Delta Air Lines Flight 2196

The flight crew was conducting an instrument landing system approach to runway 9R. When the airplane was about 3.5 miles from the runway threshold and at an altitude of about 1,230 ft above ground level (agl), the first officer (the pilot flying) disconnected the autopilot, after which the airplane began to deviate to the right of the localizer course. When the airplane was at an altitude of about 500 ft agl, the first officer disconnected the autothrottle; at 300 ft agl, he began correcting to the left to return to the center of the localizer course. When the airplane reached the decision altitude (200 ft agl), the airplane was drifting toward the taxiway N extended centerline, which was parallel to, and about 650 ft to the left of, the runway 9R centerline. Radar data indicated that the airplane was 1 mile from the runway 9R threshold at the time that the airplane aligned with taxiway N. When the airplane was at an altitude of 120 ft agl and was 600 ft to the left of the runway 9R centerline and 50 ft to the right of the taxiway N centerline, the first officer initiated a go-around after the captain’s (the monitoring pilot) command. The airplane descended to about 50 ft above the western end of the taxiway before it began to climb. Engine power increased while the airplane was above and aligned with taxiway N. The airplane was then vectored for an instrument landing system approach for runway 10. The flight crew subsequently landed the airplane uneventfully. Another airplane (a Boeing MD-88) was taxiing westbound on taxiway N at the time of the incident approach. According to radar data, the airplanes, at their closest distances, were separated by 286 ft horizontally and 257 ft vertically. The approach became unstabilized when the first officer improperly adjusted the airplane’s heading and flew outside of the localizer course. The airline’s procedures indicated that an approach would be considered to be stabilized if it maintained, among other things, a “lateral flight path while in the landing configuration.” The manual warned that, if a stabilized approach could not be established and maintained, pilots were to initiate a goaround and not attempt to land from an unstable approach. Also, the airline’s procedures indicated that an approach should not continue below the decision altitude (200 ft agl in this case) unless “the aircraft is in a position from which a normal approach to the runway of intended landing can be made.” Thus, the flight crew’s actions were not consistent with company procedures. Further, when the airplane reached the decision altitude for the approach, the flight crew failed to call for a go-around and execute, in a timely manner, the initial steps for a go-around. Specifically, flight data recorder data showed that the takeoff/go-around switch was not selected until 4 seconds after the airplane reached the decision altitude and that a total of 12 seconds elapsed between the time that the airplane reached the decision altitude and the thrust lever began advancing toward go-around power. These delays caused the airplane to descend about 150 ft below decision altitude and come within about 50 ft of an occupied taxiway. The 1052 hourly weather observation for the destination airport indicated, among other conditions, 1/8 mile visibility, mist, patches of fog, and an overcast ceiling at 300 ft agl. The flight crewmembers received this observation about 1057 (9 minutes before the incident). Thus, the crewmembers were provided with sufficient information to understand the weather conditions that the flight would encounter during the approach to the airport. The captain and the first officer reported no history of sleep disorders, and a review of their sleep histories revealed that they received adequate rest during the 3 days preceding the incident. Further, sleep opportunities for the captain and first officer were aligned with local nighttime, so circadian disruptions were not an issue. Thus, the captain and the first officer were not likely experiencing fatigue during the incident flight.

Verbatim NTSB analysis and probable cause from the NTSB dataset