Storms? Not really. Evidence is starting to point elsewhere...
The development of a Loss of Control scenario at night in weather with an autopilot disconnect and perplexing/conflicting instrument indications? .... how far is that from an incipient UNRECOVERABLE attitude? The answer is not very far at all. It's measurable in mere seconds, particularly if Mach Crit and/or stall speed intervene to further confuse the issue..... or if the pilot's reaction and initial control response is incorrect (as in: rolling the wrong way).
And that's where the power of surprise and the differing impressions/reactions and actions/disagreement of those seated at the controls comes into play. Once the nose drops, speed increases and the g comes on, the two junior pilots would be quite out of their element and the disorienting dynamics that ensued...totally beyond their experience..... particularly if yaw and or high AoA was to then induce some engine asymmetry to compound the problem. Attitude flying just isn't available "out the window" when in cloud at night, so it's the first priority to "go out the window" (i.e. priority one.... "fly the jet" is fatally disregarded because of the system alert distractions). INITIALLY, following autopilot disconnect, even though the pilot immediately implements manual side-stick control, the ATTITUDE CHANGE CAN BE QUITE INSIDIOUS as the pilots try to concentrate on making sense of the conflicting array of aural and visual alerts and aural alarms that they are suddenly presented with. Low perceptibility roll-rate thresholds are a major cause of loss of control at night.
We could extrapolate further here and comment upon some other imponderables (that are never covered in flight simulator sessions):
a. Cruising in Ci/CS cloud, as the airspeed probes became gradually clogged with ice crystals, overcoming the pitot-heating capability, would the system have opposed that apparent airspeed loss by auto-thrust increments - resulting in the aircraft flying faster than what was displayed? i.e. dangerously accelerating towards a coffin corner encounter with its control compromising compressibility effects?
b. Would the engines, operating at higher thrust at a high cruise altitude, become more vulnerable to compressor stalling (N over root t exceedance) during any yaw asymmetry or high AoA (i.e. whatever happened after autopilot kick-out).
c. Because the three probes were the BA variety and equally affected, there'd be no initial prospect of there being sufficient disagreement between systems to trigger any alert. So much for triple redundancy eh? However, ultimately the trending discrepancy between thrust and airspeed and trim would have triggered a tripping threshold and the autopilot would have clicked out (see d. below). That would possibly have been the FIRST indication to the pilots (otherwise concentrating upon the weather radar display) that they'd suddenly had some type of system malfunction. Just "what" wouldn't be clear and would never be sorted by them, as the situation rapidly deteriorated. At this point the ACARS would've robotically started spewing its ether data, but not in any coherent manner or useful order. There'd be no time for a distress call under this scenario....
d. At this juncture, insufficient attention to airspeed and attitude is a crucial factor in what happens next. The airspeed may have appeared "normal" (or slightly low) but may have actually been 30 or 40 knots faster. Why "slightly low" all of a sudden? At a certain point. when the pitot heat has been overwhelmed by ice crystal accumulation, the rate of clogging increases exponentially. It's the same physical process that allows large hailstones to form. As it falls, the hailstone increases its surface area which permits it to coalesce with even greater amounts of freezing water and thus exponentially increase its size and mass during descent. In other words, all of a sudden the pitot tubes become almost totally clogged and that's likely what took the FMGS parameters into imbalance or quite out of tolerance, precipitating the autopilot trip-out. What's the pilot likely to do at this point. noting the airspeed to be "low"? He increases power (engine compressor stall likelihood increases) and lowers the nose to pick up a safer speed. But if he's already close to Mach Crit, that might be all it takes to put him into that dreaded speed regime.
e. Dreaded? My only experience with it was during a descent from 43,000feet in a trainer. I thought that I'd half-roll and pull-through to get down quickly and back into some circuit practise. "Alt & Comp" flown dual had been quite boring, except for the max rate descent. However in a jet that pitched UP upon encountering compressibility (or Mach Crit), hitting that airframe pecadillo whilst inverted made for a quite eventful ride. Inverted, it kept pitching up (which was actually now DOWN into an inverted lower nose attitude) for the next 25,000 feet of height loss. Quite disconcerting when you're a bit bereft about what to do next and simultaneously encounter roll reversal. Luckily you run out of Mach eventually at the lower levels. But if the AF A330 had encountered Mach Crit, penetrating it deeply with a high power set, how would the pilots have coped with the ensuing pitch-up? (assuming that jet pitches up and not down). And what was the longitudinal pitch-trim state anyway - once the autopilot had disconnected?
f. How does the A330's system design compensate in longitudinal pitch trim in such a spurious airspeed circumstance? Whilst on autopilot, does the THS (hoz stabilizers) move and the elevators oppose and hold the (nose up or down?) resultant trim forces? Would the aircraft have been in trim when the autopilot self-disconnected? Or would it have been trimmed for a much slower speed and therefore pitched UP/down upon disconnect? I don't know, I'm just posing the question. In the unfathomable world of malfunctioning flight-control automation, nothing would surprise me. But I wouldn't be the first pilot to disconnect an autopilot and be stunned by what forces it had been holding due to an unalerted system trip (Varicam C/B).
g. So assuming the above scenario has more or less "nailed it" as far as pitot-related developments go, what may have happened next? As said (or inferred) at the outset (above) once you lose it in roll and bury the nose and start pulling g, you end up in a self-sustaining spiral that can be destructive. Clean jets accelerate so fast once the nose is below the horizon. However, given the concentration of the sea-floor debris and the damage analysis of the impact attitude, I'm persuaded that a pitch-up/stall/spin entry and high-rate descent would've been the AF447 follow-through to its high level LOC. As the nose pitched up, if one engine had stalled or flamed out (and especially if the other thrust lever was not immediately idled) a spin entry would've been de rigeur (as the French say). Recoverable? Not really. Think of the vertical spin axis and the resulting centrifugal forces in the cockpit. Even if they hadn't been totally disoriented, there'd have been precious little by way of experience or instrumentation upon which to determine, select and hold the control inputs required for possible recovery. Large B/A ratios in a multi-engine high aspect ratio spin require spin recovery control positions to be set and held for quite a period in order for the yaw/pitch/roll coupling to be effectively countered. We're talking in excess of a minute here. They'd not have been "a propos" that specialist technique.
The lesson for manufacturers and operators [and pilots in particular] is that once a system defect becomes apparent across a certain model (A340/A330 in this case), investigate and extrapolate it into worst-case scenarios and then take the pessimist's course of action. Take the ample precedents as a fortuitous "heads up" threat to safety and just fix it; don't sit on your hands and budget for future modification action or interim alert crews with underwhelming safety bulletins. The Silent Voices from the Tombs always mouth the same words: "Lip-service".
Would I blame the pilots or the weather? Not really, they were set up - as were all A340/A330 crews and pax. AF447 was just the unfortunate first crew to thread the needle.