Professional Pilots on the San Francisco Crash

On the "FLCH trap" and other possibilities


Let's say it again: after an airline disaster, it often takes a very long time to figure out exactly what went wrong. 

In part that is because in most of the world airline travel is now so extraordinarily safe. Unlike car or motorcycle crashes, or fatalities with small amateur-flown airplanes, first-world airline crashes are now so rare that they usually don't fit a known pattern. Thus the most famous disasters often turn out to be case-of-one tragic instances of many things going wrong simultaneously, in previously unforeseen ways. Patrick Smith's Cockpit Confidential, which I mentioned yesterday, includes a dramatic analysis of the worst-ever airline accident, the collision of two fully loaded 747s on the island of Tenerife in 1977, which stresses the everything-wrong-at-once origin of the crash. (You can also find his account of the crash here.) Also, as evidence comes in over the months and years, hypotheses about a crash's origin can change -- for instance in the much-discussed case of the doomed Air France flight 447.

So: it may take a while to know for sure about Asiana flight 214, though probably less time than in most other cases, and with a higher degree of eventual certainty, because the flight crew (along with most passengers) is fortunately still alive. As we wait to hear what the pilots did, saw, and thought, what's a reasonable range of analyses to consider, without just rolling around in the speculative mire? I've heard from a number of professional pilots, who work from these (apparently) known facts:

  • The airplane undoubtedly "landed short" of the runway;
  • The skies were clear, the winds were light, and there appeared to be no challenging weather conditions;
  • There is no report so far of wind-shear circumstances that might have suddenly pushed the airplane off its intended glide path;
  • There is no report or evidence so far of engine failure or other mechanical/fuel-flow problems that might have kept the airplane from reaching the runway;
  • The Air Traffic Control tapes released so far are full of "emergency!" "help on the way!" discussions after the plane's impact, but there appears to be no mention of problems before the plane crashed. And, significantly:
  • The glide-path indicators for the "Instrument Landing System" on runway 28L, where the plane was headed, had been out of service since June*, because of construction at the airport. This meant that the pilots would not have had a dashboard-instrument indication of the proper path to follow through their descent; they would have to judge with their eyes. In theory that shouldn't matter; every pilot everywhere originally learns to land a plane on a "visual" basis. But it's a complication;
  • And as an extra complication, runway 28L's PAPI system -- the Precision Approach Path Indicators, a set of red and white lights that indicate whether a plane is too high or low on its final approach -- was also out of service. [Update: conflicting reports here. The NOTAM saying that PAPI is out of service was issued after crash, presumably because the impact itself damaged the lights. Other reports said they were not in service. I don't know at the moment and will clarify when I do.] the In normal circumstances, a crew like this would have the ILS glide-path indicator, and the PAPI lights, and their own eyes all as redundant guides for the right path of descent. It seems as if these pilots had eyes only.  [See update above.][Update-update: The PAPI lights were working. See this release from the FAA.]
Thus we have the hypotheses, which I am sharing because they illustrate the way informed observers weigh the (as yet incomplete) evidence at hand. First, from a reader whose name and background I know, and who has been an Alaska bush pilot and airline pilot with some 23,000 hours of flight experience [a lot]:
I just learned the COMBINED flight time of the three [Asiana] captains in the 777 is about 10,000 hours, according to the airline. [Update: Apart from this reader's tone, the number turns out to be wrong. The 10,000-hour figure applied to some of the individual pilots in this flight, not combined.] 

I have more time than that asleep in flight...

They could not estimate a visual approach unaided.

They got low.

They tried to spool up [suddenly increase engine power] but didn't have time. 777 takes 5 seconds to power up, another 5 seconds to correct the sink rate.

They rotated [raised the nose] to stop the sink and got very tail low.

The rest is obvious.
Again: it's possible that this is unfair or mistaken, but the reader has standing to speak and lays out his reasoning. 

If you really are interested in details, you could check out PPRuNe. This is a site for professional pilots, whose postings anyone can read. For an example of the nitty-gritty of the discussion, consider this post
The 777 can catch you out with with what is known as the "FLCH trap."

When you are above the glide slope and need to get down in a hurry Flight Level Change (FLCH) is a useful mode to use. Normally you transfer to another mode like glideslope or vertical speed, or you switch off the flight directors.

However in this situation the glideslope was off the air so the ILS would not have ben selected or armed. If the flight directors were left on and the plane was descending at a high rate in FLCH the autothrottle would have been inhibited and would not have put on power so the thrust levers would have stayed at idle. 

If the Asiana was a bit high (quite normal for SFO) then regained the visual glideslope, the rate of descent would have decreased and the speed would have started slowly reducing but with the thrust levers staying at idle the 777 would now be in the same situation as the Turkish 737 at AMS, ie speed decreasing below Vref and not being noticed.

The 777 has autothrottle wake up, ie when the aircraft approaches a stall the power comes on automatically to almost full power. This gives pilots great confidence however autothrottle wake up is inhibited in FLCH. 

So 777 pilots will be looking at this scenario and wondering if Asiana were in FLCH with flight directors on, too high, stabilised late and did not notice they were still in FLCH and that the autothrottle was not keeping the speed to Vref plus 5 untl too late.

Just a theory but I think it far more likely than engine failure, radalt failure or autothrottle failure and I suspect when the events are unravelled this will be what has happend.
Let's hear from another PPRuNe pilot, on a compounding factor. I won't explain all the references, but essentially he is explaining why many things could have gone wrong at once. "Lining up the holes" alludes to the image of redundant aviation safety measures as a series of slices of Swiss cheese. Each of them has its weak points, or holes -- but if you lay down enough slices, odds are the holes won't all line up and you'll still have some coverage everywhere:
SFO and their notorious ATC instructed 'slam dunk' visual approaches [in which the planes are kept high, then ordered to descend quickly before landing] from downwind have resulted in so many incidents at our airline that it is a regular item in recurent simulator training. 

Throw in the lack of visual or electronic glideslope guidance and the holes are lined up. True, you can set up an LNAV/VNAV profile but this requires a bit of heads down time in the box at a busy phase, not easy unless you are expecting the manouvre.
And from another pilot, in response to the "slam dunk" point:
Well said - I couldn't agree more. 

I would agree that as a professional pilot we should of course be more than capable of rising to the challenge of such an ATC imposed "slam dunk" approach - however please consider these factors that could all conspire to affect pilot performance

>a long 10 to 12 hour flight
>middle of the night body clock time
>to an airport that you may be not so familiar with (long haul pilot roster - you may only visit the destination once in two or three years)
>a slam dunk procedure that would be a challenge at the best of times (I bet even the short-haul/ domestic colleagues get it "not quite right" on occasions.

For what it's worth - I am of the opinion that slam dunk approaches for "Heavy" jets like the B777 these have no place at a major international airports.

In a "heavy" jet it's always (in my experience) a challenge to "get down & slow down" and become stabilised on this particular approach at SFO - something that sometimes ATC fail to appreciate.

Throw into the mix this runway allegedly not having any functioning ILS or even visual vertical reference guidance system - then it all adds to the possibility of "an accident waiting to happen."
You can find more at the site if you're interested. Finally, I see just now that Patrick Smith has put up his own analysis, which properly stresses (a) the rarity of airline accidents in general, (b) the likelihood that early speculation will turn out to be wrong, (c) the professionalism, bravery, and skill of the cabin crew in getting so many passengers safely off a burning plane, but also (d) this:
Looking at some of the footage [eg photo at top of this post], I was appalled by the number of passengers who chose to evacuate the burning aircraft *with their carry-on luggage.* We've seen this in several on-the-runway evacuations in recent years. Lugging your carry-ons down the aisle in the middle of an emergency, when a few seconds can mean the difference between life and death, is unpardonably reckless.
Whatever else might prove mistaken in the analyses above, this last point certainly rings true.
* Here is the text of the "NOTAM," or Notice to Airmen, announcing the limited ILS status. The opaqueness of the terminology is unfortunately typical of the Telex-era legacy coding of aviation announcements, but professional pilots would know what it means. In essence it says that at SFO airport the ILS glide path would be OTS WEF -- "out of service with effect from" June 1, 2013:
"SFO 06/005 SFO NAV ILS RWY 28L GP OTS WEF 1306011400-1308222359
CREATED: 01 Jun 2013 13:40:00