Brock McEwen has released a new reverse-drift analysis of the MH370 debris that has been found in the western Indian Ocean. The executive summary is below.
Broadly speaking, Brock’s new paper supports the conclusion of his earlier work on the subject, and also parallels the findings of GEOMAR and Météo France, as I’ve written about earlier–namely, that reverse drift analysis suggests that the debris did not originate within the current search zone.
In conducting his analysis, Brock has erroneously included objects found in the Maldives which did not come from MH370, but my understanding is that the inclusion of this bad data did not materially change his results.
The Australian is reporting that “Despite finishing his term as the head of the ATSB without finding MH370, [Martin] Dolan said he remained hopeful the aircraft would be found” and believes the search should continue. The full story is behind a paywall but Amanda Rose has provided a screenshot here. Also of interest in the article is the assertion that, due to bad weather, the search might stretch on through October.
Meanwhile the New Straits Times says that “The ministerial tripartite meeting on the Malaysia Airlines Flight MH370 will be held on July 19, Transport Minister Datuk Seri Liow Tiong Lai said Friday… Liow reportedly said that the meeting would deliberate on the next course of action regarding the search for the aircraft, which went off radar on March 8, 2014, with 239 people on board while on its way from Kuala Lumpur to Beijing.” China, Malaysia and Australia have long said that the search will end after the current 120,000 sq km search area has been scanned, but some observers hold out hope that the rash of recent debris finds will encourage officials to press on.
@buyerninety
Agreed drift trajectories are chaotic.
For lack of better material to date, I’m basing my conclusions largely on the adrift website, as the IG papers do. Nevertheless, it suggests the Roy find in Mossel Bay is the most inconsistent with the current search area (relatively spoken).
Having said that, I have also looked up other drift models available in the public domain and none seems to be consistent with that specific find and its presumed origin that far on the arc.
I nonetheless agree we are not talking about exact science here, given the length of time passed since the disappearance of MH370.
@Nederland
I doubt your statement that Godfrey’s remark; ‘the time of arrival was the fastest given possible by the Adrift model’ would favour obviously a relatively southernly starting point.
On the contrary; this would imply it favours more northernly starting points IMO.
Simply because if he had chosen the slowest possible arrival times those would have needed more distance to cover and therefore need to start more southernly.
As I suggested before, using reverse drift models brings in a misconception. It adds chaos the other way around leading to nowhere (or everywhere which has the same result). It’s a factual turning around of facts. Debris started moving somewhere on or near the 7th arc, not from there points of landing.
Does anyone sees this misconception also?
@George Tilton
Yes as you mention with that example it is surely possible. It still cann’t be ruled out completely but I must assume now it’s very unlikely this happened in favour of the more likely possibility the SDU reboot was triggered by the human intervention of at least one main AC bus was activated again at 18:25.
I think more likely/probably explanations should be favoured above more complicated ones as long as no other information support the complicated ones more covincingly.
@Ge Rijn
A reverse model yields an ensemble of possible starting points which is exactly what you would expect from such a model. Just as a forward models yield an ensemble of possible end points. It is completely symmetrical in a Markov sense. There is no intrinsic problem with a reverse model.
I do see a misconception. Yours.
One thing I noticed about Brock’s graphs depicting various models …. Is it was very low chance of the debris coming ashore on Reunion or Rodrigues islands. They are located in a very low probable for debris collection it seems as they aren’t in a conduit for debris…
@jeffwise – A few months ago, in addition to Brock, one of this peanut gallery’s posters (perhaps sk999 or MuOne) posted a drift analysis. I thought it was well done. I also thought I had bookmarked it but I didn’t. Is there a search method to find it?
@LaurenH,
To reduce your search effort, it wasn’t me.
@Ge Rijn
Drift runs roughly counter-clockwise in that part of the Indian Ocean.
That means the further south you have your drifters originate, the more do you have your parameters set to ‘faster/fastest’ to become consistent with finds such as the Roy debris in Mossel bay.
Lauren H,
Most everything I have posted is available here:
https://docs.google.com/document/d/14hleZyx1pUPL44yaeHKt6jnSQ3DbgRq2zibbKkFLq2c/edit?usp=sharing
@Ge Rijn
I am not advocating that my example did happen…crew efforts trying to resolve a possible thrust problem could cause a reboot of the SDU…that certainly is human intervention.
Human intervention can be nefarious or accidental…I lean to accidental…
Just an observation here, but in the past 2 1/2 years or so of this search, the only person/persons to find any pieces of the plane, have NOTHING to do with the data!!!
@all
Brook McEwen’s forward drift study has a chart on page 41 showing debris, excluding the Maldives, calculated using tighter definitions of nearness.
There is only one place on the chart where there are three consecutive high probability areas in an East West drift direction near the 7th arc.
This would be the minimum one would expect to see if debris started near here and before it had a chance to disperse.
This area is up near Christmas Island.
DennisW has shown us a graphic by a Russian analyst using sea level sensors located at four land based positions triangulating a disturbance round the time MH370 would have been at the 7th arc.
This point is up near Christmas Island.
A flightpath coming round below Sumatra at a constant speed fits the BTO’s like a glove without the need to change directions at the ping rings to make it fit.
The final location is up near Christmas Island.
This is surely reason to consider the area between Christmas Island and the coast of Java as a possible area for a ditching of MH370.
@Nederland
I interpretate ‘fastest possible times of arrival’ as the earliest possible time debris came ashore.
In case of the Mosselbay piece it was found in April but turned out to have come ashore at least 4 months earlier in December.
If you take the moment of finding in April as the slowest possible arrival time it needs to have started more southernly than taking the faster arrival time in December.
Maybe it’s only a different way of interpretation by me of ‘fastest possible arrival times’.
@DennisW
The GEOMAR study and the latest Brock MCEwen study show the intrinsic problem with reverse drift studies over such distances and time in such a big chaotic system IMO.
And the more debris landing points you add in such a model the more chaotic and useless it becomes in predicting a specific starting area, for chaos gets added the other way around. There lies the intrinsic problem in this case with reverse drift modeling IMO. There is nothing wrong with reverse drift modeling but the more chaotic a system is in itself and over time and distance the less useless it becomes for predicting starting points. Especialy when you start adding more debris landing points.
Those models show that clearly IMO especialy the latest Brock McEwen model.
@Bugsy:
Those persons have yet to find the flight recorders.
@All
Update on MH370 Search
http://www.news.com.au/travel/travel-updates/incidents/mh370-search-hits-another-setback-with-lost-equipment-poor-weather-and-sick-crew-members/news-story/b8b1888396e248e1a68fe2353f96c8df
@Ge Rijn
The methodical approach really should be clarified to be useful imo.
It could mean the respective debris arrived several (how many?) months prior to first discovery on each shore. But this does not seem to be possible as I couldn’t work out a way to have the Roy piece arrive in Mossel Bay long before December 2015. Broadly speaking, the further north you have it originate (discounting areas too close to Java), the more likely it is to make it that far. So, it very much looks like that you need to adjust your parameters to get a result like the one in the Godfrey paper. Obviously without a proper discussion of these parameters, the paper is of little value. It only shows that MH370 is unlikely to have crashed in the current search area and probably came down much further north, but everyone seems to be aware of that now.
Which “Godfrey” paper are folks referring to? Are you commenting on Brock’s paper or another thing entirely?
@Pail Smithson
It’s referring to the following papers:
http://www.duncansteel.com/archives/2652
It’s related to the Brock McEwen paper for it builds on a previous drift studies also of Brock McEwen mentioned in a previous paper of Godfrey:
http://www.duncansteel.com/archives/2631
The discussed difference is (IMO) those are based on forward drift models and GEOMAR and the latest Brock McEwen model on reverse drift modeling.
IMO there is an intrinsic problem with reverse drift modeling in predicting starting areas; it just predicts too many possible areas to be usefull caused by this intrinsic problem of adding chaos the other way around.
@Paul Smithson
http://www.duncansteel.com/archives/2652
@Erik Nelson,
You said: “Is there any way to make a “single turn FMT” at IGOGU towards ANOKO onto a 180 heading fit the rest of the satellite data ?”
The answer is no. You can’t hit the 19:41 arc with any roll mode or LNAV.
@Ge Rijn
I understand ‘fastet possible time’ is the time any debris could have arrived at the respective places at the very earliest (regardless of likelihood of this happening, as long as it’s other than zero), in relation to the points of origin covered in that paper.
The paper does not go further north than 28S, 100E, but, I feel, a point further north would match the evidence known thus far better than the ones discussed in the paper.
E.g.
21S, 104E
http://adrift.org.au/map?lat=-21&lng=104
Please consider the following speculative scenario. And please recall, according to FI, the Vietnamese stated that they tracked MH370 from IGARI as far as BITOD, and that the aircraft vanished from their radars only “after BITOD” (paraphrase).
So, perhaps, co-pilot Hamid was the acting pilot flying, with pilot Shah manning communications (including XPDRs ?), out of KLIA to IGARI-BITOD… where some crisis developed, which resulted in the aircraft turning 270-degrees, all the way around to starboard, i.e. the “Lido loop”. After looping around and crossing its previous flight path, between IGARI – BITOD, the aircraft stabilized, its flight-path straightened, and it headed NW at something like a 285-degrees, for ballpark 30s. During this period, Captain Shah took command of the aircraft, thusly implying co-pilot Hamid took over handling communications, as they assessed & evaluated the situation, and decided on a course of action. Then, the acting pilot flying turned the aircraft SW, initiating diversion towards Penang, back towards and into the KLIA FIR.
A long looping turn to starboard might be consistent, with an external electrical fire, on the left wing and/or left engine, extinguishing which would require depowering all electrical circuits on the left side of the aircraft, e.g. L MAIN AC bus. Also required would be to fly high & fast, to blow out the flames, and prevent them from re-igniting from smoldering wiring. Also, a turn to starboard would push any flames on the left wing outboard, away from the fuselage, engine, and left wing fuel tank.
A left wing and/or engine fire, could also explain the allegations of Mike McKay and Kate Tee, who claimed to observe fiery explosions & ghostly-glowing aircraft more-or-less consistent with the times & places MH370 is known to have been between 17:25 – 19:00.
Moreover, the “Lido loop” at IGARI – BITOD, could explain the supernaturally sharp turn, in the published radar track from MH370, which many have noted resembles the separate tracks of two planes crossing. Additionally, the extra distance travelled by the aircraft, could account for VictorI’s observed ~40s offset, between the actual radar returns, and those of his best-fit M0.84 LRC simulation.
So perhaps a LEFT wing and/or engine fire and/or electrical fire, and rescyulting “Lido loop” emergency turn maneuver, could account purported ground observations (McKay, Tee), and inferred de-energizing of the LEFT side electrical systems ? I would guess, that knowledgeable experts, would know what electrical systems route through the wings, especially the left wing and the outboard tip of the same, in order to circumscribe the range of consistent possibilities, to a manageable few.
In sum, if everyone keeps discussing problems on the LEFT side of the aircraft, e.g. L MAIN AC bus, then perhaps that implies an underlying cause of those symptoms, i.e. some crisis with the LEFT wing and/or engine ??
@Dr. Bobby Ulich
I agree, according to my crude MS Excel model, I find that something more-or-less like
NILAM-IGOGU and/or ANOKO-BEDAX-180S
best fits the BTO data. Thanks for the clarification, and please let me know, if any obvious explanation for ANOKO-BEDAX-180S exists.
@Ge Rijn
Actually, the model gets better as more debris is found since additional finds place additional constraints on where the debris could have originated. Likewise, the places (that have been well searched) where debris has not been found provide additional constraining information. I do agree that the drift information is far from precise.
@Bugsy
Yes, I know how frustrating it can be to see analysts changing their predictions with a frequency one normally associates with changing underwear. It is the nature of enquiry – hypothesis testing. I think the peer review in evidence on this site, while sometimes a bit brutal, is healthy.
I do think that there was misplaced overconfidence in the original SIO analytics.
@DennisW
A forward drift model gets better when more debris is introduced to refine a possible starting drift area and exclude others because it only includes places where debris is found. That’s why these latest forward models excluded a crash area under ~34S because the debris finds tell this. And because no debris is found on WA shores (and others as Sumatra/Java) they predict a crash area around 30S or some more north.
Reverse drift modelling only gets worse IMO if you add more debris places, for you only add more chaos modeling back resulting in a tremendous area of possible starting points.
IMO this latest model of Brock McEwen shows this clearly.
I would also like to read his opinion about this.
At last it’s his subject that’s discussed on this topic 😉
@Ge Rijn
My understanding of Brock’s work is that he did not use reverse modeling, but invoked repetitive forward modeling from various start points and accumulated the results of that forward modeling. Similar to watching poker on TV. The probabilities displayed for each hand to win are derived by playing many thousands of hands using the known remaining cards, and keeping track of the results.
Just to throw in that I would by no means discount the Geomar study. Rather than based on a website, it is state of the art research and a lot of very serious thought went into it. (For example, how do we know the adrift website’s duck represents an item submerged in the water?). The methodological approach is also clear, as it does take into account further debris finds, which are along the route the debris field takes:
”The debris that were recently found at other locations in the southwestern Indian ocean fit this interpretation. “The ocean currents through the Mozambique Channel and along the South African coast are extensions of the route that passes by La Réunion”, says Prof. Biastoch.”
http://www.geomar.de/en/news/article/mh370-wo-koennte-man-suchen/
You can easily verify that the model fits all other finds as well if you enter a given point of origin into the adrift website.
Not to the least degree, the Météo-France study comes to very similar conclusions.
@Ge Rijn
Brock McEwen’s latest model (30 Jun) is based on Adrift forward data, as is Richard Godfrey’s (2 Jun). The main difference, as I understand it, is that Brock’s is far more general, using start points all over the IO whereas Richard’s uses a restricted range of the 7th arc. Hopefully Brock will elaborate.
@Dennis W
Sorry I have partly duplicated your post about forward drift – I was watching tennis and got distracted.
@AM2 @Jeff Wise
Then I’m confused for Jeff Wise clearly states in his text on this topic: ‘Brock McEwen has released a new reverse-drift analysis’.
I understood therefore he used Adrift forward drift data in this ‘reverse-drift analysis’.
@Ge Rijn
It is a “how the problem is posed” issue. By definition, if you a known a starting point and the question is where does the debris go, then it is a forward model. Likewise if you know an end point and the question is where did the debris come from, then it is a reverse model. Forward and reverse do not refer to the algorithmic methodology in this case.
I do not fault Jeff’s use of the term “reverse model”.
@Nederland @DennisW
Offcourse I won’t (and cann’t) discount the state of the art science applied by those scientists from GEOMAR (and others).
I only take the freedom to question the usefullness of taking found debris points as basic input in a drift study calculating back in time introducing all chaos it possibly finds on its way back resulting in a chaotic pattern of multiple possible starting points and search areas.
I see an intrinsic problem in working that other way around in predicting a more specific search area for reasons I mentioned.
If this it not true I rather like to read arguments that can convince me.
@Ge Rijn
You are not listening.
@Ge Rijn
Although the overall effect of Brock’s modelling is to see where the debris might have come from, the forward method used has a large number of hypothetical crash locations and follows the potential drift from each of those locations etc..
You wrote “I see an intrinsic problem in working that other way around in predicting a more specific search area for reasons I mentioned”. I can’t see a model using forward drift method or one using reverse drift method being able to predict a more specific search area. What seems to be the case is that these recent models do add more doubt to the current search area (as Jeff said). I wonder whether they will consider using the remaining funding to move the search after the weather improves at the end of winter in Oz (Sep or Oct…).
@DennisW @Brock McEwen
I understand the methodology is the same but IMO the outcomes are the oppisite of eachother.
A forward-drift study from a specific chosen area (in this case near the 7th arc somewhere) can predict where debris can be expected to be found and where not.
In this case the latest forward drift studies done this way related to the found debris till now points to an area north of ~30S till around ~28S.
A reverse-drift study can not predict a specific search area in the same way IMO.
On the contrary; the more found debris points you introduce the more possible starting point and possible search areas come out of it. In other words chaos increases the other way.
It’s this point I like to have clearer. It might show stubborn but Brock McEwen states in his paper; ‘peer review of all assumptions and analysis is warmly welcomed’.
So I invite him to comment on this maybe ridiculous thinking of me.
I would be gratefull.
@Ge Rijn
The problem with all drift simulations is that none can define a suitable search area. This is because there is just one impact point but multiple varieties of probabilities (major or minor) in each model. All they can say is it is unlikely that MH370 impacted, say, south of S30. It may therefore be worth considering (imo) that the purpose of the flight may not have been to bury the plane as far away as possible, but to consider other scenarios such as an intended but failed landing (for whatever reason), in which case there aren’t very many possible destinations in that area.
You’ll get the same results if you choose some point along the arc, for example, any drift study I am aware of precludes the current search area (or one even further south) as inconsistent with some finds such as the one in Mossel Bay. I’d like to see some more professional drift modelling done to see whether or not other potential areas (such as those identified in May or June 2014) are generally consistent with the debris finds, something I’d expect – but then again what’s the point as the ATSB seems unwilling to redress their current search strategy.
@DennisW
I think you don’t see my point yet.
If it turns out to be nonsens you’ll hear it from me to admit it.
I’d read the chart above the following way:
Every coordinate is assigned a probability (applying a forward drift model) that it may be the point of origin for debris arriving at the destinations known so far. The bigger the orange spot, the bigger the probability. Obviously, it doesn’t make sense to search for the plane unless one puts trust in the Inmarsat data. I don’t see a problem with the fact that the intersection of the probability field with the seventh arc isn’t the strongest, as long as there is a reasonable intersection.
The two reverse drift models, on the other hand, (Geomar and Météo-France) take only the flaperon find into account. So there is no confusion. The argument is that this was the earliest find and therefore gives the best idea of where MH370 could have crashed. Other debris finds are an extension of the proposed route. You can, however, perhaps discount parts of that area if they turn out to be irreconcilable with forward drift models including finds as the one in Mossel Bay. So, definitely it’s a good idea to compare charts.
@Nederland
‘whats the point as the ATSB seems unwilling to redress their current search strategy’.
I don’t know where exactly anymore I read it but they and other officials are also reviewing all serious independent sources of information.
I’m quite sure someone there is also reviewing this blog of Jeff Wise.
And then it matters IMO.
Sometimes (often 😉 it’s nonsens and sometimes something valuable comes out.
And it keeps some pressure on.
So IMO there is a point in keep on going discussing anything possibly sensible.
@Ge Rijn
I agree.
Just Geomar et al don’t seem to be very happy that the ATSB haven’t or barely have responded to their models to this time, however much they collaborated with others and refined their analysis.
Otherwise, I’d say we’d have more reliable data in hand to discuss.
@Ge Rijn: my study was not 13 (or 16, or 9, or 5) reverse drifts, averaged. It was 745 forward drifts, ranked (with rankings based on how well each explained all finds).
As soon as I saw Jeff’s use of the word “reverse”, I corrected him (see p1 comments).
I understand the confusion, because as soon as one hears of a study which uses subsequent debris finds to indicate an original impact point, the mind immediately leaps to the assumption that reverse drifts were used.
I chuckle a bit at the whole debate, though, because the “stink” on reverse studies – as I’ve said before – is not intrinsic, but rather very specific to IO currents. As Dr. Pattiaratchi explained to me, the “many streams into one river” effect SW of Sumatra means that reverse studies – in this application, but not necessarily in others – spray out more dramatically than do their forward counterparts.
The other advantage I saw in choosing forward over reverse was strictly an artefact of my data source: I was able to ensure all 745 forward drifts started in March – had I tried to base rankings on 16 reverse drifts, each “starting” from the location of each find, I’d have been unable to set the month to the actual month of each find – only December is available.
I hope this clarifies.
@Dennis: re: “reverse”: actually, I agree. It is in aggregate a reverse analysis, built by stitching together 745 sets of forward drift data. No need to quibble on either side.
@Brocj McEwen
Yes thanks, this is clarifying.
At first I missed your correction to Jeff about the word ‘reverse’. So that’s on me..
My interpretations of your study would have been different then.
Nonetheless it seems you share my point of view that reverse drift studies can have a tendency to spray out more dramatically than its forward drift counterparts in certain areas.
I still see reverse drift studies causing a ‘one river into many streams’ effect resulting in an abundance of possible starting points and areas, and forward drift studies causing a ‘one river into many streams’ effect causing an abundance of possible but also impossible landing points of debris depending on where you start your forward drift analysis.
You stated in your study you did not limit yourself to the 7th arc and offcourse there’s no objection to investigate something like that. But IMO your results , conclusions and suggestions imply to dismiss the 7th arc. And if you decide to do that you are left with almost half of the IO as a potential search area in your latest study which renders every further search effort useless IMO.
But if that is what the truth has to tell then it must be so.
@Erik Nelson
“So, perhaps, co-pilot Hamid was the acting pilot flying, with pilot Shah
manning communications”
No ‘perhaps’, rather it’s always been considered high probability that
Shah was training/testing the Co-pilot (acting pilot) during the flight
as Co-pilot was transfering to flying 777 from his previous experiencce
flying A330’s. Per FI;
“On the day of the flight, he was operating his last training flight
before he was scheduled to be checked out on his next scheduled
flight.” (Not sniping, merely stating.)
Do we have a misunderstanding of terms? I thought it was agreed the
‘Lido Loop’ was a false picture generated by a Russian PR firm. That aside,
my opinion was that the aircraft HAD probably done a right turn due to the
co-pilot, with fire jetting out at the level of his right knee, jumping
out of his seat, thereby pushing hard down against the right rudder pedal.
(Possibly a column forward push also occuring). Pilot Shah then naturally
correcting left back onto course, followed quickly by a (continuation?) left
turn as Shah headed the aircraft towards Penang(?) airport. (I’m still
pondering why not Kota Bharu – ATC hours of operation? Emergency services
hours of operation?)
hXXp://aip.dca.gov.my/aip/AIP2016/html/eAIP/WM-AD-2.WMKC-en-MS.html#AD-2.WMKC
hXXp://aip.dca.gov.my/aip/AIP2016/html/eAIP/WM-AD-2.WMKP-en-MS.html#AD-2.WMKP
What I took from Brock’s paper is that based on the location of the confirmed debris finds, it would be far less likely for the starting point to be anywhere along the 7th arc then it would be for other points in the Indian Ocean. This is not to say it is impossible for the starting point to be somewhere along the 7th arc considerably north of the current search zone. Just that the 7th arc is less likely when compared to other hypothetical IO starting points not limited to the Inmarsat data.
@all Has anyone looked at this?
We know this was Hamids final check flight. Could Shah have set up a situation to test Fariq that got out of hand? In other words could Shah have set up a small issue and Fariq went too far. Possibly simulated a fire, shut down the transponder and ACARS as part of a test?
I know it’s crazy and dangerous but so is everything about mh370. I just hadn’t seen it brought up in the forum.
@Brock McEwen
Have to correct a small thing. When I turned around your analogy; ‘many streams into one river effect’ I meant to say; ‘one river turning into many streams effect’.
And I understand you didn’t dismiss the 7th arc with your study. The effect is only it renders the 7th arc and with it the Inmarsat data almost of no use IMO.
Another issue I like to suggest something about is your remark that you have been unable to set the month to the actual month of each find for only december is available.
You mean the Mosselbay-piece I assume based on its abundance of barnacles in the december 2015 picture.
I suggest then there is another piece available; the flaperon. It also had an abundance of barnacles suggesting it did not landed a long time before the date of finding in july 2015.
For the flaperon to be the piece with barnacles found closest to the 7th arc area and the Mosselbay-piece the furthest I think it can be reasonable to assume all debris found till now has landed within the timeframe of july 2015 and december 2015.
That’s all. Hope I don’t bother you to much with my writing..
@Ge Rijn
@Brock McEwen
The “ROY” piece was first found heavily barnacled, then re-found some time later barnacle clean.
Is it possible to infer from that interval, a rough “time frame for cleaning by predators” which could be used to determine a prior “latest beaching date” for other items which are barnacle clean when found, and would that help matters ?