Air France 447: the facts and what's behind them

This is what the Air France A330's trajectory looked like during its last few minutes, starting when everything was still fine...

The BEA's simple three-page factual summary is here. It doesn't attempt to judge, it just reports facts. Conclusions will come with the final report.

From this point I will number the paragraphs so I can refer back to information already mentioned. The paragraph numbers are not intended to tally with those on the diagram - it has its own key directly below it.

1. On page 1 the report sets the scene on the flight deck. The captain has gone for a rest, and has been replaced by a supernumerary First Officer. That is standard for long flights.

2. The aircraft is cruising at FL350 (35,000ft) and there are storm clouds in the area, as there always are in the inter-tropical convergence zone. The two copilots are aware of them. The pilot flying (PF) briefs the copilot who has just taken the captain's seat that they are in choppy air but cannot climb above it because the aircraft's weight and the relatively high outside air temperature means that they are about as high as the aircraft can safely go until it has used up more fuel.

3. A few minutes later the PF makes a sidestick control input which raises the nose and causes the aircraft to climb rapidly to 38,000ft. There was no reason to climb, the PF did not announce an intention to do it, and the aircraft was not cleared by ATC to do so. The natural result of climbing without an increase in power is a loss of speed. But we'll deal with that shortly.

4. The problems for the pilots began when the autopilot(AP)/autothrottle (AT) disconnected. The disconnect occurred because there was a temporary disagreement between two independent airspeed sensors (pitot tubes) about what speed the aircraft was flying at. When there is a disagreement between two inputs to the flight control computers, the computers do not adjudicate, they abdicate control to the pilots.

5. The drill for pilots at that point, according to flight manuals for Airbuses and Boeings alike, is to leave the power where it is and to continue to fly straight and level. That way the aircraft's speed remains the same as it was, no matter what the airspeed indicators are showing. When the AP/AT trip out the aircraft is fully in trim and the power stays where it is.

6. The same logic that caused the autopilot disconnect also caused a change to the flight control law, taking it from "normal" to "alternate" law. This robs the pilots of the flight envelope protection that is automatic in "normal" law, but otherwise the aircraft flies in the same way. The main protection they lost was protection against stalling.

7. The PF verbally acknowledged the fact that the AP had tripped out and that alternate law was in force.

8. Meanwhile the BEA report says that the airspeed disparity that caused autopilot/authrottle disconnect, lasted for less than a minute.

9. The crucial moment in this flight came 11s after the autopilot had tripped out. Because of the light turbulence, the aircraft rolled slightly to the right when no longer controlled by the AP, and the PF naturally made a sidestick input to bring the right wing back up. The problem is that, simultaneously he pullled the stick back, pitching the nose up. There is no indication of why he should have done that (see paragraph 3), but a few minutes before this action he had stated that they should not climb (see paragraph 2). He said: "So we've lost the speeds", and then: "alternate law", so I suspect it was a slightly panicky reaction to the airspeed disparity and the AP/AT trip-out, but it was not a logical reaction.

10. The result of the PF's flight control input was a dramatic climb at 7,000ft/min and a similarly dramatic drop in the airspeed because of the climb. The aircraft reached 38,000ft, way beyond the height at which it could have sustained stable flight. Almost immediately upon applying the nose up demand the stall warning sounded twice.

11. When the stall warning first sounded the PF maintained nose-up pitch, but as the speed dropped with increasing height he applied a nose-down input and the rate of climb dropped from 7,000ft/min to 700ft/min. The aircraft reached a maximum height of about 38,000ft, and at that point the angle of attack was 16deg, which is well beyond the stall.

12. The stall warning sounded again when the aircraft began to descend, and the pilots selected maximum power on both engines (TOGA). At that height this would not have had the dramatic effect it would have had at a low altitude. But the PF still maintained nose-up inputs, and the trimmable horizontal stabiliser was gradually trimming further and further nose-up because those inputs were maintained.

13. The horizontal stabiliser eventually stopped at a 13deg nose-up setting, and the report says it stayed in this position for the remainder of the flight. At about the same time the airspeed disparity that had caused the confusion resolved itself, and the airspeed on both ASIs increased to 185kt in response to the high power.

14. The captain re-entered the flightdeck 1min 40s after the AP/AT disconnect, and about that time the stall warning stopped because the recorded speeds become invalid. The BEA explains that this occurs because, when the indicated airspeed drops below 60kt, the angle of attack measurements become unreliable and are rejected.

15. The aircraft by this time was descending through 35,000ft with an angle of attack exceeding 40deg, a nose-up attitude of 15deg or less, and a rate of descent of about 10,000ft/min, and the aircraft was oscillating in roll up to 40deg each way. The pilot's response, says the report, was to pull back the stick to the stops (full nose-up demand) combined with full left roll demand for about 30s.

16. At some point the pilots had taken the throttles out of the TOGA detent and set them to idle thust. The PF declared he had no valid instrument indications. The BEA does not explain this, but it may have been caused by the very high angle of attack.

17. Then the PF made some nose-down inputs, the angle of attack decreased, and the airspeed readings became valid again, causing the stall warning to recur.

18. The BEA observes that whenever the angle of attack readings were valid, they exceeded 35deg, so the aircraft was deeply stalled during its entire descent.

19. The aircraft, at impact with the water, had a nose-up attitude of 16.2deg, it was 5.3deg left wing low, and had a vertical speed of 10,912ft/min.

If you struggle to understand how professional pilots can become so confused, I admit I do too, but history shows it happens sometimes. Read this to understand where today's pilot training might be going wrong.

And if you put the key word 447 into the search window for this blog, you will find that I have listed some examples of previous accidents that contained many of the factors this one does.