JAL incident at Haneda Airport

[xcris]

Originally Posted by Iron Duck

There's what the Captain says, and the evidence of what his aircraft did. The evidence is that the aircraft taxied straight onto the runway and lined up, then waited for 30+ seconds, and then (perhaps) started rolling in the final second before the collision. From this we can infer that its crew believed not only that they were clear to line up and hold, but also to take off. The 30 second wait? Spoof instructions to line up and hold, and then to take off? Only the CVR can reveal this, but whatever happens its contents are going to embarrass someone, and this is difficult.

Right in all said. But what if they realized at the last moment that they were not permitted to enter the runway and tried to move? The CVR will clear this.
When I said "facts" I meant the [ATC] comms transcripts revealed (so far). The Captain, from his first statement - which was a bit premature, IMO - put emphasis on the fact they were cleared for t/o. None of the transcripts helped him (so far). Ofc, with the benefit of the doubt.

[FollowTheSupper]

ADS-B Position Accuracy

Yes, indeed!

In case the wrong impression of ADS-B capabilities might have been gathered from earlier posts, the inset in the accompanying screenshot, clearly shows that the Estimated Position Uncertainty (EPU) for the final ADSB position of the A350 (as downloaded from the ADSB Exchange website), is in the second-highest accuracy band (i.e. EPU <10m). Since the highest accuracy band for ADS-B is defined as EPU <3m... then the correct answer to Maninthebar's query about ADS-B derived position accuracy (at post #597) should effectively be: "only between 3m and 10m".

As the span of the horizontal stabiliser on a Dash 8 (at approximately 10m) is comparable to this error limit, then this also supports EXDAC's rejection of using the ADSB-derived position as reliable information to determine the lateral relative alignment at the time of collision (post #631).

I will speculate that physicus's response of: "A few centimetres" (post #601) is merely the result of confusing the available precision implied by the standard data string format (in which latitude and longitude values are each resolved to ±0.000001° - see the red-underlined figures in the accompanying image), and the actual accuracy of the position derived by the receiver from the GNSS signals.

At Tokyo's latitude, a change of 0.000001° in latitude equates to a displacement of approximately 12cm... while in longitude, (which is more relevant to lateral position on a northerly runway), the same angular change represents a displacement of just under 8cm. These geographic coordinate resolution values are therefore both equivalent to only "a few centimetres" of displacement. However, as is typical of digital systems, the ADS-B data output format appears to retain its 6-decimal-place resolution throughout, regardless of any variation in the accuracy of the GNSS-derived position which is being parsed into it. It should be noted that the error can (and often does) vary significantly, and that the FAA guidance for approving ADS-B equipment (see document link provided in EXDAC's post #631), only requires that the horizontal position error should remain below 0.05nm (or approximately 90m, i.e. wider than most runways!), for a unit to be acceptable for use within controlled airspace.

Despite the (perhaps comforting?) "digital" nature of the signal processing, it is emphasised that a position solution from any GNSS equipment is not an "exact science"... and is only an "estimation" of the true position... based on being the "best fit" to the available satellite signal data. When speed of computation is set as a priority over accuracy of result (as is generally the requirement in a moving vehicle), then the lower accuracy of the resulting position is unlikely to provide much forensic insight.

In special circumstances. GNSS technology is potentially capable of sub-millimetre relative measurement over short ranges (e.g. <100m)... albeit when assisted by differential techniques to post-process long-duration, static observation data acquired in wide open spaces (i.e. away from spurious reflected signals), using dual-frequency, survey-grade antennae (with the Phase Centres calibrated to sub-millimetre level). Unfortunately, in the real-world, the effective accuracy of most GNSS equipment in routine civilian use, will be somewhat more disappointing... possibly by 3 or 4 orders of magnitude, at times!

Not to discourage the computer-modelling community from their valuable contributions... but I well remember the rather sarcastic cry of old (from engineers looking out of the window as the reality of events diverged from the model... mostly due to a poor appreciation of the error sources): "Well the numbers are displayed on the screen and they're in colour... so they must be right! (... and remind me again, just how much we're paying for this equipment?)"

[jumpseater]

Originally Posted by sunnySA

I’d be interested in 3 things.
1/. listening to the instructions from the Ground controller prior to handing over to the Tower controller.
2/. for the Dash 8 details time out of the system (finish) as they reached the holding point ?
3/. Intercom communications between Tower and First responders including the Fire Fighters at the airport. In particular was only the A350 mentioned or were both aircraft mentioned in the initial communicate to all out the fire trucks?

My thoughts
1/ Probably routine with no surprises. The departure clearance (airborne route) could be given by "delivery" or "ground". The DH8 then taxi's and when established on C taxiway gets transferred to tower frequency.
2/ Don't understand what point you're making, needs clarification
3/ Clarification first.
Airport fire services =RFFS
External public fire/medical/police = first responders
Accident occurs crash alarm activated. RFFS emergency deploy to accident location. RFFS will get immediate basic known detail on first call from tower controller. Likely: a/c type, location ,nature of accident as best known, and these days wind direction. Wind direction to ensure that people are out of any hazmat trail direction. At that first call they have seen an accident but may not be aware of two aircraft involved, here we can assume ATC didn't know DH8 on runway. RFFS crew will aim to get to site as fast as practical. One fire station (Wes)t is about 3km from C5, East station about 1km. Think about their (RFFS) task, they have to get PPE on, in vehicles and start briefing on the move with info received from ATC. From a standing start, they are manning up a 8.0T vehicle and driving it 3km and are expected to do this in around 3 minutes.. . Due to apron movements and it being at night they are unlikely to be given an unrestricted route, they will need de-conflicting from that traffic with the most direct route. Probably co-ordinated between ground and tower as to who gives that route/pathway to accident site. Ground and tower also stopping any aircraft ground movements instructing go-arounds for inbounds. RFFS also need to see and avoid other vehicles and airport personnel whom are on the apron or service roads. Depending on when it was realised there were two accident debris locations all vehicles may have been sent to C5 initially due mass of flames and obvious accident at that location. It then becomes clear there are two locations, a rapid decision has to be made to split the RFFS team and to what degree. It may be that two sites were immediately apparent, in which case the deployment workload above is in effect doubled as two teams are being sent and separately co-ordinated to two different locations. On arrival at the location they will determine offensive or defensive action and deploy appropriately, bearing in mind they will have passengers, walking wounded etc to avoid whilst deploying. Their prime job is RESCUE, the total fire extinguishing is secondary. Their job is to save lives and make as many escape routes as practical.

External first responders will arrive as fast as practicable to known rendezvous points at an airport. They will then need escorting by trained airport staff to the accident location/s. That may take a bit of time as you'd ideally need a number of vehicles rather than trying to do it in single numbers.

[xcris]

Originally Posted by aox

C is a taxiway and not also a runway

You will be surprised! In some AD charts (i.e Civil Aviation Bureau, Japan EFF: 18 Jun 2020) on pages regarding the special attention taxi routes you'll find the runways depicted by letters: A-RWY, B-RWY, C-RWY (that's 16L-34R) and D-RWY respectively.
Maybe in the "internal culture" @ Haneda this runway designation might be trivial, since keeps coming on? Maybe someone, current with RJTT, can help with an explanation?
i.e. excerpt from AD 2-45:

[tsumini]

Not convinced RH engine #2 was running (as assumed here several times) when evacuating RH front exit door. Unable to post photo due not enough posts.

[JP7000]

Both Sankei and Yomiuri newspapers report that "more than 3" (Yomiuri) and "all six" (Sankei) crewmembers on the Dash-8 had their headsets on as part of routine procedure and were listening in on the communications with Tokyo Tower. The surviving captain insists that all crewmembers agreed that they had been given permission for take off. Sankei (mostly Google translate with my corrections) article below. (DANGER: Please note that this is a double-translation so the technical terminology can be imprecise, note in particular the ICAO-compliant official term "runway holding position"= 滑走路停止位置 as defined in Japan in 2012. In Japanese "holding" is usually "待機" and "stop" is "停止" so a layperson would read this as "runway stopping position" but I suppose there is some logic behind the reason why this word was chosen. The Ministry transcript and newspapers are therefore correct to use this term.)

"It was revealed through interviews with inside sources on the 5th that in the accident at Haneda Airport where a Japan Airlines plane collided with a Japan Coast Guard aircraft, the captain of the Japan Coast Guard aircraft (39) was interviewed after the accident and explained that he had "confirmed with other crew members" about the aircraft's approach to the runway. At the time, everyone on board the Japan Coast Guard aircraft was able to hear the communications, and it is possible that the Japan Coast Guard aircraft may have misinterpreted the air traffic control instruction of ``1st (scheduled takeoff)'' as a priority takeoff. It was also revealed that the air traffic controller at Runway C, where the accident occurred, explained in an interview with the Ministry of Land, Infrastructure, Transport and Tourism that he was "unaware of the Japan Coast Guard aircraft's incursion into the runway." The Japan Transport Safety Board will analyze the voice recorder recovered from the Japan Coast Guard aircraft and investigate the situation at the time in detail (it is reported elsewhere that this device built by Honeywell will be analyzed with the help of NTSB of the United States. Japanese officials were unclear whether the voice recorder will be flown to the United States for analysis or NTSB officials would come to Japan instead)

"According to communication records, at 5:45 p.m. on the 2nd, an air traffic controller instructed the Japan Coast Guard aircraft, "No. 1. Please taxi to the runway holding position (Japanese ICAO compliant official term 滑走路停止位置) on C5." After eight seconds, the captain radioed back the instructions, but the aircraft continued onto the runway and stopped on the runway for about 40 seconds. It collided with a JAL plane that landed immediately after that at 12:47. According to coast guard officials, there were six people on board the coast guard plane at the time, including the captain and the five who died. All of them were wearing aviation headsets with integrated headphones and microphones, allowing them to hear radio communications with the control tower. Normally, instructions from air traffic control are given to the captain and co-captain, and the captain explained in previous interviews that the aircraft had "entered (the runway) with permission." He said in a subsequent interview that he had "confirmed (the control instructions) with other crew members." According to the Ministry of Land, Infrastructure, Transport and Tourism, there is no special term in the air traffic control terminology that indicates the order of takeoff and landing that encourages priority takeoff. The Japan Coast Guard plane was scheduled to deliver relief supplies to areas affected by the Noto Peninsula earthquake, but there was no mention in the communication records that permission was given to enter the runway."

Note: It is maybe my irresponsible observation that the Ministry of Land Infrastructure Transport and Tourism (which overseas air traffic control etc.), and the Japanese Coast Guard do not exactly see eye to eye now concerning this issue. But until the Noto earthquake situation with many missing persons is solved, the JCG must concentrate on more important rescue and relief efforts. Concerning this accident the Tokyo Metropolitan Police have launched an inquiry due to possible "criminal negligence" ("crime of corporate manslaughter") with "unspecified accused".

[waito]

Time to update the swiss cheese list

Originally Posted by andrasz
From what we know so far, a classic case of Swiss cheese that will likely make it into future textbooks. Investigation will take months to years, but I doubt any significant new information will be added other than a more elaborate analysis of the individual holes:

• 34R was T/O only runway previously, JL516 was the first to use it for landing. The MA722 crew may have had a mental picture not to expect any landings on this runway. For same reason, they may not have expected the need to hold short.
• Usage of "Number one" by ATC may have further reinforced the MA722 crew, missing the "abeam C5" or misunderstanding it as hold ON RWY abeam of C5 (which is exactly what they did). While usage of English in all ATC comms in Japan is commendable, in this case it probably just added to the confusion.
• Stop bar and taxiway lighting INOP.
• JL516 and MA722 on different frequencies
• MA722 was sitting aligned on the runway for 45+ seconds. Probably Japanese culture at play, it is impolite to challenge authority or appear impatient, they were likely patiently waiting for ATC to clear them for T/O.
• DL taxied past MA722 a good 20-25 seconds before the collision, it was already out of their view and expectantly on its t/o roll as the A350 was approaching, so no extra set of eyeballs to wave off JL, as it was done in SFO.
• The combination of HUD and LED lighting probably prevented the JL crew from seeing MA722 even at close range, I'd wager they never saw what they hit.
• Visibility of a DH3 from the rear in the dark is practically nil. The only visible tail light is white, probably flooded out by runway lights, and the flashing orange upper ACL is obscured by the high tail from the approach angle. The otherwise fully warranted SOP of not using strobes until the beginning of T/O roll did not help in this case.

The only important question remaining is why did the approach controller not notice that 34R was occupied ? Even if MA722 was not where it was supposed to be, surely HND has SMR, there should have been both visual and aural warnings as a measure of last resort.

From my list, the two bold probably the primary factors:

Certainly the Dash-8 crew's missing of checking the approach path, where they clearly should have seen the A350.
The discrepancy of the Dash-8 crew in reading back an instruction to proceed to Holding Point C5 only, but still intruding on a runway with mindset to takeoff. Now we could argue they were disoriented and not aware of having entered an runway - active Runway - Rwy 34R. But then sitting 45 seconds on that runway and not finding out what it is and that they entered it despite a hold short instruction was another big glitch. Let's see.
The situation of degraded airport lights versus Tower operating a complicated scenario and not extra checking? What Radar tech was available?
How about transponder replies from Dash-8?

I want to add a contributing factor:
Missed opportunity by ATC (Tower) to detect the Rwy intrusion: by primary/secondary response, by Alert system (which was stated to have functioned?) or by visual inspection (landing light and strobes on Dash-8 were on)

So IMHO its mainly Human factors, and we now need to wait for the investigation results, what affected those. We can only speculate, but I will not take part on that. Maybe I will reject weird speculation, let's see.

The Coast Guard Crew were on a honorable, demanding mission with priority, so I see absolutely no reason to premature bash any of them!!!

[xcris]

L.E. For those still in the "LETTER-RUNWAY" dilemma:
Surprisingly, I found this on Wikipedia: "Both the Tokyo Monorail and the Keikyū Airport Line added stops at the new terminal, and an international air cargo facility was constructed nearby. The fourth runway (05/23), which is called D Runway, was also completed in 2010, having been constructed via land reclamation to the south of the existing airfield. This runway was designed to increase Haneda's operational capacity from 285,000 movements to 407,000 movements per year, permitting increased frequencies on existing routes, as well as routes to new destinations."
Source: https://en.wikipedia.org/wiki/Haneda_Airport
There is a reference, too, to an old article: https://web.archive.org/web/20111008...h-report05.pdf
Conclusion: Indeed, at Haneda, the runways to be designated by letters is common practice.
See below for yourself!
Now, I am curious if this common practice sometimes is taken to the comms between (some) controllers and (some) domestic flights?

[woocash]

VasAviation video starting here is a great visual explanation why the crew of the dash was reinforced in thinking they have takeoff clearance:

You Tube

They had no idea about JAL516, and they heard JAL166 being talked to continue due to one departure ahead. Seems like these otherwise normal comms may have reinforced their misjudged clearance

[AirScotia]

Originally Posted by JP7000

Both Sankei and Yomiuri newspapers report that "more than 3" (Yomiuri) and "all six" (Sankei) crewmembers on the Dash-8 had their headsets on as part of routine procedure and were listening in on the communications with Tokyo Tower. The surviving captain insists that all crewmembers agreed that they had been given permission for take off. ".

I find it odd that the permission had to be queried and confirmed with the non-flying crewmembers. It suggests to me there was some uncertainty as to what they'd heard.

[Fly-by-Wife]

Originally Posted by tsumini

Not convinced RH engine #2 was running (as assumed here several times) when evacuating RH front exit door. Unable to post photo due not enough posts.

So what would explain the smoke, flames and sparks coming out the back of it? As is clearly visible in numerous clips showing the evacuation - e.g.

You Tube

[kontrolor]

that's why taxiing instructions do not include any kind of mention of the runway except at the beginning. Line up and hold has long been replaced by line up and wait, hold short was not used already when I was TWR atco some 25 years ago. Plus, DH8D crew seems never to report their position, not even when lining up. Have they done that, they would have probably lived to tell the tale as there would be enough time for A350 to initiate go-arround. Increasing number of runway incursions on busy and complex airports point to only one conclusion - ADSB neeeds to be mandatory (as 8.33 channel radio station) and all TWR controllers should be trained in ground survelliance and there should be software in place, detecting faulty positions of aircraft. There is no more excuse for not doing this. Crew under pressure, visibility issues, complex taxiway systems....all this is contributing factor not only to this, but also to almost all simmilar cases...

[xcris]

Originally Posted by AirScotia

I find it odd that the permission had to be queried and confirmed with the non-flying crewmembers. It suggests to me there was some uncertainty as to what they'd heard.

This and many others. I am starting to think that how that flight was operated (MA722) and what happened there was SOP at all :/

[tsumini]

Much windier than most people realize. Same direction as engine exhaust. Fuel source of fire is behind engine, not from the engine. Wind blowing around engine. People coming down the slides have clothing blowing pretty hard too.

[Jasonbay]

Originally Posted by waito

17:47:27:

CCTV2 shows: Dash-8 starts to make a small move forward less than 1 second before impact, which coincides with the later statement of the Dash-8 Captain to just have started the takeoff.

Originally Posted by Easy Street

The Dash-8, or at least its fuselage and cockpit, hardly moved. I don't see any evidence of parts of any significance being accelerated to 100kts. The impacts, firstly on the stabiliser and then on the wings, were almost certainly so energetic as to disrupt the structure (in layman's language, rip the tail and wings off) before any sustained accelerative force could be transmitted to the fuselage. Essentially the fuselage was left sitting where it was as the A350 ploughed on through. The instantaneous acceleration (jolt) would have been massive, but very brief. I suspect the deaths of the other crew members will either have been due to being directly contacted by the A350 NLG and/or the damage it caused, or by being incapacitated/trapped and then burned. The captain's survival is certainly miraculous.

Originally Posted by DIBO

Indeed, see attached mp4 (via .zip as per upload requirement), the red circle marks the Dash8 with nosewheel taxilight and vertical tail (LH side) clearly visible. A350 approaching with nosewheel light and LH wingroot landinglight clearly visible.
Dash8 taxilight does not move until completely overran by A350 'nosewheel light'. And even then only moves 'slightly', about the same time the A350 nosewheel light seems to tumble and disappear.

I don't see any support for the recurring idea (besides the JCG pilot's statement) that the DH3 was commencing its takeoff roll. First and foremost, the LDG/APP lights are never turned on. Second, as DIBO points out, the A350 has basically overlapped with the majority of the DH3 by the time any lateral movement is seen.

See frame-by-frame analysis of the SinglePreference CCTV footage (www.youtube.com/watch?v=6NbVdIoJsHY) here: www.youtube.com/watch?v=JLNivIxZRJw

F-32: DH3 tail illuminated by A350 lights, as previously noted.
F-31: The DH3's strobes flash. We use the mid-point to estimate the position of the DH3's wingbox, as a longitudinal reference point to assess relative position with the A350.
F-22: Maximal illumination of the DH3 tail by the A350 probably indicates the instant right before collision.
F-20: It appears that the DH3 vertical stabiliser (or a part thereof) translates forward while the DH3 taxi light remains motionless. As Easy Street notes, this probably means the tail sheared off. It's essentially caught between a rock and a hard place. It can't stay in place because that would mean cutting through the much larger and more massive A350 fuselage. It can't collide inelastically with the A350 nose and translate forward while remaining attached, because that would mean instantaneously accelerating the DH3 (which is massive, though not as massive as the A350) to match the A350's speed. The only way to square the circle is for the stabiliser to shear off.
F-17: Reflection from an object occupying the same space as the vertical stabiliser, even after the A350 NLG has passed through, suggests that the stabiliser sheared at a point other than at its base attachment to the DH3 fuselage.
F-16: We note that the A350 NLG is now in the same longitudinal plane as the DH3's wings. This corresponds to an overlap of about 12m between both aircraft, longitudinally.
F-13: The DH3 taxi light begins to move.
F-11: The A350's upper and lower anti-collision lights flash. Comparing against a picture of a JAL A350-900, estimate that the lower ACL is at the same longitudinal position as the rear of the DH3, which corresponds to a near total overlap of the 25.68m long DH3.
F-10: Both aircrafts' NLG lights experience a change in intensity, which suggests an abrupt shift in the lights' focal axis. This is around the time that the A350 NLG shears off. As for the DH3, its light increases in intensity potentially because of a collapse of its fuselage, aiming its light more directly at the camera.
F-07: The A350's NLG and NLG light appear to be disrupted following this frame.
F-04/F-03: The A350 left wing-mounted LDG light appears to be coming in line with the DH3 wingbox. Accounting for parallax, it is likely that the light (and the engine nacelles) are virtually in line with the DH3 wings at this time.
F-01: Just prior to ignition of the fuel leaking from the DH3's wings, we can see that the A350 landing light is clearly in line with the DH3 wingbox. The engine nacelles have likely gone through the wings at this time.

[boaclhryul]

Originally Posted by AirScotia

...permission had to be queried and confirmed with the non-flying crewmembers...

Given the imprecision of translations and the Nth-hand nature of this, how certain can we be that this query/confirmation was with NF crewmembers plural and not just PNF?

[JP7000]

http://www.nikkei.com/article/DGXZQO...00C24A1000000/
Nikkei Newspaper: Google translate with minor corrections:

"According to the ministry, air traffic controllers are not routinely required to visually observe aircraft movements after issuing instructions."

"There is one controller in charge of each runway and taxiway, as well as an assistant who monitors the situation from behind (e.g., 8 people monitoring 4 runways). Air traffic control has a ``runway occupancy monitoring support function'' that alerts the controller in case another aircraft enters the runway when a landing plane is approaching. The warning system does not give audible warnings, but instead uses flashing lights on the status screen. According to ministry officials, the device was operating normally at the time of the accident, but it appears that air traffic controllers missed this alarm (elsewhere it is reported that the operators were distracted with the scheduling of other planes on "Runway C" and missed the flashing red light)

On the 5th, Minister of Land, Infrastructure, Transport and Tourism Tetsuo Saito announced his intention to compile emergency measures in response to the accident by the end of the three-day holiday period. First, starting on the 6th, he will deploy air traffic controllers to constantly monitor the "runway occupancy alarm" device. Haneda Airport has one of the world's busiest flight schedules, with 1.5 aircraft taking off and landing every minute. The Japan Transport Safety Board plans to proceed with an investigation into issues such as the air traffic control system and operational burden.

[Swapper]

Re: the HUD debate

Can't seem to find it now but somewhere in the thread (perhaps in one of the attached videos) I have a vague recollection of someone mentioning that there were three flight-deck crew on the A350, rather than two. Did anyone else notice that or was I mistaken? If correct, it could potentially mean three pairs of eyes missed seeing the Dash 8 on the runway, one pair not having to look through the HUD.

[DIBO]

Originally Posted by AirScotia

I find it odd that the permission had to be queried and confirmed with the non-flying crewmembers. It suggests to me there was some uncertainty as to what they'd heard.

Originally Posted by xcris

This and many others. I am starting to think that how that flight was operated (MA722) and what happened there was SOP at all :/

Please, be reminded that newspaper reporting on statements from non-aviation Ministries, have produced a lot of nonsense these recent days.

Originally Posted by boaclhryul

Given the imprecision of translations and the Nth-hand nature of this, how certain can we be that this query/confirmation was with NF crewmembers plural and not just PNF?

Thank you

[framer]

Certainly the Dash-8 crew's missing of checking the approach path, where they clearly should have seen the A350.

I don’t think that’s a trigger for action or concern. It is more the norm than the exception to look up final approach and see an aircraft at 3-4nm.
Also t night, it is difficult to tell if the aircraft is at six miles or three miles. I doubt they missed checking it at all.