Korean Air Flight 801 - Aircraft Accident Report (NTSB)/Analysis

by National Transportation Safety Board

2. Analysis

2.1 General

The three Korean Air flight 801 flight crewmembers were properly certificated and qualified in accordance with applicable Korean Civil Aviation Bureau (KCAB) and U.S. Federal regulations, International Civil Aviation Organization (ICAO) standards, and Korean Air company requirements. No evidence indicated that any medical factors affected the flight crew's performance.

The airplane was properly certificated, equipped, and maintained in accordance with applicable KCAB and ICAO standards and Korean Air company procedures. The airplane was authorized to operate in U.S. airspace under the provisions of 14 Code of Federal Regulations (CFR) Part 129. The weight and balance of the airplane were within the prescribed limits for landing. No evidence indicated that the airplane experienced preimpact failures of its structures, flight control systems, or engines.

ATC personnel involved with the flight were properly certificated and qualified as full-performance level controllers. ATC radar and communications equipment were found to be functioning properly, although the FAA-maintained minimum safe altitude warning (MSAW) system had been intentionally inhibited.

This analysis examines the accident scenario, including weather factors, flight crew performance and decision-making, and other relevant factors during the approach, as well as flight crew fatigue issues. The analysis also examines the performance of ATC personnel, the effects of the MSAW system's intentional inhibition, and the timeliness and effectiveness of the emergency response to the accident site. In addition, the analysis examines Korean Air's flight crew simulator training, KCAB oversight of Korean Air's flight training programs, FAA oversight of Korean Air's operations under 14 CFR Part 129, and international efforts to reduce the number of controlled flight into terrain (CFIT) accidents.

2.2 Weather Factors on the Approach

A review of weather data indicated that variable clouds and scattered rain showers associated with a weak eastward-moving low-pressure trough were affecting the Guam area about the time of the accident and that the showers increased in intensity as they moved over the higher terrain of the island. However, Safety Board interviews with flight crewmembers who flew into Guam before and after the flight 801 accident indicated that the lights of the island were occasionally visible from as far away as 150 nm. In addition, CVR data indicated that the accident flight crew made visual contact with the island about 16 minutes before the accident (about 0126:25) when the flight engineer stated "it's Guam, Guam."

On the basis of weather data and witness statements, flight 801 was likely to have initially encountered variable scattered to broken cloud layers below 5,000 feet msl during the final approach to Guam. Ground lights were likely occasionally visible along the coastline, and it is probable that only scattered clouds existed below the airplane in the vicinity of the FLAKE intersection, located 7 DME from the NIMITZ VOR (UNZ).

Doppler radar data indicated that a heavy to very heavy rain shower was centered over higher terrain about 4 nm southwest of the airport (along the approach corridor) about the time of the accident. Weather data indicated that, although the Apra Harbor area (about 5 DME on the approach course) would likely have been visible to the flight crew as the airplane descended through 2,000 feet msl, the airplane would have entered clouds and light precipitation shortly after passing Apra Harbor. Radar data indicated that the flight likely experienced rain of continuously increasing intensity as the airplane proceeded inbound toward the airport and that the flight encountered very heavy precipitation for a short time near the outer marker (GUQQY). About 0141:48, when the airplane was near the outer marker, the CVR recorded the captain stating "wiper on."

Although a hunter on Nimitz Hill stated that it was not raining at the time he observed the flight overhead, Doppler radar data indicated light to moderate rain shower activity between Nimitz Hill and the airport. Therefore, on the basis of weather radar and observation data, the Safety Board concludes that, after the flight crew made an initial sighting of Guam, Korean Air flight 801 encountered instrument meteorological conditions (IMC) as the flight continued on its approach to Guam International Airport. Further, the Safety Board concludes that, although flight 801 likely exited a heavy rain shower shortly before the accident, the flight crew was still not able to see the airport because of the presence of another rain shower located between Nimitz Hill and the airport.

2.3 Accident Sequence

2.3.1 Description of the Approach and Required Flight Crew Procedures

The instrument landing system (ILS) glideslope (GS) inoperative, or localizer (LOC)-only, approach to runway 6L at Guam International Airport required the flight crew to maintain at least 2,000 feet from the FLAKE intersection (7 DME from the UNZ VOR) to the GUQQY (outer marker) final approach fix (FAF), which was located 1.6 DME from the UNZ VOR. After passing GUQQY, the crew was required to maintain at least 1,440 feet msl until passing the UNZ VOR. After passing the UNZ VOR, the next step-down fix was to 560 feet (the minimum descent altitude [MDA]), and the flight crew was required to maintain at least this altitude while counting up to 2.8 DME (the missed approach point [MAP]) from the UNZ VOR.

To properly fly this approach, the flying pilot's navigation receiver would need to be tuned to the localizer frequency, and the nonflying pilot's navigation receiver would need to be tuned to the UNZ VOR to provide the pilots with DME information for FLAKE, GUQQY, and the MAP and allow them to identify station passage over the UNZ VOR. Station passage over UNZ would be marked by the following instrument indications: the TO/FROM indicator in the first officer's horizontal situation indicator (HSI) would have changed from TO to FROM, and the No. 2 (double) pointer in the captain's and first officer's radio magnetic indicator would have swung from pointing forward to aft. Although station passage is not defined by DME, station passage over the UNZ VOR would have been indicated by a DME countdown to near zero and then a count up as the airplane continued away from UNZ to runway 6L.

To use the autopilot and flight director (FD) during a nonprecision approach, the flying pilot would have set the Nav Mode Switch to VOR/LOC on the glareshield instrument panel once the airplane was established on the intercept heading to the localizer outside the FAF. The FD and autopilot flight mode annunciator (FMA) on the pilots' instrument panels would then indicate NAV armed and GS blank.^[1] After localizer capture, the pilots' and the autopilot's FMAs would indicate NAV capture,^[2] and the GS would remain blank. The autopilot would turn the airplane as needed to track the localizer. (Although the nonflying pilot's navigation receiver would be tuned to the VOR, a different frequency, that pilot's FMA would also indicate NAV capture because the UNZ 243° radial approximately overlays the localizer.) The FD command bars on the captain's and the first officer's instruments would provide left and right roll commands to maintain a capture on the respective courses.

When the airplane was descending, the flying pilot would normally have reset the altitude selector to the next (lower) altitude target and moved the VERT SPEED (vertical speed) wheel to position the airplane in about a 1,000-feet-per-minute rate of descent to reach the next crossing altitude. The FD pitch command bars would move up and down to provide guidance to maintain the desired rate of descent. (In a precision approach, the command bars would provide guidance to maintain the glideslope; in nonprecision approaches, they provide guidance to maintain the desired vertical speed.) Once the airplane had reached the altitude set in the altitude selector, the FDs and autopilot would capture and maintain that altitude. At no time during the localizer approach would the flight crew have armed the autopilot or FDs to capture a glideslope signal or have referenced or attempted to track any glideslope needle indications.

2.3.2 Flight 801's Premature Descent

The Combined Center/Radar Approach Control (CERAP) controller vectored flight 801 to join the runway 6L localizer course between the FLAKE intersection and the GUQQY outer marker/1.6 DME fix. Although the flight was restricted to no lower than 2,000 feet msl in that portion of the localizer-only approach procedure (until crossing the GUQQY outer marker), flight 801 descended below 2,000 feet about 1.9 nm before reaching the outer marker.

Further, although the approach procedure specified that at least 1,440 feet msl should be maintained after passing the GUQQY outer marker and until passing the UNZ VOR, flight 801 descended below 1,440 feet about 2.1 DME before reaching the UNZ VOR. (None of the indications of UNZ station passage would have been presented on the flight crew's instrument panels at any time before impact.)

CVR information indicated that the captain was flying the airplane on autopilot during the approach. As flight 801 descended on the approach, the captain twice commanded the entry of lower altitudes into the airplane's altitude selector before the airplane had reached the associated step-down fix. After the captain heard the first officer call out "approaching fourteen hundred [feet]" about 0140:33, as the airplane was passing 5 DME at 2,400 feet msl,^[3] the captain directed the first officer to reset the altitude selector to 1,440 feet, replacing the step-down altitude of 2,000 feet before the autopilot had captured that altitude or reached the GUQQY outer marker/1.6 DME fix. Further, about 0141:33, with the flight neither having leveled off at 1,440 feet msl nor reached the UNZ VOR step-down fix, the captain instructed the first officer to set 560 feet, the MDA, in the altitude selector.

The altitude selector provides the basis for the altitude alert's aural annunciations and the autopilot's altitude capture functions. The captain's premature orders to reset the altitude selector indicated that he had lost awareness of the airplane's position along the final approach course. Therefore, as a result of the captain's commanded input to the altitude selector, the autopilot continued to descend the airplane prematurely through the 2,000- and 1,440-foot intermediate altitude constraints of the approach procedure. The CVR comments indicated no awareness by the captain that the airplane was descending prematurely below the required intermediate altitudes.

2.4 Flight Crew Performance

2.4.1 The Captain's Performance of the Approach

2.4.1.1 Approach Briefing

Korean Air cockpit procedures call for an approach (landing) briefing^[4] before descent. Also, company training instructs the flying pilot to conduct an approach briefing before descent. According to the Korean Air 747 landing briefing checklist card and testimony by Korean Air officials during the Safety Board's public hearing, this briefing should include a discussion of weather conditions, a review of the instrument approach procedure, details of the approach's execution (including the minimum safe altitude, approach frequency and approach course, the runway touchdown zone elevation, and the missed approach procedure), crew actions and callouts, and any abnormal configurations or conditions.

CVR information indicated that the captain briefed a visual approach in his approach briefing, which he referred to as a "short briefing." However, the captain also briefed some elements of the localizer-only instrument approach, indicating that he intended to follow that approach as a supplement or backup to the visual approach. Specifically, the captain's briefing included a reminder that the glideslope was inoperative, some details of the radio setup, the localizer-only MDA, the missed approach procedure, and the visibility at Guam (stated by the captain to be 6 miles). However, the captain did not brief other information about the localizer-only approach, including the definitions of the FAF and step-down fixes and their associated crossing altitude restrictions or the title, issue, and effective dates of the approach charts to be used. The Safety Board notes that the landing briefing checklist did not specifically require the captain to brief the fix definitions, crossing altitudes, or approach chart title and dates,^[5] although it would have been good practice to do so.

Further, according to public hearing testimony by a Korean Air instructor pilot, company pilots were trained to conduct a more detailed briefing than the one specified in the landing briefing checklist for a nonprecision approach, such as the localizer approach to runway 6L at Guam. According to the instructor pilot, this more detailed briefing included a discussion of the "instrument approach in detail" and a discussion of the "stepdown altitudes and how they were determined." The Safety Board notes that this information is essential for a nonprecision approach briefing.

The Safety Board also notes that the captain did not brief the first officer and flight engineer on how he would fly the descent (including his planned autopilot/FD modes and his plan to fly either a constant angle of descent or a series of descents and level-off altitudes associated with the step-down fixes), and he did not discuss go-around decision criteria. Further, although not specifically required, it would have been prudent for the captain to note the need for special caution in the UNZ VOR area (which he had described as a "black hole" in his approach briefing to another first officer about 1 month earlier).

The Safety Board further notes that, in this case, a thorough briefing was especially important because the accident captain and first officer were flying together for the first time, which is a situation that has been linked to flight crew-involved accidents.^[6]The Safety Board According to recent human factors research, a good briefing is important to develop a "shared mental model" to ensure "that all crew members are solving the same problem and have the same understanding of priorities, urgency, cue significance, what to watch out for, who does what, and when to perform certain activities."^[7] The Safety Board concludes that, by not fully briefing the instrument approach, the captain missed an opportunity to prepare himself, the first officer, and the flight engineer for the relatively complex localizer-only approach and failed to provide the first officer and flight engineer with adequate guidance about monitoring the approach; therefore, the captain's approach briefing was inadequate.

2.4.1.2 Expectation of a Visual Approach and Role of the Guam Airport Familiarization Video

The Safety Board notes that, when the captain flew to Guam about 1 month before the accident, he executed a routine ILS approach to runway 6L in good visibility, with a scattered cumulous buildup. Further, the most current ATIS information available to the accident flight crew indicated that visual conditions (scattered cloud decks and 7-mile visibility) existed at the airport. Korean Air's Guam airport familiarization video, which the captain and first officer had viewed in July 1997, noted that weather conditions in Guam allowed visual approaches most of the year and that, even though IMC is likely during the rainy season from June to November, "you [the pilot] will be guided from over Apra Harbor to the localizer. You will then perform a visual approach...." Thus, the captain may have assumed that conditions for the flight 801 approach would be similar to those he experienced about 1 month earlier. The captain's anticipation of a visual approach probably became a strong expectation after the flight crew's early visual sighting of Guam. Although the captain would likely have recognized the possibility of flight through clouds as the airplane descended from its cruise altitude, he may have assumed that the visual approach slope indicator (VASI) system would be in sight after the flight was vectored onto final approach by the CERAP controller. The VASI system would have provided visual guidance for a constant angle of descent that safely cleared obstacles.

As previously discussed, the captain's landing briefing included references to his expectation of visual conditions at the airport as well as an abbreviated and inadequate briefing for the localizer-only approach. The Safety Board concludes that the captain's expectation of a visual approach was a factor in his incomplete briefing of the localizer approach. The Board is aware that it is a common practice among air carrier pilots to abbreviate the briefing for a backup instrument approach when a visual approach is expected. Although there may be little benefit to fully briefing a backup instrument approach in daylight conditions when no appreciable possibility of encountering IMC exists, the Safety Board concludes that, for flights conducted at night or when there is any possibility that IMC may be encountered, the failure to fully brief an available backup instrument approach compromises safety. Therefore, the Safety Board believes that the FAA should require principal operations inspectors (POI) assigned to U.S. air carriers to ensure that air carrier pilots conduct a full briefing for the instrument approach (if available) intended to back up a visual approach conducted at night or when IMC may be encountered. The Safety Board notes that, although Guam was not a designated special airport requiring special training or familiarization by flight crews, Korean Air encouraged its flight crews to view the airport familiarization video. However, the Guam familiarization video gave only a generalized description of the topography of the island of Guam. Although the video mentioned some of the obstacles near the approach course, it did not specifically state that the UNZ VOR was located on a hill, the DME was not colocated with the localizer, or the final approach segment was over hilly or mountainous terrain.

Even though the airport familiarization video accurately identified some landmarks and advised pilots not to fly over a residential area and a Naval hospital (for noise abatement), the Safety Board also notes with concern that the video contained no discussion of factors that made operations into Guam challenging, such as the high terrain along the approach course or in the vicinity of the airport. Further, the presentation did not describe the complexity of the Guam nonprecision approaches, including the multiple step-down fixes, the use of two separate navigation facilities (the localizer and the VOR), and the countdown/count up DME procedure.

The Safety Board concludes that the Korean Air airport familiarization video for Guam, by emphasizing the visual aspects of the approach, fostered the expectation by company flight crews of a visual approach and, by not emphasizing the terrain hazards and offset DME factors, did not adequately prepare flight crews for the range of potential challenges associated with operations into Guam. Thus, the Safety Board believes that the KCAB should require Korean Air to revise its video presentation for Guam to emphasize that instrument approaches should also be expected and describe the complexity of such approaches and the significant terrain along the approach courses and in the vicinity of the airport.

The Safety Board addressed the issue of the classification of special airports and approaches to certain airports in connection with its investigation of the October 19, 1996, accident involving Delta Air Lines flight 554, an MD-88, at LaGuardia Airport in New York. On August 25, 1997, the Board issued Safety Recommendations A-97-92 through -94, asking the FAA to develop and publish "specific criteria and conditions" for the classification of special airports (including special runways and/or special approaches) and use these criteria to evaluate all airports and "update special airport publications." On November 13, 1997, the FAA responded that it was revising Advisory Circular (AC) 121.445, "Pilot-in-Command Qualifications for Special Area/Routes and Airports," and that the revision would address the issues discussed in the safety recommendations. On August 17, 1998, the Board classified Safety Recommendations A-97-92 through -94 "Open--Acceptable Response" pending completion of the AC. The Board recognizes that the FAA's eventual evaluation of Guam against the newly developed criteria might result in its classification as a special airport. The Safety Board further recognizes that, because the captain flew into Guam and viewed the Guam airport familiarization video during July 1997, he would have been authorized to conduct the accident flight even if Guam had been classified as a special airport.

Nonetheless, the Safety Board concludes that the challenges associated with operations to Guam International Airport support its immediate consideration as a special airport requiring special pilot qualifications. Therefore, the Safety Board believes that the FAA should consider designating Guam International Airport as a special airport requiring special pilot qualifications.

2.4.1.3 Possible Explanations for the Approach Conducted 2.4.1.3.1 Confusion About Status of Glideslope

Despite several indications that the flight crew was aware that the glideslope was inoperative, in the last 2½ minutes of the flight (beginning shortly after the airplane was established on the approach), the CVR recorded a series of conflicting flight crew comments concerning the operational status of the glideslope. About 0139:55, the flight engineer asked, "is the glideslope working?" The captain responded, "yes, yes it's working." About 0139:58, an unidentified voice in the cockpit stated, "check the glideslope if working?" One second later, an unidentified voice in the cockpit stated, "why is it working?" About 0140:00, the first officer responded, "not useable." About 0140:22, an unidentified voice in the cockpit stated, "glideslope is incorrect," followed by the captain's statement, "since today's glideslope condition is not good, we need to maintain one thousand four hundred forty [feet]." However, about 0141:46, after the airplane crossed the GUQQY outer marker (1.6 DME from the VOR), the captain again stated, "isn't glideslope working?"

The Safety Board considered whether the flight crew might have misinterpreted some cockpit instrumentation indications as a valid glideslope capture signal. During the localizer approach into Guam, both pilots' HSIs would have appeared centered; the captain's would have captured the localizer, and the first officer's would have captured the VOR radial. With VOR/LOC selected, the localizer captured, and the pitch commands set to VERT SPEED (the most likely setting), the captain's FD command bars would have shown some vertical and horizontal movement, similar to an FD that was responding to a captured localizer and glideslope. However, the raw data glideslope needles on the attitude director indicator (ADI) and HSI would not have been affected by the VERT SPEED setting; therefore, the captain's ADI and HSI glideslope needles should have been covered by "off" flags.^[8] Further, there would have been no glideslope capture annunciator on the GS bar of the FMA on top of the captain's and first officer's instrument panels.

The Safety Board also considered whether the flight crewmembers might have observed intermittent movement of the glideslope needles during the approach, thereby creating or adding to their confusion about the glideslope. An FAA navigation expert testified at the Safety Board's public hearing that spurious radio signals could cause a sporadic or intermittent glideslope indicator deviation in the absence of a valid glideslope signal. However, he stated that the glideslope off flag would still appear on the HSI and ADI glideslope needles and that, when the off flag appears, any movement of the glideslope needle should be considered unreliable. Postaccident testing by Korean Air and the KCAB confirmed that an airplane's glideslope receiver could be affected by spurious radio signals when no valid glideslope signal was being transmitted. The tests demonstrated that spurious signals could cause movement of the glideslope needle and that, when the receiver was subjected to a steady signal, retraction of the off flag was also possible. However, the Safety Board notes that these tests were conducted with an airplane on the ground and that the airplane's navigational receiver was subjected to extreme signal modulations transmitted very near the airplane's antenna. These conditions are not likely to be encountered by an airplane on an actual instrument approach.

The Safety Board also notes that the flight crew of a Boeing 727 reported glideslope anomalies on August 5, 1997, while executing the localizer-only approach to runway 6L at Guam.^[9] (The purpose of the flight was to test a newly installed GPS.) However, the captain of the 727 stated that he thought the glideslope anomaly might have been caused by the GPS wiring installation. Further, the first officer stated that he and the captain "never thought twice" about the glideslope indications because they knew that the glideslope was inoperative. The Board's investigation into the 727's maintenance history indicated that, in the weeks after the test flight, several cockpit navigational displays, including the first officer's HSI and ADI, were repeatedly removed and replaced by maintenance personnel because of anomaly reports written up by flight crews. The maintenance documents indicated that the cockpit display problems were the result of integrating the new GPS with the existing cockpit displays.

Although it is possible that spurious radio signals caused some erratic movement of the glideslope needles on the accident captain's HSI and ADI, it is unlikely that the accident airplane's navigation receivers could have been subjected to a steady spurious signal of a duration that would have resulted in a continuous glideslope needle activation and flag retraction over a period of minutes and several miles of aircraft motion. Thus, the presence of the off flags over the glideslope needles at some times and the absence of FMA glideslope capture indicators on the captain's and first officer's instrument panels should have been sufficient to convince the flight 801 flight crew to disregard the glideslope indications. Even if the flight crewmembers did see a continuous glideslope needle activation and flag retraction, it would not have been prudent or reasonable for them to rely on a glideslope signal of any sort when the glideslope had been reported to be unusable. (Korean Air officials stated that flight crews were trained not to use navigational aids, including glideslopes, that were reported to be unreliable or unusable). Therefore, the Safety Board concludes that, although the captain apparently became confused about the glideslope's status, the flight crew had sufficient information to be aware that the glideslope was unusable for vertical guidance and should have ignored any glideslope indications while executing the nonprecision localizer-only approach.

The Safety Board notes that, when a glideslope signal is not generated by the transmitter (resulting in an open frequency channel), an airborne glideslope receiver will continue to seek a glideslope signal, although navigation receiver filters are designed to block most spurious radio signals. The postaccident testing conducted by Korean Air and the KCAB involved the glideslope receiver; however, the Safety Board concludes that navigation receivers, including glideslope receivers, may be susceptible to spurious radio signals. Therefore, the Safety Board believes that the FAA should disseminate information to pilots, through the Aeronautical Information Manual, about the possibility of momentary erroneous indications on cockpit displays when the primary signal generator for a ground-based navigational transmitter (for example, a glideslope, VOR, or nondirectional beacon [NDB] transmitter) is inoperative. Further, this information should reiterate to pilots that they should disregard any navigation indication, regardless of its apparent validity, if the particular transmitter was identified as unusable or inoperative.

2.4.1.3.2 Confusion About Location of DME

About 0140:37, when the airplane was at 2,400 feet msl and descending at 1,000 feet per minute, the captain stated, "since today's glideslope condition is not good, we need to maintain one thousand fourteen hundred forty [feet]. please set it." This statement suggests that the captain was attempting to comply with the restrictions of the localizeronly approach and believed that he had passed the GUQQY step-down fix. However, the CVR recorded no discussion between the captain and the first officer about DME values or their position in relation to the next step-down fix, the VOR, or the airport.

The Safety Board considered whether the flight crew might have confused the configuration of the runway 6L localizer approach with one in which the DME is located on the airport. A review of the flight crew's training records showed that the nonprecision approaches incorporating DME provided to the flight crew during training and check rides had the DME located on the airport. A countdown/count up DME procedure, which is rarely encountered on a localizer procedure, was not included in any of the Korean Air simulator training scenarios. If the flight crewmembers had the misconception that the DME information referred to the distance from the airport, they might have believed that the airplane was much closer to the airport than it actually was (the DME was located 3.3 nm southwest of the airport) and that the airplane was well above the minimum altitudes for the intermediate step-down fixes and thus ready to descend directly to the MDA. If the captain had this misconception, it could explain why he flew the airplane and commanded altitude selections as though he believed he was at or above the altitude constraint for each navigational fix along the approach. If the other flight crewmembers shared this misconception, it could explain why they failed to challenge the captain's premature descents below 2,000 and 1,440 feet.

However, this scenario suggests strongly that the captain was not noting the definitions of the navigational fixes on the approach chart, which were clearly defined as DME values.^[10] Thus, the Safety Board concludes that the captain may have mistakenly believed that the airplane was closer to the airport than its actual position; however, if the captain conducted the flight's descent on this basis, he did so in disregard of the DME fix definitions shown on the approach chart.

2.4.1.4 Summary of Captain's Performance on the Approach

As the approach progressed without encountering the visual conditions the captain had anticipated, the captain likely experienced increased stress because of his inadequate preparation for the nonprecision approach, which made the approach increasingly challenging.^[11] CVR and FDR data indicated that, shortly after the captain appeared to become preoccupied with the status of the glideslope, he allowed the airplane to descend prematurely below the required intermediate altitudes of the approach. Thus, the captain may have failed to track the airplane's position on the approach because he believed that he would regain visual conditions, the airplane was receiving a valid glideslope signal, and/or the airplane was closer to the airport than its actual position. Regardless of the reason for failing to track the airplane's position, the captain conducted the approach without properly cross-referencing the positional fixes defined by the VOR and DME with the airplane's altitude. Therefore, the Safety Board concludes that, as a result of his confusion and preoccupation with the status of the glideslope, failure to properly cross-check the airplane's position and altitude with the information on the approach chart, and continuing expectation of a visual approach, the captain lost awareness of flight 801's position on the ILS localizer-only approach to runway 6L at Guam International Airport and improperly descended below the intermediate approach altitudes of 2,000 and 1,440 feet, which was causal to the accident.

2.4.2 Flight Crew Monitoring of the Approach

CVR evidence indicated that the flight crew seemed confused about, and did not react to, a series of audible ground proximity warning system (GPWS) alerts during the final approach. The first audible GPWS callout occurred about 0141:42, with the "one thousand [feet]" altitude call. A second GPWS callout of "five hundred [feet]" occurred about 0142:00 (when the airplane was descending through about 1,200 feet msl), to which the flight engineer responded in astonishment, "eh?" However, FDR data indicated that no change in the airplane's descent profile followed, and the CVR indicated that the flight engineer continued to complete the landing checklist. Similarly, no flight crew discussion followed the GPWS callout of "minimums" about 0142:14, and the first officer dismissed a GPWS "sink rate" alert 3 seconds later by stating "sink rate okay." About 0142:19, the flight engineer called "two hundred [feet]," followed immediately by the first officer saying "let's make a missed approach." The flight engineer immediately responded "not in sight," followed by the first officer repeating "not in sight missed approach." According to the CVR, a rapid succession of GPWS altitude callouts down to 20 feet followed, as the flight crew attempted to execute the missed approach.

The GPWS minimums callout occurred about 12 seconds before impact, when the airplane was descending through about 840 feet msl. The first officer's first statement suggesting the execution of a missed approach occurred about 6 seconds before impact. The captain initiated a missed approach and thrust began increasing about 4 seconds before impact. However, no significant nose-up control column inputs were made until just before initial impact. Analysis of FDR data indicated that, if a missed approach had been initiated 12 seconds before impact (at the GPWS minimums callout), it is likely that the airplane would have successfully cleared terrain by about 450 feet. Analysis of the FDR data also indicated that, if an aggressive missed approach had been initiated 6 seconds before impact (when the first officer made the first missed approach challenge), it is possible that the airplane might have cleared the terrain.

The Safety Board notes that the flight crew would have been gauging the airplane's height above the MDA by referring to the airplane's barometric altimeter (which displays altitude above sea level) and not the radio altimeter (which senses altitude above ground level and upon which the GPWS minimums callout was based) and that the MDA of 560 feet msl was never reached. Nevertheless, the GPWS callouts were a salient cue that should have caused the flight crew to question the airplane's position and the captain to act conservatively and choose to execute a missed approach. The Safety Board concludes that the first officer and flight engineer noted the GPWS callouts and the first officer properly called for a missed approach, but the captain's failure to react properly to the GPWS minimums callout and the direct challenge from the first officer precluded action that might have prevented the accident.

Although the first officer properly called for a missed approach 6 seconds before impact, he failed to challenge the errors made by the captain (as required by Korean Air procedures)^[12]earlier in the approach, when the captain would have had more time to respond. Significantly, the first officer did not challenge the captain's premature descents below 2,000 and 1,440 feet. The Safety Board was unable to identify whether the absence of challenges earlier in the approach stemmed from the first officer's and the flight engineer's inadequate preparation during the approach briefing to actively monitor the captain's performance on the localizer approach, their failure to identify the errors made by the captain (including the possibility that they shared the same misconceptions as the captain about the glideslope status/FD mode or the airplane's proximity to the airport), and/or their unwillingness to confront the captain about errors that they did perceive. The Safety Board notes that the captain's failure to brief the localizer approach to back up the expected visual approach could have adversely affected the flight crew's preparation for monitoring the approach. If the captain had briefed the details of the approach, including the various navigational fix definitions and associated altitude constraints, he would have enhanced the flight crew's ability to monitor the approach and challenge any errors he made.

Even if the first officer was attempting to monitor the approach, his ability to identify errors made by the captain would have been impaired by the requirement that he tune his navigation receiver to the UNZ VOR, thus forcing him to look across the cockpit to the captain's instruments to monitor the glideslope/FD status, any indications of glideslope capture on the captain's ADI and HSI, and the airplane's lateral position on the localizer. However, the first officer would have had information on his own HSI and radio magnetic indicator about the airplane's position relative to the VOR (the step-down fix for the descent to 560 feet) and the DME readings that defined the remaining fixes of the approach.

The first officer's ability to monitor the captain was also possibly hindered by the likelihood that he was using a different instrument chart than the captain for the localizer approach. The Safety Board found an out-of-date chart for this approach (dated January 19, 1996) in the cockpit. On the basis of the captain's comments on the CVR, it appears that the captain was using the correct chart (dated August 2, 1996), which included different definitions and names of DME fixes and different crossing altitudes than the out-of-date chart. Thus, if the first officer was using the out-of-date chart, he would have been hindered in monitoring the captain's compliance with the altitude constraints at the fixes.

Although the precise reason(s) for the lack of monitoring by the flight crew could not be determined, the Safety Board concludes that the first officer and flight engineer failed to properly monitor and/or challenge the captain's performance, which was causal to the accident.

Problems associated with subordinate officers challenging a captain are well known. For example, in its study of flight crew-involved major air carrier accidents in the United States,^[13] the Safety Board found that more than 80 percent of the accidents studied occurred when the captain was the flying pilot and the first officer was the nonflying pilot (responsible for monitoring). Only 20 percent of the accidents occurred when the first officer was flying and the captain was monitoring. This finding is consistent with testimony at the Safety Board's public hearing, indicating that CFIT accidents are more likely to occur (on a worldwide basis) when the captain is the flying pilot. (See section 2.8 for a discussion of CFIT accidents and prevention strategies, including monitored approaches.) The Board's study found that the failure of first officers to challenge errors (especially tactical decision errors) made by a flying captain was a frequent factor in accidents involving such errors. In addition, the study noted that, while monitoring and challenging a captain's tactical decision error, "a first officer may have difficulty both in deciding that the captain has made a faulty decision, and in choosing the correct time to question the decision." The study concluded that a first officer "may be concerned that a challenge to a decision may be perceived as a direct challenge to the captain's authority."

The Safety Board is concerned that the use of the nonflying pilot in a passive role, while the flying pilot is responsible for the approach procedure, programming the autopilot/FD controls, and monitoring the aircraft flightpath, places an inordinately high work load on the flying pilot and undertasks the nonflying pilot. The Board is also concerned that, when the nonflying pilot has a passive role in the approach, the flying pilot may erroneously consider the lack of input from the nonflying pilot as confirmation that approach procedures are being properly performed. The Board is aware that some international air carriers use the nonflying pilot in a more interactive role during the performance of a nonprecision approach, in which that pilot leads or prompts the flying pilot through the approach procedure by stating the next procedural change, including course, heading, altitude, time, visual contact, or MAP. The Board is also aware that some air carriers employ a "monitored approach" method, with the first officer as the flying pilot and the captain as the monitoring, nonflying pilot until just before landing.

The Safety Board notes that the monitored approach method provides for more effective monitoring by the nonflying pilot because captains are more likely to be comfortable offering corrections or challenges to first officers than the reverse situation. Thus, the Safety Board concludes that monitored approaches decrease the workload of the flying pilot and increase flight crew interaction, especially when experienced captains monitor and prompt first officers during the execution of approaches. However, the Board also notes that, when there are differences in aircraft handling skills between captains and first officers and the approach is not flown using the autopilot, a monitored approach with the captain as the nonflying pilot may not always be appropriate. Therefore, the Safety Board believes that the FAA should conduct or sponsor research to determine the most effective use of the monitored approach method and the maximum degree to which it can be safely used and then require air carriers to modify their procedures accordingly.

2.4.3 Flight Crew Fatigue Factors

Fatigue can be a factor in flight operations.^[14] The Safety Board examined several fatigue-related factors, including time of day, recent sleeping patterns, and the number of hours since awakening, to determine whether fatigue was a factor in the accident captain's performance. The Board was unable to obtain information on the recent sleeping patterns of the first officer and flight engineer.

The accident occurred after midnight (about 0042) in the flight crew's home time zone (which is 1 hour behind Guam local time). Research has found that this time of day is often associated with degraded alertness and performance and a higher probability of errors and accidents.^[15] The arrival time was also several hours after the captain's normal bedtime (2200 to 2300 Seoul local time) and a time at which his body would have been primed for sleep.

CVR evidence indicated that the captain was tired. At the beginning of the approach, the captain made unsolicited comments related to fatigue, stating "eh...really...sleepy." In its investigation of the 1993 American International Airways accident at Guantanamo Bay, Cuba, the Safety Board noted that individuals often tend to underestimate their own level of fatigue.^[16] Thus, the captain's comment could have reflected a significant performance degradation. Neither the first officer nor the flight engineer made similar comments.

According to his family, the captain slept his normal sleep routine in the days before the accident and had an opportunity to receive adequate rest. He also took a nap from 1100 to 1340 (Seoul local time) on August 5 and would therefore have been awake for 11 hours at the time of the accident. The Safety Board has found this time since awakening to be associated with greater errors.^[17]

For example, in its investigation of the 1993 accident in Pine Bluff, Arkansas, the Board determined that the captain and first officer had both been awake for 11 hours.^[18]

Further, the sleep history provided by the captain's family does not address the quality of sleep he received. For example, the time of the captain's reported nap corresponds to a typical physiological period of wakefulness when napping is difficult204 and of limited efficiency at reducing sleep debt.^[19] If the captain actually napped at this time, it suggests that he may have had unusual sleep needs that were not indicated by the number of hours in his reported sleep history.^[20]

Fatigue degrades all aspects of performance and alertness, and deficiencies associated with fatigue were displayed in many aspects of the captain's behavior. The captain's preoccupation and confusion with the status of the glideslope to the exclusion of other critical information, his incomplete briefing, and his failure to react to the GPWS alerts are typical of fatigue effects that were found to be present in the Guantanamo Bay and Pine Bluff accidents.

On the basis of the time of day, statements recorded on the CVR, and sleep and fatigue research, the Safety Board concludes that the captain was fatigued, which degraded his performance and contributed to his failure to properly execute the approach.

2.5 Pilot Training

The Safety Board examined Korean Air's Boeing 747 pilot training and proficiency checking program to determine what effect, if any, it may have had on the performance of the flight crew of flight 801. In training its pilots to fly the 747-200 and -300 series airplanes, Korean Air conducted 10 4-hour simulator sessions in which pilots were taught various maneuvers, emergencies, and scenarios, followed by a proficiency check in which pilot performance of certain maneuvers was assessed. The profile for each simulator training session outlined the specific airport, runway, weather, and airplane malfunction to be expected and whether the flight would result in a landing or missed approach. The training curriculum was not varied. Korean Air's Director of Academic Training testified at the Safety Board's public hearing that, at the time of the accident, the company's practice was to follow simulator scenarios exactly as outlined in the training curriculum and that instructors were not permitted to vary the scenarios. The director also indicated that the proficiency checks used the same approaches that had been practiced in the previous simulator training sessions.

Further, the only nonprecision approach practiced throughout the simulator sessions that used DME information was the VOR/DME approach to runway 32 at Seoul's Kimpo Airport. However, the DME at that airport is located on the field, unlike at Guam. No scenario was presented in which pilots were required to count down to and fly past the DME and then count up to the MAP, which was required for the Guam approach.^[21] Further, according to the airline's training syllabus, the VOR/DME approach to runway 32 at Kimpo was the only nonprecision approach that Korean Air flight crews were required to perform on their check ride.

The Safety Board notes that proper training in the execution of nonprecision approaches is essential to safe operations. The complexity of such approaches and the absence of precise vertical guidance create more demands on pilot skills and cognitive performance than precision approaches. An expert on CFIT accidents testified the following at the Board's public hearing:

Nonprecision approaches generally are much more complex than precision approaches. For many pilots, they are less familiar. They are more error-prone. They require [a] more comprehensive briefing. They need particularly careful and accurate monitoring, and it is possible for complex step-down procedures for steps to be missed or to be taken out of step. In other words, to get one step ahead of the airplane could be fatal. Such approaches also need much more carefully managed airplane crew and checklist management, and it is a characteristic of many CFIT accidents that they occur when the crew is preoccupied or distracted by other tasks.

The Safety Board notes that the Air Line Pilots Association (ALPA), in its submission regarding this accident, estimated that air transport pilots typically conduct one to three nonprecision approaches a year and practice these approaches "just as infrequently" in the simulator. In its investigation of the November 12, 1995, accident involving American Airlines flight 1572, an MD-83 that crashed in East Granby, Connecticut, while on final approach to Bradley International Airport in Windsor Locks, Connecticut,^[22] the Board found that even relatively minor errors in the monitoring of the execution of a nonprecision approach can lead to an accident.^[23]

The Safety Board is concerned that the repeated presentation of a single nonprecision approach scenario throughout simulator training (to the exclusion of all other kinds of nonprecision approaches) provides insufficient training in nonprecision approaches. Specifically, the repetition limits pilots' opportunity to understand and practice the flying techniques necessary to perform the different kinds of nonprecision approaches and limits their ability to successfully apply these techniques to novel situations or unusual approach configurations encountered in line operations, such as the localizer approach at Guam. Further, Korean Air's reliance on the same approach for both training and checking resulted in an inadequate evaluation of a flight crew's ability to execute the varied nonprecision approaches that might be encountered in line operations.^[24] Therefore, the Safety Board concludes that Korean Air's training in the execution of nonprecision approaches was ineffective, which contributed to the deficient performance of the flight crew.

In addition, on the basis of the history of similar accidents involving U.S. air carriers, the Safety Board concludes that U.S. air carrier pilots would benefit from additional training and practice in nonprecision approaches during line operations (in daytime visual conditions in which such a practice would not add a risk factor). Therefore, the Safety Board believes that the FAA should issue guidance to air carriers to ensure that pilots periodically perform nonprecision approaches during line operations in daytime visual conditions in which such practice would not add a risk factor.

2.6 Air Traffic Control Factors

2.6.1 Controller Performance

Safety Board investigators evaluated the performance of the CERAP and Agana tower controllers to determine whether their performance played a role in the circumstances of the accident. FAA Order 7110.65, "Air Traffic Control," prescribes the ATC procedures that controllers are required to follow. The investigation revealed three deviations from those procedures on the part of the CERAP controller.

The CERAP controller failed to provide the flight crew with a position advisory relative to a fix on the final approach course when he cleared flight 801 for the approach. If such a position advisory had been given, as required by paragraph 5-9-4, the pilots might have been prompted to cross-check their radar position with the cockpit DME and other navigational aid indications, thereby improving their situational awareness. In addition, the CERAP controller did not inform the flight crew or the tower controller that he had observed a rain shower (described by the CERAP controller as a "cell" during a postaccident interview with Safety Board investigators) on the final approach path, as required by paragraph 2-6-4. Although the pilots should have been aware of the weather situation because they were using on-board weather radar, their decision-making might have been aided if the CERAP controller had provided his weather observations.

The CERAP controller also failed to monitor the flight after the frequency change to the tower controller.^[25] As a result, the CERAP controller did not immediately recognize that the airplane was overdue. (Paragraph 10-3-1 states that a controller who has any reason to believe that an aircraft is overdue should immediately take appropriate action.) If the CERAP controller had been properly monitoring the flight on one or both of the radar displays he had available to him (the en route display and/or the terminal display), he might have observed flight 801 disappear on final approach. Also, the controller might have noticed the approach path warning (low-altitude MSAW alert) that was generated on the en route radar display,^[26] which began about 6 seconds before impact and continued until at least 23 seconds afterward. These actions would have resulted in an earlier notification of the accident to emergency rescue personnel and possibly an earlier emergency response. (See section 2.7 for a discussion of the emergency response.)

Further, if the CERAP controller had been monitoring the flight on the terminal radar display, which was located to his immediate right and would have been clearly visible to him,^[27] he might have seen the airplane descend prematurely toward high terrain and have been able to alert the flight crew and prevent the accident. This radar display would have shown the flight descending through 2,000 feet msl while almost 7 miles from the airport and outside of the outer marker. The radar display would have also shown the airplane crossing the outer marker almost 800 feet lower than the established crossing altitude of 2,000 feet.^[28]

Although the CERAP controller told Safety Board investigators that he did not continue to monitor the flight because he was engaged in other duties about the time of the accident, the ATC transcripts indicated no activity during that time. The transcripts indicated that the controller instructed the flight crew, about 0140:42, to contact the Agana tower. The controller then made a radio transmission to another aircraft about 0140:54. From about 0141:14 to 0141:30, the controller had a conversation with another controller at a different center, and about 0142:05, he acknowledged a transmission from another aircraft. However, the transcripts indicated no further activity until 0143:49, when the CERAP controller called the Agana tower with a flight plan. Thus, the ATC transcripts indicated no activity during the time period beginning 21 seconds before and continuing until 1 minute 23 seconds after the flight 801 crash (which occurred about 0142:26). Therefore, the CERAP controller should have been able to monitor the flight during this time. If the controller had done so, he would have had an opportunity to warn the flight crew of the flight's premature descent and possibly prevent the accident.

The Safety Board concludes that the CERAP controller's performance was substandard in that he failed to provide the flight crew with a position advisory when he cleared the flight for the approach, inform the flight crew or the Agana tower controller that he had observed a rain shower on the final approach path, and monitor the flight after the frequency change to the tower controller. It could not be determined whether the absence of the CERAP controller's procedural errors, singularly or in any combination, would have prevented the accident or reduced its severity. However, the Safety Board concludes that strict adherence to ATC procedures by the CERAP controller may have prevented the accident or reduced its severity. Therefore, the Safety Board believes that the FAA should develop a mandatory briefing item for all air traffic controllers and ATC managers, describing the circumstances surrounding the performance of the CERAP controller in this accident to reinforce the importance of following ATC procedures.

2.6.2 Intentional Inhibition of the Minimum Safe Altitude Warning System at Guam

Since February 1995, the Guam ARTS IIA MSAW system^[29] had been intentionally inhibited by the FAA from providing low-altitude alerts inside a 54-nm ring around the ASR-8 radar antenna. The system was inhibited because it had been generating what air traffic controllers believed to be numerous false alerts, or "nuisance warnings." Thus, at the time of the accident, the MSAW system was only available (uninhibited) in a 1-mile-wide band around the ASR-8 radar site, between 54 and 55 nm. Korean Air flight 801 crashed approximately 3 nm southwest of Guam International Airport in an area of rising terrain that would have been covered by the MSAW system if it had not been inhibited.

FAA technical staff and Safety Board investigators conducted a postaccident simulation using the original parameters intended for the system. The simulation results indicated that, if the MSAW system had not been inhibited inside the 54-nm radius, both a visual and aural low-altitude alert would have been generated on the ARTS IIA monitors in the CERAP facility about 0141:22, as the airplane was descending through 1,700 feet msl. Accordingly, the Safety Board concludes that, if the ARTS IIA MSAW system had been operating as initially intended, a visual and aural warning would have activated about 64 seconds before flight 801 impacted terrain, and this warning would have likely alerted the CERAP controller that the airplane was descending below the minimum safe altitude for that portion of the approach.

Flight 801 was under the control of the Agana tower controller at the time that the low-altitude MSAW alert would have been issued by the ARTS II system in the CERAP facility. The Agana tower was not equipped with a functioning terminal radar display. Therefore, for the crew of flight 801 to have received a low-altitude advisory, the CERAP controller (who was still responsible for monitoring the airplane after he initiated a frequency change to the tower controller) would have had to relay the alert to the tower controller, who would then have had to convey the alert to the flight crew. Given the prevalence of CFIT accidents, controllers would be expected to vigilantly monitor the system and provide timely notification to either another controller or a flight crew when an MSAW alert indicates the existence of an unsafe situation. The Safety Board concludes that 64 seconds would have been sufficient time for the CERAP controller to notify the Agana tower controller of the low-altitude alert, the tower controller to convey the alert to the crew of flight 801, and the crew to take appropriate action to avoid the accident.

Because of its periodic evaluations of air traffic facilities, FAA quality assurance staff knew as early as July 1995 that the Guam ARTS IIA MSAW system had been inhibited. The inhibition was cited in a 1995 FAA facility evaluation report but was only classified as an "informational" item. The FAA conducted no followup activities after the 1995 evaluation to determine whether corrective action had been taken to restore the MSAW system to the full service for which it was designed. In April 1997, the FAA conducted a second evaluation of the Guam facility, but the FAA's report on this evaluation did not even note that the ARTS IIA MSAW system was inhibited. Thus, the FAA missed two opportunities not only to recognize that the MSAW system was inhibited to the extent that it was rendered almost completely useless but also to take corrective action. An appropriate corrective action could have prevented this accident. Therefore, the Safety Board concludes that the FAA's quality assurance for the MSAW system was inadequate, and the agency's intentional inhibition of that system contributed to the flight 801 accident.

As previously noted, in this accident there would have been sufficient time (64 seconds), if the MSAW system had generated an alert in the CERAP facility, for the CERAP controller to have relayed the information to the tower controller. However, under different circumstances, an aircraft descending below the minimum safe altitude may not generate an MSAW alert as far in advance, so controllers may have significantly less time to react. In those cases, it would make a critical difference if the MSAW alert were provided directly to the airport tower.

The Safety Board has long been concerned about the issue of aural MSAW alerts in towers. As part of its investigation into the January 1995 Beechcraft A36 accident, the Safety Board found that the FAA did not have a policy regarding the installation of an aural MSAW alert at low-density ATC towers equipped with D-BRITE radar displays. As a result, the Safety Board issued Safety Recommendation A-95-120 on November 30, 1995. Safety Recommendation A-95-120 asked the FAA to develop a policy that would require the installation of aural MSAW equipment in those visual flight rules (VFR) terminal facilities that receive radar information from a host radar control facility and would otherwise receive only a visual MSAW alert.^[30]

In June 1996, the FAA stated that it was feasible to install the aural MSAW alert in 112 VFR towers. In July 1997, the FAA stated that 69 of 112 ATC facilities did not have remote displays with aural alarms and that aural alarms at these facilities would be installed by February 1998. In May 1998, the FAA stated that the aural alarms at these 69 remote sites would be operational by the end of that month. However, in March 1998, at the Safety Board's public hearing, the FAA's Deputy Program Director for Air Traffic Operations indicated that the new projected completion date for the installation of aural alarms in VFR towers was April 2000. In October 1998, the Safety Board expressed its concern to the FAA that VFR tower controllers who have visual representation from a distant host radar may not receive an aural alert when aircraft under their control, or with whom they are in radio communication, descend below the minimum safe altitude. The Board asked the FAA to ensure that all VFR tower controllers with visual representation from a host radar would in fact receive such warning. Pending further information from the FAA, Safety Recommendation A-95-120 was classified "Open--Acceptable Response."

On September 29, 1999, a representative from the FAA stated that the agency's management had indicated that the Agana tower was currently receiving aural MSAW alerts. At an October 7, 1999, briefing attended by the FAA Administrator, the Safety Board Chairman, and staff from both agencies, the FAA indicated that 69 MSAW aural alarms had been delivered and that 51 alarms were to be delivered. The FAA expected that the acquisition of these 51 alarms would be completed by October 2000 and that their installation in VFR towers would be completed by April 2001.

On October 12, 1999, the FAA Program Director for Serco Aviation Services told Safety Board staff that the Agana tower has the capability to receive an aural MSAW alert but that, unless the Guam CERAP transfers responsibility for the aircraft's data block, the tower will not receive the warning. The official added that the CERAP does not currently transfer responsibility for the aircraft's data block to the Agana tower; therefore, the tower does not receive an aural MSAW alert.

On October 14, 1999, the FAA Program Director for Air Traffic Operations confirmed that Agana tower was not receiving aural MSAW alerts. In an October 15, 1999, facsimile, the program director indicated that the tower "has the software and hardware capability in place to receive aural alarms." The director further stated that the FAA had issued a policy "to ensure that the facility that is in direct radio communications with the aircraft receives the aural alarm" and that the policy would become effective by November 15, 1999. In a followup telephone conversation with the Safety Board's Director of the Office of Aviation Safety, the program director indicated that a national policy would be issued to ensure that procedures similar to those being implemented at Guam are followed at other VFR towers.

On October 25, 1999, the FAA indicated that the MSAW aural alarms for the ARTS IIA system at Guam were reconfigured on October 24, 1999. The FAA stated that, in the event of a low-altitude alert for an aircraft operating in the vicinity of Guam International Airport, aural alarms will be simultaneously generated at the CERAP and the Agana tower, along with visual low-altitude alerts on the radar displays at both facilities.

On November 2, 1999, the Safety Board received a copy of draft FAA Notice N7210.485, "Minimum Safe Altitude Warning for Remote Tower Displays." According to the notice, facility managers at ATC towers that have aural alarms for MSAW are to ensure that "the operational support facility has adapted the software functionality to ensure the aural alarms operate in the ATCT [air traffic control tower]" and that "aural alarms are received in the ATCT upon transfer of communications." The FAA indicated that the effective date for this notice would be February 1, 2000.

The Safety Board is concerned about the delay in the implementation of Safety Recommendation A-95-120.^[31] In addition, the Safety Board is especially concerned that the FAA, until it received queries from the Board, was apparently not aware of, or not addressing, procedural barriers that prevented the installed equipment from being used as intended. However, on the basis of the FAA's apparent continued intention to fully implement this recommendation, it remains classified "Open--Acceptable Response."

2.7 Emergency Response

Although a fire station was located about 1 mile from the accident site, the first emergency response equipment (dispatched from a different fire station about 3 ½ miles from the accident site) did not arrive on scene until approximately 52 minutes after the accident. Safety Board investigators attempted to determine the reason(s) for the slow emergency response and the extent to which it could have been reduced or avoided.

Because of the air traffic controllers' delayed discovery of the accident, ramp control personnel, who were responsible for emergency notifications, were not aware of the accident until 0158, about 16 minutes after the crash occurred. As discussed in section 2.6.1, if the CERAP controller had been monitoring the flight more closely, this delay might have been eliminated or reduced.

After being notified of the accident by Guam airport ramp control, the Guam Fire Department (GFD) communications center dispatched GFD Engine No. 7, which was stationed about 3 ? miles from the crash site, at 0207. However, Engine No. 7's departure from the station was delayed by 12 minutes because its brake system needed to be recharged with air. Engine No. 7 departed the station at 0219, and its en route response time was 15 minutes. Engine No. 7 was the first emergency response vehicle to arrive at the VOR access road (at 0234, 52 minutes after the accident). The nearest fire station to the accident site was the U.S. Navy Federal Fire Department (located about 1 mile from the accident site). According to Federal dispatch facility logs, that station was not notified of the accident until 0234. The station's Engine No. 5 was then immediately dispatched and arrived at the accident scene at 0239 (a response time of 5 minutes). The Chief of Staff for the Commander, U.S. Naval Forces, Marianas, notified Navy "first responders" to stand by after she learned of the accident at 0216. However, the Navy had not yet received a request for specific Federal firefighting and medical resources; therefore, it would have been inappropriate for the Chief of Staff to have dispatched these resources.

1. Armed is indicated with white letters on a black background. Blank is indicated by an absence of letters.
2. Capture is indicated with black letters on a green background.
3. It is not clear why the first officer made this callout at this altitude. It is possible that the first officer may have intended the callout to mean that only 1,000 feet remained before reaching the 1,440-foot stepdown altitude, or he may have confused 2,400 feet with 1,400 feet on the altimeter. It is also possible that the first officer may have believed that the DME was located on the airport and that the airplane was approaching a DME value at which the airplane could descend to 1,440 feet. (See section 2.4.1.3.2 for a discussion about possible DME confusion.)
4. The approach briefing is called a "landing briefing" on the Korean Air checklist card.
5. Testimony by Korean Air officials at the Safety Board's public hearing indicated that these items were taught in company flight crew training.
6. 191 National Transportation Safety Board. 1994. A Review of Flightcrew-Involved Major Accidents of U.S. Air Carriers, 1978 Through 1990. Safety Study NTSB/SS-94/01. Washington, DC, pp. 40-41. 190
7. 192 Orasanu, J. "Decision-making in the Cockpit." In Cockpit Resource Management. 1993. Ed. E.L. Weiner, B.G. Kanki, and R.L. Helmreich. San Diego: Academic Press, p. 159.
8. The Safety Board notes that the raw data localizer and glideslope needles and off flags on the first officer's ADI and HSI would have been out of view because his navigation receiver was tuned to the VOR.
9. The Boeing 727 flight crew stated that no glideslope flags were visible and that the ADI glideslope needle was "centered."
10. An FAA flight check of the VOR and DME transmitters conducted on August 7, 1997, determined that the systems were functioning properly and within prescribed tolerances.
11. Human factors research has shown that a common decision-making error, especially in high stress and workload situations, is for people to tend to ignore evidence that does not support an initial decision. Human "operators tend to seek (and therefore find) information that confirms the chosen hypothesis and to avoid information or tests whose outcome could disconfirm it," which produces an "inertia which favors the hypothesis initially formulated." See Wickens, C. (1992). Engineering Psychology and Human Performance, 2nd Edition. Columbus, Ohio: Charles E. Merrill.
12. Korean Air's cockpit training procedures instructed the pilot flying a nonprecision approach (with the autopilot engaged) to program the autopilot/FD controls, including VERT SPEED and ALT SEL (altitude select), unless that pilot directed the nonflying pilot to do so. In addition, flight crews were trained that, while executing the approach profile, the flying pilot was to initiate all heading, course, and altitude changes, including selection of the step-down altitudes. The role of the nonflying pilot was to monitor and challenge if the flying pilot failed to follow proper procedures.
13. 198 A Review of Flightcrew-Involved Major Accidents of U.S. Air Carriers, 1978 Through 1990, pp. 47-49 and 55-59.
14. Gander, P.H., Gregory, K.B., Miller, D.L., Graeber, R.C., Connell, L.J., and Rosekind, M.R. (1998). "Flight Crew Fatigue V: Long-Haul Air Transport Operations." Aviation, Space, and Environmental Medicine, 69 (9, Section II, Supplement), pp. B37-B48. 199
15. Akersdet,T (1998)." Shift Work and Disturbed Sleep/Wakefullness." Sleep Medicine Reviews 2,2(2), pp 117-128.
16. Akerstedt, T. (1998). "Shift Work and Disturbed Sleep/Wakefulness." Sleep Medicine Reviews, 2(2), pp. 117-128. National Transportation Safety Board. 1994. Uncontrolled Collision With Terrain, American International Airways Flight 808, Douglas DC-8-61, N814CK, U.S. Naval Air Station, Guantanamo Bay, Cuba, August 18, 1993. Aircraft Accident Report NTSB/AAR-94/04. Washington, DC, p. 135.
17. A Review of Flightcrew-Involved Major Accidents of U.S. Air Carriers, 1978 Through 1990, pp. 23-28.
18. National Transportation Safety Board. 1994. In-flight Loss of Control, Leading to Forced Landing and Runway Overrun, Continental Express, Inc., Embraer EMB-120RT, N24706, Pine Bluff, Arkansas, April 29, 1993. Aircraft Accident Report NTSB/AAR-94/02/SUM. Washington, DC.
19. Dinges, D.F. "Napping Patterns and Effects in Human Adults." In Sleep and alertness: Chronobiological, behavioral, and medical aspects of napping. 1989. Ed. D.F. Dinges and R.J. Broughton. New York: Raven Press, p. 181. Lavie, P. "To Nap, Perchance To Sleep--Ultradian Aspects of Napping." In Sleep and alertness: Chronobiological, behavioral, and medical aspects of napping. 1989. Ed. D.F. Dinges and R.J. Broughton. New York: Raven Press, pp. 114-117.
20. Investigation determined that, on July 27, 1997, the captain's personal physician diagnosed him with bronchitis and prescribed a medication that could be used as a sleeping aid. On July 28 through 30, the captain flew an international round trip between Korea and the United States. The combined effects of the captain's illness and his long trip across numerous time zones were likely to have provided disruptions to his sleeping schedule that might have continued to affect him at the time of the accident.
21. 207 During training, Korean Air pilots performed two different NDB approaches; each was performed once, and neither incorporated DME. The pilots also performed the localizer approach to runway 14 at Kimpo once and the VOR/DME approach to runway 32 at Kimpo five times. The localizer and VOR/DME approaches used a DME that was colocated and frequency paired with approach navigational facilities located on the airport. Thus, the pilots were exposed to four nonprecision approaches during their training, and the VOR/DME approach to runway 32 at Kimpo was the only approach performed more than once.
22. National Transportation Safety Board. 1996. Collision With Trees on Final Approach, American Airlines Flight 1572, McDonnell Douglas MD-83, N566AA, East Granby, Connecticut, November 12, 1995. Aircraft Accident Report NTSB/AAR-96/05. Washington, DC.
23. The Safety Board found similar nonprecision approach-related factors in its investigation of the 1995 accident involving an American Airlines Boeing 757 on a nonprecision approach to Cali, Colombia; the 1990 accident involving a Markair Boeing 737 on a nonprecision approach to Unalakleet, Alaska; the 1989 accident involving a Flying Tigers Boeing 747 that crashed while executing an NDB approach to Kuala Lumpur, Malaysia; and the 1989 incident involving a USAir Boeing 737 executing a localizer backcourse approach to Kansas City, Missouri.
24. The Safety Board notes the Korean Air simulator training now incorporates a variety of approach scenarios, including approaches in which the DME is not colocated with an on-airport navigational facility and approaches involving countdown/count up DME procedures. Also, the simulator training now includes approaches likely to be encountered in the airline's domestic and international operations.
25. 211 The controller was required to continue monitoring the flight because radar service had not been terminated in accordance with paragraph 5-1-13.
26. The CERAP controller was monitoring the EARTS (en route) radar scope, which had a functioning MSAW capability. However, this MSAW capability was based on a different algorithm than the disabled ARTS IIA MSAW system (see section 2.6.2). The ARTS IIA MSAW system compares the airplane's trajectory with the ILS glideslope. The EARTS system uses a single altitude (based on the lowest MDA for all nonprecision approaches to the runway) from the FAF to the point at which MSAW processing terminates (usually 1 mile from the runway threshold).
27. The Safety Board recognizes that the en route radar display was set to a range of 265 nm and therefore could not be used for effectively monitoring the final approach. The terminal radar display was set to a range of 60 nm and displayed the final approach course for runway 6L.
28. Although the CERAP controller told Safety Board investigators that his last observation of the target of flight 801 on the terminal radar display was when the airplane was 7 miles from the airport at an altitude of 2,600 feet, FDR and radar data do not support his statement. The data indicated that, when the CERAP controller instructed the flight to contact the Agana tower, the airplane was at an altitude of about 2,200 feet and maintained a continual descent. Therefore, the airplane was probably farther than 7 miles from the airport when the CERAP controller last observed the flight.
29. Although the CERAP controller told Safety Board investigators that his last observation of the target of flight 801 on the terminal radar display was when the airplane was 7 miles from the airport at an altitude of 2,600 feet, FDR and radar data do not support his statement. The data indicated that, when the CERAP controller instructed the flight to contact the Agana tower, the airplane was at an altitude of about 2,200 feet and maintained a continual descent. Therefore, the airplane was probably farther than 7 miles from the airport when the CERAP controller last observed the flight.
30. For more information on this safety recommendation, see section 1.18.2.2.
31. The Safety Board is also disappointed that the aural MSAW alert could not be installed on the D-BRITE system at Guam in a timely manner because the FAA did not certify the system until April 1998, more than 1 year after it was delivered.