Armchair air crash investigator – Asiana 214

Asiana Airlines Flight 214

Photo by David Eun.

At 18:26 UTC on Saturday July 6th, 2013 – Asiana Airlines Flight 214, a Boeing 777-200ER, crash landed at San Francisco International Airport. At the time of this writing, there have been 2 fatalities confirmed. This post is nothing more than speculation as to what occurred during the crash of Asiana Airlines Flight 214, and what caused the crash. I am not an air crash investigator and not associated with any aviation accident investigation agencies. This post is about what I think happened and not necessarily what will be discovered that actually happened.

Listening to LiveATC recordings, everything appears to be normal – there have been no indications of any mechanical or human issues and weather conditions were excellent; Asiana 214 heavy was cleared to land on runway 28 left. Moments later, ATC can be heard scrambling to divert and hold all aircraft coming in for landing or about to take off, the voice of the pilot of Asiana 214 can be heard but is unintelligible; ATC informs the pilot that emergency crews are on the way.

Various witness statements that I have heard throughout the day:

The very first thing I did after I heard that there has been a major crash at SFO was check Twitter. Sure enough Flightradar21 had confirmed that there had been a crash and they had the log of the final moments.

Asiana Airlines flight OZ214, Boeing 777, HL7742 has crashed during laning in San Francisco

— (@flightradar24) July 6, 2013

I also took a look at Flighaware’s tracklog for Asiana 214.

Asiana 214 Crash Tracklog

After looking at both sets of information, here are some things I thought about:

So we have a Boeing 777 coming in a bit hot and high on final approach – perhaps descending quicker than usual to get on the correct glide slope, but at the last moment we have a sudden change of attitude as the aircraft attempts to ascend with a low rate of climb and low airspeed.

So… what happened next? Well…

Asiana 214 Crash Runway Debris

Eyewitness accounts and this image, show that the Asiana Boeing 777 struck the seawall before skidding across and off the runway. Further images of the aircraft and of the runway show major pieces of the tail section, namely the horizontal and vertical stabilizers, as well as the landing gear and one engine had been torn from the aircraft and came to rest on the runway.

Runway diagram from Wikipedia

Right away, it is apparent that the aircraft landed not even close to where it should be. That strip of pavement right after the water is not the runway, but the blast pad, a section unusable to aircraft. This section is followed by the displaced threshold, which is used for take off, taxi, but not landing. Only after you pass these two sections do you get to the actual usable runway for landing which is indicated by the vertical white stripes and the runway numbers. On runway 28L, the distance from the beginning of the blast pad to the runway threshold (the vertical white lines) is some 300 feet.

Putting it all together

Four reds – you’re dead

So how did an advanced jet airliner with an experienced crew, on a clear day, land short of the runway? To answer that, you need to understand a few things about how airline pilots land aircraft.

Most, if not all landings done by airline pilots are landings using the ILS. ILS stands for Instrument Landing System and its purpose is to provide landing guidance to the aircraft and crew should they not be able to see the runway visually. Different airports have different types of ILS systems, some only provide vertical guidance others only horizontal guidance. But most major international airports provide extremely precise ILS systems that both provide vertical (glide slope) and horizontal (localizer) guidance and could work in sync with the aircraft’s autopilot to land the aircraft automatically. Now its important to understand that the majority of time, landings, at least in the final stage are done manually, even when the ILS is operational.

At the time of the crash of Asiana 214, it is my understanding the ILS systems on 28L/28R were inoperative, a look at the NOTAMs confirms this and the cause of this inoperation was because the guidance systems had to be adjusted due to a change in touchdown location because of some work being done on the runway.

With the ILS inoperative the only other way the pilot could land would be visually – that is, looking out the window and using your eyes (looking at the instruments to cross check would be a good idea too) to gauge distance and height. Landing visually requires obviously, the ability to see the runway out the window without any obstructions from clouds or other weather, this is known as VMC (visual meteorological conditions). VMC prevailed at the time of Asiana 214′s landing. With a visual approach and landing you still have some tools you can use, namely the PAPI/VASI.

PAPI Explanation

PAPI (and VASI), which stand for Precision Approach Path Indicator and Visual Approach Slope Indicator, respectively, are a set of lights that assist the pilot when landing visually. A runway either has PAPI or VASI lights, not both. The lights change color depending on your angle relative to a specific point on the runway.

And the cause of the crash was…

I don’t know – everyone is just speculating at the moment. At the time of this writing, the NTSB is in possession of the flight data recorder and cockpit voice recorder – once those are analysed we will have a much better understanding into the final moments of the flight.

Nevertheless, I am an armchair air crash investigator so I will offer up some theories starting from most unlikely to most likely:

  1. Icing in the FOHE – Another crash of a Boeing 777, British Airways Flight 38 had a similar setting. The plane crashed short of the runway on final approach, the plane declared an emergency only seconds before hitting the ground. The cause of the crash was the inability of the engines to produce increased thrust due to icing in the fuel oil heat exchanger (FOHE). The aircraft, which was already slow and low due it being on short final, could not maintain its airspeed and was at risk of stalling – only by lowering the nose and increasing the descent rate to make up lost energy was the crew able to avoid a stall – at a cost of landing short of the runway. The reason why a repeat of this accident is extremely unlikely is because it was found that only Rolls Royce engines had this design defect. The Asiana Boeing 777 did not have Rolls Royce engines.
  2. Unawareness of inoperable ILS – The ILS on runway 28L/28R was not functioning due to runway repairs. More specifically the runway repairs had to do with changing the touchdown position. Whether or not they could pick up the ILS frequency is another question. But if they could pickup the ILS frequency and proceeded with an ILS landing, either using the autopilot or just following their instruments, it is a somewhat plausible theory that the crew proceeded with an ILS approach using erroneous information from an inoperable ILS which caused them to come up short of the runway. The problems with this theory is that: it suggests that they could even pickup the signals from the ILS, that the crew did not review airport documentation and NOTAMs, and that they simply did not look at all out the window and see that they were too low, either by their own visual perception or through the red lights they would most likely see from the PAPI lights.
  3. Confliction with autopilot/auto-throttle – The Boeing 777 is an advanced aircraft, and advanced aircraft typically have an auto-throttle (A/T) feature. As the name suggests, A/T allows the pilots to select a speed they wish to hold and the engines will automatically adjust thrust based on the aircraft’s attitude. Assuming (and this is mostly guessing) that the A/T was on and they selected a slow speed (so they could descend quick as they did) and that the throttles when physically moved have no effect when the A/T is engaged, it is possible that when the crew realized that they were too low, began a go-around by pitching the nose up and putting the throttles forward. However if the A/T negates any physical inputs from the throttles, it is possible that the crew forgot that the A/T was engaged, realized that they were too low, attempted to push the throttles forward, got no response, realized that the A/T was engaged, disengaged the A/T, pushed the the throttles forward, but at that point it was too late – the engines would not spool up in time and the aircraft impacted the seawall.
  4. Just a bad approach – Looking at the track log, the pilots were in a rush to get down. As I wrote before, not a nosedive by any means, but they did come in hot and high. They also bled off quite a bit of speed on their way down too. I believe the pilot flying, in an attempt to “catch” the glide slope, descended too quickly and inadvertently cut through the glide slope – going from too high to too low. From the pilot’s point of view, he would see the PAPI lights quickly go from four whites to four reds. At this point, I believe the aircraft was dangerously low to the water and the pilots realized this. However, this realization came too late – the aircraft was too slow to be able to climb. The pilot attempts to initiate a go-around, the nose goes up and the descent rate decreases. The pilot unsuccessfully attempts to climb, as the engines did not have time to spool up and provide the neccessary thrust to raise the airspeed. The aircraft then strikes the seawall – tearing off the landing gear, tail section, one engine, and most likely causes damage to other control surfaces. The aircraft climbs a bit however, the final nose up input followed by the impact most likely caused an aerodynamic stall which, combined with a loss of directional control from the loss of the tail section and engine, causes the aircraft to crash into the ground. The aircraft skids onto and off the runway where it makes a near 180 degree turn swerves about 40 degrees to the left on the grass and comes to rest.

Again, I am no air crash investigator. These are just my conclusions reached through all the data available and my own aviation knowledge. My conclusions rely on data from sources like FlightRadar24 and Flightaware, which are not 100% accurate. The true answers will most likely be found in the black boxes that the NTSB has in their possession at the moment.

NTSB press conference:

Target speed for the approach was 137 knots. #Asiana 214

— NTSB (@NTSB) July 7, 2013

The throttles were advanced a few seconds prior to impact and the engines appear to respond normally. #Asiana 214

— NTSB (@NTSB) July 7, 2013

Sound of stick shaker began approx. 4-sec prior to impact. #Asiana 214

— NTSB (@NTSB) July 7, 2013

Call to go around made approx. 1.5-sec prior to impact. #Asiana 214

— NTSB (@NTSB) July 7, 2013

  • John

    As a fellow armchair investigator, I think #1 (FOHE icing) can be ruled out. Not only is it the wrong engine model, and the RR engines have been fixed, but BA38 flew through the arctic circle in January. This flight flew over the Pacific in July.

    It does get cold up at 35,000 feet, but BA38 was basically the coldest possible situation for any jet engine. Plenty of other 777′s fly (even with RR engines, and even prior to the fix) without icing problems.

    • Birdstrike

      RE: ” BA38 flew through the arctic circle in January. This flight flew over the Pacific in July.” There isn’t much difference, really, because it’s about 60 degrees below zero even over the Pacific at jet altitudes. Seoul to SFO is a long flight. Maybe FOHE is in fact not a factor, but it wouldn’t be because of the route of flight.

  • Steve

    I did a similar analysis as you (see : I think you’re wrong about the climb, though. The last radar return is probably a Mode C return after the aircraft has crashed, and hence completely unreliable. If the aircraft were at 200 ft and climbing at only 85 knots, the results would have been much worse.

    • Moma Vujisic

      I agree about the altitude – there’s no way the aircraft could of been at 200ft. However, I believe that at the last moment when the pilot was pulling up after they struck the seawall – that they could of possibly climbed a tiny bit, stalled, and fell. The final readout has a +120 FPM – I think it’s possible that the 200ft altitude reading was just that of Flight Aware’s systems interpreting the data and projecting an altitude. Of course, I could be totally wrong and the aircraft could of just pancaked on the runway – we’ll see.

  • Marcel

    Even with ILS out, you still have the RNAV (GPS) approaches (which even the little Cessna I trained in could do). In addition, even visual approaches have set distances/altitudes that keep you on glide-slope, so the PAPI/VASI lights are not your only indicator.

    • Moma Vujisic

      You’re correct – the reason I only included the PAPI/VASI was that in my opinion, it would be the main tool, if not the only tool a pilot would really use when making a visual approach on a clear day.

  • Linas
    • Moma Vujisic

      Thanks for this – I have not seen an aerial overview of the crash site yet (news media keeps showing a zoomed in image of the aircraft). I heard from another source that the aircraft turned 180 degrees and could not verify myself.

      • Steve

        On the news they called it a 360 degree turn so perhaps it did spin horizontally, just greater than the 180 degrees.

        • Moma Vujisic

          While nothing can be ruled out, I seriously doubt there was enough energy to make the aircraft spin 360 degrees. And if there was, I would definitely think that there would be significantly more damage both the aircraft and to the passengers.

    • TheRealKorbenDallas

      The cellphone video available on CNN, albeit unclear, shows quite a bit of rotation with one wing lifted into the air. The plane skids with only nose and the left wingtip touching the ground. It certainly does not look like just a “34-40 degree swerve”. It does look like the plane made a 180 degree turn and then some.

  • Mike Czarny

    nice write-up

  • Ron

    Another possible contributing factor is spool-up delay. It can take several seconds for a jet engine to develop full power from flight idle once the throttles are pushed forward. If your approach isn’t stabilized you can end up low and slow and not realize it until it’s too late to recover.

  • pogue972

    CNN has just published some low quality cell phone video of the crash & The NTSB are answering questions on the crash right now, but I’m sure the interview will be available later somewhere (via NBC)

  • Rob

    If the captain was not at the controls, based on your analysis, would there have been any practical moment to take over the aircraft from the co-pilot and try to solve the problems with the approach?

  • Pingback: Asiana Flight 214: A Pilot’s Perspective | Marty Andrade


    Weren’t the PAPIs NOTAMED OTS also?

    • John F. Miller

      Per NTSB briefing, the PAPI was operational at the time of the crash, but was destroyed as part of the impact.

  • Jack Hamm

    What upsets me most is in the photo at the top: people with luggage. Talk about wasting precious seconds when a fire is imminent.

  • Chris

    Great write-up – written with intelligence rather than speculation. Of the many forums and discussions taking place on this incident, I haven’t heard anyone mention one intrinsic “trait” of the 777: It’s VERY “touchy” on approach. The aircraft has two huge engines that produce an enormous amount of thrust. Pilots – especially younger/less experienced pilots – often have difficulty on approach because of this. 777′s typically “porpoise” on approach, as just the slightest touch on the throttles will produce significant thrust which will, in turn, pitch the nose up. The converse is also true: the slightest reduction in thrust can cause the nose to drop. I imagine… also in my “armchair” analysis… this is a contributing factor – especially if the pilot is fairly new to the 777. I would imagine that when pushing the throttles forward at the last second (to slow/stop the descent) caused an exaggerated nose-up attitude, further compounded by the knee-jerk reaction of the pilot to pull the nose up. This resulted in the tail-strike against the seawall which catastrophically damaged the aircraft. Although I haven’t done a thorough analysis of the data, it would be interesting to see a simulator re-creation where even though the aircraft was below GS, the pilot maintained the required pitch angle and continued to land. I would think the point of impact would have been about the same, even though they may have lost the landing gear (clipped the seawall) but still able to land on the belly. Obviously I’m no expert, but I’m really leaning towards pilot error all the way around.

  • Guest

    The cellphone video available on CNN, albeit unclear, shows quite a bit of rotation with one wing lifted into the air. The plane skids with only node and the left wingtip touching the ground. It certainly doe snot look like just a “34-40 degree swerve”. It does look like the plane made a 180 degree turn and then some.

  • Sharat

    your understanding of how the A/T works is a bit flawed – the engines will respond to manual over-rides even when the A/T is engaged. The A/T does not negate any manual input as such. It is just that unless the manual input is sustained, the A/T will ultimate override the previous input and revert to doing what it is commanded to do – hold the selected airspeed either from the FMC or MCP.

    • John F. Miller

      You seem to know something about the A/T. What would it take for a crew to either a) believe the A/T was engaged when it was not or b) set the A/T to a speed significantly below the intended approach speed. When I herd the NTSB report on the BlackBox auditions, the think that caught my attention – beyond the slow air speed – was that the throttles were set a idle until seven seconds before impact. It immediately made me wonder if the crew assumed the Auto Pilot had throttle control when it did not.

      • Phasy

        My knowledge is from the 737-800, however they’re very similar airplanes. When you disconnect the A/T in the 738, an audible click is heard from the switch moving and the A/T disconnect warning light blinks until you disable it. The button is on the throttle, so your hand would have to already be on the throttle to do this (or the MCP but that would be way more deliberate). It would be impossible to fail to notice this. The speed selector switch is of-course on the MCP directly in front of you. You wouldn’t set this below final approach speed for any reason that I know of, even when your speed is too high. When you don’t use the A/T, you still set the A/T speed on the MCP, as it is also indicated in the speed ticker and used to fly the approach. Final Approach Speed set is also a checklist item.

        The throttle being at idle makes me think of Turkish Airlines 1951 that crashed at Schiphol, Amsterdam in 2009. This plane had a faulty radio altimeter (measures the exact distance to the ground with radar, not air pressure). Upon starting the approach, the radio altimeter indicated -6 feet, causing the plane to close the throttle (retard) for landing. The crew started their descent high and so had plenty of leftover energy, they understandably failed to notice it until the final moments.

        The A/T systems are incredibly complex, they have various phases and states of operation, data sources, control inputs and warning systems. Like the author, I also tend to think this had something to do with the A/T. However, it’s not just cruise control, where you set a speed and the computer does the rest. It’s a bit trickier. :)

  • WWW

    Combine the data from FlightAware with the LAT/LONG and minimum altitudes of the GPS approach plate. (I plotted them on Google Earth to get a visual). Also look at the previous flights of Asiana 214 into KSFO. You will see that the airspeed and rate of decent at 2,400′ was “normal” compared to previous flights. Unfortunately, the accident aircraft was relatively high all the way down the glideslope while its airspeed decreased to near stall speed approaching the ramp. (The plane was coming down while the nose was rotating up). With the downward velocity vector, the lack of lift (low airspeed) and engines at low thrust, the plane was already committed into the ground. Pulling back on the controls did nothing more than rotate the tail into the ground.

  • DJ

    Good analysis by an “Armchair Investigator”. True & correct facts will be revealed by the NTSB.

  • machadolab
    • Zero Cool


  • Johnny54

    The video appears to show the aircraft sliding while making a counterclockwise rotation of approximately 330 degrees. This would be consistent with the tail breaking off. Without the downward air pressure from the tail section the nose would be driven further into the dirt. As the aircraft spun near the 180 degree point it caught some lift under the rear of the main wings and that lifted the tail. Another contributing factor to the tail rising might have been the air rushing thru the gapping hole in the empennage. Some very lucky passenges to survive this accident.

  • Pedro

    I’m sorry to criticize. It’s a good effort but it has some inaccuracies:

    1) You may be right since I haven’t plotted the vertical descent profile. But how do you know the aircraft is hot and high? The fact that you have a high V/S doesn’t necessarily mean you’re high. It only means (at least) that you’re vertically FAST. So they were at least “hot”. But to find out if they were “high” you’d have to calculate the actual flight path angle. And to do that you must cross-check Ground Speed VERSUS Vertical Speed – not only Vertical Speed.

    2) 160 Kts on final approach is perfectly acceptable even on fully normal operations. It may be a “Decelerated Approach” as opposed to a “Stabilized Approach”. It’s common practice given appropriate FMA (Flight Mode Annunciator) modes.

    3) “Blast pad” is unofficial lingo. The correct one is Clearway.

    4) Avoid using the term Glide Slope when the NOTAMs state that the Glide Slope is U/S. The correct term in non-precision approaches such as this Localizer-only approach is Glide Path.

    5) “Different airports have different types of ILS systems, some only provide vertical guidance others only horizontal guidance.” I’ve never heard of Glide Slope ONLY approaches! But aviation can be exotic at times. Can you provide with an example? As far as partial ILS goes, I’ve only heard of LOCALIZER-ONLY or LOC-DME approaches…

    6) “With the ILS inoperative the only other way the pilot could land would be visually” Not true. They had localizer. So it categorizes as a non-precision approach. Which implies a higher MDA (Minimum Descent Altitude) than the Full ILS DA (Decision Altitude). You don’t need VMC (Visual Meteorological Conditions) to land with an NPA. You need appropriate visual references at MDA/DA – which may be VERY different from VMC.

    7) “However if the A/T negates any physical inputs from the throttles” This is simply FALSE. You can easily override the A/T back-driving of the Throttles by moving them manually. If you do that, however, the A/T will slowly drive them back to the Thrust Lever Position calculated as required to react to the speed increase caused by your manual input (should be close to or at Idle position) IF you haven’t signalled to the Auto Flight system that you intend to Go Around (such as pitching up to 15º or, correctly, press the TO/GA switch before you do that).

    7a) Having said that, I do agree with you when you say the A/T operation (or lack of awareness to its operation) may have a played a crucial role in the crash.

    8) “At this point, I believe the aircraft was dangerously low to the water and the pilots realized this. However, this realization came too late – the aircraft was too slow to be able to climb.” Although what you say is true, it does require some clarification: Going low is, obviously, unsafe. However, being low itself isn’t necessarily critical enough to cause a crash if you’re not terrain restricted (over sea). If you fly the correct approach speed correctly (in this case 137 kts) it should cater for more than enough margin for small-to-moderate vertical corrections or even an abrupt terrain avoidance maneuver. The problem was that they were low AND slow.

    Anyway, I appreciated reading your article, and it was my intention to constructively criticize. :)

  • Pingback: Asiana Flight 214: A Pilot’s Perspective II | Marty Andrade