As I’ve discussed in earlier posts, by its own calculations the ATSB has already searched most of the high-probability areas of the Indian Ocean seabed in its quest to find the wreckage of MH370. The only remaining area of relatively high probability that has not been searched is a stretch along the inside the 7th arc.
(In the image above, the area that had been searched prior to the release of the ATSB’s December 3 report is outlined in black.)
Yet this is not where the search is currently underway. According to ship-tracking conducted by Mike Chillit, Fugro Discovery has spent the weeks since the ATSB issued its report searching an area 40 nautical miles beyond the 7th arc, in the pale blue “low probability” area of the ATSB’s heat map.
It’s hard to understand why.
UPDATE 12-22-2015: I was delighted to learn that Richard Cole is back on the case, paralleling Mike Chillit’s work by collecting and collating ship-movement data in order to understand what areas have already been searched. Richard has given me permission to reproduce one of his charts, which shows the situation much more clearly than my amateur effort above. I’ve outlined the area already searched in light blue. One thing I notice looking at this is that the unsearched high-probability area near 87.5 N 37.5 S hasn’t even been bathymetrically scanned yet! Click to enlarge:
Mark Fox – You aren’t completely alone. If the SDU hadn’t fired back up as MH370 flew away from radar coverage, the Maldives would be the search area and there would be no SIO search at all. Satellite engineers look at the reboot and look at the eyewitness accounts from Kudahuvadhoo and say – I’ll take the reboot thanks. But what about two years later with the issues banking up? Some dig in and some relax their thinking. If the effect of the reboot was to send the search down to the SIO, was it also the intention? To me a spoof involves BTO as well or don’t bother. Is anyone game enough to say it can’t be done?
The reboot is the difference between the Islanders being credible witnesses and being considered a bit bonkers.
@Matty
Why fly a perfectly good airplane full of PAX to that area?
It is less defensible than the current search area since the required spoof implies intent. You cannot blame it on some improbable combination of aircraft failures.
Dennis – I think there are plenty of why’s from wherever you look at it. Why fly it anywhere apart from Beijing? We don’t know but when planes have been diverted in the past it has always been bad news. Many reasons are given for the causes of a reboot but a scenario involving human interference is apparently ruled out. I always found this strange as there is so much evidence of interference/intent. Some will stake their considerable technical expertise in saying it’s virtually impossible but who in this forum can claim to have ever been employed to engineer such deceptions? Such people exist. The ‘why’ is beyond our field of view.
Dennis – My concern with the whole thing is this: the reboot is a source of mystery and conjecture, but it is solid enough to discount every other avenue? Is that a good position to be in? And yes I know…..it’s the best we have – but for how long?
@Matty
“The ‘why’ is beyond our field of view.”
Ah, taking the “IG Defense”?
It has become almost as famous as the Nuremburg Defense.
Best wishes to all for 2016.
Here are some of my theories for today (generally I’m still sitting on the fence):
1. The Maldives sighting was of a plane carrying the Royal couple, taking them to DG at first light for safety and taking in a little bit of sightseeing on the way. This would be after the fate of MH370 was known to authorities.
2. The flaperon may well have been an original on 9M-MRO but the wording, sometimes quoted as “correspond” or “correspondait”, is too vague for my liking if it means the same in French as English (see Oxford Dictionaries: “Have a close similarity; match or agree almost exactly”). There was wing damage to that plane in 2012 (http://aviation-safety.net/wikibase/wiki.php?id=147571) and it is unclear whether a flaperon was replaced then and stored… The French have not gone as far as saying THAT flaperon was on THAT plane THAT day (I wish they would). Even so, it seems very possible that the flaperon was planted whether the plane crashed in water or landed somewhere.
@Jeff – you have said more than once that it could have been planted. Have you any information from your many sources? This could be the key to unravelling the whole mystery.
@ no one in particular
I spent part of the day re-reading “Bayesian Methods…”. My conclusion is that experience really does matter.
@Matty
“My concern with the whole thing is this: the reboot is a source of mystery and conjecture, but it is solid enough to discount every other avenue? Is that a good position to be in? And yes I know…..it’s the best we have – but for how long?”
for long enough to search the whole reachable 7th arc and its surroundings
StevanG – Hopefully, yes, but is anyone betting the farm on finding it?
@Matty — The SDU could have been rebooted as the result of deliberate action, or it could have rebooted as a result of the left AC bus being isolated for another reason: perhaps, for instance, a suicidal pilot wanted to turn off the Cockpit Voice Recorder. However, unless the entire Inmarsat data set was cooked up as a scam from the beginning, any explanation of what happened to the plane must somehow deal with how those values were generated. So far, the only explanations anyone’s been able to come up with are a) the plane wound up on the 7th arc in the Indian Ocean, or b) the values were spoofed in an almost inconceivably sophisticated hijacking operation.
I feel that it’s not really useful to posit any theory that simply ignores the Inmarsat data set (or, for that matter, ignores the reboot, as “accidental fire” scenarios do)
@Matty
I’d invest some of my own money if someone decides to search a bit northern, it’s simply a value bet for me (although I wouldn’t request it to be paid back if the plane gets found there).
@StevanG
An Interesting dynamic, if a vast sum of money was independently sourced, could a private search take place on the “high seas”?
Can’t see why it couldn’t, the good thing is that search on the “high seas” should be much cheaper per square mile than in the current search area.
Here is a link to a nice article on how Kalman filters work. It cuts out much of the esoteric math jargon and works out a very simple example (with lots of simple graphs) that matches quite closely the MH370 path analysis.
http://www.bzarg.com/p/how-a-kalman-filter-works-in-pictures/
excellent post however Kalman filter didn’t really work in MH370 case so it’s time to move on and scrap obviously wrong assumptions
Jeff – I’ve felt for a while that if anyone set out to spoof data they would start with the BTO’s or not at all. What if for instance the BTO’s revealed a continuous path to the west? Would we have even had a BFO analysis at all? Or what if the BTO’s in such an instance didn’t line up with expected BFO values? I reckon they would have been explained away or attributed to the inherent vagaries. The BTO’s were the obviously foreseeable component. Sounds like a stretch, but to me it’s a bigger one to say never.
StevanG – If it’s in the water then somewhere further up the arc and largely intact is looking more likely all the time. Being 60 metres long would it be viable to follow the arc with a bathy and look for long hard objects? Sounds a lot cheaper.
First of all, hope everyone had a good Christmas and New Year (and ‘have a Merry Christmas!’ to any Russian posters).
@Brock, VictorI, Matty
You have no need of praise from amateurs like me, but I salute your objectivity and open-mindedness in still entertaining the Maldives angle. Such an open mind is what the search for MH370 needs.
@all
Jeff and Susie are both right. This may be the year. But deep down, I wouldn’t want the whole saga to turn out to be a technical malfunction/lithium battery fire. In fact, I’ll feel a sense of total emptiness if that happens.
I know it seems cold and heartless to ‘root’ for one scenario, after all this isn’t a video game. But allow me to say something totally insane: as long as the plane hasn’t been found, the faintest chance of survivors and the plane having landed still remains.
Remember the phantom cellphone ringing? The NOK were clearly alarmed by it, Jeff Kagan on CBS later dismissed it. But I’ve always wondered why we had such a reaction from the NOK. After all, didn’t they know how their own phone providers work? Here in the UK at least, a cellphone normally rings when its being received at the other end (and goes straight to answerphone when it isn’t). Not sure about the US/Australia/China though.
Yes, I’m clutching at straws here and yes, we all presume 239 people lost their lives. On the other hand, we still cannot say for sure with just a lone flaperon for evidence.
On a slightly different note – since Blaine Gibson’s visit to Kudahuvadoo in July 2015, another independent investigator, Sergio Cavaiuolo, has recently visited a neighbouring Maldivian island (in December 2015). There he found three fresh witnesses who stated a low-flying plane appeared to circle their island (around 7am in the morning) before flying off.
But Jeff is right. Eyewitness testimony can be faulty. What is slightly annoying is locals can’t seem to tell the difference between a Maldivian airliner and a Malaysian one. But they saw something, without doubt.
And also, without doubt, another Malaysian airliner did fall from the sky before the end of July 2015 (exactly as predicted by a former pilot). A hydrophone signal was heard on the seafloor on the same day as MH370 vanished. A duplicate of MH370 did lie in a Tel-Aviv hangar a full four months before MH370 had even departed for Beijing. An American Reaper fin did wash ashore in the Maldives roughly the same time as the flaperon. The ’coincidences’ just keep adding up.
Nothing should be disregarded. The Maldivian sightings may very well not have been MH370, but can we really say for sure that a low-flying aircraft on the very same morning that MH370 vanished had NOTHING to do with its disappearance? Isn’t it foolish or something even worse to throw away clues that might actually lead us to the truth?
@Susie Crowe
You are an amazingly brave and altruistic soul to have gone through such a deep personal loss and yet still find the time to offer your help on here.
I always enjoy reading your posts as I’m sure many others on here do. Your gentle nudging and keeping of the peace in an often heated forum does not go unnoticed. You offer a great deal on here, whether through pursuing officials for answers or your contributions to the forum.
@Matty
“Being 60 metres long would it be viable to follow the arc with a bathy and look for long hard objects?”
sure it would be, although the probability it’s right on the arc isn’t exactly huge it’s still more probable than current search area
Matty “If it’s in the water then somewhere further up the arc and largely intact is looking more likely all the time. Being 60 metres long would it be viable to follow the arc with a bathy and look for long hard objects? Sounds a lot cheaper.”
Steven said half of it: The width of the cross track error of descent and the corresponding sink to the bottom from the arc is one issue.
The other, being the length of the arc.
To do as you suggest will make the current search effort and cost pale into insignificance. That said, it it really all about the cost? But the buck has to stop somewhere.
Rightly or wrongly, that’s why the powers that be analysed what evidence they had and came up with probabilities for search areas. They cant search it all.
You can take them to task on their AP hypothesis, fuel exhaustion, drift models and some obvious other falsehoods to build such probabilities, but to search a major length of the 7th arc isn’t going to happen in my lifetime.
SIO,CI,Kuda,SCS or any other place bordering the oceans of the world all suffer the same fate. No wreckage. Your choice of theoretical terminal location really comes down to the probability of how you view the scant evidence at hand.
That is bonkers.
What width they cover with one pass? They have spent north of $100M in searching of current search area, if one pass of the 7th arc north of current search area costs a million (roughly), they could cover 100 of those widths or somewhat less if we include overlay.
The DST Group analysis released in December 2015 is looking back and not forward. It merely regurgitates the analysis that led to the priority search area defined in June 2014. What needs to be done is to take account of the knowledge gained since, i.e. the negative results of the area searched sofar. That is what Metron did in the search for AF447 and what led to the finding of the crash location. That is what needs to be done in the MH370 search, in particular for the end-of-flight scenario. Maintaining the hypothesis of uncontrolled descent and therefore a probability distribution that is symmetrical about the 7th arc (figure 10.9 of the B.M. paper) is not compatible with the known facts anymore.
@Gysbreght, I’m no statistician, but my understanding is that adding independent new information to a Bayesian analysis will generally shifted the probability distribution. It’s interesting to note in the Dec 3 report how the addition of a) BFO data, and b) drift analysis altered the probability distribution previously derived from BTO values: it didn’t. In my interpretation, that’s because neither is compatible with the calculated probability distribution. It’s as if you’re trying to figure out whether your wife moved to LA or San Diego, and then a postcard turns up with pictures of icebergs: it doesn’t help you decide between the two Californian cities because it’s compatible with neither.
[People with an actual solid understanding of probability, feel free to correct me.]
I guess I’m making a separate point from your observation about the need to include the negative findings of the subsea search, which is also an important one. It’s disappointing that the “Bayesian Methods” book didn’t include a proper Bayesian analysis.
@Matty, A BTO spoof requires access directly to the SDU; when I wrote “The Plane That Wasn’t There,” I believed that the hijackers wouldn’t be able to do that. Gerry Soejatman has set me straight about that, and indeed the two Ukrainians were sitting almost directly under the SDU access hatch. On the other hand, it remains true that few outside Inmarsat knew that the company had only just begun to record BTO values, which until then was generally not done in the industry. So it’s reasonable to imagine that even very sophisticated hijackers might now know they would need to spoof the BTO.
If both the BTO and the BFO values were spoofed arbitrarily, then the Inmarsat data doesn’t really tell us anything at all, and the plane could have gone pretty much anywhere in the world. I’m still not impressed by the Maldives eyewitness accounts, though. Why, having carried out such a fantastically complex plan, would you let your plane get eyeballed at low altitude by villagers? And what could you possibly be trying to achieve?
@jeffwise: In my opinion, the fundamental weakness in the Maldives theory is that if MH370 was seen at the time claimed, the plane either landed or crashed soon after. There is not evidence of either. The entire set of satellite data might have been somehow altered, but how do you hide a B777?
@Jeff: in a Bayesian analysis, the insensitivity of the distribution to the addition of new data can mean one of two things: either the new data is compatible (postcard is of SD Zoo), or the new data carries little weight (postcard is blank). The “book” explicitly states that it down-weights the drift analysis – ostensibly to reflect a much greater confidence in the signal data than in the drift result.
The failure of search leaders to truly refine their search area by attaching material weight to either surface (your point) or sea bottom (Gysbreght’s point) search results are both consequences of an abiding trust in the signal data*: if what Inmarsat published in May/’14 is forever an axiom, then “the odd pdf of that odd pdf” will never be falsified – we’ll just have to keep deeming our search results increasingly unlucky.
In my opinion, both this abiding trust and those who peddle it should be examined.
* Many will rightly point to the additional assumptions lying between the pdf and the pdf – fine, but this makes search leaders’ determination not to move the search even more inexcusable.
@Jeff
Kalman Filtering, Bayesian Inference, Monte Carlo, particle filters,… all great tools, but none are applicable to the MH370 terminus problem IMO. I believe the disappearance of the aircraft was the result of deliberate human actions. Just like the flying of planes into the twin towers. You would not use a Kalman filter to explore and refine that outcome. It was not stochastic, and the towers were not struck because they were the tallest structures around, and therefore most likely to suffer from aircraft impact.
All of us bring a lifetime of bias to any problem solving situation. If you are a carpenter you look for a nail to strike. If you are a techie, you gravitate to the numerical data. Sadly, the MH370 terminus problem has been turned into an episode of “Numbers”, and lines have been drawn in the sand by most of the parties involved.
The reality is that the numbers cannot be used to predict the terminus. I don’t care what tool you plug them into. The numbers can only be used exclude terminal locations.
@Brock McEwen: The DSTG put no weight on the drift analyses, relying upon the satellite signal data, the radar data, and (incorrectly) bounding the performance of the aircraft. I suspect that the DSTG would claim that they ARE including the results of subsea search in this way: after the higher probability areas are exhausted, it will use the Bayesian logic to determine whether to re-scan previous areas or scan lower probability areas. Remember–there was a 10% probability assigned to scanning and missing the aircraft.
The authors of the DSTG book may have impressive CVs, may know a deal about Bayesian analysis, but their understanding of B777 flight control dynamics is amateurish.
Have a close look at Fig 6.2, purporting to show Magnetic Declination in the SIO region. It is complete nonsense. If they have used the data behind this chart in their analysis it is no wonder that the current “hot spot” is confirmed.
@Brian Anderson: Good find on the figure in the DSTG report showing the incorrect magnetic declination.
So now we have but another reason why their analysis favors straight paths: in addition to their incorrect lower speed cut-off of M0.73 and their assumptions about stochastically distributed maneuvers, their calculations use values of magnetic declination that are too low.
I doubt that the assumed “flight dynamics” and the lower speed cut-off have much effect on the resulting pdf on the 7th arc. Perhaps allowing the BFO residual error to drift down more than 10 Hz is more significant. I wonder how the “hot spot” would shift if the BFO values were given more “weight” in the relative “BTO and BFO measurement weighting”.
@Gysbreght: With the assumed error band on the BFO data, the DSTG paths essentially ignore the BFO data except for bounding the time of the turn south to be between 18:28 and 18:40 UTC. In fact, with lower speeds, there are paths ending much further north than the search zone, especially if multiple turns are allowed.
@Gysbreght
I can get anywhere on the 7th arc from North of Christmas Island to the Southern extreme of the current search area while satisfying BFO and BTO. Sooner or later people are going to come around to the idea that the BFO data has almost no roll to play in terminus location. You could probably argue that BFO narrows it down to the Southern hemisphere. Highly restrictive flight dynamics are the only thing that allow you to stick a pin in map.
@Victor: re: 10% miss rate: good point – though if they’ve done this, one would expect all unsearched regions of the distribution to have, all else equal, nine times the density of their searched counterparts.
Since this “hole in the donut” is not apparent, the only way I can imagine “Bayesian methods” could have incorporated the 10% stat is on a more macro level: a mathematical equivalent of the following argument:
“Signal data must be right, but we’ve searched enough pdf to rule it out, therefore publish ORIGINAL pdf, since best bet is to search it all again.”
If so, why not just admit this?
That’s the IG’s canon. Sadly, they ended up empty-handed at the southern extremity of the reachable part of the 7th arc.
@DennisW, I agree with 90 percent of what you’ve been saying today but I don’t think we have to consider the flight dynamics “highly restrictive.” In order to go slower/further north, the plane would have to follow a curving trajectory, which no autopilot mode will generate. To get north of broken ridge, in fact, we have to imagine that whoever was at the controls was manually making a number of intermittant course corrections to the left, in a way that mimicked an arcing path, yet by coincidence crossed the ping arcs in a way consistent with normal speeds and autopilot modes (i.e. straight and fast). This would really be one helluva coincidence and I can hardly fault the ATSB in setting it aside when deciding where to spend the taxpayers’ money.
PS I’m aware I’m being a broken record on this but I feel like many of us are close to being on the same page.
I agree with the intermittent course corrections to produce an arcing path, but why “normal speeds and fast” ? Wouldn’t the speeds be lower?
@Gysbreght, poor wording on my part. The relative spacing of the ping arcs is such that they match a course as straight and fast as planes usually fly. If the plane was actually flying slowly in a curve, it generated this pattern only by sheer chance.
Perhaps Brock can assign a probability to “sheer chance”?
What is the probability that someone was controlling the autopilot?
@Jeff
My CI path is actually comprised of three straight paths – not because I think that is the path the plane actually took. It is because I am lazy, and stopped screwing around with it as soon as I convinced myself it was feasible. As I have stated before, I am concerned about feasibility not about reconstructing a flight path from ambiguous data. What would be the point in doing that? Autopilot modes have absolutely nothing to do with the generation of BTO and BFO. Those values (BTO and BFO) are dependent only on aircraft position and velocity (and time since the satellite motion contribution is time variant).
Of course, BFO is also influenced by system error such as drift in the oscillator chain.
I don’t blame the ATSB either. They hitched their wagon to the wrong ponies (as you did early on). The notion of a likely fixed AP flight path after the FMT because “that is the way pilots like to fly airplanes” still causes me to chuckle in the shower.
The ATSB was also victimized by some unfortunate timing. Had the flaperon been found before the underwater search started or had the ATSB been sensible and waited to start the search until some debris was recovered, the landscape would be very different today. Retreating now is virtually impossible. It would be far better (from a public relations standpoint) to continue searching where they are searching until June and then pulling the plug, than deciding to search somewhere else based on a reconsideration of all the available data.
I am anxiously waiting for another show to drop – the French publicizing the conclusions derived from the flaperon forensics.
@Gysbreght
Yes, as you know there are an infinite number of flight paths that can satisfy BFO and BTO. Because a path is straight (whether or not it is fast) does not create a reason to regard that path as somehow an “anointed” path. It has the same status as all the other paths. In fact, there are an infinite number of straight paths that fit the data. These paths gradually shift to the NorthEast as speed is decreased.
The path taken before the FMT was certainly not a straight path.
OK, for the record, I’ve been working on a report that is still in draft stage, but one part has been looking at where along the 7th arc the BFO is minimized. To constrain parameter space (but in a way that is reasonable), I assumed constant Mach along a route and varied the heading at each ping ring crossing so as to match the BTOs. Slower speeds end up further North – we all agree. The routes are curved – we all agree. The minimum of the BFO is around -30 to -32 deg latitude, and the appalling systematic slope in the BFO residuals that many have commented on (see, e.g, Baysian Methods, Fig. 10.5 bottom) is largely eliminated. Further, a route near this minimum also happens to fall close to a magnetic track route. Magnetic track routes are also curved, as was pointed out long ago on a different forum. I haven’t been able to figure out how to make them align better, but still working on it.
Also, for what it’s worth, I have looked through records of past aircraft diversions to see how the planes were flown. In every case where that information is available (4 total), the planes were on autopilot.
@Brock McEwen: Let me be clear. I don’t believe that the PDF contours they published reflect knowledge of the search. I do suspect that as they venture into lower probability areas, some of the highest probability areas will be re-searched, and Bayes theorem will help them decide where. This also might explain why there seems to be some areas recently re-searched. There might be something in the scans that warrants a re-scan rather than blindly continuing into lower probability areas.
That said, since the aircraft debris likely scattered as it sank to the seabed, it would be likely that if they scanned and missed debris, there would be other debris found on nearby scan passes. Based on this, to find any debris in the search area at this point would be very low probability.
@jeffwise
“In order to go slower/further north, the plane would have to follow a curving trajectory, which no autopilot mode will generate.”
it’s true if we assume the altitude was constant, but we don’t have that data, actually fairly reliable radar calculations involving the end of radar track (at 18:22 or 18:28 forgot now) show the airplane descended to FL260 from crusing altitude and that trend likely continued
we also don’t know if autopilot was engaged(for most of the time), it’s just an assumption
I do however agree with you it would be a huge coincidence the BFO&BTO just matched straight line to nowhere but I had bigger coincidences happen in my life.
@Gysbreght
“What is the probability that someone was controlling the autopilot?”
quite high I’d say, if nobody was controlling the plane at the time of impact we would have lot more debris some of which would have to wash ashore somewhere
Brock has calculated it somewhere, if their was only one piece floating and if the chance for noticing it was only couple of percent, if you have hundreds of such pieces the chance gets quite close to 100%
@Gysbreght: I’ve resolved not to criticize anyone’s speculation in 2016, so I will merely a) applaud your efforts to come up with something plausible, and b) attempt to respond to your specific request for a probability. Since math isn’t likely to help, here (would need to generate millions of possible paths, including millions of randomized possible turns, and examine the distribution of “straightness” among the small percentage of paths which fit the BTOs), I offer an analogy to help characterize what I believe the probability to be:
Suppose you are given a wooden ball, into which many odd-looking etchings have been carved. When you happen to view it from a particular orientation, the etchings magically resolve into a beautiful face. I would conclude that this face was the artist’s subject, and not what any other orientation produced.
This analogy exaggerates how I’ve always felt about paths which curl by just the right amount to offset the effect of their slower speed – but not by much.
Regardless: I dearly hope I have your support for a public demand for fuller disclosure of radar, signal, fuel/performance, and flaperon data/analysis; the secrets such disclosure would reveal may well prove me wrong, and you right.
Sk999,
What do you mean by “I haven’t been able to figure out how to make them align better, but still working on it”?
Align with what? Better than what?
Jeff,
Re: “If the plane was actually flying slowly in a curve, it generated this pattern only by sheer chance.”
Then what? I have already shown that a mechanically-primitive flight mode does exist, and it satisfies BTO & BFO due to wind and Coriolis. In addition it fits 23:15 BFO, while “classic” AP deviates considerably. By chance.
Besides, another class, the magnetic-heading-AP mode, also has solutions with the same-order residuals as the “classic” AP.
After visiting the Inmarsat website, I’m curious. What data was transmitted between the plane and the satellite throughout the whole flight? Were there any other types of communication, phone, internet access, VoIP phone calls or GSM or SMS messaging? Inmarsat’s website says: A single Inmarsat installation enables a wide range of uses in the cockpit and the cabin. These include safety communications, weather and flight-plan updates, as well as passenger connectivity for email, internet access, VoIP telephones, GSM and SMS messaging.