On Feb. 23, 2019, at 1239 CST (1839Z), Atlas Air Inc. (Atlas) Flight 3591, a Boeing 767-375BCF, dived from 6,000 ft. into a shallow, muddy marsh area of Trinity Bay 34 mi. east-southeast of George Bush/Houston Intercontinental Airport (KIAH) while on approach. The captain, first officer and a non-revenue pilot riding in the jump seat died, and the cargo aircraft was destroyed on impact.
After a 16-month investigation, the NTSB determined the probable cause was “the inappropriate response by the first officer as the pilot flying to an inadvertent activation of the go-around mode, which led to his spatial disorientation and nose-down control inputs that placed the airplane in a steep descent from which the crew did not recover.
“Contributing to the accident was the captain’s failure to adequately monitor the airplane’s flightpath and assume positive control of the airplane to effectively intervene.
“Also contributing were systemic deficiencies in the aviation industry’s selection and performance measurement practices, which failed to address the first officer’s aptitude-related deficiencies and maladaptive stress response. Also contributing to the accident was the FAA’s failure to implement the Pilot Records Database [PRD] in a sufficiently robust and timely manner.” (See “Pilot Records Redux,” page 57.)
According to the NTSB’s investigators, here is what happened to Flight 3591. This information comes from testimony at the Safety Board’s hearing into the incident and investigators’ reports:
Atlas operated the aircraft as an FAR Part 121 domestic cargo flight for Amazon.com Services LLC.
The accident airplane arrived at Miami International Airport (KMIA) from Ontario, California, International Airport (KONT) at 1941Z on Feb. 22. The accident flight crew previously had been on duty 6 hr. and 24 min. on the day before the accident and completed 24 hr. and 17 min. of rest before the scheduled report time of 1438Z for the departure of GTI3591 at 1608Z on Feb. 23. (We’ll use UTC in this report because of the 1-hr. difference between EST and CST.)
The accident flight departed Miami about 1633Z on an IFR flight plan to KIAH. The first officer (FO) in the right seat was the pilot flying (PF) and the captain was the pilot monitoring (PM).
The subsequent departure and climb from KMIA and cruise at FL 400 to the Houston area were uneventful. About 1715Z and again at 1717Z, GTI3591 requested and received KIAH ATIS weather information Papa.
Papa was recorded at 1713Z and showed winds 320 at 18 kt., gusting 42 kt.`; 8-sm visibility with mist; scattered clouds at 1,600 ft.; broken clouds at 2,000 and 2,900 ft.; and temperature/dew point 19/12C. The remarks section included wind-shear advisories in effect at the airport.
About 1759Z, GTI3591 checked in with the White Lake Ultra High Radar controller and was advised to expect light chop. The controller then provided revised routing after the GIRLY intersection and cleared the flight to KIAH via the LINKK1 arrival.
About 1809Z, ATC issued the flight a descent to FL 340, and 10 min. later it was cleared to descend via the LINKK1 arrival. At 1825Z, the captain advised ATC they were beginning their descent.
About 1830:37Z, GTI3591 checked in with Houston Approach with Information Sierra and reported descending via the LINKK1 RNAV arrival. (Sierra was recorded at 1752Z and showed winds 320 at 14 kt., 9-sm visibility, few clouds at 2,000 ft., scattered clouds at 3,300 ft., broken clouds at 6,000 ft. and temperature/dew point 21/12C. The remarks section did not include wind-shear
advisories.)
ATC instructed GTI3591 to fly the Runway 26L transition as the aircraft was descended through 17,800 ft. about 73 mi. southeast of the airport.
At 1834:08Z, the controller advised GTI3591 that there was light to heavy precipitation west of the airport.
At 1836:24, GTI3591 advised ATC they would fly to the west side of the weather — a small frontal area of moderate cells. The controller responded that that would be OK, but there were a “bunch of departures,” so he needed the flight to descend and maintain 3,000 and to “expedite” the descent. At that point the aircraft was 48 mi. southeast of KIAH at 10,000 ft. The captain did not respond, so the controller repeated the descent clearance and told GTI3591 to “hustle” down to 3,000 ft. The controller added the Boeing crew could expect vectors northbound for a base leg to Runway 26L. The captain responded to ATC and confirmed the descent to 3,000 ft.
About 10 sec. later, the FO extended the speed brake and asked the captain (PM) to lower the slats, which he did. The captain then turned his attention to setting up the FMC for the approach. Both the autopilot and autothrottle were engaged and remained engaged for the remainder of the flight.
The FO then commented that he was experiencing a potential failure of his attitude director indicator/horizontal situation indicator (ADI/HSI) display information, and then made a comment about using the electronic flight information (EFI) switch. The captain took control of the airplane and the FO cycled the EFI switch, which returned the display to normal.
At 1837:18Z, the controller instructed GTI3591 to turn to a 270-deg. heading, and the captain confirmed the heading when the airplane was about 40 mi. from KIAH and descending through 8,500 ft. Shortly afterward, according to recorded data, the captain transferred controls of the airplane back to the FO, who became the PF and the captain became the PM again. The captain continued setting up the approach into the airplane’s FMC.
Meteorologists determined the airplane was beginning to penetrate the leading edge of the cold front about 1838:25Z where the crew probably encountered associated wind shear and instrument meteorological conditions (IMC). As the aircraft penetrated the weather, the FDR recorded load factors consistent with the airplane encountering light turbulence.
At 1838:31Z, the airplane’s go-around mode was activated. The accident flight was about 40 mi. from the airport and descending through about 6,300 ft. MSL toward the target altitude of 3,000 ft. MSL.
Investigators said this location and phase of flight were inconsistent with any scenario in which a pilot would intentionally select go-around mode, and neither pilot made a go-around callout to indicate intentional activation. The go-around mode activation seemed to have been accidental.
The FO, who was the PF, was probably holding the speed-brake lever as expected in accordance with Atlas Air’s procedure, theorized investigators. The inadvertent activation of the go-around mode, they said, likely resulted from unintended contact between the FO’s left wrist or watch and the left go-around switch due to turbulence-induced loads that moved his arm.
When the go-around mode is selected, the autothrottle system begins to increase power and the aircraft begins to accelerate and pitch-up to 5 deg.
At 1838:37Z and while the thrust of the engines was increasing, the controller informed GTI3591 that he would turn the flight northbound for a base leg in about 18 mi.
The FDR recorded the speed brakes retracted to the near-zero position, and the captain then responded to ATC, “Sounds good.” ATC then advised GTI3591 that it was clear on the other side of the weather and they should have no problem getting to the airport. The captain responded, “OK.” That was the last transmission from the aircraft. The airplane was about 35 mi. from KIAH and descending through about 6,000 ft.
About a second after this ATC transmission, the FO made an expression of surprise followed by a comment related to airspeed. Three seconds later, the FO exclaimed, “We’re stalling,” and 4 sec. later, “Oh, Lord, have mercy on [me] myself.” The cockpit area microphone picked up someone calling, “Pull-up.”
Despite the presence of the go-around mode indications on the flight mode annunciator and other cues that indicated the airplane had transitioned to an automated flight path different from what the crew had been expecting, neither the FO nor the captain were aware that the airplane’s automated flight mode had changed.
Within seconds of the go-around mode selection, the FO probably made a large nose-down control input (overriding the autopilot), forcing the airplane into a steep dive from which the crew did not recover. Only 32 sec. elapsed between the go-around mode activation and the airplane’s ground impact, said the Safety Board.
At 1839:39Z, ATC lost radar contact with GTI3591 about 34 mi. from KIAH at an altitude of about 5,800 ft.
About 16 sec. after the captain told ATC “OK,” the FDR stopped recording data with the airplane descending at an airspeed of about 433.5 kt. with the autopilot still engaged. The airplane never stalled.
Security camera footage captured the final stages of the flight and showed the airplane in a steep descent into the Trinity Bay marsh. Both pilots and the jump-seater were fatally injured and the airplane destroyed.
NTSB’s Findings
The investigation determined that whatever went wrong happened in the cockpit — other factors were eliminated including ATC services, maintenance of the airplane structures, powerplants and systems, airplane weight and balance, and Atlas policies and procedures. Nor was evidence found that the crew was impaired due to medical conditions, alcohol or other drugs.
“Given that the first officer [FO] was the pilot flying and had not verbalized any problem to the captain or initiated a positive transfer of airplane control, the manual forward elevator control column inputs that were applied seconds after the inadvertent activation of the go-around mode were likely made by the FO,” said the Safety Board.
The FO likely experienced a pitch-up somatogravic illusion as the airplane accelerated due to the inadvertent activation of the go-around mode, which prompted him to push forward on the elevator control column.
“Although compelling sensory illusions, stress and startle response can adversely affect the performance of any pilot,” said the Safety Board, “the first officer had fundamental weaknesses in in his flying aptitude and stress response that further degraded his ability to accurately assess the airplane’s state and respond with appropriate procedures after the inadvertent activation of the go-around mode.”
The investigators discovered the first officer had a history of training performance difficulties at multiple employers “and demonstrated a tendency to respond impulsively and inappropriately when faced with an unexpected event during training scenarios.” That suggested to the Safety Board that the FO had an inability to remain calm during stressful situations — a tendency that may have exacerbated his aptitude-related performance difficulties.
The inadvertent selection of the go-around mode and the FO’s sudden push-over occurred while the captain was setting up the approach and communicating with ATC, therefore his attention was diverted from monitoring the airplane’s state and verifying the flight was proceeding as planned, which delayed his recognition of and response to the FO’s unexpected actions that placed the airplane in a dive.
“The captain’s failure to command a positive transfer of control of the airplane as soon as he attempted to intervene on the controls enabled the first officer to continue to force the airplane into a steepening dive,” said the Safety Board. “The captain’s degraded performance, which included his failure to assume positive control of the airplane and effectively arrest its descent, resulted from the ambiguity, high stress and short time-frame of the situation.”
NTSB members believe the first officer’s skill deficiencies went unidentified by Atlas in part due to the FAA’s struggle to develop the PRD. “Had
the FAA met the deadline and complied with the requirements for implementing the PRD as stated in the Airline Safety and FAA Extension Act of 2010, the PRD would have provided hiring employers relevant information about the first officer’s employment history and training performance
deficiencies.
“The first officer’s long history of training performance difficulties and his tendency to respond impulsively and inappropriately when faced with an unexpected event during training scenarios at multiple employers suggest an inability to remain calm during stressful situations — a tendency that may have exacerbated his aptitude-related performance difficulties.”
The Safety Board added the first officer’s repeated use of incomplete and inaccurate information about his employment history on resumes and applications were deliberate attempts to conceal his history of performance deficiencies and deprived Atlas Air and at least one other former employer of the opportunity to evaluate his aptitude fully and competency as a pilot.
“Atlas Air’s human resources personnel’s reliance on designated agents to review pilot background records and flag significant items of concern was inappropriate,” said the Safety Board, “and resulted in the company’s failure to evaluate the first officer’s unsuccessful attempt to upgrade to captain at his previous employer.”
Employers — airline, charter and business aviation operations — that rely on designated agents or human resources personnel for initial review of records obtained under the Pilot Records Improvement Act of 1996 should include flight-operations subject matter experts early in the records review process, according to the board’s report.