At about 11:28 a.m. PDT on July 6, Asiana Airlines flight 214 crashed at the approach end of runway 28L at San Francisco International Airport (KSFO). Less than 24 hours after the accident, it’s way too early to know what happened, but there are some signs that the aircraft was not flying a stabilized approach at any part of the approach to land, and ended up low and slow just before impact. Below, I outline the data that’s available at this time, and why I think that’s a possibility. But first, a disclaimer: There’s not nearly enough data to determine the cause of the accident. The below is speculation, based on the data available to me, which is not even a few percent of the data that will be available to the NTSB. Most speculation this early in the process is wrong, precisely because so little data is available. So anything below that may appear to be stated as a fact is really a conjecture.
First, the conditions at KSFO were ideal. The METAR 32 minutes before the accident was
KSFO 061756Z 21006KT 10SM FEW016 18/10 A2982 RMK AO2 SLP097 T01780100 10183 20128 51005
That is, there was a 6 knot crosswind to runway 28L, with good visibility. The winds aloft were light, so wind shear doesn’t appear to be a factor. Twenty seven minutes after the accident, the METAR was
KSFO 061856Z 21007KT 170V240 10SM FEW016 18/10 A2982 RMK AO2 SLP098 T01830100
Still a light crosswind, but with more variable direction. However, given how light the winds were, this would not likely have had a big impact on the landing. So weather doesn’t appear to be a factor. Indeed, the weather was near perfect at the time of the landing.
There are reports that the ILS glideslope was unavailable, but that the PAPI (precision approach path indicator) for runway 28L was functional. There are some conflicting reports that the PAPI was unavailable as well. Aircraft were making visual approaches to 28L and 28R.
Using radar data from FlightAware, I plotted the altitude and airspeed of AAR214, as well as that of UAL852, another 777 which landed successfully only 10 minutes beforeAAR214. Both flights were long distance international flights (UAL852 originated at Heathrow, London), probably both full of people but without much fuel, so both aircraft would have had more or less the same weight on landing, and therefore the same reference approach speed, 
. It’s instructive to compare the landing profile of these two flights. The first plot below is the altitude vs. distance from the touchdown zone for both aircraft. UAL852 is in black, and AAR214 is in red.
Note that UAL852 is on a nearly constant 3.2 deg glideslope. The PAPI for 28L is nominally 2.85 deg, but because a 777 is a large aircraft with a large cockpit to wheel height, it would be typical to fly the approach a tad steeper than the standard glideslope.
On the other hand, AAR214 is 500 feet or so above the glideslope until about 4 nm out. At 3 nm out, the aircraft is quite high, on a 4.48 deg glideslope to the touchdown zone, or about 50% higher than it should have been. From that point, the aircraft descends rapidly, presumably to acquire the correct glideslope. At about 1.5 nm out, the aircraft crosses and then descends through the glideslope.
At the last reliable radar return, the aircraft is at 100 feet, 100 feet below glideslope. Note, however, that the radar returns are quantized to 100 feet, so the result may not be very accurate. Nevertheless, you can see that the descent rate on short final is very high, perhaps twice what would be expected for a stabilized approach.
It gets worse. The plot below is groundspeed for the two aircraft.
Because the winds are light and generally a crosswind, the speeds shown are probably within a few knots of the actual airspeed. Since we don’t know the weights of the aircraft, we don’t know what target approach speed () was. However, you can see that starting about 6 nm from touchdown, UAL852 slows from about 190 knots to about 145 knots. A typical (V_{\text{ref}}) for a 777 loaded with passengers but not much fuel is about 145 knots, so that makes a lot of sense. So UAL852 flew a stabilized approach, on the glideslope from 12 nm out, and slowing to for full flaps about 3 nm (1 minute) out.
On the other hand, AAR214 was never on a stabilized approach. Until about 30 sec before touchdown, it was high and fast. Only 3 miles out, it’s 20 or 25 knots too fast, and 500 feet high. As a result, the pilot no doubt reduced power to intercept the glideslope from above. 1.5 nm out (nominally less than 40 sec from touchdown), he’s finally on glideslope and at , but with a high sink rate on low engine power. If he applied power at that point, the engines would take some time (a few seconds) to spool up, and he would further sink below glide slope, slow down, or both.
The situation can be appreciated more precisely (but more technically) by looking at the total energy of the aircraft, that is, the sum of the potential energy due to altitude plus the kinetic energy due to velocity. The total energy is given by
![\[E=m g h + \frac{1}{2}m v^2\]](http://flyingprofessors.net/wp-content/ql-cache/quicklatex.com-892af72ffcccf01c69bb621721edc7b4_l3.png "Rendered by QuickLaTeX.com")
where 
is the mass of the aircraft, 
is the acceleration due to gravity, 
is the height of the aircraft, and 
is the velocity. Because we don’t know the weight of the aircraft, it’s convenient to normalize the energy by 
, yielding the energy height
![\[h_E = h + \frac{v^2}{2g}\]](http://flyingprofessors.net/wp-content/ql-cache/quicklatex.com-c5428df3a4e67dd00d09ac71ba39e69a_l3.png "Rendered by QuickLaTeX.com")
The plot below compares the energy height for the two aircraft:
Note that the energy of UAL852 decreases at a steady rate until about 6 nm out, where the rate of energy dissipation increases, because the aircraft is slowing. At about 3.5 nm out, the rate decreases, because the aircraft has hit its target approach speed and stops slowing down.
AAR214 has a much different trajectory. At about 3 nm out, the rate of energy dissipation increases a lot, because the aircraft is both too high and too fast. As a result, the power is reduced significantly, perhaps even to near idle, in order to simultaneously slow the aircraft and get it down to the glideslope. At about 1.5 nm out, it has about the right airspeed and altitude (and therefore energy), but the energy continues to decrease precipitously. If the pilot added enough power at this point, a safe landing might have been possible. But it takes several seconds for the engines to spool up, and the pilot may not have added enough power or done so early enough, so both the altitude and airspeed continue to decrease below their desired values. Indeed, at the last radar return, AAR214 would have been near its stall speed, and unable to pull up.
These data are entirely consistent with the eyewitness observations, which indicate that the aircraft approached steeply, and then tried to pull up when it got too low, but was unable to. It’s also consistent with observations that the engines were powering up just before impact.
Whatever the reason the pilot flew this approach profile, it’s clear that he never had the aircraft established on a stablized approach. FAA Advisory Circular AC 120-71 states that:
An approach is stabilized when all of the following criteria are maintained from 1000 feet HAT [height above touchdown] (or 500 feet HAT in VMC) to landing in the touchdown zone:
- The airplane is on the correct track.
- The airplane is in the proper landing configuration.
- After glide path intercept, or after the FAF, or after the derived fly-off point (per Jeppesen) the pilot flying requires no more than normal bracketing corrections2 to maintain the correct track and desired profile (3° descent angle, nominal) to landing within the touchdown zone. Level-off below 1000 feet HAT is not recommended.
- The airplane speed is within the acceptable range specified in the approved operating manual used by the pilot.
- The rate of descent is no greater than 1000 fpm. If an unexpected, sustained rate of descent greater than 1000 fpm is encountered during the approach, a missed approach should be performed …
- Power setting is appropriate for the landing configuration selected, and is within the permissible power range for approach specified in the approved operating manual used by the pilot.
It appears that there is no point in the approach where the approach is stabilized. Indeed, the rate of descent at 600 ft was 1320 ft/min, well above the allowable descent rate. Standard procedure at most airlines would have required the aircraft to go around at that point. However, it’s not clear that standard practice conforms to standard procedure, and pilots may be reluctant to initiate a go-around at a busy international airport on a clear day.
One other important factor may be that the glideslope signal of the instrument landing system (ILS) was out of service at the time of the accident. It’s standard practice to use the ILS glideslope even when on a visual approach. Indeed, FAR 91.129 (e)(2) requires that, “Each pilot operating a large or turbine-powered airplane approaching to land on a runway served by an instrument approach procedure with vertical guidance, if the airplane is so equipped, must: (i) Operate that airplane at an altitude at or above the glide path between the published final approach fix and the decision altitude (DA), or decision height (DH), as applicable.” Normally pilots would use the ILS glideslope for this purpose. They could have used GPS for vertical guidance (there is an RNAV (GPS) PRM RWY 28L approach), but may not have done so.
It’s far too early to know the cause of the crash of Asiana Airlines flight 214. However, early indications are that it might be due to an unstabilized approach, which is a leading cause of approach and landing accidents. If the cause does turn out to be an unstabilized approach, it will be relatively straightforward for the NTSB to make that determination. The data recorder and cockpit voice recorder will have ample data to determine what decisions the crew made on final approach, what the state of the aircraft was throughout the approach, and what control inputs were applied.
UPDATE:
At about 4:45 EDT today, the NTSB held a press conference, and it appears most of our conjecture is correct. The target approach speed was 137 kt (not 145). Just before impact, the power was indeed at idle, and the airspeed dropped “significantly below 137 knots and we’re not talking about a few knots.” Seven seconds before impact, a pilot called for increased power. At 4 seconds before impact, the stick shaker actuated, indicating incipient stall. At 1.5 seconds, a pilot called for a go-around, much too late, obviously. So it appears (so far) that our analysis is more or less correct.
UPDATE 2:
Below is a screenshot of the last part of the final approach path:
Here is a link to the KML file for those who may want to visualize more of the flight path.
Thank you for a nice presentation of the facts. As a commercial pilot, I get frustrated with news media attempts to explain what happened… getting it wrong all the while. Your information is clear and well explained. Thank you.
Very nice work, particularly with the lack of background in transport jet flying. Just one technical detail to share: “because a 777 is a large aircraft with a large cockpit to wheel height, it would be typical to fly the approach a tad steeper than the standard glideslope.” I think what you meant to say is that it APPEARS to be high on glideslope when you’re looking out of the cockpit windshield of a jumbo jet; the actual glidepath is (or should be, anyway) the same for any aircraft using that runway’s glideslope or papi.
High-energy approaches have always been tolerated more in the USA than in other countries, partly because of our “let the captain fly his damn plane” mentality of allowing for technique and judgement in unusual situations. In Europe, a high-energy approach is considered a serious matter even if it all turns out ok.
In this case, it does indeed appear that the crew conducted a high-energy approach thinking that they would get it all worked out a few hundred feet above touchdown, and somewhere along the way, by miscalculation, ended up lower and slower than they wanted to be and didn’t have time to spool-up the engines for a recovery or go around. Usually when a high-energy approach results in an accident, it’s not like this but the opposite: see Southwest at Burbank, Southwest at Chicago Midway, and American Airlines at Kingston, Jamaica. All ended up off the other end of the runway after touching down long and fast, not smacking into the breakwater BEFORE the runway, out of airspeed and power. That’s what makes this one special.
Thanks to accidents like these, it is now becoming standard practice in the industry to have the entire contents of the Digital Flight Data Recorder uploaded via satellite every 24 hours to analyze flight profiles, configuration, everything. The pilot then gets a nastygram from management if he flies an approach a little high or fast or disregards a cockpit warning. If this trend continues, the day will come that we aren’t even allowed to land the plane manually except in an emergency.
Thanks for the kind words and additional insight. One small nit: You say
My understanding is that large aircraft fly a PAPI with three white/one red instead of two white/two red, which as you note, puts the wheels of the aircraft at the same height over the threshold. But flying 3W/1R puts the eyes of the pilot 0.25 degree higher, all the way out the approach. That is, flying 3W/1R puts the aircraft cockpit a little higher over the threshold, but significantly higher far out on the approach. The nose of the aircraft is flying a 3.25 deg glideslope, and therefore so is the rest of the plane.
Again, thanks.
Flying a 3/1 PAPI is incorrect procedure, probably left over from VASI days.
From FAA docs: “The glideslope of the PAPI
must provide the proper TCH for the most demanding aircraft height group using the runway.”
All aircraft should use a 2/2 PAPI approach unless they are operating at an airport which normally doesn’t handle that (large) aircraft type.
“If this trend continues, the day will come that we aren’t even allowed to land the plane manually except in an emergency.”
Isn’t this mentality, in part, what contributes to this problem? I know that flying a 777 is a lot more complicated than flying a F-33 Bonanza, but at some point, looking out the window and flying a visual approach is a skill that you have to have if you’re in the pilot’s seat. At some point, doesn’t the over-reliance of flight directors, auto pilots, auto throttle, auto landings, etc. degrade the flying ability of air crews?
You’ve got at least three pilots, all sitting up front, and they don’t realize that they’re too low and slow in time to go around? How can that happen to guys that are supposed to know how to fly? This isn’t somebody bombing around the local airport in a C-182 with a new ticket and 75 hours TT?
A guy in a C-182 and 75 hrs is probably newly licensed or in training and soloing. But I am sure this pilot knows what the correct sight picture is for a landing that is going to be a greaser. Usually for most pvt pilots, the “best” they are is right after licensing, unless they go on for other ratings.
I too am amazed that the descent was never stabilized, especially for a plane as big as this one. Someone forgot the basic principles of inertia and they got way behind this airplane.
Yep – more manual flying is the answer, not less. Aside from the obvious, this Asiana one is looking like a CRM issue to me.
The 777 flies the exact same glide path as all other airliners.
Scott
777 pilot
Thanks for the quick note and id’ing yourself as a 777 pilot! It seems like a little but it said a lot.
Hi Scott,
As far as “throttle response” and “time to spool up” with the Pratt (?) engines in the 777, could you elaborate a little on how quickly a half, 3/4 or full power TOGA might have arrested the 800-900 fpt sink rate they carried into the last half mile final (from FlightAware)? Am left wondering at what point, beyond which, the 777 engines would not have been a solution given sink rate, airspeed and altitude. I’d read those engines are “fairly quick” to “spool” at 3-4 seconds, but I don’t know how that equates to “actual thrust” (as in a possible lag after full spool to full power?) Have only flown a little SEL. Thank you.
Thank you for the physics lecture. Your charts are quite telling. Only now have I fully considered the amount of lifting force (energy) needed to check the high rate of descent (literally falling) of such a large object, which must be achieved prior to a safe landing. It appears as though the aircraft didn’t have sufficient airspeed to generate enough lift to slow the high rate of descent. IMHO, the pilot’s ego kept him from aborting.
Best example of that one is the well reviewed video of the Thunderbird ejection / crash, in which the airplane came out of a loop with too much downward momentum. Even at full thrust and correct attitude, it was not enough to recover.
Same thing likely happened here (no loop but too much downward momentum). Flightaware had a rather scary reading of 87 kts at 200ft. Way too slow for a large ship like the 777.
I believe the 87 kt reading occurred after impact. The location indicated is actually beyond the final resting point of the fuselage.
This is the kind of information I’d like to see on TV. Very informative! Thanks for the clear info.
Very nice and insightful explanation – I wonder if you had this nailed before NTSB could get the “go team” off the ground. 🙂
Keep in mind that being brief can cause an appearance of arrogance or impoliteness. Neither is the case with my response.
I’m a pilot with limited experience. My very 1st thought when I heard the nature of the crash was to low & too slow. This would cause a nose high/tail low attitude. Loss of control due to stall and impact. I would guess every single pilot out there thought the same thing right off the bat. Landing 101. We all learned it the first time we were allowed to take controls during training. I know it and I’m sure the NTSB knew it right away. They are still going to investigate and essentially keep their mouths shut.
That was my first thought, especially after seeing the video. I just recently made my first solo going for PPL so I’ve been doing a lot of touch and go’s lately.
I would also add that un-stabilized approaches are not tolerated at my airline. The Asian carriers in my opinion are much more likely to have Capt’s with god syndrome.
I don’t want to sound stupid but I only know how I would handle a bad approach. When you say “not tolerated” does that mean that the expectation would be that a go around decision would have been made many, many seconds before this pilot seems to have?
Hey guys,
A very informative post !
As a B777 Captain for a major Asian carrier we have very strict stabilised approach criteria. If not stabilised completely by 1000 ft AAL we go-around, no questions asked.
KSFO is always a difficult airport with numerous types of approaches and congested ATC. These arrivals are made so much more difficult when landing after 15 hours in the air, and when your body clock is on your home time zone.
One thing that I don’t yet understand is why the auto thrust did not power up to maintain speed. Even if disarmed, it has an auto wakeup function that should work in any modes that they would fly an approach in.
But maybe it was turned off altogether for the visual approach.
Anyway, I’ll have plenty of time to ponder that point on my way to….KSFO…tonight.
Regards.
A a B777 pilot, can you fill us in on the workings of the Auto throttle in a high energy approach?
Would the AT need to be fully disabled so that engines could be idled during approach?
From your flying experience, could the fully VFR approach procedure have them miss the checklist to enable auto throttle?
And before we throw these experienced pilots under the bus (as everyone seems to be doing), what could have failed that is not very obvious on the first pass of the FDR?
Cheers all!
777 has an autothrottle “wakeup” feature that will bring the throttles almost to max near the stall.
However….
As one pilot very brilliantly puts it, there is something called the “FLCH Trap.” FLCH (flight level change) autopilot mode with FD on will actually disable the wakeup feature. See what he wrote here.
FLCH is a very good mode for changing flight levels, at higher altitudes, but it is not an approach mode.
However, when being vectored for an approach and the aircraft is too high, an easy way to recover is to set a lower altitude, select FLCH, speed brake etc. this give a high rate of descent, and as long as the pilot has the G/S armed the aircraft will capture it. The G/S acts as a safety net.
In this case, the aircraft did not level out, and we know the GS was not active. I have seen several times in this situation where the ALT can be inadvertently set to zero. In that case, the auto throttle will not wake up – as it is performing exactly as directed. If zero is set in the ALT window, with FLCH selected, there is no need for any thrust.
FLCH is what is referred to as a speed on elevator mode. The pilot sets the airspeed, and the elevators raise or lower the nose to maintain that airspeed. However if the flight directors are not followed, for instance if the nose was raised, the speed would decrease.
Very nice analysis indeed, as factual as can with the little info available. In regards to the autothrottle wake up function as mentioned before by Nev : It is unfortunate, but there is a possible situation where it doesnt work. With the autothrottle disarmed, the auto wake up of the autothrottle will indeed function. In SPD or thrust, same thing. However, when doing a visual approach, there is a reasonable chance the FMA told them the autothrottle was in HOLD, and in that state, the auto wake up of the autothrottle system will not come on. It is disabled in that case. Bit of a catch, been told by numerous trainers, tried it on the SIM, and it will not come on when autothrottle system is in HOLD. Only a first officer on the 777, prone to mistakes like anybody else, but fairly sure of this.
Hi Paul, I don’t think the wakeup function will work in FLCH or TOGA mode. That said, I haven’t seen or tried it. Nev.
Exactly! I too have been wondering the same thing. It is amazing with the amount of automation in a ship such as a “trip seven” there was not any “machine intervention” that would have alerted these vets with tens of thousands of hours logged to make such a rookie mistake.
Thanks for the Physics breakdown , as a Physics/Science student I like to tell my friends that most things can be explained through physics and/or chemistry. Your analysis seems correct and corresponds with the recent NTSB news conference report. Your analysis does add the important information of what the plane was doing during its landing decent.
I also agree with above posters about these “high energy approaches” and that such approaches should be avoided in non-emergency situations. It does seem if high energy approaches were restricted then there would be a reduction in crash landings , such as this AAA214 and other crashes where the planes run off the runway.
When I see accidents like this I always ask the question what automatic override actions could the plane have taken in the last critical time span to avoid the accident. The 777 was aware of the incipient stall speed. It was also probably aware of the momentum caused by the steeper glide path. Finally in crash landing of planes, there is a tendency to pancake. Would a shearing structural design such as those we have on high quality automobiles on unsymmetrical front impacts lessened the damage?
I don’t think Shearing structural design – “crumple zones” would help Pedro, as the aircraft in question hit tail first and a wing lifted significantly, engines ripped off their mounts – you can see that in the video and photos on MSM ( Main stream Media). You also see that the fuselage is significantly “oval- shaped ” at the tail, and more or less still cylindrical and with less damage at the nose from initial impact, it looks like more damage was done to the nose from friction, as it looks like it was trying to curl under the fuselage due to ground friction. This is noted by the bowing of the fuselage window lines in front of the wings all the way to the nose. Beyond that observation Aluminum reacts significantly different from Steel under the same forces, it is much more malleable, and aircraft are less predictable than cars, where impact points can be from any direction in a crash. Most cars are designed to crumple in the front and rear only, with strengthening bars put in the doors.
Considering the high vertical impact velocity, the subsequent skidding & plowing, and the way the aft end came up and around at the end and dropped from perhaps 60 ft, it is hard to fault the airframe’s crash-absorbing capabilities. That is a rugged airframe.
A little known fact about about the crash behavior of aluminum is that, even under very high crash loads that greatly exceed the metal’s yield point, the metal grains do not have time to permanently deform (yield) before the load is gone. It’s behavior remains almost completely elastic because the overload is so short. So even if a part of a fuselage gets momentarily crumpled into a fraction of its original size, it will immediately spring back almost to its original shape, with little evidence of what it has been through. Controlled crash tests by NASA with high speed cameras have documented this fact.
This makes it difficult to look at a post-crash airframe and know how much punishment it took, so we need to be cautious in interpreting what we see there.
It did have a shearing design which likely did lessen the damage. The gear and engines are both designed to shear off before they can rip the wing and wing fuel tanks apart. They seem to have fulfilled that task admirably. Both main gear and engines separated from the plane yet the wings (and, critically, the wing box) stayed intact.
To your first question, the ability to monitor airplane total energy and report it to the pilot has long been available, though I’m not sure of its use in transport aircraft. The display is typically looks like a glideslope or fast/slow indicator, with a vertical bar and pointer which centers in the middle of it. Even if such a display weren’t presented to the pilot, it would likely not be technically difficult to incorporate an EGPWS warning based off it. I could envision a future warning saying “Too Low, Terrain…. Power Power”.
At one of the news conferences yesterday, the NTSB chair said “inside the aircraft there’s significant structural damage.” She was explaining that the plane held together enough to make the crash survivable.
It doesn’t seem like there is much room on an airframe to put crumple zones. Probably better to keep it strong and work on better passenger restraints.
Yeah there is some crumple built in but mainly via the wings, gear etc. The seatbelts have some give in them but i’m not sure about the seats. In modern light aircraft the seats will absorb lot of vertical energy – some are rated to 26Gs.
Thank you for the analysis and posting.
If the pilot would have had at least a HUD with an FPV, he would have apparently seen it lined up with the “-5” degree chevrons most of the way down the chute, should the data provided above be correct. This kind of approach is fine for fighter jets with massive excess thrust and available AOA/G margins, but never in a transport.
The current paradigm in ADI displays as well leaves the pilot with only the yellow and red band of an outsourced colorful speed tape and perhaps pitch limit indicators to tell him/her of impending stall – which in the vast majority of transports provides no tactile warning, hence the stall shaker. All three of these indications (airspeed, PLI, shaker) are one dimensional concepts. Meaning, the only information they provide is deviation from nominal.
Try driving down the freeway in a car without a spedometer and see how hard it is to precisely maintain 65 mph based solely on what you can glean from the odometer. Or from aural commands from someone sitting in the passenger side yelling only “slow!” and “fast!” repeatedly in your ear. This is the concept of deviation. And it is a relic of the past with us today simply because the civil aviation community has yet to find a way to incorporate what the military has been successfully tuned into for decades – the concept of displaying “energy state” not via airspeed, but rather AOA.
But regardless of any need for a paradigm shift in display symbology, few things vindicate a crew from chasing a high and fast close-in at the ramp. The acrobatics required to pull that off are graphically displayed above; there is only ONE point of intersection between the flight paths of both 777s. Imagine how hard it would be to show up configured correctly on-speed on-altitude at THAT single point.
Not a wise move. But then again, ALPA has been arguing with the FAA for years about bad vectoring on the LDA approach leading to un stabilized descents…..
Thank you. I am not a pilot, but found your presentation clear and informative. Once more data is available, I hope you have the time to add to your analysis. This is far better than anything I have seen on the news.
The first broken chain of events could have saved them. With the ILS out of services then should have used the 28L GPS approach. That would have placed them on the proper glide slope and would have enabled a full auto pilot controlled landing.
I would bet most comercial operators require coupled approaches even when cleared the visual. With such a large aircraft making large adjustment is unaccetable.
Too bad.
….the only other possibility being an erroneously computed VREF due to incorrect pilot entry of aircraft ZFW and CG, but if the data above is correct, it would seem like the classic case of getting behind the jet due to a host of possible factors, leading to an destabilized push over, followed by the command decision to overcorrect a “high and fast” into a “low and slow”, ending up well into the back side of the power curve, where the only option is to descend (or pull into shaker and risk stall).
Many pilots are criticising pilots/airlines for not wanting to fly visual approaches and depending too much on ILS and here we see an experienced captain training his FO to get some practice and he crashes! Dammed if you do , dammed if you don’t. You can’t win!
An experienced captain training his FO is standard practice. An experienced captain (and FO) failing to bail when the situation gets out of hand is not.
One mystery remains: how did the airspeed get so low? Did the thrust stay at idle? How? The 777 autothrottle and engine combination is very responsive in approach mode and should have maintained the selected airspeed regardless of the aircraft vertical profile. The attempted go around would have been possible if the speed had not been so low. In fact, if the ADS data is correct it almost looks like the correct vertical path was attained briefly around 300′ followed by increased rate of descent and incipient stall.
Bear in mind that Mode C altitude reporting is quantized to the nearest 100 feet, and that the radar returns occur about every 12 seconds. The last three data points are 400 ft, 300 ft, and 100 ft. But that could really mean 400 ft, 251 ft, and 100 ft, a perfectly linear decent. It could even be 449 ft, 251 ft, and 149 ft, a decreasing sink rate rather than an increasing sink rate. You can’t read too much into the sparse data available.
Just a small point which would affect your basic data assumptions:
I think you will find the ‘flightaware’ data is not derived from radar position. It is ”ADS-B” data which is provided by the aircrafts nav and air data systems and transmitted via an extended transponder function to suitable ground stations. The altitudes are referenced to a standard atmosphere so need to be corrected for QNH. You may be correct about altitude rounding. The speeds (I think) are groundspeed not airspeed. Someone else may be able to verify just how accurate or coincident the positions are. Having said all that I think the plot looks pretty accurate.
Hi,
Mode c is indeed in 100 ft increments,but mode S and ADS can ve upto 25 ft increments if the transponder is fed by alt allowing 25ft increments…
Bart.
Sure, but it’s report out on FLightAware quantized to 100 ft. That’s the point I should have made.
So, what you are saying is we need error bars on the charts? Having one line drawn naturally makes your eyes believe it is the true physical reading.
Error bars would be appropriate. What’s really important is understanding the limitations of the data. I’m working on a post on just that issue and the error the New York Times made using the same data.
latest news is that pilot in command had less than 50 hours on type. one can believe a pilot with little 777 experience could fall prey to the “FLCH trap”.
according to the NTSB the engines were at idle during the descent which seems pretty darn dangerous. being in FLCH mode could explain that.
For those who haven’t seen it, here’s a good article talking about what 777 pilots think, including the FLCH trap
I think this could be the reason behind the slow near stall speed , I went to the link on the full article and seems that FCLH trap is the most likely reason given that the glide slope indicstors were out.
I am not a pilot, but I found your information very interesting. I am wondering if it would be possible to annotate one of your graphs (or create a new one) which would show the status of the PAPI lights during the descent. I am assuming that at some point the indication was 0 red/4 white since he was well above the glide slope (or at least 1 red/3 white). Clearly there would have been 2 red/2 white lights when the glide slope was crossed. If the pilot had applied power at that point I am assuming that the landing could have been salvaged. Could the crash have been prevented if power was applied at the 3 red/1 white point (was there enough time for the engines to spool up)? Knowing that his descent rate was high (and the runway was long) I would think the pilot would have begun applying power no later than the 2 red/2 white point to reduce the rate of descent, although there is speculation that the pilot may have expected automation to increase power. However, at the 3 red/1 white point it should have been obvious that manual intervention was necessary. Of course, it is easy to criticize from my computer chair and my intention is not to denigrate the pilot, but to imagine what may have gone wrong and how it could have been prevented…
It’s a little more work than I want to do to annotate it (at least right now), but each zone in the PAPI is (I believe) 0.25 deg wide. That means that for part of the approach, when the aircraft was more than 1.5 deg high, the PAPI would have been 4 white. Of course, well before they hit the sea wall, the PAPI would have been 4 red. The problem is that the resolution of the altitude data is poor. If I’ve calculated correctly, at 1 nm out, a change in one PAPI light (0.25 deg) is only 25 ft. But the resolution of the Mode C altimeter is 100 ft — more than the full width of the PAPI system. So it’s hard to say when recovery was last possible. But even if they had applied power 10 seconds earlier, I think they might have had a chance. But part of the point of the post is that they weren’t in a stabilized approach at 500 ft, and so should have done more than just add power at that point, they should have gone around at that point and tried again.
Thanks for your reply. I see that the resolution of the data may not support what I was requesting. I am now concerned about what is considered a “stable” approach. It seems to me that looking at your data (and the NTSB data) that at 500′ the plane could have been considered to be “in the gate” by the pilots since the airspeed was 134 knots and it had intersected the glide slope. Yes, the descent rate was too high by 100-200 fpm but maybe that wasn’t considered enough to warrant a GA at that point? The problem was (and perhaps this is your point) that although they seemingly passed through the gate at 500′ that was the only gate they passed through. A “stable” approach should require passing through at least two gates to confirm that the trajectory is correct. It may just be unfortunate that they managed to cross the guide slope at almost exactly 500′ which (I understand) is the last gate in the approach where a GA decision is normally made. It may have led them to believe they were on the correct trajectory, but in reality their altitude and airspeed deltas were such that they almost immediately departed from the correct trajectory after the 500′ gate. I suppose this shows the value of the descent rate indicator–this instrument shows more about what is going to happen in the future than the other instruments which only show what is happening right now…
A stabilized approach requires more than hitting a gate in space. It requires that
Arguably, AAR214 failed items 1, 2, 5, and 6. It wasn’t on the correct path because its motion was not along the glidepath. It needed a large change in pitch to acquire the glide path. The sink rate was 1320 fpm at or about glideslope crossing. The power was at idle, which isn’t appropriate.
The real point of a stabilized approach is that the pilots will only have to make small corrections the rest of the way to touchdown. The pilots of AAR 214 still had big corrections to make at 500 ft. They needed to add just the right amount of power, and arrest their sink rate.
Also, compounding, ABC News now reporting that this was the pilot’s first approach at SFO:
https://twitter.com/DavidMuir/status/354076450007154689
I know that that fact has been a big topic on the news and Twitter, along with the pilot’s relative inexperience in the 777, but I’m not sure that’s very important. Yes, he had only 43 hours in 777s (and probably significant simulator time as well), but he had thousands of hours as captain of 747s. He should know how to fly a stabilized approach. How many hours in type did the captain of the first commercial 777 flight have? It might have been close to zero, but with substantial time in a simulator. Many the airline pilots could weigh in on this?
A second pilot with more 777 flying experience was acting as an instructor. According to the NTSB, at 34 seconds before the crash, the aircraft speed was 134 knots. The speed was supposed to be at least 137 knots. It takes at least 5 seconds to power up the engines. So at about 39 seconds before the crash, the instructor should have reminded the pilot to increase power to the engines. But according to the NTSB, the engine power wasn’t increased until 8 seconds prior to the crash. That leaves a 30 second period where neither the pilot nor the instructor realised that there was a problem with the airspeed.
So I agree that the limited 777 flight experience of the pilot is not that big a deal. The pilot acting as instructor had considerably more 777 flight experience, and didn’t catch the problem either.
<>
that’s the scary part.
It turns out that my analysis above is off. The pilots state that they believed that the autothrottle was set to 137 knots. Either they were mistaken about this or the autothrottle malfunctioned.
You are absolutely right.
(MD11 FO)
Thank you for posting a thorough analysis. I’m a student of powerplant lag in boosted race cars and helicopters; I didn’t realize the spool-up in a commercial heavy was that much.
Now that we know the V at seawall impact was 120 knots, I get a one-third G deceleration sliding on the runway for 2300 ft with a spike of a half-G at 8.5 sec after impact when the airframe tilts up to cause braking.
Some early commenters speculated that 214 came in right of centerline based on a streak or debris pattern on 28L’s apron. Could it be that the starboard engine made those marks? Also, there are different interpretations of “spin” and”cartwheel” from witnesses and viewers of the Hayes video. I can’t tell if the plane slid without yawing, or yawed through most of a circle before it came to rest.
Good analysis. A few more factors. It sounds as if one of the pilots was on IOE (initial operating experience) meaning he was new to the aircraft but not to flying into SFO. The charted visual approaches at SFO will eat your lunch if you get behind early.
Great summary and analysis! This is exactly the kind of in-depth reporting that the major news media (CNN for example) should be bringing to the public, rather than hour after hour of essentially on-air babble. I totally understood everything you presented. (I should, given my education in aeronautical engineering and flight mechanics!) Well done.
Nice work, Steve. You put a lot of work into this.
Of course, this was an unstable approach. Any approach that has the tail of your airplane dragging in the water (what appeared to happen, from my perspective, after watching a video) prior to reaching the runway is unstable.
I’ve flown with Korean airline pilots. I’l be blunt. They are not the best pilots. They are VERY good at doing things to near perfection, when it is something they are trained for. When a situation develops that falls outside of what they are trained for, they are ar$es over elbows. Visual (figure it out yourself) approaches are not something that they are well trained for.
I’m not sure what happened, but I’d bet a $1000 to win $1 that it was pilot error. They obviously did not figure it out too well.
How much stigma, and/or employer/FAA censure is attached to a go-around?
Mentioned in passing, in the MSM, was that the pilots at this airline used to be
mostly (all?) former military, but had come down in recent years to
approx. 50/50 former military and civilian-trained. I wonder how that plays into the “Captain with the God syndrome” cockpit scenarios …
This is a good analysis, but I wish you’d pulled the relevant NOTAMs as well:
!SFO 06/005 (KSFO A1056/13) SFO NAV ILS RWY 28L GP OTS WEF 1306011400-1308222359
!SFO 06/004 (KSFO A1053/13) SFO NAV ILS RWY 28R GP OTS WEF 1306011400-1308222359
!SFO 07/046 (KSFO A1326/13) SFO RWY 28L PAPI OTS WEF 1307062219
So, the GS has been out of service for several weeks now, as they install new antennas. But the PAPI was in service, until right about the time of the accident, which suggests to me that they may actually have struck it.
Thanks. I had seen the NOTAMs, but the post was way too long as it is. I concluded from the NOTAMs that the PAPI was working, but destroyed by the crash. From the aerial shots, the aircraft obviously slide across the PAPI, so this makes sense. However, on some forums, airline pilots have been saying that the PAPI on 28L has been in and out of service but not NOTAMed. So I’m still agnostic on this, but I’m sure the NTSB will (or has) figured it out already.
in yesterday’s briefing the NTSB chair was careful to say that the PAPI was working at the time of the crash, and NOTAM’ed immediately after the crash – the airplane apparently hit the PAPI lights and destroyed them.
The PAPI was definitely off the day before the crash. Can’t comment on the day of the crash.
Thanks for the great analysis. I also believe you are right, but no one of you are mentioning lack of engine thrust (engine failure) to be a possible contributing factor? As I see the radar plot this could also have been the case, if they when they were finally stabilized inside distance 2 Nm tried to add thrust but with no response. Therefore similar to the BA 777 at LHR in 2008.
Also I just wanted to give my input on PAPI. The PAPI is usually adjusted to the largest A/C type operating, to cross the thresshold at a given screen height and mainwheels to touchdown 305 meters from threshold. Therefore you can check the GAD charts or PAPI info on landing chart to see the screen height usually in brackets on EAG and Jeppesen plates. As I remember a 3 degrees path to touchdown 305 meters down with a B777 gives a screen height (eye crossing height) of 66 ‘ at threshold.
By the way, what a great website!
Best regards from Scandinavia.
Carsten
This post was written before the NTSB briefing, but now that we’ve heard the briefing, we know that the engines were operating normally. The real question is why the pilots (or autopilot, depending on the flight mode) didn’t apply thrust earlier. Also, why didn’t the pilots recognize that the were slowing precipitously? As in any accident, there’s a chain of events. Every answer brings a new question. The point of my post (before there had been any real data analysis by anyone) was to show that the approach was unstabilized. Unstabilized approaches account for a large percentage on accidents / incidents on landing, some sources say more than half. So understanding that it was unstabilized is key to the next question, Why was it unstabilized? There’s obviously some bad decision making or airmanship, but there could be some human factors / aircraft systems issues as well.
Yes, I got the bit about the PAPI a bit wrong. What I knew that is correct is that at some smaller airports, the PAPI is not designed to accommodate larger aircraft, and there have been accidents where the wheels of a larger aircraft touched down before the threshold as a result. But of course that wouldn’t happen at KSFO.
Steve,
Just FYI, the 137 knot target approach speed the NTSB mentioned is almost certainly the indicated airspeed displayed in the cockpit. (she said it came from the CVR recording) Groundspeed wouldn’t be used or discussed in the cockpit. Airflow over the wings (IAS) is all that matters here. It’s also standard on transport category aircraft to add 5 knots to that target speed as a buffer. (or more if there’s a crosswind or any risk of windshear etc) There was a slight quartering crosswind, so it isn’t inconceivable that the added 10 knots or so to the target. That would put it up in the same general range as the UAL aircraft that was observed at 145 knots GS.
Of course you’re right. Just to be clear, I used groundspeed as a surrogate for airspeed because airspeed isn’t reported on the radar data. And because the METAR at the time indicated generally a crosswind, the correction from groundspeed to airspeed near the surface would be small. But the graph should be labeled with the actual variable used.
Also, I looked at the winds aloft forecast, but by the time I did my analysis, it was the forecast for the next period. I can’t seem to locate the actual winds aloft forecast valid at the flight time. Any pilot out there know how I can do that?
I find it absolutely remarkable what you have been able to put together in short order from publicly available sources. Although, as you have said, this “is not even a few percent of the data that will be available to the NTSB” it is a very useful picture of some of the events leading up to the accident. Of course what was occurring in the cockpit or the heads of the pilots remains speculation, so the why’s of ending up in this unstabilized approach will be of considerable interest, but that we can see this information this quickly and clearly is pretty amazing.
Thank you. It actually didn’t take very long to put the data together, but then I do that for a living. I downloaded data from FlightAware, and did a little numerical work in MATLAB, which is a scientific programming language. What irks me is that the expert talking heads on TV could have looked at the same data in only a few minutes and concluded the same thing:
This is (part of) the data from FlightAware I used. Note that it shows the aircraft descending at 1320 ft/min at only 600 ft. That should be reason enough to go around at that point. At 300 ft, the descent rate has been reduced to 840 ft/min, but the airspeed has dropped to 123 knots, 14 knots below (V_{text{ref}}). Go around! This analysis (including preparing the graphics for the post) took only ten minutes. Why did none of the experts do this simple thing?
I did hear an analyst on CNN reference the FlightAware data, and he came to the same conclusion you did. Seems pretty straightforward.
Good analysis. The accident has much in commom with the Turkish Airline crash in Amsterdam (http://en.wikipedia.org/wiki/Turkish_Airlines_Flight_1951):
-“Training” flight
-Un-stabilized approach
etc.
Interesting and very informative to me, being a captain, the ocean going ships type.
Pilots got their “tech problems”- we got a lot of unqualified people calling themselves “captains” – you should read some of our accident reports.
Cheers from Spain- Claus.
It’s possible that the pilot used spoilers to slow down and forgot about it when he tried to level the plane for the landing. That would explain the high pitch attitude we saw in the amateur video.
This is quite possible. Normally you will get an EICAS message “speedbrakes extended” to warn if you have landing flap and speedbrakes at the same time. It is possible that they did not have landing flap, although not likely.
It would seem that the AF crash over the Atlantic, the commuter crash near Buffalo and this Asiana flight all shared the same feature: the pilot pulled the yoke out rather than pushed it in.
Based on this site and my own flight training, I would guess this is how it went:
At 100’, Asiana was doing 109 knots—well below stall speed. The pilot then yanked the yoke forward, momentarily bringing the plane up to about 200’, but dropping the speed to an incredibly low 85 knots. Had the pilot dropped the nose as he increased thrust, he still had enough altitude to overfly the runway and do a missed approach–but lose face in the process. Other more experienced pilots were in the cockpit, letting him hone some experience. They were probably letting him make his wrong moves, telling him to correct them at the last moment, assuming things would work out.
But with the plane’s nose up and the tail way down, it slapped into the seawall and then the rest of the plane snapped to the ground, skidded around and came to a stop.
See my Update 2 above. It shows the flight path on Google Earth. We’re talking about the last two data points here. Note that the last data point is well beyond the point where the plane came to rest, and 200 feet in the air. Although the aircraft transponder and encoding altimeter might have been working after the crash, I don’t think it can be deemed reliable.
Is the ADS altitude data corrected for QNH? If not it will be about 100′ high.
Be wary of ‘experts’ on Pprune etc. I have around 8000 hrs on the 777 and fly for an Asian carrier. Most of the online opinions and talking heads apply to different aircraft or operations or are just plain wrong.
If, for whatever strange reason, the thrust levers remained at idle, then when the pitch was increased to intercept the correct glide path from above the speed decay would have been rapid. It is possible the Instructor Pilot who was performing monitoring duties momentarily had his attention outside. If disconnected Autothrottle wake up doesn’t occur till just before stick shaker and is inhibited below 50′ anyway. On approach the engine spool up time is only a few seconds as ‘approach idle’ is higher than ‘ground idle’.
Very interesting and detailed explanation. BUT… why did the pilot try to land by himself instead of relying on the auto-pilot ? Isn’t strange ?
Autoland requires the ILS to be fully operational.
Steve, looking at FlightAware it is interesting to note the same flight (214) the day before (7/5) did a go around on the same approach. Of course, I do not know the reason but it is clear by the data they aborted the landing. The day prior to that (7/4) the same flight appears to do another non-stabilized approach with a rapid decent to 800. Arriving high and fast on this approach appears common. None of the prior 777 flights even approached 137 kn. All of them maintained 145 plus.
A few things. First, we don’t know the weight of those aircraft, so their reference speeds could be higher. Second, they sometimes use 28R with a slight tailwind. All we can see from FlightAware is groundspeed, not airspeed. Third, the target approach speed is actually a kew knots (typically 5) above (V_{text{ref}}). It may be higher in gusty conditions. If I get a chance, I might look at the 7/4 data.
Are high energy approaches at high density airports tolerated more in the US compared with Europe to increase the landing rates?
Excellent discussion, worthy speculation. I flew nothing larger than corporate jets & have been out of the cockpit for a couple of decades, so I’m a stranger to things like “autothrottle,” & in my day PAPI was VASI, right? But the charts you posted appear to tell the whole story — too high on final, pulled way back on the power to increase descent rate, overshot the proper glide path, too late pushing the power back up & probably some increased pitch resulting in a stall at the threshold. Not much question it’s going to be pilot error & it will be interesting to see what follows.
I landed at SFO a couple of times in the last four weeks on both occasions the PAPIs were out of service. I am not yet convinced by NTSB claims that PAPIs were functioning but obviously the pilots will be able to confirm. The situation at SFO has been of some concern all summer with the lack of g/s and high feed ins in order to be noise compliant. I can’t explain the lack of reaction to a very slow approach speed and I can’t explain why they were so low. However a contributory factor would be the absence of glide path info. For me this remains a crucial piece of the jigsaw. Four reds will ALWAYS illicit a response by a pilot to a below glidepath situation, I have a nagging feeling that the PAPIs were not available. I hope I am wrong.
I wonder if you could do the same glide analysis for AAR 214 for the previous day 5-July. From the speed/altitude graph presented on FlightAware, it’s slope is more similar to AAR214 from 6-July… That is, until about 500ft before it actually did do a go around.
On July 5, AAR214 went around at 1900 ft. The go-around happened 5 nm from touchdown at 1900 ft, which is between the two cases above. It was also going slower at go-around. I’m guessing that it was a tower-initiated go-around due to a conflict, but I’d have to listen to the tapes to know for sure.
Is it confirmed that AAR214 on the day before also did a go-around? i.e. on both July 4 and 5 ?
Also, was it the very same plane on previous days?
The flight track on FlightAware definitely shows that it went around, from 5 miles out, which would usually indicate that it was tower initiated. I doubt it was the same aircraft, since it’s a ten or eleven hour flight each way. There’s not enough time in a day to fly both ways and do the turns at each end.
Thank you.
And on July 4th also?
Thank you, a very sage and informative analysis. As with all aviation accidents, there will not be a single point cause – the man-machine interface and the sterility of the flight deck in the descent are going to be interesting; time to resist lazy, single-point-cause speculation as the media seems happy to seek.
I see parallels with the ’09 Schiphol/Turkish Airlines 737-800 accident – Situational Awareness, crew culture and role dominance are going to be central to the investigation alongside more tangible mechanical fault exclusion. KSFO a hard-work field around which to operate in a heavy.
Pilot, physician and aviation human factors researcher
From SCMP today: Airline pilots with experience flying the Boeing 777 or flying into San Francisco said the Federal Aviation Administration notified pilots in June that the ILS was turned off. Pilots were also warned that the thresholds, or beginning, of runways 28 left and right had been moved.
A white line that previously designated the end of the runway was blacked out and a new line painted further west, said Rory Kay, a training captain for a major airline who landed at San Francisco the day before the crash.
The change in the runway line might have added an element of confusion to the landing, he said.
All Boeing 777s, like most modern airliners, have cockpit computers that use GPS to create a glide slope for landing that is nearly as good as the ground-based ILS, said Bob Coffman, an American Airlines captain who formerly flew the 777.
It would be standard procedure for pilots to create their own glide path before landing, but the computer’s database relies on where the runway usually begins, he said. Moving the threshold would invalidate the computer-generated slope, he said.
But pilots also receive FAA notices on ILS shutdowns and movement of runway thresholds in a pre-landing briefing, so the Asiana pilots should have been aware that they were going to have to rely more heavily on visual cues, pilots said.
I flew into SFO the 4th on the Golden Gate 6 STAR and did the same ILS Loc Z 28L with the glide slope inop. This is the normal approach for the runway. The PAPI was on and normal. I have as many questions as most of you do. Let’s clear up a few things, a visual approach usually is fully automated and programmed and briefed before we start down for landing. The controllers at SFO are great and move many planes in a short period of time, but it’s up to the pilots to say “unable” if they hang you up to high. I have two theories and after viewing thousands of flights from Flight Data Recorders or QAR’s I think I have seen most of the common errors. I think the FP flying pilot thought this aircraft was going to capture the glide slope at 3100 but did not because the glide slope was out. After he figured out it did not capture he did what Birdstrike said. LvL Change trap that commands the auto throttles to go to idle and opens the speed window. Why did he not arrest his descent with power? At around 1100 feet the EGPWS would have alerted all the pilots with “Sink Rate” about 3 times followed by the ” Woop Woop Pull Up” until leveling just above the water. The second therory is a little long but I have seen it many times. With only 43 hours in the aircraft he had to have an Instructor Captain in the right seat. I think the ICAO rules state the PIC has to have 50 hours or more with credit for extra landings. If this is the case the right seat pilot may have assisted with the automation in an attempt to save the approach. Now here is the question? Did the left seat Captain think the instructor took the plane? Were both pilots waiting for the other to correct the descent? With the engines at idle and unspooled it would have taken time to get thrust for go around. Thanks for the site Steve. We need to talk.
A training pilot was in the right seat and is responsible for the flight.
I agree with most of the above. However, for grins, let’s look at another scenario;
See the image below:
I was sitting here doing some research, and stumbled across something that might be of interest to you guys, or, you can refer this information to the appropriate people for further research.
Take a look the attached jpeg image. There are two pictures in the image. The one on the left was taken on Saturday( Day of accident), the one on the right is a Google Earth Image, Date Taken unknown. (Disregard the aircraft in the right picture as this is a random picture takes from satellite) The runway layout is the important factor.
Look the Thresh hold and Touch Down Markers. Apparently, at some point, this area was moved down the runway. Possibly for Crossing Height Displacement, or to allow better clearance if someone taxied into position, who knows. But there is now a Displaced Thresh hold where there was not previously. Thus, ILS, Runway Markings, and a lot moved forward about 500 to 1000 ft.
That said, I don’t know if the 777 was on the ILS, or, onboard data type of approach (RNAV / WASS) approach. The ILS “Should” have taken care of the low approach. However, if there is old data in the FMS, it “Could” have appeared to be too low, and they punched out of the AP due to an unexpected visual presentation. I am not sure if they use a Data Base, or Synthetic Vision derived for a Data Base, or they were using the actual ILS Raw data. At any rate, if there was old data being used, they would have landed about 500 to 1000 feet shorter than the currently published approach.
Also, if on a WASS Glidepath, but seeing a different VASI indication, “Might” have lead to confusion, and a quick change in profile to correct the developing situation.
Here’s the link you reference
The ILS glideslope was out of service, in part because they are moving the thresholds, so they weren’t using that. I’d be surprised if there were old data in the FMS. In any event, it’s hard to see how moving the threshold (and touchdown zone) 270 ft or so down the runway could cause them (even with some confusion) to land short of the old threshold.
The data base is checked before every flight and the next revision data base is loaded in the FMS ready for the active date.
It was just a theory about the displaced thresh hold.
I still like the Unstabilized approach “Theory”. Years ago, when flying in and out of Atlanta, we used to get a “Slam Dunk” for 26 R out of 11,000. I’ve seen many guys in the right seat not able to cope with such a clearance. They were not trained in flying a visual approach, especially from such steep profiles. They were trained 1,2 3, do this then that. They never got the real world sensation of non-standard approaches. It all goes back to initial training. 61 and 141 / 142 are not training students as we used to train them. (There is good and bad with that statement thought). I investigate a lot of accident / incident data. We all know there many factors, but a lot comes back to initial training habits.
That said, I see this a lot on flight checks. What to do when something goes non-standard! Scenario based training is just coming into play in a lot of areas. Students are not getting good training. FACT! I just presented 3 seminars on Pilot Deviations and Runway Incursions in the Tampa and Sarasota Area. Interestingly, 49 out of 50 deviations had not been a part of any safety programs, or self – recurrent training since their certificate issuance. Thus, we go back to the Buffalo Accident, and the 1500 Hour proposition.
I have not seen the preliminary data on the SFO Accident. PIC, SIC hours, experience, training, crew rest, scheduled hours, and many other “Human Factor” input data, etc etc. I think this is where we will start to see a possible trend in causation. Once we have more data, / information, we can then, start to build a clearer picture of events leading up to the accident.
Dennis
It is disheartening that the airplane deviated so far from the proper approach parameters despite clear airline operating procedures, established go-around limits, available engine and flight path automation, and windows to look out. It is bad enough that some of this automation was apparently disabled, but also one must seriously wonder if each of the pilots in the cockpit somehow thought one of the other pilots was flying the airplane. In any case, it is hard to argue that anyone in the cockpit was paying attention.
It seems likely that the final NTSB recommendations will include comments about cockpit discipline and crew coordination.
In modern airliners with lots of automated systems, what percentage of “flight hours” do experts think — roughly — are “live-time” where pilots control the aircraft?
Out of a 10 hour flight, perhaps 1/2 hr live time?
So about 5% of pilot-hours stated are “actual” flight hours for longer hauls? Does that seem reasonable?
Excellent observation. Being that this plane is used for long haul transpacific flights then 43 hours in this plane could mean the pilot only had 4 landings total and none at SFO. I still think that the final analysis will show that mistakes were made due FLCH trap that pilots all thought the air speed throttling was being automated when it wasnt.
Thank you — can some 777 pilots speak to the confusion of FLCH and automated throttling?
e.g. see thread at:
http://www.flightgear.org/forums/viewtopic.php?f=40&t=13614
My husband sent me this link to get a ‘great explanation of the Asiana crash’. Your interpretation of the information you were able to glean from multiple sources was the best I’ve heard/read yet!
My husband is a 17 year veteran of the Air Force (10 active duty and 7 Air National Guard) and a 7 year commercial pilot. His Bachelor degree from the Academy was in Behavioral Sciences (Human Factors) and he has a Masters in Aeronautical Science and a Masters in Human Factors in Aviation Systems Specialization. With that said, we’re always talking about the potential breakdown of CRM (cockpit or crew resource management) on a plane.
I’ve been listening to everything on the news, reading all of comments left before me here (all insightful and very interesting!) but today I heard someone bring up the Korean/Asian cultural aspect that *MIGHT* have played a role in why 3 other pilots who were also on the plane didn’t speak up in time. When I brought this up to my husband, he’s mentioned that it was a big problem with Korean pilots especially through the 90’s and that Americans were sent over to train their pilots in all aspects of CRM. And by cultural…I mean that a co-pilot might not speak up because it would be ‘disrespecting’ their elders, and you just don’t do that.
Malcolm Gladwell (author of Blink and The Tipping Point) actually wrote about this in one of his books. Check out this piece from the Wall Street Journal in 2008 talking about Korean pilots in particular: “Malcolm Gladwell on Culture, Cockpit Communication and Plane Crashes” http://blogs.wsj.com/middleseat/2008/12/04/malcolm-gladwell-on-culture-cockpit-communication-and-plane-crashes/
I’m just introducing another idea into the mix. It’s just so hard for me to understand how a pilot with over 10,000 flying hours (granted most of it was on other planes, with only 43 hours on the 777) could make such a grave error with a co-pilot and another set of pilots on the plane as well. NO ONE spoke up until a few seconds before impact?!! The NTSB is saying that 7 seconds out was the first indication that the pilots recognized the need to increase air speed. Passengers and even people watching on the ground recognized the plane as too low and too slow…so how did the pilots not know?!
My husband also sent me a small sample of Korean Air Carrier history (read: crashes) and it’s not good.
Anyone have any thoughts on the cultural aspect of the pilots in this accident?
Thanks for reading even though I’m not a pilot!
I’ve seen comments like this, too. While the NTSB will certainly look at crew resource management, I think (since you asked) that it’s unfair to the pilots in this case to speculate about what role their culture may have played in this accident. For one thing, we don’t know that it’s an issue. For another, if there are cultural issues at play, then it is the responsibility of the airline and the corresponding regulatory agency to effect training protocols that mitigate this issue. The right question to ask is first, How effectively did the crew manage their resources and coordinate their actions? If there was a problem there, then the NTSB needs to find the root cause, whether it’s training, cultural, or otherwise.
Thank you for your post and kind words.
Great analysis.
It would be interesting to run the same analysis on previous landings of Asiana Airlines Flight 214, to see if unstabilized approaches are common.
regards
Three qualified flight crew failing to decide to abort the landing untill far too late. I would agree that there is a senerority issue here. But also a lack of live flight time. New pilot to the 777′ new airport to the pilot(s), landing aids out of order, 150m rather than 300m per-landing zone.
Whilst I agree with the secondary cause being an unstabilised flight path. Is the primary cause too many vairiables for a pilot new to the 777? Pilot overload?
Your post makes me think of the KLM Tenerife accident in 1977, when captain Van Zanten, a very senior captain, wasn’t challenged enough by other crew members…
Excellent information with informative use of charts to visualize and clearify gathered facts.
It would be interesting to collect a larger number of “landing profiles” and show them in the same graph, showing how many 777-approaches in fact are stabilized. Maybe 10-15 is enough to see a pattern?
I’ve done this on other aircraft and the numbers are actually 1000+ to 1 unstable. When we put these in a flight path overlay with the unstable paths highlighted it makes you think “why continue”? I’ll try to find one with all the data de identified for you.
I’m curious about which criteria you used to identify an unstabilized approach and arrive at the 1000-1 figure. One of the criteria mentioned in the post is a “sustained” rate of descent exceeding 1000 fpm. Would this be enough to classify the approach as unstabilized in your analysis?
Also, do you know what fraction of the unstabilized approaches resulted in go-arounds?
If any 777 pilots are browsing, can they weigh in on whether this complicated procedure/bug has been fixed?:
http://www.flightgear.org/forums/viewtopic.php?f=40&t=13614
QUOTING:
“Re: B777 autopilot / FLCH is actually V/S mode
by Cascade on Sat Oct 01, 2011 10:51 am
If you mean a horizontal air speed, I can help you.
The B777-200ER CAN maintain a minimum airspeed automatically, included while climbing and/or descending to a target altitude.
Ensure that the cockpit switches “A/T ARM L” and “A/T ARM R” are switched on. Then, click on the “A/T” button IN THE COCKPIT (not the button in the AP popup window!).
Now the Boeing maintains a minimum selected speed, which can easily be edited in the popup window “Autopilot control” (F11).
After a climb or a descend in FLCH mode, the ALT HOLD neatly turns on automatically.
While approaching, you can easily maintain a minimum air speed automatically, depending on the current weight, flaps. Even better, when approaching automatically (APPR mode) the minimum
Note:
Pressing the “A/T” in the “Autopilot control” popup window (F11), always enforces a full throttle or an idle throttle, nothing in between, ignoring that A/T ARM Left and Right switches are switched on (while
clicking the “A/T” in the COCKPIT, does take the A/T ARM Left and Right switches into account).
May be someone can reproduce this and report it as a bug?”
(F11) is figure 11 in the manual. It’s the speed window that commands the auto throttles to maintain if the auto throttles are on or armed. In one of the NTSB tweets they said the Auto Pilot was disconnected at 1600 feet. Remember the auto pilot and auto throttles are different controls with different ways to disconnect. You can fly it in the mixed mode…that my Boeing instructor did not recommend. I’m not sure if that has changed in the past year. I know one incident in the past few years referenced the mixed mode was one of the probable causes.
Thank you!
I see where your getting your info. FLCH is not a term we use. LvL CH or Level change is. It means that the power is at idle for descent or full power for climb. The website you reference is for Microsoft flight simulator or X-Plane. Both are great programs and fun to play with.
Thanks — but would you agree that the mixed auto-pilot/auto-thrust system is confusing in 777 and _may_ have contributed to the crash?
Seems like pilots were unaware that engines were at idle, and may have thought that auto-speed maintenance (auto-thrust) was engaged when it was not(?)
The problems you mention were, as I understood it, discussions of a bug in the Microsoft Flight Simulator rather than one in the actual aircraft.
777 pilots’ comment?
No idea about MS.
On the aircraft, FLCH is a speed on elevator mode. Ie the elevators maintain the airspeed set by the airspeed bug.
VS will maintain a set vertical speed in Ft/min, and it will maintain an airspeed until either it is unable too. For instance in a steep climb, the airspeed will decay, or in a steep descent the airspeed will run away.
Why continue with an unstable approach?
A. The pressure/culture in some airlines/countries clouds the judgement of the pilot.
The fear of admitting you have got the approach wrong, the fear of asking the senior pilot if you can go around.
Perhaps because you are under the mistaken impression that auto-throttle is working? (see, e.g., my comment above yours)
There is some long out of date information on the Mandelbaum effect, resting accommodation, and dark focus that would predict that under certain visual conditions and for specific people, the visual field would be distorted such that an airplane would land long and hard. Quoting Roscoe, 1979: “When pilot make approaches and landings with any type of imaging flight display projected at unity magnification, they tend to come in fast and long, round out high, and touch down hard. These landing characteristics occur because the imaged runway appears smaller, farther away, and higher in the visual field than it does when viewed directly from the same approach path on a clear day.” Further, Roscoe notes that this type of effect can occur even without an intervening display: “When searching head-up from the pilot’s seat, the instrument panel appears in the dim periphery; the pilot sees mainly empty space through a windshield that reflects glare and may be dirty or scratched. These conditions suggests that pilots can unknowingly be subject to the “Mandelbaum” effect.” The observed behaviour in this plane crash (as well as the earlier crash when a Japanese Air Lines plane actually landed in the bay short of the SFO runway) fall directly into what was described by Roscoe.
This was initially researched by Stanley Roscoe at New Mexico State University and others but was rejected without sufficient testing by the community. In fact, the rancor was sufficient that no one even bothered to perform any testing that might disprove the effects described by Roscoe. The work is described in numerous papers and publications but many of these are out of date and not available on-line but, as a result of some work I did in this area a number of years ago, I have a fair collection from work I was involved with. Three of the key references are:
Stanley N Roscoe, “When Day is Done and Shadows Fall, We Miss the Airport Most of All”. Human Factors, 1979, 21(6), 721-731.
Owens DA, “The Mandelbaum effect: evidence for an accommodative bias toward intermediate viewing distances”., J Opt Soc Am. 1979 May;69(5):646-52. Retrieved from http://www.ncbi.nlm.nih.gov/pubmed/479967, July 8, 2013
D. Alfred Owens, “The Resting State of the Eyes”, American Scientist, Volume 72, July-August, 1984, Pgs 378-387.
I am not working in this area at present but would hope that someone might look into this in an effort to identify what I believe to be a real effect and a possible (and avoidable) cause for the recent crash.
This is especially the case for approaches over flat sea with few references.
I think it’s strange that the cultural factor hasn’t been discussed. A Korean employee would rather run a plain into the ground (or watch it crash) than question a superior. Korean Air has worked hard on this, not sure about Asiana. From Wikipedia: “The last fatal accident, Korean Air Cargo Flight 8509 in December 1999 led to a review of how Korean cultural attitudes had contributed to its poor crash history. Following the review, Korean Airlines began hiring predominantly Western pilots and since that time safety has greatly improved, and the airline ranks among the best in the 21st century.”
I’m a little bothered by the statement, “A Korean employee would rather run a [plane] into the ground (or watch it crash) than question a superior.” Perhaps that is or was a problem with some or even many Koreans. It’s unfair to paint all Korean pilots now with that broad brush. Further, US pilots are not immune to this problem, although the industry is certainly aware of the importance of crew resource management.
And it *has* been discussed and is likely being looked at.
Just have to shake my head at the media types. This from Fox News; “214 heavy means that the aircraft was low and slow?”
Did they really say that?
Fox News can — at times — be low and slow themselves.
Steve, thanks for all the time put into this! A quick question (which may be painfully obvious to those with instrument ratings 😉 ) — from what point did you calculate distance when you did your analysis?
It seems that it might’ve been 28L’s lat/lon from AirNav, but without being able to get exact numbers off your graphs, I can’t really tell. Thanks!
Everything is measured from the PAPI. In VFR conditions, that would be the aim point of the approach. Perhaps should have been the aiming point markings, which are 260 ft closer. It’s not a huge difference.
Right. So, uh, where does one go to look up the lat/lon of the PAPI specifically? 🙂 Like I said, all I’ve been able to find so far is a lat/lon for 28L on AirNav…
Thanks!
Oh, sorry. I used Google Maps, and picked the point off by hand.
Ah, OK. Thanks!
News reporting landing gear hit first, and instructor pilot was in command of plane.
What was situation with auto-thrust?
Tweet from:
Andrea Mitchell ?@mitchellreports 13m
Instructor pilot told NTSB didnt realize auto pilots werent maintaining speed on approach
http://www.latimes.com/local/lanow/la-me-ln-asiana-pilots-knew-plane-was-in-trouble-ntsb-says-20130709,0,5967736.story
The LA Times is reporting the following info from NTSB, it is very consistent with the data presented here.
“…The pilot said the crew thought the auto throttle was maintaining speed but it was not.
…
Officials said Lee and his more-experienced instructor pilot sitting next to him didn’t discuss the predicament. Cockpit voice recordings show that the two didn’t communicate until less than two seconds before the plane struck the sea wall and then slammed into Runway 28L.
Officials said the Asiana jetliner had fallen more than 30 knots below its target landing speed in the seconds before it crashed, even as the crew desperately tried to apply more engine power.
But even before that, the aircraft had departed from a stable and planned approach to the runway, failing to keep up with its intended speed of 134 to 137 knots at 500 feet over the bay.
…
The Washington, D.C.-based Flight Safety Foundation, which advocates for airline safety, said in a recent published report that 97% of the time, pilots do not abort a flight from an unstable approach. The reasons they most often cite are their experience and competency to recover.
”
Steve – Do you think it would be worthwhile to investigate the frequency of unstable approaches and how often they result in a go-around? I would think that a coder with access to the FlightAware API could mine the data pretty quickly given the right parameters.
Fliifast indicates in a post above that he’s done this, and the ratio of stabilized to unstabilized approaches is 1000 to 1. He doesn’t say how often they result in a go-around, but he would know.
FLCH is a very good mode for changing flight levels, at higher altitudes, but it is not an approach mode.
However, when being vectored for an approach and the aircraft is too high, an easy way to recover is to set a lower altitude, select FLCH, speed brake etc. this give a high rate of descent, and as long as the pilot has the G/S armed the aircraft will capture it. The G/S acts as a safety net.
In this case, the aircraft did not level out, and we know the GS was not active. I have seen several times in this situation where the ALT can be inadvertently set to zero. In that case, the auto throttle will not wake up – as it is performing exactly as directed. If zero is set in the ALT window, with FLCH selected, there is no need for any thrust.
FLCH is what is referred to as a speed on elevator mode. The pilot sets the airspeed, and the elevators raise or lower the nose to maintain that airspeed. However if the flight directors are not followed, for instance if the nose was raised, the speed would decrease.
Very sad flying in here tonight and seeing the hull lying there.
Sorry, my reply above was in reply to an earlier question.
I would be very interested to hear your comments re: NTSB statements after interviewing 2 pilots on 7/9/2013.
I’m not sure what to think; it would be interesting to hear one of the ATPs give their opinion.
Steve,
Incidentally, regarding your comments about the angle of the glide path flown in a heavy jet. If the previous aircraft was flying in VNAV- very common on a visual approach with no GS indication, then VNAV uses barometric altimetry calibrated for ISA. So SFO summertime, noon, on a clear day, you would expect to fly a steeper than 3degree path. I notice that the PAPIs were off the previous day. So maybe they were not on for the day in question either.
Nice presentation and thanks for the info.
I would like to bring out one point that, although it MAY have been stated in one of the responses already (correct me if I’m wrong… I just didn’t see it), but it has not been discussed enough… TURN THE AUTOPILOT OFF, LOOK OUT THE WINDOW AND FLY THE AIRPLANE THE OLD FASHION WAY. This APPEARS to be a classic case of too much automation and heads are DOWN screwing around with the automation. This does not surprise me and the NTSB findings in the end… will not surprise me. Please watch this excellent presentation from American Airlines that was produced back in Apr 1997. http://www.youtube.com/watch?v=h3kREPMzMLk (copy and paste or just click on the link its called “Children of the Magenta)
My back ground is 38 years in flying, 35 in big airplanes (707, 727, L1011, 747, 747SP, 744) and I’ve been into SFO more times than I care to admit as it dates me. It’s a challenging airport from an ATC and traffic standpoint, but not difficult and is just another airport. It requires common sense flying and skills. The water surrounding the airport has very little effect on depth perception and to use that as a “reason” for what happened or even influenced the event is unacceptable… look at the runway, not the water.
Most airlines using aircraft with full “Glass cockpits” require the pilots (through policies and procedures) to turn the Autopilot (A/P) on very early (even as low as 250 feet) and most of the pilots I fly with don’t turn it off until around a 1000 feet (after the aircraft is stabilized). That’s about 72 seconds of “hand flying” the airplane. At the end of the day though, this is not going to give experience to anyone let alone a pilot. Many accidents have been caused because they didn’t FLY THE AIRPLANE and let the A/P do the work… beyond its capabilities. The opening statement given in the YouTube video defines the issue accurately; “…autopilot dependency…” Please watch the video in its entirety.
Remember, we’re pilots… not “gamers”, not “technical crew”, not “computer operators” and NOT BUS DRIVERS… We’re pilots and to let society (and engineers, airline management or even your spouses and girlfriends/boyfriends or MEDIA) tell us otherwise, just moves us one step closer out of the cockpit. Learn to be an excellent PILOT and don’t let anyone take that away that from you… you’re not hired to be an excellent computer programmer (in the cockpit at any rate).
(By the way, I knew the pilot in the right seat on the JAL flight that landed short of the runway… they were flying a new Flight Director and the captain was not flying it correctly and they had only had “differences training” on the new system with no real flying application yet).
Indeed evidence points to automation being issue:
http://www.hurriyetdailynews.com/pilots-relied-on-automatic-equipment-in-plane-crash.aspx?pageID=238&nID=50413&NewsCatID=358
But remains to be seen if auto-throttle malfunctioned.
Even if it did, it does not excuse the (apparent) lack of monitoring the airspeed.
Some interesting details here:
http://www.afp.com/en/news/topstories/asiana-pilot-couldnt-see-runway-us-crash
To Steve and everyone who has participated in this forum-
First off, what a breath of fresh air to read something online that is informative, grounded in facts and logic.
Second, thanks to Steve for putting this information together. I have absolutely no background in physics or the hard sciences (I’m social sciences), but I could glean something from the discourse, probably because it was civilized and focused on an exchange of information.
I have read all of the comments above. Several times cockpit culture or Korean culture was pointed out as one of the possible culprits for this tragedy. Deference to seniority in Middle Eastern or Asian cultures is probably well understood to be problematic even in situations where deference could be deadly, so I would assume somewhere in flight education , the manual/instructor hammers home this point, right?? Any pilots know about this? And, can you give me an example of how it is explained/taught?
many thanks!
Good stuff guys.
Here are some follow-on thoughts from an US international airline guy with decades of experience.
Autopilots are of little use when high and fast on an approach.
A manual landing is a near emergency for these Asian airlines. They just do not do them.
In an attempt to salvage the high fast approach to a hopeful landing, the pilot clicked off the autopilot switch on the yoke and evidently clicked off the the autothrottles switch on the thrust lever, whether intentionally or inadvertently. I’d bet $100 he did not announce it on the CVR when he did it, as he should have. Heck, he may not even have known he did it. The autothrottles remain armed but not engaged when this is done. The pilot must control thrust. The pilots unfamilarity with the airplane was a significant factor.
The autothrottle disconnect switch may be used before landing and is used to disconnect the autothrottles in terrain avoidance and windshear recovery maneuvers.
Evidently, none of the pilots in the cockpit were aware the throttles were in manual.
They were probably occupied with getting the aircraft down to an acceptable glide path, in an attempt to “save” the landing.
No one was paying attention to thrust. That is not uncommon due to the constant use and reliability of the autothrottles.
I learned and observed as an LSO on aircraft carriers in the Navy that most low/short landings occur following a high/fast start. This event was a classic failure in that regard.
In my view, the reasons for this accident were a combination of;
1) an inexperienced new captain to the 777, coming off an highly automated computer controlled Airbus 320 in which he seldon, if ever, manually operated the throttles,
2) an over-tasked, inexperienced first-time line-check instructor,
3) a young relief pilot on the jumpseat, culturally intimidated and not willing to point out the failings of his superiors,
4) another experienced pilot, probably a captain, riding in the cabin rather than the cockpit for the landing and finally
5) a high fast start to the approach that in an attempt to “save face”, should have been
discontined before 1000′ and finally and crucially,
6) the non-recognition of the disconnected autothrottles
were the reasons for this accident.
Interestingly, if the autothrottles had been engaged, it may have been a non-event-
a very below average approach to an likely safe landing.
Regards
Vey well thought out. Interestingly, they may have had the AT engaged, but if they were in FLCH, ALT set to zero…….(it happens!) and raising the nose through the F/Ds, then there would be no AT response, and no AT wakeup.
….if the industry is unwilling to give up its romance with automation, I say we create a civil LSO qual and post them at the TDP of every runway in use, requiring aircraft to check in directly from approach frequency. Dual runway approaches would require a complicated bolter waveoff pattern though. I have to add that the carrier pilot, trained to manage glideslope with POWER, is by virtue of that fact always attuned to the position of the throttles. The modern civil ADI provides, as I said above, only a one dimensional readout of airspeed, without a velocity vector or E-Bracket with which the pilot can monitor both approach angle AND energy trends. I’m sure Rockwell Collins and Honeywell and Thales are composed of good people, but their creations are ergonomic disasters. Their monopoly on the market, coupled with the basic civilian ignorance about the ergonomic simplicity of some military cockpits, will ensure “pilots” today keep earning flight hours when not even flying, and keep growing less capable in instances when they need to.
Non-pilot here.
From the discussion it seems that there is a hodgepodge of automated systems (auto-pilot, auto-throttle, ILS, stall warnings) that are meant to assist the pilots, some of which can be disconnected or be non-operational at any given moment. So it is the pilot’s responsibility to be aware of which particular combination of systems is operational at any given moment and act accordingly.
I would think that this increases cognitive load on the pilots, and that they would benefit from an automated “landing monitor” that cannot be disconnected and would give warnings. It could work much like a car’s GPS, but it would use speed, altitude, and landing coordinates to project the flight path and throw red flags.
From Steve’s charts, a monitoring system would have noticed the speed continue to drop below 140 knots at around 500ft, and presumably thrown a red flag well before the stall warning kicked in.
Does a system like this exist in commercial aircraft?
Javier:
Yes, warning systems exist for low and slow in addition to flap/gear/speedbrake misconfigurations. The problem with this flight is that they did not have a computer monitored glideslope (ILS) so they would not get a “too low glidelslope” warning.
If they set in 420′ or whatever the GPS approach minimums were for decision height they also would have got the resultant “approaching minimums and minimums” automatic callouts and had some backup warning. They probably put in field elevation or nothing so they got no minimums automated callout.
Many warnings are disabled when automation is disconnected.
Thank you for the response.
I understand the need for being able to disconnect automatic controllers since the pilots are ultimately responsible for flying the plane. But it seems to me that it you’d still want a system that simply monitors the flight parameters and warns the pilots well before the plane is about to crash. I’m talking here about a system that *cannot* be disconnected like the other systems.
There has been a lot of speculation about pilots not communicating properly due to cultural issues, and about their failure to monitor speed, sink rate, etc. Those are things that software can do very well.
Steve mentioned in a comment above that this flight likely missed 4/7 criteria for a stabilized approach. In an age where everyone has a GPS capable of turn-by-turn navigation, it’s pretty surprising that this sort of accident of inattention can happen on a major commercial carrier.
There is no excuse for not actively monitoring speed, no matter the settings.
Unless that instrument was malfunctioning, I have a sense of where this may go.
Randy et al,
First of all – I agree whole heartedly with the concept that pilots need to be more “in control” of their machines. There has been a definite push in technology that has definitely made pilots take a step back towards monitoring, and that is a bad trend. Fortunately I’m currently flying an aircraft that requires more proactive management – but we all know how good most of these aircraft have become at pretending to manage themselves.
We’ve managed to shift accidents to different causes. But we’ve also greatly reduced their number – so automation isn’t all bad. Its just like the Airbus debate – yes we’ve lost (or nearly lost) a couple of aircraft because of the Airbus system, but we’ve saved many more thousands because of it too.
I’ve watched the “Children of the Magenta” video many times over the years, and much of it is instructional.
However, one thing has always bothered me. AA have one of the worst records for runway excursion of any airline in the world.
Having a look at your list of aircraft that you’ve flown (we have one in common), I can see that you’ve probably done a mix of short and long haul flying. I’m willing to bet most of your skills came from the short haul operation where you get the opportunity to practice those skills. Also, you most likely flew the formational part of your career in the days when hand flying was considered routine. I’m sure you’re very aware that on a long haul operation, you may only do 2 or 3 landings a month – hardly enough to keep proficiency, especially when you may need to hand a landing or two over to George.
Sadly, these days it is not practical to hand fly in a lot of cases. I generally prefer to hand fly take offs to around about 10,000ft – but these days I’m limited in my ability to do that. Airways are now constructed so that the margins for error that used to exist simply don’t any more. Many of our departures require a level off at 5000′, with overflying traffic at 6000′. Experience tells us that automation is FAR safer in this environment than hand flying level offs in a ‘sporting’ aeroplane at 5000′. We know this because almost all ‘near miss’ conditions have been hand flown. It is perfectly reasonable to expect pilots to use automation where it has been shown to have a significant safety benefit.
Add into this noise abatement. At some airports we are instructed to use the Autopilot because it guarantees a closer following of the LNAV track – any deviation results in higher noise penalties and financial imposition on the airline, including possible suspension of landing slots. Its reasonable for an airline to expect the use of the autopilot in these circumstances.
Long haul pilots simply don’t get the opportunity to fly approaches like they used to. In some cases, thats because they don’t try to maintain their skills, but I really think that is the minority. Most pilots are actively trying to maintain their currency and skill set, and do try and take the opportunity to practice that.
Perhaps Ironically, the pilots here could have prevented screwing the pooch two ways – either by the traditional method – looking out the window and just flying the machine (which, in this case, seems to have failed them), or better automation management in the first place (V/S – 800 anyone?).
Unfortunately too often we see knee-jerk reactions after accidents like this. I hope the whole aviation community takes a deep breath and recognises this for what it is – another unstable approach accident, and also one in a training environment (a la Turkish in AMS and numerous others), and continues to study why experienced, qualified, and otherwise good pilots can make such monumental foul-ups.
I beg of everyone do NOT simply blame the pilots as being incompetent, or just fall back on the old chestnut of “Asian (Korean) pilots balls it up again”. This is a systemic problem that exists throughout all of aviation. We need to get to the root cause of how so many aircraft are being crashed in this specific environment. And if it is as simple as poor handling skills, then don’t just fall back on “more hand flying”, because we live in the real world, not an ideal one. We need to find out how to improve manipulative skills while not throwing away the benefits we’ve achieved from great automation advances.
Thanks for this insightful, thoughtful analysis. All I can say is pretty much what everybody else has said – it was clear to me the moment I watched the video of the crash what happened. Nobody was flying the airplane and they simply flew the thing into the ground. I’ll never forget my first flight instructor pounding it into my head: FLY THE AIRPLANE FIRST!! Everything else is a distant second in importance. It astounds me that in 2013 we seem to still have the same basic problem that has plagued us for decades: pilots failing to exercise the first and foremost responsibility of being a pilot: to actually *fly* the aircraft. All the wonderful gizmos, flashy lights, beepy things, and auto-everything are wonderful, but sometimes you just have to be a pilot. Oh, and if it’s not Boeing, well, you know the rest 🙂
I bet the port engine got knocked off early and flew ahead on its own (in the aftermath it was discovered far forward), while the starboard engine, finally spooled up and still attached, is what pushed the tail-less plane up and around, causing the lofty ground loop we see in the video.
The girls killed were sitting together in the last row. Perhaps their seat belts were not securely fastened — another passenger sitting with them was reportedly unharmed. One fell out the tail early, but the other: the one discovered just forward of the left wing (the one might’ve gotten run over by a fire truck) — I bet she was ejected out the tail by centrifugal force during the ground loop and flung forward, landing ahead of where the plane came to rest. How else could she have gotten there?
Great analysis. I did a similar analysis, but way less detailed for turkish 1951 back in 2009. In that crash, the pilots didn’t notice the engine were idle either, and crashed a few hundred meters in front of the landing strip at amsterdam airport. A major issue there was that the radio altimeters controlling the auto-thrust were showing a negative height and therefore commanding thrust to go to idle. The pilots were even aware of the failing altimeter, but not aware of its effects on auto-thrust. Can anyone comment on the 777’s dependency of auto-trust on the altimeter? Could the same thing happen in the 777?
Except for a rare weather phenomena, such as a microburst near the ground when you are low and slow, or a mechanical problem, there is NEVER an excuse to land short. Am trying not to overstate that and will try to explain.
The cockpit of the large transport I flew for about 6000 hours was 28 feet above the ground. The landing gear was about 30 feet back from the cockpit. So, you can imagine when you round out to land there are a lot of 3 dimensional issues one must consider. It was hard to get used to. Two of our aircraft landed short and were destroyed, although no one seriously hurt. They ground it into us: The pilot of the next short landing will result in separation from the service. We caught on. Many of today’s large jets have a similar issue but pilots of both my generation and today’s, are by definition pilots because they can deal with all the vagaries. The one difference these days is that there are a lot more electronics that are capable of aiding the pilot, including automatic landing assist that many of them use. However, with the proliferation of electronics there is less hands-on time. Hands-on experience is where you get the feel of the airplane at different speeds and conditions, which is necessary to be in control.
From what I’ve learned about the SFO 777, the pilot thought he was on auto assist and was going to transition to hands-on later. It seems clear he did not have a good feel for the process. But, what really gets me, there was an instructor pilot in the right seat! The final airspeed was supposed to be 137 knots. Three different pilots are supposed to monitor that and each thought the other was doing it. As it turns out for the last few hundred feet the airspeed was down to 107 and they were way too low. That’s stall speed and the stick shaker was telling them that. By the time any of them realized it, it was too late to recover. I ask, how can any any member of the flight crew allow that to happen, let alone all three of them? I can only shake my head.
I actually am disgusted with these guys. They have plenty of experience, and landing short is the easiest thing to correct. I think it is inconceivable and unforgivable. There are some pilot errors that cause crashes that I can sort of see what happened and understand why they erred. Landing short, there is never an excuse.
The news will mention that they started their decent from 18,000′ for a straight in approach. This actually is a little harder to judge heights as one approaches the runway, and it required a steeper glide path. It sounds like the trouble started right about then. They were way above the glide path to start with and pulled the throttles way back, almost to idle, in order to cut speed from 180 to 137 and get back on the right slope. The pilots seemed preoccupied with catching up and being in the right spot and were cutting speed and altitude. Adding to their discomfort was the lack of the SFO electronic glide slope that most of them use, even in good weather. While that is an inconvenience, it isn’t that big a deal, especially if one is comfortable with hands-on. As they worked their way down it appears that no one noticed the airspeed declining. As they got closer and realized what was happening, the pilot pulled the nose up!! Classic mistake. He was already slow with almost no lift and the worse thing one can do is pull up. He put it directly into a stall. This has always been what happens to panicked pilots. It was probably too late anyway but going max power was the thing to do.
I can share many stories about how my fellow pilots ignored the obvious and crashed. But I will actually tell you a story that happened to me, and why I am alive today. I spent 4 years in Washington state flying the C-124 then was sent to Hawaii for 3 years. I accumulated quite a bit of time and got to be pretty good. I could touch down that airplane so softly sometimes, you didn’t know we landed. When it came time for me to officially check out as an Aircraft Commander, I had to go thru 2 separate flight checks; one of those was handling emergencies, landings, take-offs and various distractions in the local area, to make sure I could handle them. Once that is passed I had to go on what was called, a route check flight. This is a mission to many of the locations in the Pacific we regularly flew into: Japan, Taiwan, Phillippines, Wake Island and several other spots. It was just to let us do the landings, take-offs and the enroute planning required. This was the easy part, I could fly the hell out of the airplane and I had been to most of this spots anyway. This was to be a fun trip.
We took off headed to Midway Is. with some heavy cargo and after that we were to then head out into the rest of the Pacific. There was an instructor pilot and 2 of us to be checked out, along with the rest of the crew of 10. About halfway there we had an engine failure, which wasn’t that uncommon, so the instructor pilot took over, as he should, and we continued forward. I rode the right seat. As we got closer to Midway, a very small spot in the ocean, the weather deteriorated and I called in an emergency and asked for a straight in approach with radar guidance. (Humans at the radar site call out glide path and azimuth). C-124’s had a quirk. It was a bit underpowered anyway and with only 3 engines we had to very careful not to exaggerate any slow speed or nose high conditions, because you could get to a place that if too low and slow, it took lots of power and nose down to recover. Obviously, if we are at a low altitude and you get in this position, one cannot recover. So, during local check out we practiced recovery at 5000 feet just see how to do it. And done incorrectly, one loses a lot of altitude. I understood the phenomena.
The undercast went down to about 500′ and when we broke out I saw the runway and knew we were too low and too far away. There was nothing but ocean below us. I reminded the instructor pilot we were a little low and slow and this was confirmed by the radar people. He took no action. I waited and then said it again. Then of all things, he did not add power but pulled the nose up. Because of my recent check out, my alarm bells went off. I grabbed the controls, said, “I have the airplane” and pushed the throttles all the way forward. Much as I hated to, I also lowered the nose a little to prevent a stall, which scared the crap out of him. We may have been at 200′ by that time. We staggered forward at max power, finally struggled over the runway threshold and hit hard as the airplane basically stalled on the runway.
We taxied down the runway, told him, “you have the airplane”, and then sat back to wonder what might happen to me for taking over for the guy giving me my check, and who was a higher rank. No words were spoken as we got out of the airplane, went to base operations, checked into quarters, showered and went to the Officers Club. At the first drink he finally said, “well, you didn’t really have to do that”. Nothing more was said to me then or when we returned, but I reported it to save my ass. Fortunately, almost every other member of the crew came up to me to say thanks. There never was a doubt in my mind, or some of the crew, that I prevented an ocean landing.
There are defining moments in ones life that lead to another phase or level, and that was one for me. I gained a great deal of insight to who I was and what I could do. The rest of my tour in Hawaii I flew dozens of missions, and discovered by accident that other crew members had asked to fly with me. Apparently the word had gotten around. I therefore felt an extra responsibility but learned to love it.
The fact that essentially the same thing was going on in the 777, I find no excuse for a co-pilot not taking over, or even the one in the jump seat.
Very interesting story. I’m wondering if you would take a look at this later post, especially the last graph, and tell me what you think? The graph is something that may be unfamiliar, it’s a plot of the total energy (kinetic plus potential) of the aircraft. High and fast is greater energy, low and slow is less energy. The graph shows 138 flights, with the two highlighted flights being Asiana 214 on the day it crashed and two days earlier. They seem to be real outliers.
I’m blown away by the coincidence of the 2 Asiana flights at the high energy extreme. Makes one wonder if it isn’t the practice of that airline to be there; i.e., that’s where the company wants them. Or, perhaps the pilots themselves decide to do it that way. Also wonder if it was the same crew. Going to follow this as the NTSB goes through all of this. Thanks for the input.
With a lot of years on 3 and 4 engine airliners as a Captain I can tell you that you have done a very good analysis. Two points: the glide slope is flown from the position of the glideslope antenna on a given airplane and each airplane has a minimum Threshold Crossing Height. The pilot has to ensure the TCH for a given runway meets or exceeds that number and a good approach is typically 50′ or more over the threshold. This is one of several reasons the touchdown zone extends 3,000 feet down the runway. Secondly, the amount of power available and the time required to spool up are affected by a a high angle of attack which the crew of 214 found themselves, especially when 30-40 knots below the computed approach speed. Time and data collection may tell us why but this crew had an airplane too low, too slow, and probably behind the power curve (literally, not just the expression in the vernacular).
Cap’n John
A ‘friend’ who trained Asiana pilots sends this:
After I retired from UAL as a Standards Captain on the –400, I got a job as a simulator instructor working for Alteon (a Boeing subsidiary) at Asiana….
[snip]
Clint, I don’t want to be heavy-handed here, but this story has been sent from multiple sources. A quick web search shows that it’s all over the internet, on hundreds if not thousands of sites. I’ve removed it because (1) Everyone who cares about Asiana 214 has seen it already, (2) there’s no way to know it’s true, and (3) it uses a broad brush to paint all Korean pilots.
Looking back at subsequent data released by the NTSB. All equipment was operating “nominally” (everything was working OK) according to the flight recorders. The engines were spooled up at the time of the crash with the thrust levers full forward but the pilots manually moved them only 1.5 seconds before impact, far too late. The PAPI was indeed working correctly and the training pilot noted that he saw first three whites and one red, the normal 2 and 2, then 3 reds and one white and finally all red lights before the crash. The Auto throttle engage switches were in the engaged position when examined after the crash. The stall stick shaker engaged 4 seconds before impact. The planes nose was elevated as high as 40 degrees when the tail and the rear landing gear were ripped off by the seawall (the training pilot was heard to instruct the flying pilot to “pull up” before impact). Part of the tail remained in the water after impact. Speed was down as low as 103 knots and only reached 106 knots when impact occurred. Finally, the pilots said that they were having a difficult time aligning with the runway and flight path during the descent and were concentrating on this.
Shoot, forgot something. The ILS landing approach system was off at SFO on runway 28L because of the FAA. The FAA mandated that SFO implement the Runway Safety Area Enhancements. In their words “The Federal Aviation Administration (FAA) establishes design standards to ensure the safety of airports. These standards include criteria for runway safety areas (RSAs), which are clear areas around a runway, free of objects and structures. Similar to a truck safety ramp on a highway, RSAs are designed and maintained to enhance safety in the event that an aircraft were to undershoot, overrun or veer off a runway. RSAs also provide greater accessibility for firefighting and rescue equipment during such incidents”.
The ILS equipment was forced to be moved to a different part of the runway area and so was not available at the time of the crash.
http://www.flysfo.com/web/page/about/construction/runway/updates
numerous, good observations here. I liked this one: j32 says: (July 8, 2013 at 12:43 am)
“Good analysis. A few more factors. It sounds as if one of the pilots was on IOE (initial operating experience) meaning he was new to the aircraft but not to flying into SFO. The charted visual approaches at SFO will eat your lunch if you get behind early.”
I have flown T38s and KC-135s in USAF. Sink rate awareness was a big deal, especially in T38s. I used to teach the dangers of “getting behind” the jet when flying single engine approaches in the 38 … not enough thrust to go-around if excessive sink rate developed on an approach. It seems pretty obvious that the 777 crew let the approach deteriorate into excessive sink rate and slow airspeed while trying to correct the approach – they were “far behind” the jet.
BTW: this can happen in any airplane, even C172s and Cherokee 140s. Students pilots should be taught sink rate awareness (causes and cures)