarm had not sounded during KE801's transgression (McK-18.8). It was due to a bug introduced into the MSAW system by a software upgrade developed by the FAA Technical Center in New Jersey intended to reduce the number of false-positive alarms the system was generating, in response to complaints by the Cerap operators (McK-18.8). During subsequent testing at 191 MSAW installations in the U.S., three occurrences of software errors were found and corrected: at Cerap; also at Fayetteville, North Carolina and Florence, South Carolina, where the errors had caused no known incidents (FI-27.8). Investigators said that the lack of MSAW advisories did not cause the crash, but could have helped prevent it (CNN-10), (IHT-11.8) (FI-27.8).

Local radar service of any sort, including MSAW, is not required for performing the localiser-only ILS 6L approach into Agana.

Radio Navigation Fixes on Approach

An approaching aircraft on the full or localiser-only ILS into Agana flies a course of 063° magnetic, guided by the laterally-sensitive localiser radio beam at Agana from an Initial Approach Fix or Point (IAF) 7nm on the Distance Measuring Equipment (DME) from the Nimitz VOR, identifier UNZ, which lies directly on the approach course 3.3nm from the 6L runway threshold. Charts describing the approach in both plan and profile views, similar to the charts which pilots use when flying the approach, as well as the approach procedure are presented and discussed in detail in (Lad97-approaches).

The particular positions (`fixes') on the ground that the aircraft on the localiser-only ILS 6L approach is required to identify are three:

All these three fixes at Agana are redundantly indicated by the equipment aboard an aircraft, were this equipment to be used in a standard manner, so the procedure in effect incorporates error-detection of one navigation instrument failure (see ILS-article reference below). Since the procedure requires a go-around on detection of any instrument anomaly, it is therefore fail-safe providing it is followed rigorously and there aren't correlated multiple navigation-instrument or -signal failures. One would expect the investigation to try to answer the question whether these conditions were met.

The redundancies are:

These redundancies are not only that at least two different modes of indication occur, but also that these modes are indicated simultaneously on different pieces of radio equipment. The various equipment: localiser signal and marker-beacon signals; VOR and DME signals, are indicated as correctly transmitted by an audible identifying signal which in principle must be monitored during approach (in practice, the radio navaids are often identified once at the beginning of the approach and then the IDs are not further monitored, because of the cacophony it would cause).

Ensuring the reliability of the ground transmissions is the domain of the FAA. The required reliable transmissions are:

Additionally, to attain the redundancy noted above, the OM and MM signals must be working reliably. But they are, strictly speaking, not required for the localiser-only ILS approach.
`The DME's logs recorded no interruptions the night of the crash, but investigators want to verify that. They also are examining whether Navy antenna arrays nearby may have interfered with the DME and are questioning local pilots about their experience with the DME on approach over the last several weeks. '
(McK-18.8)
The aircraft was on course, passing very close (um, too close) to the VOR, so there is no immediate need to examine the reliability of the course indication signals from the localiser, since there is no evidence of aircraft deviation from this course.

` Investigators interviewed the flight crew of a Continental Airlines aircraft that landed at Guam about 30 min. before the accident. The pilots said they had flown a visual approach and that red flags came up on their glideslope indicators, showing that no glideslope signal was being received. The performance of the VOR, DME and the four-box visual approach slope indicator (VASI) to the left of Runway 6 Left all appeared to be normal, they said.'
(McK-18.8)
Red flags were also up, indicating no glide slope signal, on the pertinent instruments on KE801's flight deck.

Glide Path and Altitude Deviations

Had the aircraft passed the FAF at the required indicated altitude (on the altimeters), and had the indicated altitude been approximately correct, the aircraft would have had to have flown a descent gradient (`glide path') of over 7.7 degrees to arrive at the impact point. Such an angle of descent would be dangerously steep (this is between two and three times the expected glide-path angle), and the rate of descent at typical approach speeds would have been between 1,800 and 2,200 feet per minute, which would be dangerously high at this point in an approach. As noted above, `Early indications were that the aircraft was in a stable, 950-fpm. descent' (McK-18.8), which is far lower than the rate needed to descend from the FAF at minimum altitude. The evidence therefore points to the aircraft being below minimums at the FAF.

As noted above, the pilot acknowledged receiving the ATIS, which included the altimeter setting information, and `the altimeters were stuck at about 600 ft.', an altitude broadly consistent with the altitude of impact. This indicates that the altimetry was broadly correct. This suggests that, had the aircraft indeed been low at the FAF, this would have been noticeable from the altimeter readings at FAF passage. The altimeters are primary flight instruments and are thus observed with priority during instrument flight and especially on approach.

The FAF would have been passed less than a minute before the accident (it is 1.6nm from the VOR, and an aircraft flying at 96kt, far lower than any safe approach speed for a B747, would traverse this distance in 1 minute). Were the `stable, 950fpm descent profile to be a roughly correct description, pertaining since the FAF, the aircraft would have passed the FAF many hundreds of feet lower than the minimum altitude at the FAF.

Weather and Visibility

At the time of the accident, weather was reported as

(McK-18.8).
The weather reported at the airport cannot necessarily be extrapolated to the approach path. First, clouds tend to gather around and over hills, of which there were some on the approach path, notable Nimitz Hill, and secondly, rain showers were localised.
`Most witnesses told investigators that there were cells of heavy rain in the area [...]. The duty weather observer at the airport, who has worked on Guam for 13 years, said the rain showers were heavier than normal. The accident aircraft itself had requested a number of deviations around storm clouds while inbound to Guam, the approach controller who handled the flight told investigators, as had most other aircraft in the area that night. The recording from a Doppler radar unti 4 naut. mi. east of the area showed moderate-intensity rain showers along the approach course.'
(McK-18.8).
However, a witness standing on the hill very close to where the aircraft impacted said it was not raining there at the time (McK-18.8).

While windshear is normally associated with thunderstorms and violent atmospheric activity rather than with moderate-intensity rain showers, the terrain surrounding the approach course is hilly and the air flowing over it can be turbulent (McK-11.8) (McK-18.8).

The terrain under the approach course and on Nimitz hill is largely unlit, and Feith said that pilots can mistake the terrain for clouds, and the NTSB is investigating the possibility of this and other optical illusions which may have led the pilots visually to believe they were too high on approach (McK-11.8) (McK-18.8).

It has been noted that 30 minutes before, an aircraft was able to make a visual approach to the runway. That means that at the FAF, the aircraft was able to identify the runway or its environment (presumably the runway lights and the VASI) and maintain this visual contact up to landing.

The runway environment consists of a VASI and runway edge and extremity lights. The VASI consists of two light beams, one above the other, split into red and white components, beaming up along a glide path, `normally set at 3 degrees' (AIM, Section 2-1-2.a, Visual Approach Slope Indicator (VASI)). An aircraft on the indicated glide slope sees the lower beam as white and the upper beam as red. An aircraft above the glide path sees both as white, and an aircraft below sees both as red. In clear weather, a VASI is powerful enough to be visible `up to 20 miles or more at night' (AIM, Section 2-1-2.a). In weather, of course, the VASI may not be visible at all until the very last stages of approach, if at all. Seeing the VASI alone constitutes `having the runway or its environment in sight' for the purposes of completing an instrument landing from MDA or DH.

The runway does not, however, have other visual aids such as Runway End Identifier Lights (REIL), which is a pair of synchronised high-intensity white strobes, runway centerline lighting or touchdown-zone lighting. It appears from a CNN photograph that Runway 6L has an Approach Light System for aircraft on instrument approach (CNN-06a, second photograph) . I am trying to determine what type this is (Low-, Medium- or High-Intensity). This information is available in U.S. Government publications.

There is no way to determine categorically whether any of the runway or its environment lights were visible to the pilots at any point in the approach. It is possible that cockpit voice recorder evidence might indicate that the pilots believed at some point that they had the runway in sight (which wouldn't necessarily mean that they did have it in sight), but nothing has been said about this publically by the NTSB. There is also no way to determine categorically what clouds were present and where on the approach path at the time of the accident.

Social Concerns and Media Accuracy

The BBC World television service interviewed David Learmount, Flight International's Operations and Safety Editor, on the evening of Wednesday 6 August (GMT). The BBC report focused on the GS being out of service, as did many early news media reports. While this is interesting information, because had the GS been available the aircraft would undoubtedly have flown the full ILS, procedurally it is not relevant to the investigation because the GS is not part of the approach that the aircraft was in fact flying.

CNN reported in their second article that the glide slope was not working, but that pilots can use other methods, including DME, that allows them to follow a stair-step pattern to land (CNN-05a). This report was basically accurate, but omitted the procedural background concerning NOTAMs, how an approach is selected, and what the basic responsibilities are of each social branch of the system: pilot, airline management, FAA.

The second important discovery for the press was the outage of the MSAW. This was a discovery to investigators, not only to the press. CNN, the International Herald Tribune, and the BBC World television news all reported accurately that `"This is not a cause - it might have possible been a prevention," George Black [...] said (CNN-10), (IHT-11.8) (FI-27.8).

Inaccuracies

Despite their earlier accuracy, a later CNN report paraphrased NTSB Chairman Jim Hall as saying that the NTSB will be looking at reports of the `malfunctioning glide-slope system' (CNN-06). The GS was not malfunctioning, and Hall did not say that. He said, as was reported later in the article that they'll be looking at why the system was "out or being replaced" (CNN-06). The FAA took the GS out of service from July 7 (a month before the accident) until September 12 to `relocate and upgrade it' (McK-11.8). The outage was published in NOTAM form, as required. Guam Governor Carl Gutierrez `said that all airlines flying into Guam know that the system wasn't working, and that "hundreds of planes have landed without it, and a dozen or so since" the crash of Flight 801.' (CNN-06). So there was enough information in the article to limit any misunderstandings that might have been caused by the misinformation of a `malfunction'.

The Electronic Telegraph, the electronic version of a British national newspaper, the Daily Telegraph, commenting on the MSAW on 11 August, said that `The crash [...] has been blamed on a radar system [the MSAW]. Federal agents said the radar [...] was affected by a programming error.' (ET-808). The first sentence is not only false, but completely misleading, given that it specifically contradicts what Black in fact said. The second sentence is, of course, true.

Misleading Causal Assertions

Some interested parties to the accident had very early opinions as to what had caused it. A Korean Air Vice President told CNN that it was the weather and the lack of GS (CNN-06). A later CNN report had Korean Air reiterating that assertion and said that the company `angrily denied' that pilot error was to blame (CNN-08). How Korean Air can have known that, when investigators on site in Guam did not, is hard to say. In fact, the statement is highly misleading to the point of being nonsense. Here is the reasoning.

Let us suppose that the weather and the lack of GS were causes. The weather forecast, which for Pacific islands a few hours ahead is moderately accurate, and the lack of GS were known to the pilot before departure (he is required to be briefed on weather and NOTAMs for the route of flight, including the airport of arrival). Since pilot error was not a factor (this is our, or rather Korean Air's, supposition), both of the sufficiently determining causal factors of the crash were known to the pilot before takeoff. Proceeding with the flight at that point was thus dangerous. Since both pilot and airline management must accede to taking off for the flight to occur, we conclude that both acceded to departure despite the fact that the flight could probably not be safely completed (I say `probably' only because it was not known exactly what the weather would be on arrival - maybe there would be a miraculous break in the clouds for 5 minutes!) A pilot judgement to pursue a course of action which the aircraft cannot safely complete is a pilot error, as the term is normally used. Let us call the corresponding action on the part of Korean Air a management action (even though a decision may not explicitly have occurred, the choice is nevertheless made. An act of omission may still be regarded as an action when an action is required, which it is, because the airline is responsible for monitoring the pilot's training, behavior and judgement). Since pilot error was not a cause (we are supposing), it cannot have been an error of judgement on the part of the pilot, therefore the corresponding management action on the part of the airline occurred.

Was this management action causal to the accident? Had this management action not occurred, the flight would not have proceeded. Had the flight not taken off, it would clearly not have crashed in Guam. Therefore this management action is a necessary causal factor. Thus, from the supposition that weather and lack of GS were sufficient causes, and that pilot error was not a cause, we are led to conclude that a management action on the part of Korean Air was also a necessary causal factor. It follows from Korean Air's statement, then, that they themselves are partly responsible for the accident, since this management action lies within their power to make or not to make. We may suppose that that is not a conclusion they wish the public to draw from their assertion of causality. However, it does follow, and to my knowledge this has not been explicitly noted so far.

I should perhaps point out the difference between what I am asserting and what Korean Air is asserting. I have asserted, and demonstrated, that

If weather conditions and lack of GS are sufficient causal factors for the accident, then a Korean Air management action is a necessary causal factor.
I have not asserted that any one of these three factors actually is a causal factor. Korean Air, though, has done so. It has asserted the antecedent of the conditional, and therefore they are committed to acknowledging the truth of the consequent. Therefore they have asserted and acknowledged
Both weather conditions and lack of GS are causal factors for the accident, furthermore they are sufficient causal factors, and so a Korean Air management action is also a necessary causal factor.
This is a conjunction, whereas my (justified) assertion was a conditional, formed from the same basic sentences concerning what is a causal factor. It is elementary logic to see that the two assertions are different.

Some Hypotheses Discussed

It is far too early in the investigation to assert causality. Not all the facts have been made public and the NTSB has not completed its investigation in any of the significant areas in which they need to gather facts. We have already seen in the last section the trouble one may get into if one asserts causality too blithely.

But it seems that most of the suppositions as to what might have happened have their conflicts with the facts known so far. I examine three of them, and indicate their difficulties.

Associating the DME With the ILS

Nevertheless, one serious hypothesis was mooted early on - that the pilot may have thought that the DME transmitter was collocated with the ILS on the airport. Flight International said:

`Initial readouts from the cockpit-voice recorder and flight-data recorder downloaded by the NTSB indicate that the pilot may have assumed that the Nimitz beacon was located on the airfield, explaining why the 747 was some 800ft below its prescribed altitude. '
(Lew)
and CNN said:
` Investigators want to know if the pilot had complete control over the plane and simply tried to land in the wrong place. Did he, for example:
  • Miscalculate the approach to the runway, homing in on a landing guidance beacon on the hilltop where the crash occurred, mistakenly believing he was actually homed in on the runway?
  • Lower the landing gear too soon, disabling an alarm system that would have warned him the plane was too close to the ground?
'
(CNN-08).

We know that the GPWS did in fact provide the warnings, so the second of the questions reported by CNN is answered above. The first of those questions, and the Flight International comment, seem to refer to the same topic.

Some ILSs have associated DME transmitters, and the approach plate for Agana ILS 6L explicitly states that DME is required (JeppGuam). However, the chart clearly depicts the VOR as being in the middle of the approach, not at the airport, and the DME distances shown are quite unambiguous. It is hard for any instrument pilot such as myself to see how such a mistake could possibly have been made when looking at the chart.

But maybe the pilot didn't look at the chart at all (Anon)? First, this would be highly illegal, not to speak of irresponsible behavior. To determine precisely which regulations govern the conduct of Korean Air flights in the U.S., the following U.S. Federal Aviation Regulations concerning preflight briefing, operation of foreign civil aircraft, and carrier-route certifying are relevant:

14 CFR 91.103 Preflight Action.
Each pilot in command shall, before beginning a flight, become familiar with all available information concerning that flight. This information must include--
  • (a) For a flight under IFR or a flight not in the vicinity of an airport, weather reports and forecasts, fuel requirements, alternatives available if the planned flight cannot be completed, and any known traffic delays of which the pilot in command has been advised by ATC;
  • (b) For any flight, runway lengths at airports of intended use, and the following takeoff and landing distance information:
    • (1) For civil aircraft for which an approved Airplane or Rotorcraft Flight Manual containing takeoff and landing distance data is required, the takeoff and landing distance data contained therein;
[....]
14 CFR 91.711 Special rules for foreign civil aircraft
(a)General. In addition to the other applicable regulations of this part [Part 91], each person operating a foreign civil aircraft within the United States shall comply with this section.
[...]
14 CFR 121.97 Airports: Required Data.
[...]
(b) [...] each domestic and flag air carrier must show that it has an approved system for obtaining, maintaining, and distributing to appropriate personnel current aeronautical data for each airport it uses to ensure a safe operation at that airport. The aeronautical data must include the following:
  • [...]
  • (5)Instrument flight procedures
    • (i) Departure procedure
    • (ii) Approach Procedure.
    • (iii) Missed approach procedure.
  • [...]

(FAR, 14 CFR 91.103, 91.711, 121.97)
The unavailability of the glideslope was NOTAM'd, as had been mentioned. The localiser-only ILS approach was therefore the expected approach, as it had been since July 7, and it constitutes part of the `available information concerning that flight'. (The VOR DME 6L approach, an alternative approach to the same runway, is one of the `available alternatives had the localiser been unavailable on arrival. Had the DME been out of service, a different airport, the so-called `alternate', selected before departure, would have been the alternative.) Second, this pilot had not flown the localiser-only approach to Guam since the GS had been out of service. He would therefore have had to look at the chart at least to find out what the step-down altitudes were. Furthermore, investigators found in the cockpit a current set of approach plates for Guam, with the ILS 6L chart on top in the plastic sleeve which many pilots use for such plates (McK-18.8). This hypothesis, that the pilot didn't look at the chart, thus seems highly dubious.

But it is also hard to see how the hypothesis that he thought the VOR was on the airport is consistent with the facts. If the runways were somehow in sight, it would be very clear that they were still 3+ miles away. If the runways were not in sight, the pilot must have known at least the altitude of the airport or the altitude of the MDA (and thus concluded that the airport was much lower). With the DME (assuming it was indicating correctly) showing a tiny fraction of a mile to go, the aircraft was still some 400ft above runway touchdown zone elevation (and some 350ft above `airport elevation', obviously measured at a different location from the runway touchdown zone). If you think you're at the airport, there's no way to lose 400ft and land in a B747 - it would take 1.25nm to lose this 400ft at a normal glidepath of 3° (for comparison, the glide slope crosses the runway threshold at a height of 50ft to make a normal landing). Also, if you know the MDA is 560', you may infer that the runway is two to three hundred feet below that, and the same reasoning applies. The only consistent hypothesis is that the pilot would somehow have mistaken MDA for runway elevation, which would have put him at an altitude of what he thought was 100ft above the runway at the accident site. But the only reason for entertaining this hypothesis is that it's consistent with the other known facts. It doesn't make any sense from a cognitive point of view.

So the hypothesis that the pilot thought that the DME was on the airport has its difficulties in face of the known facts.

`Ducking under' on a Visual

Another hypothesis that has been entertained, privately, is that the pilot had the airport in sight earlier, was trying to perform a visual approach, and was trying to `duck under' what he thought were clouds, but that was in fact hillside. We have already noted Feith's remark that the pilots can `easily mistake the dark, undulating terrain for clouds' (McK-18.8).

However, this hypothesis doesn't stand up to much inquiry, either. We have noted that if the pilot were to have been making anything like an approach with an acceptable rate of descent, the aircraft would have been low at the FAF. This implies that he would have seen the runway lights outside of the FAF, and would have been maintaining them in sight. However, it was noted above that there were moderate-intensity rain showers along the approach path, and that the runway lights were not necessarily of exceptional visibility (McK-18.8). One could thus doubt whether the runway lights could have been seen at all through rain from moderate-intensity showers. Also, the VASI and the glideslope have the same glide path angle, so that assuming that he maintained visual contact with the VASI, the pilot would have been seeing red over red and would have known that he was too low, and that this is an exceptionally dangerous thing to do, since acceptable obstacle clearances are only guaranteed if one adheres rigorously to the approach procedures.

Unusual weather conditions

The third hypothesis is that the aircraft encountered exceptionally powerful updrafts, downdrafts, or other windshear, that significantly destabilised the approach. Evidence against this hypothesis would be the relative calm and orderly nature of cockpit activity as exemplified by the cockpit voice recording. If an outside disturbance such as windshear or severe wind conditions of any sort were to occur, one would expect at the least an annunciation of that condition from a crew member on the CVR, if not a statement saying what corrective action were being pursued. No such expression has been acknowledged to have been recognised so far on the CVR.

Conclusions

All the potential hypotheses as to what went on with the flight thus have their difficulties, their tensions with the known facts so far. But it is certain that there is a truth of the matter. Whether enough evidence can be or will be accumulated to determine that truth remains to be seen. This accident may turn out to be more difficult to explain than has been thought.

Peter Ladkin

Acknowledgements

Some perceptive comments by Herb Hoffmann on the first version of this article led to some changes in this version.

References

(AIM): U.S. Department of Transportation, Federal Aviation Adminstration, Aeronautical Information Manual (AIM), Author, annual. Back

(Airnav): Paulo Santos, The AirNav Database, http://www.airnav.com. Back

(Anon): An anonymous but very reliable non-governmental source with direct official contacts, personal communication, Friday 8 August, 1997. Back

(AW-Top10): Aviation Week and Space Technology, Editorial: The TOp 10 Air Safety Priorities Author, Jun 2, 1997, p74. Back

(BBC): BBC World Television Service, Hourly news broadcasts, Observation by PBL from the on-location film from GUAM, Thursday 7 August, 1997. Back

(CNN-05): CNN, Korean Air jetliner crashes in Guam, at http://cnn.com/WORLD/9708/05/guam/. Author, August 5, 1997. Back

(CNN-05a): CNN, Rescuers search smoldering jet wreckage in Guam for survivors, at http://cnn.com/WORLD/9708/05/guam.late/. Author, August 5, 1997. Back

(CNN-05b): CNN, Rescuers search smoldering jet debris in Guam, at http://cnn.com/WORLD/9708/05/guam.later/. Author, August 6, 1997. Back

(CNN-06): CNN, Guam rescuers: 27 survivors, no more expected, at http://cnn.com/WORLD/9708/06/guam/. Author, August 6, 1997. Back

(CNN-06a): CNN, Focus shifts to cause of Korean Air crash, at http://cnn.com/WORLD/9708/06/korean.crash/. Author, August 6, 1997. Back

(CNN-07): CNN, Alarm sounded before crash of Korean jet, at http://cnn.com/WORLD/9708/06/korean.recorders/. Author, August 7, 1997. Back

(CNN-08): CNN, Pilot error focus in Guam crash, at http://cnn.com/WORLD/9708/08/korean.air.crash/. Author, August 8, 1997. Back

(CNN-09): CNN, Photos used to identify Guam crash victims, at http://cnn.com/WORLD/9708/09/guam.crash/. Author, August 9, 1997. Back

(CNN-10): CNN, Software error plagued Guam airport radar system, at http://cnn.com/WORLD/9708/10/guam.crash/. Author, August 10, 1997. Back

(CNN-11): CNN, Investigators examine software error in Guam crash, at http://cnn.com/WORLD/9708/11/guam.crash/. Author, August 11, 1997. Back

(Coy97): Shawn Coyle, Aircraft On-Board Navigation Data Integrity - A Serious Problem, Transport Canada report, available at http://bluecoat.eurocontrol.fr/reports/Coyle_97_Nav_Database.pdf. Transport Canada, 1997. Back

(ET-808): the Electronic Telegraph, Issue 808, Monday 11 August, 1997. Back

(FAA): Anonymous FAA contact, personal communication, August 18, 1997. Back

(FAR): U.S. Department of Transportation, Federal Aviation Administration, Federal Aviation Regulations, Author, continuous. Available at http://www.faa.gov/avr/AFS/FARS/far_idx.htm Back

(FI-27.8): Flight International, Guam probe reveals US MSAW problems, Author, 27 August - 2 September, 1997, p22. Back

(IHT-06): International Herald Tribune, South Korean jet Crashes on Guam, Author, August 6, 1997, p1. Back

(IHT-11.8): International Herald Tribune, Software Flaw Crippled Guam Radar System, Inquiry Shows, Author, August 11, 1997, p2. Back

(Jepp-Guam): Jeppesen-Sanderson Company, Agana, Guam, ILS Rwy 6L approach, at http://www.jeppesen.com/guam.html, Author, August 1997. Back

(Jor-08): Mary Jordan, Cultures Clash At Crash Site As Koreans Demand Dead, International Herald Tribune, August 8, 1997, p1. Back

(Jor-09): Mary Jordan, U.S. Bolsters Korean Air Inquiry, International Herald Tribune, August 9-10, 1997, p2. Back

(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

(Kno): Brian Knowlton, 28 Escape the Crash of Airliner on Guam, International Herald Tribune, Thursday August 7, 1997, p1. Back

(Lad97-approaches): Peter B. Ladkin Flying an ILS or Localiser Approach - An Example, at http://www.rvs.uni-bielefeld.de/abstracts.html#approaches Back

(Lad97-CFIT): Peter B. Ladkin Controlled Flight Into Terrain: What Is Being Done?, at http://www.rvs.uni-bielefeld.de/abstracts.html#CFIT Back

(Lea94): David Learmount, Task force plans to halve CFIT incidents..., Flight International, 9-15 November, 1994, p11. Back

(Lew): Paul Lewis, with David Learmount and Ramon Lopez, Korean Air investigators focus on possible CFIT, Flight International, 13-19 August, 1997, p4. Back

(McK-11.8): James T. McKenna, Guam Crash Underscores Navaid, Training Concerns, Aviation Week and Space Technology, August 11, 1997, pp35-36. Back

(McK-18.8): James T. McKenna, Guam Probe Targets Weather, Altitude Alert, Aviation Week and Space Technology, August 18, 1997, pp20-22. Back

(NTSB): NTSB Public Affairs Office, Telephone Query, 15 August 1997. Back

(Phi-09): Don Phillips, No Mechanical Defects Seen, in (Jor-09), International Herald Tribune, August 9-10, 1997, p2. Back

(Phi-AW): Edward H. Phillips, Safety of Nonprecision Approaches Examined, Aviation Week and Space Technology, August 18, 1997, pp23-29. Back

(Rea): John Rea, Pacific Representative of the U.S. Department of Transportation Federal Aviation Administration in Honolulu, HI, Public Comment, reported in the International Herald Tribune, p1, South Korean Jet Crashes on Guam, Wednesday August 6, 1997. Also quoted in (CNN-05). Back

(TERPS): U.S. Department of Transportation, Federal Aviation Administration, United States Standard for Terminal Instrument Procedures (TERPS), 8260.3B, July 1976, U.S. Government Printing Office. Ordering information and on-line order forms are available from http://www.terps.com/gpo, a subpage on Wally Roberts's site Wally's TERPS Page, http://www.terps.com. Back

(Wie77): E. L. Wiener, Controlled Flight Into Terrain: System-Induced Accidents, Human Factors 19, 1977. Back