If you were leading a high-profile international aircraft investigation, in command of the world’s most qualified technical experts and in possession of all the relevant data, would you bother listening to a rag-tag band of internet commenters, few of whom actually work in the space or aviation industry, and none of whom have access to all the data?
Most likely, you’d say: certainly not! But as time goes by, and the puzzle remains curiously impenetrable, you might find it worthwhile to pay a listen to what the amateurs were saying. You might even abandon some of your own conclusions and adopt theirs instead.
This appears to be the case in the search for Malaysia Airlines Flight 370, which disappeared en route from Kuala Lumpur to Beijing back in March. From the beginning, the authorities running the investigation — first, Malaysia’s Ministry of Transport, and later the Australian Transportation Safety Board (ATSB) — held their cards close to the chest, releasing very little information about the missing plane and maintaining a posture of absolute conviction. The investigators’ self-confidence reached its apex in April, when their methodology led them to an area of ocean where underwater accoustic signals seemed to be coming from pingers attached to the plane’s black boxes. Officials assured the press that the plane would be found in “days, if not hours.” But then it wasn’t. A scan of the seabed found nothing; the pingers were a red herring (perhaps literally!). Back to square one.
Meanwhile, on the internet, a group of amateur enthusiasts had come together from all around the world to trade ideas and information about the missing flight. The group, which came to call itself the Independent Group (IG), emerged from various online comment threads and eventually grew to about a dozen individuals. This was a truly spontaneous, self-assembling crowd: there was no vetting of credentials, no heirarchy of any kind. (Full disclosure: I count myself among this group.) Basically, if you seemed to know what you were talking about and could comport yourself in a collegial fashion, you were accepted into the crowd.
While the mainstream press was reporting the ATSB’s pronouncements as received wisdom, the IG was raising red flags. IG members were among the most vocal critics of the ATSB’s contention that the accoustic pings probably came from black-box pingers. And later, after a public outcry led Inmarsat to release a trove of data received from the aircraft, and the ATSB issued a report explaining how it had come to identify its current search ear, the IG dove into the new information with abandon, quickly identifying holes in the data and weaknesses in the official approach. In a pair of papers, the group recommended its own search area, hundreds of miles to the southwest of the ATSB’s officially designated zone.
Today, the ATSB has released an update to its earlier report, explaining why it has decided to reassess its conclusions and move its search zone to a new area — one that overlaps, as it turns out, with the IG’s recommended area. (In the graphic above, the white bracket shows the ATSB area; I’ve added a yellow dot to show the IG area.) Needless to say, this has caused elation within the ranks of the IG, who see the move as vindication of their methods, and indeed validation of their combined efforts over the last few months.
A few observations on the new report:
— One of the reasons the ATSB gives for the shifting of the search area is the recognition that Inmarsat data related to an unsuccessful ground-to-air telephone call attempted at 18:40 indicated that the plane had already turned south at that time. The IG had been basing its analyses on this data point for months.
— Since the June report, the ATSB has improved its BFO model by taking into account various factors — such as temperature shifts caused by the Inmarsat satellite passing through the Earth’s shadow and the mis-location of the Perth ground station in an important Inmarsat algorithm — that IG member Mike Exner has been working through in detail for months.
— The ATSB has fundamentally changed its approach in how it is assessing the plane’s likely path. In its June report, the focus was on what I call the “agnostic” approach: it generated a large number of flight paths based on as few initial assumptions as possible, then graded them based on how well they fit the timing and frequency data received by Inmarsat. This resulted in a population of potential flight paths that fit the data well, but did not make any sense in terms of how a plane might be flown. Some of the routes, for instance, involved multiple changes in heading and airspeed. Today’s report explicitly excludes such flight paths. The ATSB and the IG alike now assume that the last several hours of the flight were conducted without any human input — the crew were presumed to be incapacitated by hypoxia or other causes, so the plane flew on autopilot until it ran out of fuel and crashed. This has been the IG’s starting point for ages, and the fact that the ATSB has now adopted it is a major reason for why the two group’s search areas have now converged.
— You can see in the graphic above how an emphasis on matching the Inmarsat data will tend to lead you in one area (“Data error optimisation”) while an emphasis on routes that comport with real-world autopilot functioning will lead you to another (“Constrained autopilot dynamics”). To be sure, they overlap, but the peak area of one is far from the peak area of another. I think it’s important to realize this, because it helps us to understand why it has been so hard to get a handle on where MH370 went, why the official search area keeps moving, and why knowledgeable people have been furiously debating possible flight paths for months: the BFO and BTO data just do not match up that well. In order to arrive at its recommended area, the IG has been willing to accept much wider deviations from Inmarsat data points than the ATSB has been comfortable with.
— Finally, it’s worth nothing that the ATSB approach is superior to the IG’s in one important regard: it is at heart statistical, looking at families of potential routes rather than proposing and assessing one at a time. There is a tendency, as an individual–and I have fallen into this myself–to cook up a solution, run an analysis, and to be so impressed with the result that one wants to shout about it from the rooftops. (Ask me about RUNUT some time.) The IG has come up with a search area essentially by pooling together a bunch of individual solutions, each of which is generated by a different set of procedures and different set of assumptions. It’s a herd of cats. To really move the ball forward a more rigorous approach is needed, one that takes each procedure and sees how it would play out if the assumptions are methodically modified.
The upshot is that, since the early days of the investigation, the attitude of search officials has changed radically. Once dismissive of amateurs’ efforts to understand the incident, they have clearly begun to listen to the IG and to turn to it for insight and ideas. Indeed, you could say that since the release of Inmarsat data and the issuance of the ATSB report in June, the search for MH370 has become effectively crowdsourced: a de facto collaboration between the professionals and a spontaneous assemblage of knowledgeable experts.
UPDATE:
The overlap between the ATSB’s analysis and the IG’s is more evident in the image below, courtesy of Don Thompson. It shows the fan of values calculated by ATSB to match likely autopilot settings.
Luigi: I assume that you see where I am going with this. The IFE phones can be deactivated from the flight deck. Meanwhile, a bit more than one hour after the aircraft is diverted, we have the SDU being powered up once again. And the next best place to be if you are a pilot locked out of the flight deck and unable to defeat the reinforced door is which such place? A rhetorical question, of course.
My point is that if we are going to toss around the possibility of the E/E bay being compromised, we might as well once again begin looking at the two people among those aboard the aircraft most likely to have the technical skill to enable them to manipulate the servers/buses/systems in the E/E bay – and perhaps restore power to the SDU.
@Rand
If you’re saying that locked-out flight crew might have reactivated the SDU from an equipment bay in an effort to get out a distress call, I’d have to say it doesn’t seem that plausible to me. I think the pilot in the cockpit would have the ability to deny comms anyway, and it would represent an exercise in futility.
As I understand it, the satcom reboot was pretty much coincident with the plane’s retreating from Malaysian radar coverage into international waters. Presumably it would also be pulling away from military radio comms range, to bring in your interesting suggestion from the other day. One obvious interpretation would be that the pilot brought up the satcom to keep a channel open for communication should the command authority on the ground have a last minute change of heart. Not saying that’s the right interpretation, but it seems a natural idea looking at the basic fact pattern involved.
@Lauren,
@Bruce,
Another reason the satellite and aircraft searches for floating debris were negative may be that they were not looking far enough west.
My predicted terminus is 83.5 degrees east. The March air search maps I have been able to find recently don’t seem to go that far west.
Does anyone know if 83.5 E longitude was searched at any time by air or by ship?
On the subject debris:
As Miles O’ Brien noted in the recent NOVA doc, there were 3000 pieces of debris found from AF447. And we don’t even have a shoe from MH370.
@Matty, Geysbreght:
“Inmarsat are saying that the BTO and BFO logged for the message at 00:19:29 can be trusted, and those at 00:19:37 should be disregarded.”
“it would be a lot more comforting if they could explain exactly why they have to be disregarded other than being anomalous in some way.”
PING (no pun intended)
So @Airlandseaman, if “the IG believes the -2Hz value is good”, why is Inmsarsat saying disregard? Put another way, is there something that Inmarsat *knows* that the IG doesn’t?
@Matty: “An empty plane will use a lot less fuel?”
PING.
So now, let’s cue Clive Irving: https://twitter.com/nihonmama/status/480995706254024704
And when you’re finished reading, ask yourself these two questions:
1.WHY did he say this?
2.WHO was his (likely) source?
@Rand, Luigi:
Excellent question Rand. I agree with Luigi’s answer and let me add:
1. Remember – not only do MAS 777’s have SAT phones in business —
https://twitter.com/nihonmama/status/468969785309802496
— but MAS DELETED this info from its website:
(ARCHIVED): https://twitter.com/nihonmama/status/486370278230855682
2. Oxygen – would they have time to to get to it? My cousin told me that at FL25, the pax could potentially survive a depressurization (ie., Total Useful Consciousness or TUC would be longer). But above FL40, TUC would be seconds. Let me also say that even though the *certified* max altitude for a 777 is 43,1000, when I asked my cousin (in March) about the reported climb to 45,000 (later dismissed), he did not push back and say that altitude was beyond the capability of the airplane. But he did that if someone wanted to take out everyone it the main cabin in seconds, a brief climb to 45K altitude would do it.
3. Remember, per the Air NZ cockpit lockout story (which made this known to the public), there’s an alternative method to access the cockpit.
https://twitter.com/nihonmama/status/486370278230855682
If I prioritized the likely response of MH370’s pilots in a lockout situation (and based on training and standard operating procedure for such), I suggest that this would have been the first step.
@Rand, Luigi, Matty, Gysbreght, Airlandseaman:
Just posted a response to all yours which does not appear here. If it doesn’t pop up soon, I’ll retry.
Gysbreght
It is to be expected that the BFO values will drift for a few minutes following a *cold* start (like the logons at ~1600 and 1822). But the logon at 0019 followed a only a brief 1 minute power outage, during which time the OCTCXO would not have drifted more than a few Hz max. Further, there is no credibility for the assumption that a value at 001929 is good, but the value 8 seconds later is bad.
Nihonmama
No. Inmarsat does not know anymore than we know about the 001937 BFO data. In fact, it appears they understand less than we do, in this case.
Initially, they said they did not understand at all what was going on at 0019, and dismissed all the 0019 data. Later they decided it was good after all, except, for the 001937 BTO value, which they remain uncertain about, but without giving any reason. Apparently, they thought the data was corrupt. It only looks corrupt for the reasons I have given. They never considered the possibility of vertical velocity causing the -2 Hz until we suggested it in private emails some time ago, or if they did, they never embraced it. It is not the first time they have had to play catch-up. Look how long it has taken Inmarsat and ATSB to essentially come to the exact same conclusion the IG did 4 months ago on the most likely 7th arc “crossing point”.
@Nihonmama and Rand –
I think any fuel issues can be ruled out, but a refueling cannot be ruled out.
Overall, though I don’t think there’s much room for any significant change in the fuel assumptions, if there is, I think it’s pretty much back to the drawing board.
At the moment, the fuel on board, performance characteristics, and BFOs/BTOs as currently understood all coincide with a fuel starvation event at 00:19. Any break in any of these conditions would likely break the other conditions. Likewise, in a few days, the search will either be ineffective, or the assumptions are all wrong.
@airlandseaman,
Having exhausted the -2 Hz issue, I’m still eagerly waiting for your exposition of the flight mechanics that turned a steady one-engine-out cruise condition into a sustained near-vertical spiral …
@Dr. Bobby: AMSA records suggest they were CLOSE to your spot (see 1st 2 days of search, March 18 & 19, at http://www.amsa.gov.au/media/incidents/mh370-search.asp), but not quite ON it (e84 is as far east as they got, to my eye).
Furthermore, I believe the debris (if any) was thought to have drifted WEST in the first week (I recall that being the reported net effect of wind and current in week 1, but stand to be corrected); this would INCREASE the gap between where your model would place the drifted debris, and where they searched for it…
…with AIRCRAFT. I would have hoped the SATELLITE search was both deeper AND wider, and would thus have covered your ground. Alas, just like the world’s combined radar resources near Sumatra, the world’s combined satellite resources seem to have taken a big ZE-RO on the alleged SIO debris field…
(Some hypotheses suggest a deeper truth is being hidden to save public officials from embarrassment. It is hard to imagine anything more embarrassing than the public record as it stands.)
@IG Members: I asked Mike last week for the detailed model behind his “decompression scenario” – plus whatever commentary/documentation on fuel feasibility and path reasonableness he could offer.
I see he is up to his eyeballs in defense of the IG’s 001937 BTO interpretation (which seems like a teapot tempest to me; aren’t we fussing over what boils down to just a couple extra passes of the towfish?).
So I throw it out to any IG members willing and able to comment: was there a detailed decompression scenario model I could review, was it rigourously tested for endurance to 0019, and how does the membership rationalize the (to me) odd curvature of its flight path?
Luigi – maybe you didn’t get my question. If Shah was awake at the wheel throughout and was the perp, and put it down in one big bit more or less to annoy the searchers, why did he turn the SDU back on and leave a trail all the way down south?
Luigi – Agree, an absence of wreckage and plane doesn’t validate a cyber hijack etc, but the court of public opinion will go against you if you insist the data is still good. Those numbers did become the master, but not to the public.
Mike makes the point – you need a better reason to throw away data, in fact you need a good clear one to appear credible. And I was doubtful to start with.
@Matty
>> If Shah was awake at the wheel throughout and was the perp, and put it down in one big
>> bit more or less to annoy the searchers, why did he turn the SDU back on and leave a
>> trail all the way down south?
I don’t have any strong opinion as to whether Shah remained awake at the wheel until the end. I merely propose that if the absence of debris is really a problem, that would be most easily explained by introducing the assumption that he was in control and deliberately entombed the aeroplane, e.g., by doing a belly flop onto the ocean.
As it happens, I doubt the absence of debris is problematic, and there seems to be a partial consensus among the BFO tweakers that the plane was actually in a steep dive at the end. Still, even if we discount that, this auxiliary assumption would avoid the necessity of throwing out the Immarsat ping data in toto. That is a good thing. If you accept that the pings place the terminus in the SIO but have a problem with the absence of wreckage then that position further strengthens the Shah theory, because it is highly implausible that hijackers, thieves, kidnappers or wannabe kamikazes would fly the plane to the SIO and carry out a controlled ditching. Moreover, an *aborted* hijacking, theft, kidnapping or kamikaze attack would also be unlikely to end in a controlled ditching in the SIO or anywhere else for that matter.
As far as turning on the satcom, to me that only reinforces the notion that we are looking at a situation which escalated rather than a completely pre-planned sequence of events. That has always seemed to me the most intuitive variation on the Shah theory for several reasons, and is where I tend to part company with the authors of “Goodnight, Malaysian 370.”
The debate over the end game, such as it is, is unnecessary. ATSB and the IG agree it was spiraling down. There is no debate about that (between us). The only question is the turning radius and verticle speed. At 400 kts and 25 deg bank, the radius would be 5NM. At 400 kts and 70 deg, it would be 0.8 NM. No matter what the radius, it came down, as ATSB and the IG agree, close to the 7th arc (1-5 NM). BTW…the verticle speed at 00:19:29, which BFO value the ATSB does believe in, corresponds to a verticle speed of -4800 ft/min…still a very high rate of decent…much faster than what would be consistant with a 25 deg bank.
@airlandseaman:
Just for completeness: at 5 degrees of bank the radius would be about 25 NM.
Mike: thanks for your clear and succinct explanation of the process concerning the end data and the concomitant flight dynamics.
Nihonmama: I can’t view twitter from China (jerks) and I am going from memory re the Air NZ incident in July, but there is indeed an alternative means of entering the flight deck beyond someone on the inside unlocking the door. Regardless, there any number of means of physically preventing entry to the flight deck (the shoe that you referenced could probably do the trick). On the flip side, I should say that there are very few doors that can stop someone determined to enter: a thick flat bar of hardened steel would do the trick on any ‘hardened’ cockpit door, I would imagine. Cockpit doors are not designed to prohibit all forms of entry. Rather, they are designed to INHIBIT most reasonable forms of forced entry. The facts are that most people have no intention of making an unauthorized entry to the flight deck, and then much less an ill-intentioned entry. I have talked my way onto a flight deck of an international flight (and was in the jump seat for the landing), yet I had no ill intentions; I merely was charmed by the FO and threw it right back at her, and we both mutually agreed that it would be a fun exercise. The Captain was quite literally cajoled into surrendering the idea. Ok, he had no choice, really.
My point is that all is continuous, nothing is discrete – save for the location data. But I digress…
The IFE sat phone system can be deactivated from the flight deck. The flight deck includes an interface that accesses the satphone terminal in the E/E bay. Thus, the E/E bay is superior to the flight deck in terms of the electronic logic.
I am beginning with the high probability that the official pilots were most likely the two most technically capable people aboard the aircraft to be able to manipulate things from the E/E bay.
From here, I can’t imagine any higher level of response for a pilot denied access to the flight deck than to enter the E/E bay via the absurdly available Happy Meal hatch located under a square of velcroed carpet in the galley. As for the issue of oxygen, denying the hijack-operator of emergency oxygen (by way of defeating the tanks that are located there) on the flight deck would be one means of removing the hijack-operators from their position of control of the flight deck. The next step, then, would be to re-enter the flight deck through the barricaded door, as the ‘other means’ would have been rendered inoperable. The effort to reenter the flight deck would have then have failed; perhaps someone forgot their hardened bit of steel.
Again, the official pilots for the flight are far and away the most likely to have the technical skill re the E/E bay, and thus is it no less valid to speculate that one of the pilots manipulated things in the E/E bay as opposed to a member of the crew, a passenger or a stowaway. Any sort of spoofing of the data or remote operation scenarios are in turn less probable than what I have described (ie, ingress to the E/E bay in an attempt to regain control of the aircraft).
We do have the SDU being rebooted at 1825, while there were no communications from the aircraft, despite reported calls to the flight deck satphone terminal. The logic, then, is that the party on the flight deck had no interest in answering the calls, while the party in the E/E bay had no means of interfacing with the radio or the SDU and initiating contact with the ground. There is a service terminal in the E/E bay; perhaps there is an outbound text message that was redacted from the data logs – this is the sort of thing that I believe should be chased relentlessly and hunted down until it is bled dry. There must be secondary hints buried in the data, or at least, one can hope that there are.
The location science is grounded in discrete data found in the XYZ domains, but with regards to forming a view as to ‘how’ the flight terminated in the SIO, there is no ‘starting point’ in the logic of any of this that produces a firm conclusion; likewise, there is very little data available to validate any sort of conclusion. Thus, ‘beginning’ with the official pilots being able to leverage the vulnerability of the E/E bay is just as valid and perhaps just as probable as speculating that a hijacker gained access to the E/E bay. Moreover, it is vastly more than probable than any version of a cyber hack for a variety of reasons, including simplicity.
We have:
-the deviation at IGARI from the intended flight path
-the aircraft apparently flying to the point of fuel exhaustion
-a flight terminus that is less than likely in terms of probability to be the intended destination of the aircraft
-the SDU rebooted at 18:25
-the apparent vulnerability of the E/E bay, in that it can be accessed by way of the galley hatch (or the external hatch)
-the reported vulnerability of various communication and reporting systems, with several systems open to manipulation/hacking
-a flight that terminated with the higher probability of a lack of human input
-Generalized obfuscation on the part of the Malaysian government
-redacted data logs
-no reported VHF communication from the aircraft post 17:19, despite the aircraft remaining aloft for hours
If you plug all into a probability tree together with another 60 or so other secondary and tertiary binary choices with assigned probabilities, multiplying across, then I wonder whether you will find the same highest couple of values (and their respective integrated, holistic views, devoid of any invalid logic) that I did. I don’t, of course, know what happened to MH370, but utilizing a probability tree is an interesting exercise. It also provides for the reconciliation of all elements.
I think I’m loosing it; it’s business development season and I need to get back to work. I hope to Jehovah that some you, at least, find this missive at least somewhat entertaining or thought provoking enough that you can help solve this puzzle on behalf of the NOK and All. We have the data and little else: is there a hint buried in there somewhere that would lead to a deeper investigation of the circs?
Apologies, no time for an edit.
Today for the first time in many months I queried a few people I know about MH370. Normal people they are, ones who haven’t obsessed about it and don’t read it up daily. What happened there you reckon? Almost all said it was sitting somewhere in one piece, and are paying no attention to the search. Pretty clear where public perception is going to head if it doesn’t show up.
Who’s up to date with the search anyway? How far into the priority areas are they?
Dear ATSB: re: “close to a NW point at 1912”, Appendix A, June 26 report:
1) Exactly how and why did you model this extra flight distance?
2) Exactly when and why did you abandon this extra flight distance?
A new Addendum #3 is available at:
https://drive.google.com/file/d/0BzOIIFNlx2aUMjhBWmk4ZlBBNEk/view?usp=sharing
It discusses speed, speed slope, residual errors, BFO data, Inmarsat route, aircraft range, and debris search area.
Your comments and questions are always appreciated.
@Bobby Addendum #3 looks excellent to me, especially for its simplification of the route in Figure 5-1 and the treatment of range.
Not sure I agree that your proposed location was not searched or that no debris was found. Given the generally eastern drift identified by the buoys deployed on March 20 and the prevailing generally eastern winds and currents, I believe the area where debris would be most expected may well have been searched.
At least 122 floating objects were sighted by satellite that could have drifted from your end point. http://www.nytimes.com/2014/03/27/world/asia/missing-malaysia-airlines-flight-370.html?hp&_r=0
None of the debris associated with these satellite images was recovered per Martin Dolan:
Bruce Lamon said…
Mr. Dolan, thanks for taking these questions.
Early on the Chinese (March 22) and the French (March 26) published satellite images of floating objects of significant size found in the vicinity of the initial Australian search area. Were any of these objects ever relocated or otherwise analyzed to see if they came from MH370?
JULY 14, 2014 09:14View all comments by Bruce Lamon
Martin Dolan, Chief Commissioner (author) said…
No recovered debris has been identified as originating from MH370, and the search effort has not found any debris associated with the satellite images or other visual detections.
JULY 17, 2014 11:27
(The search for this debris was halted after just a few days on March 28 on the grounds that MH370 was flying faster than previously thought, and therefore ended farther north.)
Mike: in support of Matty’s query, could you please update us with your informed view of the activities of the SIO search at present? A projected timeline regarding what we can expect to see developing in terms of these activities would also be helpful. Thx.
[Test]
@Bobby Ulich:
Thanks for the primer on statistical analysis, the principles of what you state are clear. However, the question is what is the actual noise on the MH370 BFO measurements. Fig 15 of the recent Ashton et al paper shows the predictions and data for the Amsterdam flight earlier on 7th March. Measuring off the graph (after 17:45UT when the data is changing rapidly), the standard deviation of the error is 2.3Hz (one sigma).
The MH370 data can also be examined for periods when the aircraft is stationary (before take-off) or on a steady course (around 17:07UT). The standard deviation of the R-channel BFO data was 1.7Hz in the first case and 1.1Hz in the second case (in this later case the standard deviation of a larger number of T-channel data is 1.4Hz). It would appear therefore that a large part of the error in the Amsterdam BFO data is random noise. In any case, the error figure of +/-7Hz in the text of the paper and 5Hz (one sigma) in the original ATSB report are both too pessimistic.
On this basis fits that achieve a r.m.s. errors of ~2.7Hz (as in Table 9 of the Ashton et al paper) are not over-fitting the data as you imply. A fit that achieves an r.m.s. error of 5Hz is actually rather poor, as the chi-squared test indicates.
Oh, now they tell us:
“THE satellite communications company central to the MH370 search has acknowledged there remains “significant uncertainty” about the final location of the Boeing 777.”
http://mobile.news.com.au/travel/travel-updates/malaysia-airlines-mh370-search-company-immarsat-cast-doubts-on-planes-location/story-fnizu68q-1227091656903
One interesting note: they deduce that the plane was still operating, as the SDU requires navigation inputs to steer the antenna.
Though some folks earlier may have been barking up the wrong tree looking for signal strength to show power availability, it does appear that those redacted values could cast some light on the situation.
Presumably, the ability to steer the antenna is not unlimited. Does anyone know the mechanism for steering it? Is it physically steered? What is the average lag time between navigation changes and antenna steering? I would think it could be within tolerances most of the time, but the signal strength would decrease if the antenna wasn’t steered optimally.
Is this reasonable? If so, the extra data could help, especially at the turns and in the dive.
[First posted almost 7 hours ago]
@Rand:
Thank you for the missive. Taking in all that you’ve said…
And my apologies for sending you those stories embedded in tweets – some of the comments are as interesting as the article itself. Are you not using a VPN to leap over the GFC (Great Firewall of China)? 😉
For your convenience, the direct links.
1. Who Gagged the Search for MH370? http://t.co/tLky3PZrti
The full tweet:
“Clive Irving in .@thedailybeast references alleged ‘report’ that I’ve not heard mentioned b/4 – re #Indonesia #MH370”:
“Another report, called in to the FAA and passed by them to the FBI alleged to have evidence that the 777 had landed at Banda Aceh on the northwestern tip of Indonesia where it was taken to a mysterious hangar and all the passengers and crew executed.”
2. Now deleted MAS webpage showing SAT phones in every business class seat:
http://t.co/427fMjI5uB
3. Air NZ cockpit lockout (Ben Sandilands aka @planetalking): http://t.co/fQRqVYA2tp
And, as if there isn’t enough Kabuki theater going on already, here’s more:
Inmarsat cast doubts on plane’s location
http://bit.ly/ZtGWxv
Now what da ya think is triggering Inmarsat’s “significant uncertainty” comment? Three words: Sir Tim Clark.
As JS astutely noted on 10.11:
“I find it odd that Tim Clark’s comments are now at the top of the news. I don’t believe he has a scientific basis for them, but I believe he is “placing a bet.” His comments come well-timed, just as the highest priority search areas have apparently failed to yield results. It is definitely a fourth quarter bet, but good bets are often driven by insider knowledge.”
Insider knowledge. And shoes dropping (or about to). Again, watch this space.
And Rand, your tale of flirting your way into a jump seat on a flight was so entertaining (and totally believable), that I saw it all as a short film – and decided to score it – from this album, which I’ve been listening to for the last four days (TURNED ALL THE WAY UP):
https://www.youtube.com/watch?v=Cup512wPONE
All of the fear and tragedy in the world notwithstanding, international air travel is still a beautiful thing.
My thoughts and prayers are with the families and loved ones of those on MH370 and MH17. Let us hope that the ‘wisdom of the crowd’ helps to uncover to the truth.
JS
9M-MRO had the Ball AirLink conformal phased array antenna system (HGA). It is steered using IRS data delivered via ARINC 429 cable to the SDU. This nav data is used both to steer the antenna and to compute the Tx Offset frequency that partially offsets the L band Doppler. The October 8th Update includes the redacted Receive power levels. They show that the received signal strength was fairly constant (within a few dB) over the entire flight, including the 00:19 “partial ping”.
Rand: Equator has completed 8 survey passes over the IG end point, and appears to have started back for port a few hours ago. Meanwhile, Discovery (search vessel) is near Perth, conducting sea trials with the new underwater gear. We expect it to head for the IG area very soon. Phoenix is searching now NE of our spot. It is making tracks with only 1NM spacing, so it will take quite a while to cover the previously surveyed area.
Bobby: You have some good points about the statistics, but it is a mistake to treat the BTO Jitter like white noise. It is not white noise. It is the subject of an inquiry underway with ATSB. The source is believed to be in the firmware of the Square Peg Channel Units or the Honeywell/Thales AES. Hope to pin point the source and get a better handle on how best to weight the jitter.
Rand:
I posted a response to yours a little over seven hours ago. It may (or may not) appear. 😉
@JS
Gave the Journal of Navigation article that came out last month a once-through. Generally, reading it didn’t alter my feeling that attempts to draw fine-detailed inferences from the limited dataset should be treated with caution. It looked like the major part of the BFO variability over the course of the flight came from the only partially compensated doppler shift from satellite to groundstation. I think this would be the same regardless of the trajectory of the plane — that might also be true of some of the other components (not sure). The electronics on the plane apparently do a reasonably good job of compensating for the doppler shift introduced by the plane’s motion vector relative to the satellite, hence shifts introduced by the latter would be pretty much second-order error terms due to the idealized model used in the compensation — suggesting that the BFO error would be relatively insensitive to variations in the modeled position, speed and heading. Also, it seems that there were quite a few booby traps hidden within the details of the signaling path, and some seem poorly understood while there remains the possibility that others remain uncovered. Finally, the precision of the BTO-based distance estimates is no better than several miles, and that is assuming the data points derived from a stationary plane sitting at the airport are truly representative, which might be optimistic.
@Nihonmama, What are you doing, that some of your posts go up automatically and some don’t? You are the only person that I’m having to approve posts for.
@jeffwise
A while back it seemed to me that there might be an issue with posts that contain more than one “http:” URL string — I was wondering if there was some kind of filter. Haven’t had a problem since limiting myself to one URL per post, but that might be a coincidence.
@JeffWise:
I’m doing what everyone else is doing.
@Luigi:
I think you’re right.
@Bobby: excellent paper, as always. Comments:
1) Regardless of whether they agree with your theory, model, or result, EVERY serious scientist should read (and digest) your point about over-fitting in pursuit of error minimization (sections 4.1 and 4.2, in your latest).
2) I’m glad you’ve considered the IG’s lighter=slower effect. But I have three concerns with your decision to test a range of speed deltas, and select the one which minimizes errors:
a) isn’t the IG saying this effect MUST be added, whether it hurts fit or not?
b) speed delta=0 is the best fit for YOUR path – but is it 0 for OTHER paths?
c) don’t you undercut 4.1/4.2 by calibrating the delta to “fit, not physics”?
3) Thank you for bringing hard numbers to this site re: fuel analysis. Can I please ask you, though, to “dumb it down” for us non-experts? I propose a two-column table:
Column 1: constant KTAS at FL350 (300 to 520, in 20kt increments)
Column 2: time of MH370 fuel exhaustion (HH:MM UTC)
(Feel free to suggest refinements/clarifications, e.g. altitude as time-varying vector).
The IG can then do the same, and we would then have a clear, consistent, and comprehensive basis for comparison.
If any piece in this puzzle SHOULD be nailed down by now, it is 777-200ER endurance as a function of constant KTAS (for a given weight, fuel, altitude, and PDA); that it remains murky is, frankly, ridiculous.
4) Bobby, I wwould be forever in your debt if you could run the ATSB’s “highest probability” path (circa April & May) through your model, and report back your results. Several paths were published in the “Preliminary Report” at end April – the path I’m hoping to have tested for fit/feasibility is the one leading to where they searched for pings (at s21).
It’s the northernmost path – the one that flies at 323 KGS for the last several hours.
A key sensitivity test on this would be to attempt later turns than are graphed, to see whether such a scenario can be wrestled into feasibility.
The IG’s “decompression scenario” is another “323 KGS to s21” path on which I’d love to get your professional opinion.
Thanks again for your efforts, Bobby!
Brock
World Out Of Control
https://www.youtube.com/watch?v=duZ17mNXjds
@Bobby: in acknowledgement of the strong relationship between speed and altitude, I suspect we’ll only get to the bottom of the “maximum endurance” issue if we change altitude from a constant to a variable. I hereby amend my request to instead show your fuel model’s indicated flameout times for all possible COMBINATIONS of…
1) KTAS (300, 320, 340,…,520) and
2) Altitude (FL100, FL150,…, FL350)
…so 72 UTC time values instead of 12. If any of the combinations are literally “off the charts”, please so indicate.
Thanks again.
It’s interesting to note the logic of Sir Tim Clark, as revealed in the Spiegel interview. Despite the fact that he is an industry insider, he is apparently hung on the conclusion that the aircraft was under human control for the entire duration of the flight. He seems to have reached this conclusion by way of the evidence indicating that the transponder and ACARS systems had been deactivated intentionally.
This is illustrative of how a lack of a holistic perspective combined together with quite valid logic can lead one astray when confronted with a paucity of information. Again, how is it that an intended deviation at IGARI ended hours later as a flight at fuel exhaustion? A process of reconciliation, elimination and probability assessment is required to obtain even the fuzziest image of what then may have even occurred, given the thin physical evidence.
Whistle blower, this is your cue!
JS – I would expect some more lines like that from Inmarsat the further we go with no plane. When the sea search was called off was a good time for them to put it all on the table and say this is what we believe and why. MAS might own the data but they don’t own the analysis. They will cop some stick in the press if it’s fruitless.
@Matty – Perth,
“MAS might own the data but they don’t own the analysis. They will cop some stick in the press if it’s fruitless.”
Fully agree, but I think your “they” should be the Malaysian government. Everyone else got an out:
@IG, @ATSB, @Bobby
Even if the BTO/BFO modelling was perfect, its prediction is only as good or bad as the applied initial conditions, read the 18:22 radar data.
I have heard of confidence in that that location and time of the radar track data is good and is indeed that of MH370.
If nothing is found, I predict that that data (and confidence in it) will be re-examined. And some serious pressure will, should, be applied to the Malaysians to back up, maybe even open up, their radar data for independent analysis.
To me the elephant (too sharp a corner in the turn back portion of the radar track) is still in the room. In my 16 years career of geometric modelling and 5-axis CNC path generation, I came across many similar “kinks” in early stages of my developments.
They were a sure indicator of me having made a mistake somewhere. Without fail, I would sooner or later discover a mistake, correct it, and find an expectedly smooth-ish result.
The gist of it is, if there is a discontinuity (physically impossible sharp turn) in the early radar track, there is no way that the later track is that of MH370.
It may soon be time to apply the by now sophisticated and solid BTO/BFO path modelling to the last known position and (incoming) path tangent at that sharp corner.
Maybe that would would put MH370 Into Cambodian airspace, fly past Kate’s stern, etc., etc…
Cheers,
Will
MuOne – If Inmarsat are still holding some cards close it seems ATSB are complicit? What’s been signed? The Australian govt are forking out for this episode so what sort of considerations are there to accommodate the nondisclosure and how did this come together? There is no more money after this and so far it’s being spent to the commercial benefit of Inmarsat – if the search is successful. If not maybe serious blowback.
Radar data – now that would be interesting, then turn it onto Indonesia as well. My guess would be some previously unknown quality issues would turn up. I thought the public would go mental when the sea search was called off but they always gave the impression that they were onto it and the plane would show up. IF….the white flag goes up it could go pear shaped for Inmarsat in the PR sense.
Hi Matty -Perth,
I just read the Inmarsat paper submitted 4th Sep, accepted 14th, and it sounds like they are quite openly reporting to their best knowledge and state of evolving analysis.
Many items of what they are saying seem to parallel the public IG/DS evolution of analysis and refinement. So, since that seems to indicate two independent sources arriving at similar analysis and conclusions, I like to give ISAT the benefit of doubt that they are doing their best and not holding much close to their chest of their own accord..
Same can’t be said of the Malaysian (and yes the Indonesian) government. Malaysia’s publicly released radar track is one image in the ATSB report and one slide in the presentation to the NOK, no source data, no background info on how that might be derived, no explanation on how it was assembled from different radar installation sources, no info on the track resolution, no info on how the gap in the malacca straight was filled, no explanation for the sharp corner, etc.
I still think, it will be the Malaysians (Indonesians, and maybe Thai?) that deserve the stick, for holding back on radar data release.
Cheers
Will
MuOne – OK, I was under the impression that just how Inmarsat came up with modeled BFO data for north/south comparison was withheld? And that all parties have used essentially the same means to get where they are? But I’m not a cruncher and have drifted away from the ins and outs lately.
The Malaysian/Thai/Indon radar I’m prepared to guess could be in need of a tune-up.
Nihonmama: Thanks for all the good dope.
Why, aren’t you the Asia-savvy one! My VPN account expired over the summer, and I haven’t bothered resubscribing, now that I am in Tokyo most of the time. Now that I am presently in Shanghai, however, I am desperate to watch the boobtube video that you have scored. Really? FYI, China is now blocking access to VPN sites! And, our office VPN router is on the fritz; no tube action for me…
Is not Inmarsat merely issuing caveats regarding the search in an effort to manage expectations? The search is going to take an inestimable amount of time, and the analyses are undergoing constant refinement, which in turns leads to adjustments in the designated priority search areas. Managing expectations, then, would be an important element of the process of ensuring that people remain open and interested enough to support another two years of search activity at the cost of an additional couple-hundred-very-large?
As for Tim Clark, my view is that his perspective is grounded in his rather solidified thinking on MH370, as opposed to it being a product of his being well informed. But then I am as full of erroneous thinking as I am of anything else, so let me know if you find anything to further substantiate any general momentum on the inside indicating doubt in the SIO conclusion.
Matty: your doubting nature now has you at the point of being nearly grumpy. Cheer up, bud, and let’s hope that something gets sucked up out of the pond soon to indicate that the search is at least generally on the right track. For, other than taking an active role in the search, there is nothing to do at this point but cling to the meager seat cushion of hope that the makings of a damn interesting story will actually produce a favorable ending – and the location of the aircraft. Call me naive, brand me a peanut gallery cheerleader (I do have a soft spot for short, flared skirts), call me Ismael for all I care: this thing is being run on quite a bit of tireless hope. My guess is that this the reason for your returning here day after day, which is likewise why I would imagine that we will be asking you to play host at a gathering in Perth one bright and warm Aussie summer afternoon. With a bit of luck, that day won’t prove that far off; summer arrives soon in your parts. Or perhaps the few extra beach goers to soon be arriving on your wonderfully desolate beaches with assist in the discovery of at least one seat cushion…aye, one can only hope for such bittersweet things.
@Matty & @Rand,
“Ask you to play host at a gathering in Perth one bright and warm Aussie summer afternoon”.
I’m in. I’ll bring a bottle of Honey Whiskey and a bag of lemons, you sort the ice. Mind you, I’ll be coming by train…
PS.: LG has expressed interest to try my recipe. After all, I think, we are engaged? ;o)!
Cheers
Will
Here’s an update on various 7th arc predictions and search plans.
https://db.tt/oz59bn3F
@airlandseaman,
Hi Mike.
Just for the record, I am not “treating the BTO jitter like white noise.” I have done an analysis which demonstrates that the BTO noise in flight may be as low as the quantization noise (see Figure 3-1 in Addendum #2). I believe this quantization is actually done when the data are recorded at the ground station. There may be additional, non-negligible sources of timing jitter in the satellite and aircraft. My route fitter allows the handshake positions to vary but always within an upper limit on the RMS radial error (I am now using 0.67 NM, consistent with quantization noise only). So I don’t assume any particular noise spectral density shape for the BTO residuals, just an upper limit on the RMS value which is quite tight.
In your post on October 14th at 1:55 PM, you referred to Inmarsat when you said “Initially, they said they did not understand at all what was going on at 0019, and dismissed all the 0019 data. Later they decided it was good after all, except for the 001937 BTO value, which they remain uncertain about, but without giving any reason. Apparently, they thought the data was corrupt. It only looks corrupt for the reasons I have given.”
In fact, Inmarsat gave a very specific reason for not accepting at face value the 00:19:37 BFO data (in your post you said BTO but I assume you meant to say BFO). Inmarsat said “Detailed analysis of BFO samples taken from other flights show a high degree of consistency for the signal messaging frequencies, with the exception of those that were performed immediately after the initial logon process. This called into question the BFO measurements after the log-on sequences at 18:25 and 00:19.” In other words, Inmarsat said they did not trust the 00:19:37 BFO because their analysis of other flights indicated those secondary log-on frequencies were unreliable. Their reasoning had nothing to do with the actual value of the 00:19:37 BFO number for MH370.
@Richard Cole,
The chi-squared test is of limited utility for the MH370 data. First, there aren’t many events with BFO data after the turn. Second, one has to assume a probability density function in order to perform the test, but we don’t know the PDF of the BFO noise.
As you correctly point out, there are various data sets where one can compute a RMS value of short-term BFO variation, and the resulting values don’t seem to agree very well. The Inmarsat paper did not present any BFO noise statistics (unless I missed it). I don’t think this is an oversight.
The Inmarsat +/-7 Hz “accuracy limit” implies that differences between predictions and measurements are acceptable up to 7 Hz. In other words, they are putting an “upper limit” on the maximum allowable BFO difference. Inmarsat said “. . . it is important to note that agreement is only achieved with +/- 7 Hz accuracy during this flight, and to assume better accuracy for the measurements taken on MH370 would be unrealistic.” My interpretation of their conclusion is that Inmarsat is allowing for the possibility of both random reading errors and systematic errors affecting a BFO value (hence the use of the term “accuracy,” not “noise.” This is certainly a reasonable approach for them to take.
Perhaps we are making an assumption (that read noise is the only component of BFO residual variation) that is too simplistic. If the BFO data also contain some measure of systematic error, then there may be two sources of BFO variation (besides the obvious one of the actual aircraft position and velocity): (1) a random, short-term single-reading noise component, and (2) a systematic error that depends on aircraft location/velocity. For instance, suppose the navigation data input to the airborne terminal has (small) errors. The calculated compensation frequency shift won’t be exactly correct, and you will get a systematic error. It won’t be apparent in the short-term jitter (and it won’t change sitting on the ground). It is obvious that, at some level, systematic errors must exist in the BFO data. The question is, “How large are they?”
I will note that the airborne terminal compensation is up to 900 Hz. Systematic errors of only 0.3% in this compensation process yield 3 Hz BFO error. I don’t think systematic errors of this magnitude are out of the question here. Why would anyone design into this system more accuracy than that, especially when the actual downlink Doppler alone varies by several hundred Hz? Remember the BFO compensation was utilized just to limit the instantaneous bandwidth requirements on the satellite and ground station equipment, not to provide data with which to accurately locate a downed aircraft.
Inmarsat has not attempted in their paper to separate BFO read noise from systematic errors. We have been neglecting the latter. If we follow Inmarsat’s lead on this point, we reach two conclusions: (1) all BFO residuals less than 7 Hz are acceptable, and (2) there is not a simple correspondence between RMS BFO residual value and the acceptability of the route as being consistent with the MH370 BFO data.
@Matty, MuOne:
“If Inmarsat are still holding some cards close it seems ATSB are complicit?
“I still think, it will be the Malaysians (Indonesians, and maybe Thai?) that deserve the stick, for holding back on radar data release.”
I’ll ask again: to the degree that information is being withheld (and that appears to be the case — and by more than one “party” here), how would that be done party WITHOUT the complicity of the other stakeholders? Put another way, if Malaysia and Indonesia are the only culprits here, why haven’t they been fully outed? There are competing interests here. And the aviation industry lives (or dies) on certainty — and the flying public’s belief that it’s statistically safer to travel by air than drive a car or cross the street on foot.
Let’s put aside the prospect of civil liability and potential future lawsuits for a moment. If you are Boeing or RR, and one or both of them know (and I suspect they do) what happened to this plane, and it (may) have nothing to do with catastrophic failure of the airframe or engines, how long is it going to be in these manufacturers’ interest to stay quiet and just ride along on the Malaysia-is-jerking-our-chain-and-we’re-powerless-to-stop-them bus? Yes, everyone has to wait for the outcome of the investigation – which assumes that the plane is found. But what if it’s not?
This is, in part, why I wrote about WSJ’s changing story — (see
Posted September 16, 2014 at 11:20 PM MH370 Search Area Still Too Far North, Independent Experts Suggest) — because that looked and smelled like somebody’s leak. And it got shut down.
Very clearly, the “authorities” – whether we’re talking about Malaysia, ATSB, AAIB (UK), or the FAA – have convergent interests here. But they have divergent interests as well. In that regard (and before we put ALL of the obfuscation around MH370 in Malaysia’s lap), do people recall what Duncan Steel said about the French (BEA) getting stonewalled?
Remi Jouvet, a member of the French delegation sent to Malaysia, said “The plane was made invisible intentionally.” But UK’s AAIB refused to give MH370 ping data to the French who were INVITED by the Malaysians, because of their expertise gained as a result of AF447. So the French left Malaysia empty-handed.
http://www.duncansteel.com/archives/549#comment-903
My complete tweet here re this because it contains additional comments:
https://twitter.com/nihonmama/status/467174124427554816
(Rand, my apologies – hopefully you can read the tweet when you’re next in Tokyo or when your VPN is back working).
As for the Thais — and their waiting 10 days to share radar data because they “weren’t specifically asked” — that is a textbook example of a “high context” statement. I invite people to ponder what’s really going on there.
And MuOne, I hear you (and Ron) and agree (even though I’m of the opinion that MH370 may have landed in Indonesia) — the radar track sharp corner IS a huge elephant that should not be lightly dismissed. As stated previously, I (personally) have no doubt that my dear government (US) at minimum knows or has a ‘hand’ in this story. A possible unfortunate ‘happening’ east of Malaysia (read: in the Gulf of Thailand or South China Sea) would, for a thousand reasons, be as much a reason to obfuscate — and coordinate silence around MH370’s actual fate — as it would if we’re talking about a possible terrorist-related incident that various intel agencies might want to keep under wraps.
@Brock,
I appreciate your comments on my Addendum #3.
Here are my responses on the issues you raised:
1. I found the courses that best fit each of the speed slopes. They cluster very tightly near 192 degrees. You can pick whatever slope you want. It does not move the end point very much at all.
2. Fitting speed slope to the IG’s course at 187 degrees is pointless because the speed variations are so large (> 20 knots) and increase with time rather than decrease.
3. It is reasonable to look for a speed slope (“delta”) because the speed control mode is unknown at present, and the fitted slope could potentially help identify which speed mode was used. Insofar as I know, there is no identified route that fits -3 knots per hour better than just a constant speed. So no, I’m not actually fitting a slope to find the “best” route. I was simply evaluating this possibility following the suggestion of the IG.
4. If I understand it correctly, I believe the IG’s position is that, if Long Range Cruise were selected, then one would expect to see a few knots per hour speed decrease. On the other hand, if “Constant Mach#” were selected, then the speed would be very stable until the aircraft reached significantly colder air far from the equator.
5. The only Boeing range chart I have is the one in Figure 7-1, and is only valid for Mach # 0.84. I cannot use it to predict maximum range for other speeds or altitudes.
6. There was a post on this site by “Lauren H” on October 12th at 2:25 p.m. that listed some Boeing fuel burn numbers in tabular form. They showed that at FL300 the extra fuel required to travel an extra 800 NM was 10,800 kg in 1:53 time. This gives a fuel burn rate of 5.73 metric tons (MT) per hour at 425 knots and 13.5 kg of fuel burned per NM. At FL400 the same table gave 9,600 kg to fly an extra 800 NM at 1:41 time difference. At the higher altitude, the speed is now 475 knots, the fuel burn rate is 5.70 MT/hr and the fuel burned per NM is 12.0 kg. My proposed MH370 route requires an average fuel burn rate of 6.1 MT/hr. It is consistent with the Boeing burn rates of 5.7 MT/hr even with PDA’s of 3 % and an additional allowance of as large as 4% for added fuel consumption for a speed higher than 475 knots. Thus, insofar as I can tell, my proposed route is consistent with the Boeing chart I used and with the Boeing range table information from Lauren H. As an aside, I would not trust anybody’s range calculation (mine included) to be better than +/- 2 %, considering we are guessing the PDA’s and the altitude and the speed control mode.
7. A 158 degree great circle route ends at 21 S and intersects the 7th arc at 104 degrees E longitude. The average true air speed of this route is 378 knots. The minimum leg speed is 119 knots and the maximum is 441 knots. It is impossible to fly this route with any single setting of speed control and match the BTOs.
@Rand:
You crack me up.
And — I interact on Twitter with numerous folks in China whose gobs are flapping constantly. And about everything. All via VPN. Might a new provider for you be in order?
As for Inmsrsat managing expectations… PING. That’s the book that hasn’t been written yet.
Re Tim Clark:
RunwayGirl also tweeted about this story and made an (interesting) comment:
“EK have been skeptical since the *very* beginning.” https://twitter.com/RunwayGirl/status/521659947244077057
Off-topic music note (because we wouldn’t want you to be desperately waiting to watch YouTube):
The song I selected for your short film is “La Source” from the stupendous Swing Out Sister album “Where Our Love Grows”
http://swingoutsister.com/albums
[If you somehow don’t know them, NB their second – “Kaleidoscope World” – released weeks after I moved to Tokyo. An absolute masterpiece and one of many reasons for their devoted (worldwide) following and music royalty status in Japan.]
They’ve been writing the soundtrack to my life since they burst on the scene in 1987. I strongly suspect they’ve written some of yours too. 😉
[Jeff: have two posts preceding this one. Each contains two URL’s]
Turkey wasn’t elected to the UN Security Council.
But guess who was? Venezuela, Angola, New Zealand and MALAYSIA.
http://www.theguardian.com/world/2014/oct/16/venezuela-united-nations-security-council
#geopolitics