The indicated airspeed was on target, and the approach path lights showed that the airplane was just slightly high as it descended below 500 ft, the point at which the stability of a visual approach typically is judged. These indications were deceptive, however, and the flight crew did not recognize that the airspeed was decreasing rapidly and that the airplane would soon descend below the 3-degree glidepath.
There were other signs that the approach was not stabilized: The thrust levers were at idle (the engines were not spooled up properly), and the descent rate was higher than it should have been.
By the time the crew realized that a missed approach was in order, it was too late. “The airplane did not have the performance capability to accomplish a go-around,” the U.S. National Transportation Safety Board (NTSB) said in its report on the subsequent accident.
The Asiana Airlines Boeing 777-200ER clipped a seawall bordering Runway 28L at San Francisco International Airport, slid down the runway as it shed parts and became airborne again momentarily before coming to a stop in flames off the side of the runway.
Three of the 291 passengers were killed, and 40 passengers, eight of the 12 flight attendants and one of the four flight crewmembers were seriously injured in the accident, which occurred the morning of July 6, 2013.
The NTSB concluded that the probable cause of the accident was “the flight crew’s mismanagement of the airplane’s descent during the visual approach, the PF’s [pilot flying’s] unintended deactivation of automatic airspeed control, the flight crew’s inadequate monitoring of airspeed, and the flight crew’s delayed execution of a go-around after they became aware that the airplane was below acceptable glidepath and airspeed tolerances.”
The report said that factors contributing to the accident were: “(1) the complexities of the autothrottle and autopilot/flight director systems that were inadequately described in Boeing documentation and Asiana’s pilot training, which increased the likelihood of mode error; (2) the flight crew’s nonstandard communication and coordination regarding the use of the autothrottle and autopilot/flight director systems; (3) the PF’s inadequate training on the planning and execution of visual approaches; (4) the PM [pilot monitoring]/instructor pilot’s inadequate supervision of the PF; and (5) flight crew fatigue, which likely degraded their performance.”
New Roles
The 777 was being operated as Asiana Airlines Flight 214, a scheduled passenger flight from Seoul, South Korea, with an estimated time en route to San Francisco of 10 hours and 24 minutes.
An instructor pilot and a trainee captain comprised the primary flight crew, backed up by a relief captain and first officer. The instructor pilot was the pilot-in-command (PIC) of the flight. He occupied the right seat and served as the PM during the takeoff from Seoul and the approach to San Francisco; the trainee captain was in the left seat and was the PF. Both pilots were relatively inexperienced in their flight roles.
The trainee captain, 45, had 9,684 flight hours, including 3,729 hours as PIC. He had no previous flight experience when he was hired by Asiana as a cadet pilot in 1994. He served as a first officer in 737s and 747s before upgrading as a 737 captain in 2005. He transitioned as an Airbus A320 captain in 2007 and began transition training to become a 777 captain in March 2013. By the end of May, he had completed ground training, flight simulator training and line-oriented flight training in the 777.
“He began flying the 777 with an IP [instructor pilot] as part of his required initial OE [operating experience] on June 16, 2013,” the report said. He had logged eight flight legs and 33.5 flight hours with IPs. “All of the approaches the PF had previously flown during OE were ILS [instrument landing system] approaches, and there was no requirement to perform visual approaches during OE,” the report said.
Investigators received mixed feedback when they interviewed three IPs who had flown with the trainee captain. One said his overall performance was above average. Another said that “nothing stood out” about his performance. The third IP said that the trainee captain had not performed well. “He said the PF was not well organized or prepared, conducted inadequate briefings, poorly monitored the operation and deviated from multiple standard operating procedures,” the report said.
The IP told investigators that he provided “extensive counseling” but “worried that the PF was not taking his feedback seriously.” However, the IP said that he was not overly concerned because he knew that the trainee captain had to complete more OE flights. (Korean aviation regulations required 20 flight legs and 60 flight hours of OE.)
The IP who served as the PIC and PM on Flight 214 was 49 and had 12,307 flight hours, including 9,045 hours as PIC and 3,208 hours in 777s. He served as a Korean air force pilot before joining Asiana in 1996. He flew as a 767 first officer and captain before transitioning as a 777 captain in 2008.
“He underwent 777 IP training in May and June 2013 and became qualified as an IP on June 12, 2013,” the report said. The captain who conducted the IP’s final check flight said that he was “very calm, followed the procedures correctly, had professional knowledge of the flight and had good capability and skill as an instructor.” Flight 214 was the PIC’s first as an acting IP.
Visual Approach
The relief pilots took over about four hours into the flight, allowing the primary crew to rest for about five hours. When the trainee captain and IP resumed their posts, the airplane was less than two hours from San Francisco.
Visual meteorological conditions prevailed at the destination, and visual approaches to Runway 28L and 28R were in progress. The automatic terminal information service noted that the glideslopes for the ILS approaches to these runways were out of service.
The relief captain already had programmed the ILS approach to Runway 28L in the flight management system. Expecting vectors for a visual approach to 28L, the trainee captain (the PF) planned to use the ILS localizer to maintain the lateral path to the runway and the automatic flight control system (the autopilot/flight director and autothrottle systems) to manage the vertical profile. The calculated reference landing speed (VREF) was 132 kt.
The airplane was nearing the airport at 1112 San Francisco time when the relief first officer returned to the flight deck and occupied the center jump seat. He served as an observer during the approach and landing.
As expected, Northern California Approach Control provided the crew with radar vectors for a straight-in approach to Runway 28L. At 1121, the approach controller asked the crew if they had the airport in sight. The PM (the instructor pilot) replied that the runway was in sight, and the controller issued clearance for a visual approach.
At the time, the airplane was descending through 6,300 ft at 210 kt. The autothrottle system was in the “HOLD” mode, and the autopilot was in the “FLCH SPD” (flight level change speed) pitch mode and in the “HDG SEL” (heading select) roll mode.
With these modes selected, the autopilot would hold the heading that the crew selected on the mode control panel (MCP) and would command changes in elevator position, effectively adjusting the airplane’s pitch attitude, to maintain the airspeed selected on the MCP (210 kt in this case). Vertical speed would be maintained with thrust; in this case, with the autothrottle system on “HOLD,” the thrust levers would remain in their current positions until manually moved by one of the pilots.
The PF announced, “I am intercepting the localizer,” and selected the “LOC” mode, which causes the autopilot to capture and track the localizer course. The selected altitude was changed to 3,100 ft, which corresponded with the minimum altitude for crossing a step-down fix just outside the final approach fix for the published localizer approach procedure.
Drifting High
The 777 was descending through 5,300 ft at about 210 kt when it intercepted the localizer course 15 nm (28 km) from the runway threshold. The PM said, “Let’s descend slowly to one thousand eight hundred feet,” which was the published minimum altitude for crossing the final approach fix, 5.4 nm (10.0 km) from the threshold.
The PF replied, “Yes, sir, I will set to one thousand eight hundred,” and set 1,800 ft in the altitude selector.
The approach controller then told the crew to reduce their airspeed to 180 kt and to maintain that speed until they were 5 nm (9 km) from the airport. The PM read back the instruction, and the PF changed the selected airspeed to 180 kt.
The reduction in the selected airspeed caused the autopilot to increase the airplane’s pitch attitude to maintain that speed. As a result, the descent rate decreased to 300 fpm. “The PF did not appear to promptly recognize that the airplane was drifting above the desired glidepath,” the report said, noting that the deviation would have been shown graphically on the pilots’ navigation displays.
The airplane was descending through 4,300 ft about 12 nm (22 km) from the runway when the PF changed the autopilot pitch mode from “FLCH SPD” to “VS” (vertical speed), selected a vertical speed of 1,000 fpm and selected the autothrottle speed (“SPD”) mode. “The airplane’s vertical speed began to increase toward the target value,” the report said. “However, a descent rate of 1,000 fpm was not high enough to maintain, let alone recapture, the desired glidepath, so the airplane continued to drift above it.”
The airplane was descending through 3,400 ft about 9 nm (17 km) from the runway when the pilot called for the landing gear to be extended. The additional drag from the landing gear would have facilitated deceleration, the report said, but the crew did not use the speed brakes or select a higher flap setting (flaps 20 at this point), either of which would have helped in managing the descent.
‘It’s Too High?’
Apparently referring to the airplane’s height (900 ft) above the desired glidepath, the PM said, “This seems a little high.” After a few seconds, the PF replied, “Do you mean it’s too high? … I will descend more,” and changed the selected vertical speed from 1,000 fpm to 1,500 fpm.
“This exchange was followed by 21 seconds of no communication between the pilots as the airplane’s descent rate increased and the airplane drew closer to the desired glidepath,” the report said.
However, the PF changed the selected vertical speed back to 1,000 fpm when the airplane was about 6 nm (11 km) from the runway, descending through 2,600 ft at 178 kt — and still well above the desired glidepath.
“By examining the altitude and distance to the runway, both of which were displayed on the instrument panel, and applying the well-known rule of thumb that a 3-degree glidepath requires about 300 ft of altitude loss per nautical mile, the pilots could have quickly estimated that they were still several hundred feet high,” the report said. “The flight crew needed to continue descending the airplane at more than 1,000 fpm to return to the desired glidepath.
“The crew’s action indicated a lack of awareness of the airplane’s position relative to the desired glidepath and of cues in the cockpit that could have alerted them to this. As a result of this lack of awareness and their early reversion to a descent rate of 1,000 fpm, the airplane remained high.”
The airplane crossed the final approach fix at 2,250 ft — 450 ft high. The crew selected flaps 20, and the PF set 150 kt in the airspeed selector. He also entered 3,000 ft, the published missed approach altitude, in the altitude selector.
At the time, indicated airspeed was about 4 kt higher than the maximum speed for selection of flaps 30, which would have allowed a steeper descent while maintaining the selected airspeed. “Clearly, the airplane’s excess altitude increased the difficulty of achieving a stabilized approach,” the report said.
The report noted that when Boeing test pilots later attempted to conduct an approach from this point in a flight simulator, they had difficulty achieving a stabilized approach before reaching 500 ft above ground level (AGL). “In fact, they found it impossible to do so without exceeding maximum descent rates published in Asiana’s FOM [flight operations manual],” the report said.
Mode Confusion
“The flight crew’s difficulty in managing the airplane’s vertical path continued as the approach progressed” beyond the final approach fix, the report said. Aware of the need to lose the excess altitude, the PF changed back from the “VS” mode to the “FLCH SPD” mode, believing that this would increase the descent rate.
Instead, however, selection of the flight change speed mode at this point caused the autopilot to command an increase in pitch to slow the airplane to the selected airspeed, 150 kt. Simultaneously, the autothrottle system responded, as designed, by entering the thrust mode and moving the thrust levers forward to attain the selected altitude, 3,000 ft.
The PF reacted to the unexpected pitch-up by disengaging the autopilot, moving the thrust levers to idle and manually pitching the airplane nose-down. The manual change in thrust lever position caused the autothrottle system to enter the “HOLD” mode, which effectively disengaged the system. Neither the pilots nor the observer noticed this mode change. “This is not surprising in light of human factors research demonstrating that pilots frequently do not notice mode changes on the FMA [flight mode annunciator], especially those that are unexpected,” the report said.
The report noted that the PF had not announced any of the mode changes he had made, which reduced the ability of the PM and the observer to cross-check and monitor his actions.
The airplane was descending through 1,300 ft at 165 kt and 1,000 fpm when the PM said “speed.” The PF replied, “Target speed one three seven” (VREF plus 5 kt) and selected 137 kt on the MCP. The PM then said, “It’s high,” and the PF responded by manually increasing the descent rate.
“By this point, the flight crew should have been able to clearly see the precision approach path indicator (PAPI) lights,” the report said. “The PAPI indication would have been four white lights, showing that the airplane was significantly above the PAPI glidepath angle of 2.98 degrees.”
The observer called out “sink rate, sir” three times as the descent rate increased to 1,500 fpm and then to 1,800 fpm. He later told investigators that he made the callouts because Asiana required that the descent rate be no more than 1,000 fpm below 1,000 ft. “He further stated that the PF and PM were slow to respond to his sink rate callouts, but they did respond, and the sink rate decreased,” the report said.
Decision Time
Indicated airspeed was 137 kt when the airplane descended at 1,200 fpm through a radio altitude of 500 ft about 1.3 nm (2.4 km) from the runway. The thrust levers were still at idle, and the engines were at 24 percent N1 (low-pressure spool speed). The PAPI lights showed the airplane to be slightly above the 3-degree glidepath.
“Although, at 500 ft, the airplane met some of Asiana’s stabilized approach criteria … (including being on target airspeed, in the landing configuration and on the correct flight path), it failed to satisfy other criteria,” the report said. “It was descending at greater than 1,000 fpm, and the thrust setting was not appropriate (it should have been about 56 percent N1 speed).
“Because the approach was not stabilized at 500 ft AGL, the flight crew should have conducted a go-around. Either the pilots did not notice that these parameters exceeded stabilized approach criteria or they believed that the deviations were minor and could easily be corrected. In either case, the crew’s decision to press ahead was not unusual, as industry statistics indicate about 97 percent of unstable approaches are continued to landing.”
As the airplane descended below 500 ft, the PM said, “Landing checklist complete, cleared to land … on glidepath, sir.” The PF replied “check.”
Red Lights
About five seconds later, all the PAPI lights turned red, indicating that the airplane was significantly below the glidepath. At this point, the 777 was 219 ft over San Francisco Bay and 0.7 nm (1.3 km) from the runway, descending at 900 fpm and about 130 kt.
“Both the airspeed indication, which was more than 5 knots below target approach speed, and a PAPI indication of four red lights required a go-around, but the flight crew continued the increasingly unstabilized approach,” the report said.
The PM, referring either to the airspeed or the airplane’s position below the glidepath, said, “It’s low.” The PF, still under the impression that the autothrottle system would adjust thrust to maintain the target airspeed, apparently took the PM’s callout as referring to the glidepath; and he responded by pulling back the control column. The pitch attitude increased from 5 degrees to 7.5 degrees, and airspeed decreased further.
Aural and visual master warnings were generated when airspeed decreased to 120 kt. When airspeed dropped to 114 kt, the PM called out “speed” and moved the thrust levers forward. Airspeed continued to decrease, however, and the stick shaker (stall warning) activated at 103 kt.
“At this time, the airplane was about 0.35 nm [0.65 km] from the runway at 39 ft RA [radio altitude], the descent rate was about 700 fpm, the N1 speeds for both engines were increasing through about 50 percent, and the pitch attitude reached about 12 degrees nose-up,” the report said. “The airspeed then began to increase.”
The PM called out “go around,” and the PF responded. At this point, however, the airplane lacked the performance capability to accomplish a go-around, the report said. The main landing gear and aft lower fuselage struck the seawall about three seconds later, at 1128 local time.
“Video from airport surveillance cameras showed that following the initial impact, the tail of the airplane separated, the airplane slid along the runway, and the rear of the fuselage lifted up, tilting the airplane into about a 30-degree nose-down angle,” the report said. “The airplane pivoted counterclockwise about 330 degrees before impacting a second time and coming to rest off the left side of the runway, about 2,400 ft [732 m] from the initial seawall impact point.”
The 777 was destroyed by the impact and subsequent fire. Two of the three passengers who died in the accident had been ejected from the airplane after it struck the seawall. In addition to the fatalities and serious injuries, 134 passengers, two flight attendants, the relief first officer and the PM sustained minor injuries; 114 passengers, two flight attendants and the relief captain were not hurt.
‘Faulty Mental Models’
“In postaccident interviews, the PF made several statements that indicated he had an inaccurate understanding of some aspects of the airplane’s autoflight system,” the report said. For example, as demonstrated during the approach to San Francisco, he believed that the autothrottle system was “always working” and would maintain the selected airspeed even after a manual change of thrust lever setting with the autoflight system in the “FLCH SPD” mode.
Interviews with other Asiana pilots and instructors revealed similar misunderstandings. Investigators also found deficiencies in Boeing’s documentation of the autoflight system and in the airline’s training on the system.
Moreover, the report cited human factors research showing that due to the complexity of autoflight systems and subsystems in airplanes such as the 777, “faulty mental models” of how they work are fairly common among pilots.
Improvement of 777 autoflight system training was among the specific recommendations included in the NTSB report (ASW, 9/14). The safety board also called for an expert panel to be convened to evaluate methods of training pilots on automated systems and to identify the most effective training methods.
This article is based on NTSB Accident Report AAR-14/01, “Descent Below Visual Glidepath and Impact With Seawall; Asiana Airlines Flight 214; Boeing 777-200ER, HL7742; San Francisco, California; July 6, 2013.” The report is available at <ntsb.gov/investigations/reports.html>. The report provides an in-depth examination of cabin safety issues involved in the accident; those issues will be discussed in the November AeroSafety World.