By MPat
(Note: A comment by reader Lauren H brought my attention to an analysis I’d overlooked by reader MPat. As Lauren H points out, it’s as timely now as it was when MPat first aired it back in March. — JW)
The potential arrival of more debris in the East African region is triggering interest once more in the currents and drift patterns in the SIO. To sense check the concept that debris could drift from the current search area to these regions I did a little research of my own, the premise being that the observed behaviour of real floating objects (and I am considering of course the buoys of the Global Drifter Program) should be a useful indicator of possible drift pathways, as a counterpoint to cell-based drift simulation models (which may be calibrated to high level drifter behaviour but typically lack the resolution to reproduce drifter movement in detail).
The full drifter database contains meta-data and trajectories for almost 19800 buoys worldwide (some 1400 are currently active). The meta-data includes timing of drogue loss, and a ‘death’ code to categorise the end of life status of buoys that cease transmitting. It is clear from this that drogues are typically lost in a surprisingly short timeframe. It is also notable that only 20% of all the buoys have ended their lives by running aground, with 66% simply ceasing transmission for undocumented reasons.
I have filtered out buoys that have at any time in their lives passed through the locality of the current search zone, based on a rectangle bounded by longitudes 88 to 96 degrees and latitudes -32 to -39 degrees. None were present in this area at the time of the crash, but I consider in any case all buoys that have ever been in this location (dates range from 1995 to 2014). There are 177 in this category. Of these, 39 are listed as having subsequently run aground. The locations at which they washed up are shown in the plot above.
Of the 39, 31 beached on East African coastlines, only 7 in Western Australia, and 1 in Sumatra. An example of 3 randomly chosen trajectories from the 31 that drifted west are shown below together with the box defining search locality :
The average time for buoys to reach their western beaching point after leaving the search box is 534 days (~ 18 months) with minimum 234 days (~ 8 months) and maximum 1263 days (~ 42 months). All but 3 were un-drogued during this journey, and those 3 lost their drogues en-route. For those arriving in Western Australia, the average time to beach was 362 days, with minimum 79 days and maximum 513 days.
If we relax the criterion that the buoys must end by running aground, and simply look at the locations where they eventually stopped transmitting after leaving the search area, we see the following three plots which display the 54 buoys that ended up west of longitude 55 deg (the longitude of Reunion Island),
the 12 that ended east of longitude 109 deg (coast of Western Australia),
and the 111 that remained in between:
Clearly the transport qualities of the ocean currents and weather systems will vary from month to month and year to year. It is also not clear how representative the buoys would be of the drift characteristics of floating debris resulting from a crashed aircraft. Neverthless I believe it is reasonable to propose from the buoy behaviour noted above across a 20 year drifting history that :
i) there is a strong tendency for objects that have been present in the current search area to remain trapped in the mid ocean gyre over extended periods
ii) a proportion, perhaps as high as 10% of robustly floating debris, might be expected to make landfall within 18 months of the crash
iii) the vast majority of the debris making landfall is likely to do so across the coastlines and islands of eastern Africa, with relatively little beaching in Australia.
For what it is worth, I have more background and analysis in a write-up that I hope to post soon.
Please also note that a vastly more expert analysis of drifter behaviour has been performed in October last year by David Griffin of CSIRO, in which he uses composite drifter trajectories to infer a likelihood function for where the MH370 flaperon may have originated. This is well worth a read.
UPDATE 79/2016: Reader Richard Cole has posted a link to a .kml file that shows the trajectories of the drifters that reached Australia. Here’s a screenshot of what it looks like if you drop the file into Google Earth. Interesting to note that the greater part of the debris winds up on the southern coast and Tasmania rather than the western coast.
@MPat: This is very interesting. Can you do the same plots for buoys that have passed by points further north on the arc? For instance, you reported results for 32S to 39S. What would the drift patterns look like for 32S to 25S? 25S to 18S?
How about using the width of the search area, draw squares along the last arc from Indonesia SW, and map out the buoys that pass through each box to grounding / end of life / current location in one map.
Very interesting to read and to see. Especialy to see that between 32S and 39S almost 25% did not land on African shores and islands but on WA and SA shores and even one on Java (the dot is on Java not Sumatra..).
This landing on Australien shores was also predicted by other drift studies that included the area south till 39S but what did not happen (anyway still no debris found there).
Therefore it would indeed be very interesting to see the proposal of @Victorl been carried out.
Also to see if and where landfall on Australian shores would stop to occure.
Good read Jeff. Notably though is that they are buoys. Meant to float. Not easily beached, unlike much of the debris. Also noteworthy is that they may not have been affected by cyclone Gillian.
The “adrift.org.au” model is based on the drifter buoy database, but from what I can tell it includes both drogued and undrogued buoys with equal weight, which suggests that the drift mode is biased to lower drift rates compared with what actual surface debris will experience. Has anyone examined this bias quantitatively?
More generally, some drift models, such as that used by GEOMAR, are based on models of surface currents and surface winds. Is anyone aware of any study that compares the predictions of such models with the tracks of the actual drifter buoys? The CSIRO article sort of tries it, but without finding a great match.
BTO and BFO predictions are so much simpler by comparison.
@Re Gijn
Yes, I agree with your observations. Some 22% of the drifters that made landfall did so in Western and Southern Australia. Hardly an insignificant number. Particularly compared with the verified 9M-MRO debris found in those locations.
Of course, it is not known how good a proxy a drifter is for flotsam. Your mileage may vary as indicated by the online info from several different agencies using the drifter data. Anecdotal evidence seems to suggest that undroged drifters are a fair representation.
MPat…What tool did you use to create these maps?
Victor – Yes am in the process of doing that. I wanted to rush out some stats on the search box first back in the spring, have had very little time to get at it since. Griffin did something similar, see an article in the same location I linked to in the original post.
Mike – nice idea but due to time constraints I am doing one northern box. Drifter trajectories mean that a similar drifter population can be involved for points close to one another on the arc.
GeRijn/Dennis – 7 of 39 beached in Australia, which is 18%. However if you look at the ultimate destination of all 177 buoys, only 11 went east towards Australia (6%), all but one of the rest becoming mired mid ocean or travelling west to Africa.
NYBanker – for these quick-look plots I used Excel. I later switched over to Spotfire. I have many figures using the latter which I will try to post if there is interest. There are some interesting trajectories including one which went west to Africa, bumped down the coast of Madagascar, then went around for another loop of the SIO. Some have been circumpolar prior to being captured by the SIO current system.
I also have :
– bar charts showing distributions of transit times from search area for beached drifters
– plots of trajectories of all SIO drifters in the post-crash time period
– comparison of drifter trajectories with the (short) trajectories of all 33 self locating datum marker buoys deployed during the SAR effort (not very enlightening)
I have started to package these in a document, but with very heavy full time work at present, the time/appeal of getting back behind the PC at home is limited…
Thanks, M Pat. It is good to see that your documented experimental data and my processing of IPRC’s model data (developed concurrently) broadly agree: that a high-energy impact between 32 and 39s predicts HUNDREDS of pieces of debris on Oz shores.
IF MH370 impacted on the 7th Arc between 32 and 39 south, and created some 10,000 pieces of robustly floating debris (per IG rough estimate), which then floated per these undrogued drifters, THEN we would expect between
7 / 177 = 3.95% * 10,000 = 395
…and
12/ 177 = 6.78% * 10,000 = 678
…pieces of debris beaching on Australian shores by long before now.
My own Australian shoreline debris analysis (April, based on IPRC data) used
10,000 x 5.80% = 580
…but sensitivity-tested down to
6,000 x 3.62% = 217
Even at these lower assumed values, my model still showed 32 to 39s to be emphatically counter-indicated as an impact point by the zero reported debris from those shorelines, since the probability of all ~217+ pieces remaining undetected for so long approached the infinitesimal.
Aside: Slide 49 of that study suggests to me that landfall probabilities on Oz shores was maximized at 2% windage, rather than at either extreme, due to some sort of “Goldilocks” effect. I’m guessing undrogued drifters are nearer the top end of the windage spectrum, and as such are, per IPRC data, slightly LESS likely to hit Oz shores than would a typical heterogenous collection of debris. But this is speculation on my part – and not at all required in order to reject 32 to 39s – so I won’t argue with anyone who speculates otherwise.
(pdf, 12Mb)
https://drive.google.com/file/d/0B-r3yuaF2p72Rzh1cU8wVGFZc2s/view?usp=sharing
@M Pat
Thank you. Yes the percentage is actualy lower and quite lower when all 177 buoys are taken in. Still 18% out of 39 or 6% out of 177 is quite significant I think if you compare it to the actual situation: no debris found yet in Australia (or Sumatra/Java).
What suprises me is the amount of buoys that stay ‘trapped’ in the ocean.
It could indicate that at least ~90% of MH370 floating debris is still floating around (I think).
But the probability that (quite) a number of pieces landed somewhere since but still are not found is also realistic to assume IMO.
Can you (do you want to) allready reveile which northern box latitudes you’re working on?
@WennisD 😉
Can you tell me which verified 9M-MRO debris is found in those locations (Australia)?
Or do I misunderstand you.
@M Pat
I was wondering. Is it difficult or a lot of work to split this box in two?
One part from 32S to 35.5S and one from 35.5S to 39S? And then seperate the landed buoys according those half boxes (f.i. by 2 different colors for the dots)?
@M Pat
Thanks for your excellent preliminary write-up. I look forward to seeing more when you get the time to continue.
I would be interested to look at the trajectories of those short-lived SLDMBs that were dropped. Can you pls point me towards that data?
I admire the expertise of participants above ,I am offering the original location of mh370 ending its journey in or around Madagascar and have done since day 3 , You may scoff. but dowsing gave me this result ,It has been the tool for findng missing people efficiently in the past ,I have seen one conclusion on your site that “small pieces of the plane can only have come from close by “so someone is onthe same wavelength as myself ,It has never been found because that isn’t wanted by those I n command ,it might reveal to the world what has already been disclosed online( before the rapid removal of the report) — the WHO ,WhY and When of this enigma ,Do treat the evidence of Maldivians with respect ,They saw the plane veering southward towards Maagascar ,As the 4th biggest island in the world and remote what better place for an experienced pilot to touch down ,All that is needed apparently is an underwater search in the waters surrounding Madagascar , MG
@M Pat
I forgot to argument my suggestion to you to split the your box in two. I assume it’s obvious to you but anyway maybe it’s not to others.
I think it would be interesting if by doing this allready a trend would show the number of buoys landing in Australia would decrease from 39S to 32S.
@Ge Rijn
I was being “tongue in cheek”. I do not know of any such debris.
@DennisW
Thanks for clearing. I guessed so but wanted to be sure.
@Jeff Wise @others
I am temporarily stuck at home which declares my many posts partly (for it’s so interesting).
Hope I don’t consume too much oxigen here now.
If so don’t hesitate to slow me down.
So, another comment..
In @M Pat’s second picture with the three random chosen buoy-trajectories I see allready an indication that the more southern buoys passing the box could be prone to reach Austalian shores.
The red one is passing on the edge of 39S and comes quite close to Australia while the other two pass in great distance.
Offcourse this is just 3 buoys but that’s what makes me so querious if a devide was made between the northern and the southern half of this box including all the landed buoys.
Now I’ll stop for a while I promise..
@CharlieBBucket, Thanks. I realized after I posted it that the headline is kind of inaccurate, as it wasn’t historical debris that we’re talking about but historical drifters, but I figure you guys will cut me some slack (as with my description of Brock’s analysis as “reverse drift” btw).
@Ge Rijn, Don’t self-censor, all your points have been well taken.
@MPat, thanks again for your analysis, my apologies for missing it the first time around. Let me know if you want to add or update.
@Jeff Wise
Thank you. I’m thinking you’re doing a great job.
@Jeff
Au contraire. Brock’s analysis is, indeed, reverse drift. The direction is defined by the unknown variable not the algorithm.
@DennisW
Not to get picky but the ‘misperception’ was not mine afterall.
I told you I would admit if it was so.
Forget it and move on 😉
@mG
We accept your offer of “the original location of mh370 ending
its journey in or around Madagascar”, which you found by dowsing. In which specific location is it?
We need your location, as we were very disappointed that poster
‘MG’ posting on May 26, 2016 at 4:06 AM, did not give a specific
location.
@Ge Rijn
Don’t flatter yourself. My remark was not directed at you.
@all: I attempted to use the debris – plus reams of forward-DRIFT data – to reverse-ENGINEER a data-indicated distribution of impact points.
Moving on.
@Brock – Excellent analysis. Thank you for the hard work.
From what I see it sure looks like something is wrong with the impact in the SIO idea….
@All
What Brock McEwen’s and Meteo France’s drift analysises seem to be suggesting to us is that there is a huge disconnect between where the Inmarsat data says MH370 crashed and where the physical evidence says it did..
So far, three scientific Ocean drift studies (Brock McEwen’s, Meteo France, GEOMAR) seem to suggest that MH370 crashed much further north. The Meteo France report stating that the flaperon most likely originated from “close to the Equator near Indonesia”
Although we can debate the accuracy and reliability of these drift studies there is still enough scientific data here to consider the possibility that MH370 may be 1000s of miles away from the 7th arc and may have indeed crashed close to the Equator near Indonesia.
If this is the case then we have to explain logically why this may have occurred. We know that the BTO calculations are rock solid based on the data that was given to us so if MH370 is not along that 7th arc then we know there is something wrong with the Isat data.
No way did all the brilliant scientists and mathematicians in the world screw up on the BTO math which is essentially simple trigonometery as some have stated previously.
So if we accept the premise that the BTO arcs are correct based on data given to us then we must ask ourselves ,in this scenario, was the data corrupted by some hardware issue on-board the aircraft? Was the data completely fabricated or edited as Brock McEwen has suggested, or is it valid data (actually transmitted from a plane) but the source of this data was not MH370?
In other words was a decoy aircraft used to fly into SIO and transmit those Isat signals from another plane’s SDU that was spoofing MH370’s AES ID? Although this scenario may be extremely unlikely to have occurred this possibilty may explain why debris from another aircraft is showing up in Souhern Australia and why debris from MH370 has only shown up in Africa. It also of course explains why there is an appartent disconnect between the Isat data and where MH370 may have actually crashed.
I would love to hear some thoughtful opinions on how such a disconnect can occur if this is the truth, MH370 is 1000s of miles away from the 7th arc.
Cheers,
Ken S
@Ken S,
You suggest that there is no way for all the brilliant scientists to be wrong about the BTO data.
The problem with that statement, though, is that there is an information bottleneck, within ISAT. There may be no disagreement that trigonometry and speed of light generate the 7th arc. Millions can agree that based on these BTO values, the arcs must be correct.
But millions did not witness the initial conversion of the raw, large 6-digit BTO values into the smaller 4-5-digit values that are in the public domain. Our trust in that process requires faith in ISAT’s explanations.
I don’t want to belabor the point, but I feel it’s important not to assign universal support for the part of the process that has not received universal review.
The BTOs in the public domain are based not on trigonometry but on human statements. The arcs are based on trigonometry used against values that are based on human statements, and as a result the human element must be considered a weak link, however remote.
Should we ever learn that the arcs are wrong, it won’t be because millions of scientists messed up the trig. It will be because millions of scientists got the wrong data.
@Ken S: The simplest explanation is that the plane crashed close to the 7th arc but not where was searched. As many people have explained, the current search zone was defined based on a number of assumptions, including a turn to the south by 18:40 UTC inferred from the value of BFO at that time and the assumption of level flight, and then no pilot input after the turn. I see no evidence at this time that “MH370 is 1000s of miles away from the 7th arc” as some insist.
@Ken S
There is not an intrinsic disconnect between the ISAT data and the drift data. The ISAT data cannot, by itself, be used to determine a flight path. One has to invoke additional constraints to derive a terminus. In the case of the current search area the additional constraint was the assumption of an AP controlled flight path of some sort (many different choices exist for such a path).
Additionally it is clear, as the ATSB has stated, that the debris findings are “consistent” with the current search area. What this means in ATSB speak is that drift data does not rule out the possibility that the plane terminated in the current search area. It says nothing about the probability that it did so.
At the end of day semantics have to be looked at very carefully in order to draw any conclusions relative to this event. Frankly, I have seen nothing like it in my personal experience, except in the domain of politics The filtering process needed is daunting.
@JS, @Ken S, I haven’t beat this drum in a while, so I’ll give it a thump again. The fact is that it turns out to be infinitely easier to tamper with BFO data than it is with BTO data. Indeed, I happen to think that the fact that the SDU was rebooted three minutes after the plane left primary radar coverage is a strong indication that the BFO data was indeed tampered with, since this is the only reason (apart from the debris) to think that the plane went south. Also, it turns out that MH370 was vulnerable to such a BFO hack to a very unusual degree: it was a Boeing of 777 model or later, with a Honeywell-Thales SDU, a once-per-half-hour ACARS subscription, flying under the coverage area of a satellite low on fuel, etc etc. Therefore I would argue that before we’re going to look to grand international conspiracies to alter the BTO data, we should ask ourselves how confident we are about the provenance of the BFO data.
I feel the provenance of the BFO data is suspect in giving some kind of wrong direction.
@jeffwise – You need to fix the “BFO” in the second sentence of your 1:40 PM post.
@DennisW – oh, if only we had that 8th ping……..
@Dennisw
I don’t flatter myself.
Page 4, july 5th, 6.26PM.
Forget it. Another round is evolving.
@Jeff – I’m inclined to agree, that if there was a nefarious hack or spoof, it’s on the BFO side. However, if there was a bug, I still think it would be on the BTO side.
The reason I say that, without opining on any actual likelihood, is that I believe the BTO goes through more adjustments than the BFO and requires more internal ISAT data to obtain.
The BFO is the offset from an expected frequency, but the frequency is so close to the channel frequency that a listener could determine the BFO completely independently of ISAT. I’m willing to hear counterpoints on that item because it’s not my area of expertise, but based on my understanding, the channel frequencies are known outside of ISAT, and the transmission frequency can be overheard, and the BFO is merely the delta. One could say the same about any radio signal with fixed channel frequencies.
The BTO on the other hand requires knowledge of the time slot boundary, which would be generally unknowable to the world. Further, upon receipt, ISAT subtracts a constant prior to logging that is/was also only known to them.
So there are two additional “opaque” steps involved in logging the BTO values that the world of scientists can really only take for granted. Hence, if there is a miscalculation, my position is that it falls where the least eyes have been able to look. No conspiracy is required.
Thank you Jeff for publishing this excellent work from MPat!
As several readers have already requested, it would be helpful to see a larger box, say from 20S 105E to 40S 84E or indeed several smaller boxes, as yet others have suggested.
My smaller box request would be:
Box 1 – 35S 94E to 40S 84E
Box 2 – 30S 99E to 35S 93E
Box 3 – 25S 103E to 30S 98E
Box 4 – 20S 106E to 25S 102E.
Personally, I am particularly interested in the box around 34S 94E +/- 1 degree latitude and longitude.
Hi Brock – just following your approach for a moment, I seriously doubt 10,000 persistently buoyant pieces of debris. But if we stick with that figure, then for your 395 to 678 beaching in Australia, we might expect from
31 / 177 * 10000 = 1750, to, 54 / 177 * 10000 = 3050
pieces arriving on African shores. Being bullish about the provenance of the various items of debris already reported, we have found ~10 of these, which is 0.3% – 0.6% of the total. Assuming the same discovery chance on Australian shores suggests we should have found 1 – 4 pieces of debris by now. A simpler way to think about it is that if the ratio of beachings is 4:1 in favour of Africa (31 vs 7) then we might expect to have found 10 / 4 ~= 2 pieces in Australia. I think this is too marginal to be very conclusive.
I genuinely don’t have an axe to grind here, and am as deterred as anyone else at the lack of success in the search zone. But if we are to point to alternative locations to search I think we need to do it for the right reasons, and I’m not yet sure that absence of discovered debris in Australia is one of them.
Ge Rijn – the northern box was 17S to 24S. One of the things that stalled me was that there did not appear to be many historical drifters passing through that area, so the statistics were weak. I may re-try a little further south. I’m not too interested in splitting the southern box (in this area many of the drifters move along the arc itself), but just for you – below a plot of the trajectories of all 39 beaching drifters.
http://i63.tinypic.com/2vcjcyw.jpg
Paul Smithson – I’m afraid I don’t have hard trajectory data for the 33 SLDMBs, but I bagged an image of their drift paths from slide 19 of the following presentation (add the usual www) :
raaa.com.au/convention/historic/2014/_pdf/Friday/Condon.pdf
The southern and northern groups show good trends with drifter movements, while the central group was shown by drifter behaviour to lie in a region of strong eddies.
Jeff – thanks for interest, I may ask you to put up some more figures in due course it thats ok.
What seems surprising from the drift models to me is that some debris should have washed ashore in South America by now if mh370 crashed along the SIO 7th arc …
@JS
Thanks for your opinion ,just to clarify I said that all mathematicians and scientists could not be wrong on the BTO math not the BTO data.
Yes of course having the raw data in our hands would be extremely helpful at this point but for me it’s hard to imagine engineers making such a mistake on the conversion of data if indeed this was done.
@Ken
Quite a lot of us have calculated the ping rings and there is no lack of agreement. So I don’t think we have the analytics wrong. All bets are off if the time delays were systematically altered.
There is no doubt that determined third parties could spoof the ISAT data. The question really boils down to why someone would want to do this. I cannot even fabricate a reason.
@M Pat. Many thanks for the reference on the SLDMBs.
@M Pat
Thank you for that overview of the 39 trajectories.
Is it possible for you to provide a close up picture of this box with only the 7 trajectories that reach the Australian coasts?
I tried to distinguish those in the overview but resolution is to low to differentiate between individual trajectories in the box-area.
Maybe a higher resolution picture is available too?
@Ge Rijn
A .kml file that shows the trajectories of the drifters that reached Australia is at
https://www.dropbox.com/s/2px7ypnvz0kiu3w/Just_Australia_landers.kml?dl=0
Using the time sliders allows the time evolution to be viewed.
This is derived from the Griffin data (link in the file header) that used a similar dataset to M Pat (but allowed a shorter drift time).
@Richard Cole
Thank you. Slight problem; the file does not open in my Dropbox.
Any idea what could be wrong? The .kml file type is not recoqnised maybe?
@Ge Rijn
Opens in GOOGLE EARTH.
Cheers Tom L
@Ge Rijn the links ok here, the Dropbox file opens/downloads on my IPhone browser then opens on google earth. Try copy the link into another type of browser then try and download then open.
@Ge Rijn
If you download (free) program ‘Google Earth’, you wiil
probably be able to open/view .kml & .kmz files types.
I can, and I am fairly sure your computer/video card is
not older than mine. Of course, the program tries to
connect to Googles servers ‘for authentication’, but if you
temporary disable your internet connection, the program
still works without any need to connect to Googles servers
(if you had any concerns about such a connection, that is.)
The above is what I experience for Version 7 of that program.