The June 2003 fatal crash of a Bombardier CRJ100 operated by Brit Air (a subsidiary of Air France) near Brest airport in France, was caused mainly by the pilots’ forgetting to select the autopilot approach mode (appr) when they began their approach, according to the final report of the Bureau d’Enquêtes et d’Analyses (BEA). The pilot was killed and five of the other 23 occupants of F-GRJS were injured in the accident.

The pilots detected the omission only partially, the BEA said, focusing successively on vertical and horizontal navigation. They continued the unstabilized approach until the airplane reached  decision height. They decided to go around, but the airplane landed in a field, hit obstacles and caught fire.

Brest Guipavas Airport was under low-visibility procedures at 11:44 p.m., local time, when flight AF5672, inbound from Nantes, was about to begin its approach. Weather conditions allowed an ILS Category 1 approach. The air traffic controller first asked the crew to prepare for a holding pattern, and just before entry into the pattern, she cleared AF5672 for the approach. The BEA lists the controller’s change of strategy as a contributing factor in the accident.

Autopilot Approach Mode Not Activated
While the pilots completed two steps that are prerequisites for activating the appr mode, they did not activate the autopilot approach mode, which would have captured the ILS localizer and, later, its glideslope. “There was no call for pushing the button or displaying the mode on the FMA [flight mode annunciator],” the report stated, offering two possible explanations. The captain (who was the pilot flying) might have forgotten to activate the mode because of his workload. Or the pilot’s swift action on the button could have been too quick for the system to recognize it. The copilot did not check whether the appr mode was armed. He might have been puzzled by the repetition of messages, the report suggested.

In heading mode, the CRJ started to drift to the left because of progressively changing wind. At 11:48:51, it departed the localizer capture zone. The aircraft then flew through the glideslope and its flightpath became too high. The pilots did not notice that the airplane had flown through the glideslope because of their workload, which involved several exchanges with the controller. The controller could not spot the aircraft on her screen, but she thought AF5672 was in airspace where loss of radar contact is normal.

At 11:50 the pilot realized that the aircraft was above the glideslope. He tried several times to recapture the glideslope in both vertical-speed and altitude-capture modes. After 45 seconds, the crew put the CRJ back on the glideslope.

However, during those 45 seconds, they focused only on the glideslope at the expense of their horizontal situational awareness. According to the report, flight-director readouts might have indicated that they were still on the localizer. The flight director was in heading-vertical speed mode and was thus centered.

At 11:50:45, with the aircraft back on the glideslope, the pilot called “approach is selected loc and glide.” He armed the appr mode, but since the aircraft was outside the localizer capture zone, the mode did not activate. At this moment, the controller was busy with another aircraft on the ground. The CRJ flew below the glideslope until it crashed.

Between 11:50:52 and 11:51:14, the crew realized the aircraft was not on the localizer and focused only on the lateral flightpath, giving insufficient attention to the vertical dimension. For example, their response to GPWS “glideslope” and “sink rate” alarms was deemed inadequate in the report. When the airplane was at 330 feet, the crew disconnected the autopilot and the pilot changed the trim angle from -5 degrees to +0.6 degrees. The thrust setting was inadequate, and the crew increased it from 26 percent to just 45 percent N1 instead of the required 65 percent, the report stated. The CRJ’s airspeed therefore decreased from 130 knots at 400 feet to 120 knots at 100 feet–Vref was 132 knots.

Insufficient Pitch Input For Emergency Climb
At 11:51:16, the pilot called “go around,” likely thinking he had just reached the decision height of 200 feet. But the CRJ was far from the localizer and it was actually just 93 feet above the ground. The go-around sequence began just after the GPWS alarm “100.” But the pilot’s pitch-up input was insufficient to establish the aircraft in a climb. The first sound of impact was heard at 11:51:22.

In addition, the report stated, “It is difficult to know precisely what happened in the cockpit.” Parts of the copilot’s interview contradict what can be understood from the data recording, particularly as it relates to the crew’s actions on the thrust levers. Shortly after the crash, the copilot had even stated that the pilot was motionless, both hands on the stick and staring at the instrument panel during the last seconds before touchdown. But the autopsy of the captain did not reveal any pre-impact anomaly that might have caused incapacitation.

The BEA also cited limited communication inside the cockpit as a factor in
the accident. For example, the briefing before the approach was too short and overlooked several items. The pilot did not clearly announce his actions, and the copilot did not ask for more detailed explanations. During the approach, the report said, the two pilots were not working as a team; rather, each was focused on his own job. When they appeared to focus on the same task, they had no clearly defined plan.

Copilot-Sim Instructor “Confident” in PIC
Stress (the flight was 50 minutes late), fatigue and routine may explain the lack of communication. Another factor may have been the copilot’s role as a simulator instructor. During his post-accident interview with BEA investigators, he insisted that he did not want this function to interfere with the flight and was confident about the skills of the pilot, whom he had trained in January 2003.

Brest airport did not have a minimum safe altitude warning (MSAW) system, but a simulation showed that such a system would not have prevented the accident. The controller would have received the MSAW alarm 16 to 24 seconds before the airplane crashed. “With the needed transmission times, the alarm would probably not have helped the crew to perform an earlier go-around,” the investigators concluded.

Another system that the investigators simulated was the EGPWS–the CRJ had only a GPWS. “The EGPWS would have generated a ‘too low terrain’ alarm 15 seconds before [impact],” they discovered. The report estimated that, in that case, the pilots would probably have started their go-around at a height of 230 feet.

Notable among the BEA’s recommendations to the European Aviation Safety Agency (EASA) is one related to the ILS information layout in the CRJ100 cockpit. “Bombardier and Rockwell Collins have elected to show glide[slope] and localizer on the same display but on two separate [depictions of] instruments,” the investigators pointed out. [The ADI shows the glideslope bug, and the HSI below it includes the localizer CDI.] They suggested that “using an instrument like the head-up guidance system, which regroups both pieces of information, might have helped the crew to detect the localizer non-capture.” They therefore recommended the EASA consider requiring that approach instruments display localizer and glideslope information together.