French investigators have highlighted an extraordinary lack of communication which allowed Air France flight AF447's initial excessive deviation from its assigned altitude to go virtually unchallenged.

While much attention has focused on the failure to rescue the aircraft from its subsequent fatal descent, the deviation played a crucial role in the accident sequence because it took the aircraft to an altitude that decreased its aerodynamic stability and increased the risk of a stall.

France's Bureau d'Enquetes et d'Analyses has yet to explain fully the reason for the co-pilot's putting the Airbus A330 into a sudden climb, although it followed the disconnection of the autopilot and autothrust, and a slight loss of altitude of some 350ft.

The aircraft had been cruising at 35,000ft, within a layer of turbulence, but was unable to climb above the rough air because of a poor temperature gradient. The aircraft's flight management system had calculated the recommended maximum altitude at 37,500ft.

The co-pilot, who was flying, had already mentioned this altitude limit to the captain, who was preparing to leave the cockpit for a rest break.

When the relief pilot arrived to take the captain's place, the co-pilot also referred to the lack of margin available to take the aircraft up to 37,000ft, saying: "What we have is some [recommended maximum altitude that is] a little too low to get to three seven."

BEA states that the co-pilot mentioned that the A330 was "on the edge of the [turbulent] layer" and added that he would have preferred climbing to 36,000ft.

"Climbing to a higher level was a constant preoccupation for the crew," the BEA said. "The pilots clearly wanted to fly outside of the cloud layer, probably to limit turbulence."

As it continued to fly through the turbulence the A330 encountered the atmospheric icing which obstructed the pitot system, disturbing the speed data and causing the autopilot and autothrust to disengage.

The co-pilot took manual control but BEA says he made "rapid and high-amplitude" movements on the sidestick, which were "almost stop to stop", as well as a nose-up input which increased the A330's pitch to 11° in 10s.

While the autothrust disconnection had resulted in the thrust being locked, at 83% of N1, this automatic locking was deactivated 15s later. But the thrust levers were still positioned in the 'CL' detent, the normal setting for cruise. With the A330 under manual control the engines responded by powering up to 104% of N1.

As the aircraft pitched up, its vertical climb rate increased to 5,200ft/min and then 6,900ft/min, while the pilots discussed the loss of airspeed information. Just 20s after the autopilot disconnection the relief pilot noticed the aircraft was climbing and cautioned his colleague, who responded: "Okay, okay, I'm going back down."

"According to the three [instrument displays] you're going up, so go back down," the relief pilot added, and reiterated this again seconds later, as the aircraft climbed above 37,000ft - the altitude which the crew had previously considered too high for cruise.

"Although the [recommended maximum flight level] had been a permanent preoccupation before the autopilot disconnection, neither of the two [pilots] made any reference to it," the BEA said.

While the co-pilot made nose-down inputs which helped reduce pitch and vertical climb rate, these nevertheless "still remained excessive" and the A330 continued to climb "without any intervention" from the relief pilot, it added: "At no time did either of the two [pilots] make any callouts on speed, pitch attitude, vertical speed or altitude."

It said the "low level of synergy" between the two might have resulted from "absence of a clear attribution of roles" by the captain before he left the cockpit, and a lack of crew resource management training for the situation.

BEA stated that the relief pilot should have immediately called out the excessive parameters. "The absence of specific training in manual aircraft handling at high altitude likely contributed to the inappropriate piloting inputs and surveillance," it said.

Despite the rapid increase in pitch attitude and altitude, resulting from the flying co-pilot's inputs, the aircraft's trajectory seemed, at this point, to have been controlled. The A330 was just above 37,500ft.

But the co-pilot made further nose-up commands to the side-stick, and the stall alarm started sounding for a continuous 54s period as the A330 continued to climb, reaching over 37,900ft before - just one minute after the autopilot had disengaged - it finally lost lift.

Airbus simulated the A330's longitudinal behaviour for the period of the sudden climb and found that, as the aircraft reached about 37,000ft, there were fluctuations in the angle of attack and attitude which indicated the presence of turbulence.

But the BEA added: "After this time, this turbulence appears to disappear and the parameters simulated [by Airbus] and recorded [from AF447] are highly consistent.

"Consequently it would appear at this stage in the work that the bulk of the aircraft movements in the longitudinal axis result from the actions of the [flying pilot], with the exception of small variations that are probably due to the meteorological disturbances."

Source: Air Transport Intelligence news