As I understand, the F/O needed this autoland to complete his type rating. So that would be why they performed an autoland.
Another thing: the RADIO-atlimeter only presents a value to the automated systems (autothrottle, auto callout, if equiped). The (pressure) altimeter did not jump values, so the readout in the cockpit would remain 1950 ft minus about 750 foot per minute (acc. Google Earth path) on both screens. Just as the pilots would have expected to see. There are 2 radio-altimeters on board, of which only the left one provides values for the autothrottle system.
So if you (as a pilot) are observing what all happens or not during autoland, and you see all systems trying to correct an initially too high approach, which then results in a slightly too low approach, when do you start to intervene with the autoland procedure? Right, when you realise that the system is not going to recover the too low approach. By that time, the stall warnings just about kick in!
This is exactly what seems to have happened, if I read the evedence well.
Sadly, when the crew intervened, it was already too late to spool up the turbines and create enough thrust to regain speed and altitude. So all that remained was to pitch up, and hope to delay contact with the gound as long as possible. Thus bleeding off as much forward speed as possible before impact.
There are a couple of technical opportunities for improvement I can think of to help prevent this from happening in the future:
- a clearer warning that the radio-altimeter gives an abnormal value (calculated from value difference over last second, or percentage difference between the two readouts). Something about "landing gear needs to go down" does not directly associate with a wrong altitude readout of a gauge you don't see the readout of...
- average the readout of the two radio-altimeters to come to a value to present to the autothrottle logic
- have three radio-altimeters, and have some logic decide or calculate a value
Last two are common in redundant systems design, so I was a bit surprised to read that the AT system only got values form one of two sensors. I'm sure Boeing engineers will propose such an improvement in the near future.
With a clearer warning, chances are that a crew will pick up on it. It might (and now I start speculating here) be that on the previous occasions when this happened, the crews did not report it in the flight log (because they did not realise they were issued a warinig of such severity?), or that the airline did not pick up on those reports with corrective maintanance, or the jumping readout could be a structural design issue. Any three of these would unavoidably lead to a similar event happening at some time.
I don't think there should be anyone blaming anyone for this accident to have happened. It's just a sad culmination of factors that lead to an unfortunate event. That pretty much defines an accident.
My thoughts and sympathies are with those that lost their lives, their loved ones, and those that got injured (physically and/ or mentally).