WSPR Tracking Validator Now Believes Testing Was Not Scientific

From “FALLACY: CHERRY PICKING DATA” by Alan Brooks, Jan 2019

In the last blog article, I explained in simple technical terms why WSPR data cannot be used to track aircraft over long distances, and certainly cannot be used to reconstruct the flight path of MH370. The article concluded:

At long distances and at low transmission powers, the received signals from hypothetical aircraft scatter are simply too weak by many orders of magnitude. What is claimed to be discernable “anomalies” in signal strength attributable to forward scatter by aircraft are within the expected deviations in signal strength for long distance skywave propagation involving refraction off the ionosphere. Although aircraft scatter could be detected if the aircraft were close to either the transmitter or receiver and if the transmitted power were sufficiently strong, the detection of the aircraft requires signal processing to separate the Doppler-shifted scattered signal from the much stronger direct signal, and this data is not available in the WSPR database.

Since publishing that article, even more evidence supporting these conclusions was presented by me and other contributors in 667 blog comments, which include analyses of experimental data of HF scatter off of aircraft, and statistical analyses of the WSPR-tracking claims. I considered writing a new blog article with the updated results, but reasoned that the informed already understood that WSPR-tracking was junk science, the uninformed wouldn’t appreciate the significance of the new results, and the WSPR proponents were too dug in to do anything but continue to double down on their flawed theory.

A question often asked is “How were aircraft successfully tracked in validation tests?” Those that have studied the tests respond that the tests were not scientifically rigorous, and the positive results simply reflect the biases of the WSPR proponents, i.e., the data were cherry-picked to support the claims that historical WSPR data could be used to track aircraft.

One of the participants in the validation tests was Mike Glynn, who was an airline captain for Qantas. Mike has commented on the blog that he now agrees that the validation tests he helped conduct were not scientific. I repeat his comment below in its entirety and without edits:

Having just read this thread it’s appropriate that I comment on a couple of things.

My involvement with RG goes back to learning that he was after an appropriate flight to test his method of detecting aircraft via WSPR. I was in possession of a candidate plan, which happened to be my final flight in Qantas, although I was not aware of that fact at the time.

The flight was a ferry of a 747 with an oil leak in the number 4 engine which could not be repaired in Johannesburg and had to be flown, empty, to Sydney.

I had experience in post-maintenance air-tests in the 747 and this was considered desirable by QF.

The flight was planned overhead Perth and Adelaide then direct to Sydney, and due to the unusual routing, I thought it may have been a suitable candidate for a test of WSPR.

The kick in the tail was that we only got as far as Perth due to the oil leak accelerating during the flight and we diverted to Perth and landed with the engine still running, with the oil quantity indication bouncing off zero, but still with sufficient oil pressure to keep the EICAS quiet.

So, I contacted RG and the test went ahead. The test was not a success. RG initially appeared to be tracking the aircraft till it crossed the African coast, although there was a cross-track error of 20NM or so. He eventually reported that the aircraft had landed in Melbourne.

This was obviously incorrect, but he had been making some wrong assumptions regarding the aircraft type, weight and tracking and so we decided on another test which was a flight plan of a QFA330 from Apia to Adelaide.

I supplied RG with the details of the flight including weight, type and time of departure. We had done a search of most flight-trackers and the flight was not on the sites we checked. Only after the analysis was complete did we find a site which had the flight recorded; however, I do not believe RG found and used this site.

An informational error on my part at the beginning of the plan meant RG turned the aircraft the shortest way towards Australia (to the Right) after take-off, however there is a procedure for departures on RWY 08 at APIA to turn left due to terrain. RG had stated that WSPR does not supply a direction of turns so I accepted the error at the start of the plan due to the incorrect turn.

After a couple of days RG informed me that the flight was tracking to Brisbane.

We were preparing to stop the test at that point but the following day he stated that the aircraft was tracking to Sydney and the following day he stated that the aircraft had flown to Adelaide from overhead Sydney and landed there.

This was correct; however, no documentation was given to me to substantiate how he had arrived at this conclusion.

Considering the process so far, I wanted to do another test and had another one, an A380 flight from Sydney to an Asian port, ready to go.

RG declined another test as he wanted to start on the MH370 analysis. I wasn’t happy with this, but it was his decision.

However, my opinion remains that the test process was not scientific.

When RG produced his MH370 analysis it made little sense to me as an airline pilot. The track to the north of Sumatra is very irregular and I found it difficult to reconcile it to anything an airliner would fly.

I had not heard of the “loiter” hypothesis either, so the holding pattern was new to me.

I asked RG whether he had considered the weather in the area in his analysis and he said he hadn’t. Despite comments made about the weather analysis on this thread, the results make sense to me as an airline pilot, particularly the diversion away from the thunderstorms off the south coast of Sumatra.

Recently, however, I have revisited the WSPR track analysis. My knowledge of the characteristics and limitations of WSPR is basic, and I simply don’t have the appropriate background to comment on that with any authority.

However, as a former RAAF pilot, I was trained in the principles of radio navigation and off-airways navigation. Andrew Banks arrived at my squadron just as I was leaving and was trained in the same techniques.

In my opinion, the methodology used in the construction of the WSPR track does not conform to any known principles of aircraft navigation that I am aware of.

It is arbitrary in the extreme and, I believe, constructed only to satisfy the constraints of the only solid data available, the BTO and BFO data.

I realise now that I should have looked at this earlier. and avoided looking as if the construction of this track makes any sense from an aviation POV.

Thats my error.

I will be explaining why I believe this in due course.

Thank you for your time and understanding.


Perhaps this is a positive step towards a more scientific discussion of the flaws in using historical WSPR data to track aircraft.

133 Responses to “WSPR Tracking Validator Now Believes Testing Was Not Scientific”

  1. Thanks, Mike and Victor. After some efforts also from my side, including a five-pager in Germany‘s most renowned magazine dealing with wireless topics (circulation 32.000+ each month), it is now up to RG, who lives in Germany and who is strongly backed by the German Ham Radio Association DARC, and friends to take up the thread. They should also publish their ideas etc. in a technical-scientific magazine of some momentum and underlying some auditing of professionally recognized editors who don‘t want to loose their reputation. I did it, so it must be easy for them to do in turn.
    73 Nils, DK8OK

  2. Victor Iannello says:

    @Nils Schiffhauer: I highly doubt any respected, peer-reviewed journal would publish WSPR-tracking claims. On the other hand, the popular press is much more gullible and eager for clicks.

  3. Mick Gilbert says:

    Well, for future reference, at least now we know the sounds it makes when the wheels fall off.

  4. Victor Iannello says:

    @Mike Glynn said: I will be explaining why I believe this in due course.

    We look forward to future comments.

  5. Mick Gilbert says:

    @Mike Glynn

    Mike, you mentioned the “holding pattern”. I know that we have discussed aspects of this in different forums but recently I have gone back to look specifically at the WSPR spots used to plot/create that flight path.

    Even at a stretch, the justifications (for want of a better term) for each plot of that purported manoeuvre are fancifully tenuous. When you apply the criteria that the authors themselves conjured up for the Roanoke DA-40 exercise, that whole “holding pattern” vanishes into the ether. It is highly unlikely that 9M-MRO was within 250 kilometres of where the authors alleged it was at that time, leave alone tooling around in the fashion they suggest.

    I’ll look to post a summary of what I found when I have the chance.

  6. George G says:

    @Mick Gilbert

    You may also wish to review my “Limited Critique”, subject of

    to put it in chronological context.

    The supposed “entry into the holding pattern” is discussed on Page 7 of my “Limited Critique”

    And this extract may also be of interest:

    “It is concluded that selections were made searching for the next most likely sample consistent with expectations.”

    This extract is under the Subject “All Major Turns” on Page 8.

    This extract refers to Victor’s “Cherries”.

  7. Victor Iannello says:

    We don’t know what the flight path was between 18:28Z and 19:41Z. A holding pattern at some point during the time interval is a possibility. However, to claim that WSPR data suggests that a holding pattern occurred is more nonsense.

  8. Mick Gilbert says:

    @George G

    G’day George, thank you for the link to your paper. I somehow missed it first time around.

    And yes, I’m seeing essentially what you found regarding the total absence of anything even vaguely supportive of that purported holding manoeuvre even when using the authors’ own criteria.

    Frankly, that the authors would promote such a controversial notion (viz a holding manoeuvre, purportedly supporting the fantasy of some form of inflight negotiation) knowing that its foundations failed to pass even their own evidentiary standards has got to sit at the boundary of appallingly poor scholarship and deliberate intellectual fraud. Readers can form their own views on that.

  9. Tanmay says:

    Hello All, any update if ocean infinity or any other team is going to resume the search for MH370?

  10. Victor Iannello says:

    @Tanmay: Ocean Infinity has suggested that it might search for MH370 in early 2023 or 2024.

  11. ST says:

    @Victor – Thanks for this new blog post. Reading recent articles on the drought conditions and what it is bringing to surface, in some years it sounds like climate change might bring to light some pieces of the puzzle even to the MH370 mystery.

  12. Victor Iannello says:

    @ST: At this time, unless a whistleblower surfaces with relevant information, the missing pieces of the puzzle are deep in the ocean and unlikely to be disturbed by whatever changes may be occurring at the surface.

  13. Victor Iannello says:

    As some of you already know, a Citation 551 flying a route between Spain and Germany overflew Germany and crashed in the Baltic Sea near Latvia. After maneuvering over Germany at FL360, the plane flew straight and level towards the Baltic until fuel exhaustion. The plane then progressed into a spiraling descent. The distance between the deviation from straight flight to the impact site was around 16 NM.

    It’s interesting that after fuel exhaustion, the plane first gently banked to the right, but then leveled and eventually rolled to the left, ending in the steep banked descent.

    It certainly looks like pilot incapacitation due to depressurization/hypoxia.

    Juan Browne (@blancolirio) summarizes what we know:

  14. Paul Smithson says:

    Course of aircraft changes from ~044 to 054 over the final 90 mins of flight. Of that, about +5 can be explained by change in magnetic declination between Cologne and Gotland. The remainder presumed due to the wind (changing from W ~20kt to NW ~80kt).

  15. Victor Iannello says:

    @Paul Smithson: Which would of course suggest that the autopilot was set to hold a constant magnetic heading.

  16. Paul Smithson says:

    Interesting to see behaviour of the aircraft between roughly 18:29:30 and 18:34:15.
    – heading is maintained throughout this period
    – speed declines from 365kt to 289kt
    – rate of descent approx -240 (metres?), so about -700fpm
    – from 18:34:45 directional stability is lost (initially goes left, then into spiral right) and rate of descent increases.

    Presume A/P still engaged and one engine inop for those ~4.5 mins. But why is the rate of descent so gradual? If memory serves, this is quite a bit more shallow than the Boeing simulations for 9M-MRO end of flight simulations with one engine inop.

  17. Paul Smithson says:

    @Victor – yes, that’s my point. Its consistent with default to HH norm at route discontinuity.

  18. Victor Iannello says:

    @Paul Smithson: To fully understand what occurred at fuel exhaustion, we need to understand some basics of how automated flight is implemented. For instance:

    1. Is there an autothrottle?
    2. Is there any automatic compensation for thrust asymmetry?
    3. Will the autopilot continue to operate after the second engine flames out? For how long?
    4. Is there any automatic envelope protection, i.e., stall, overspeed, bank?

    Questions like this were answered for MH370 long ago, but without some digging, we (I) really don’t know how automation was implemented in this particular Citation 550.

  19. Paul Smithson says:

    @victor, all true and I obviously don’t know either. Just intrigued to see a profile that looks more like drift down than “normal” ROD.

  20. Viking says:

    @Victor and Paul,

    I did not follow the recent crash in any detail, but I can inform you that according to Danish news two F16 were scrambled from Bornholm to follow the plane. This means that our airforce probably knows a lot more than us about the details.

    According to the news reports the pilot was the German businessman Peter Griesemann.

  21. Don Thompson says:

    I posted a short thread to Twitter last night, further notes below:

    Citation IISP, s/n 0551-0021, reg OE-FGR, hex 440A99

    En-route LEJR – EDDK. Jerez de la Frontera, Spain to Köln-Bonn.

    Aircraft navigated a presumed route to DEPOK 1C or 1V ‘STAR’ for EDDK, turning at waypoint DEPOK for Köln/Bonn VOR/DME approach but absent any prior initial descent. Suggests that an FMS was guiding the autopilot in lateral navigation. FMS route discontinuity occurs over the VOR/DME after which time (15:52Z) no further navigation inputs, the heading at route discontinuity is held. I haven’t matched waypoints to the route prior to waypoint DEPOK.

    The EDDK STARS chart.

    Unfortunately the ADS-B equipage on OE-FGR didn’t include Enhanced Surveillance: no EHS, no autopilot mode transmitted. Squawk code remained unchanged from departure (not unusual across Eurocontrol airspace), no 7xxx series code dialled in.

    Max range for aircraft type is circa 3,700km. At approx 3,000km from departure airport, LEJR, OE-FGR began its final descent. No recovery from the descent until impact with Baltic Sea off Ventspils, Latvia.

    During the final descent, at approx 29,000ft, the aircraft turned to starboard. I suspect at this point the A/P had disengaged. Prevailing winds were from the NW. Other aircraft in the vicinity recorded wind at FL350 from 345º at 38kts and FL360 wind from 308º at 45kts. Also see Earth Nullschool for wind speeds through various altitudes.

    Media sources stated that air policing intercepts were made by French Air Force, German Air Force, Danish Air Force (@Viking, all that needs be known is that an aircraft is heading into an ADIZ/overflying territory without an appropriate filed flight plan and/or is not communicating with ATC). Swedish Coast Guard helicopters were heading for the crash scene before the aircraft had impacted the sea. Had the flight plan been filed with supplemental information for fuel endurance, that would have been available throughout the Eurocontrol network.

    Unfortunately the avionics spec of this type is subject to a range of owner/operator defined upgrades. A number of STCs exist but FMS is not untypical. After import from the US in 2020 and to operate in European airspace the aircraft will require ADS-B, VDLm2, RVSM, …

    The electric system is relatively simple: a L and R generator with auto bus tie and fall back to a battery. CitationPro over at YouTube has some useful clips on Citation IISP procedures.

    Note that the Payne Stewart hypoxia accident terminated with the Learjet stalling at its operating ceiling. The aircraft had remained in climb mode and ‘maxed out’, unable to climb further. As the first engine flamed out, the aircraft stalled, the remaining engine powered the descent until impact.

  22. sk999 says:

    Incomplete guess at events. I used the flightaware track log. Times are EDT.

    1:29:45 1st engine flameout. Speed starts dropping (355 knots). Altitude stays constant. Course is 56.

    1:30:45 Speed down to 307. Begin descent. ROC (-550). Continues dropping to around -900. Speed drops to 294 but remains roughly stable.

    1:34:45 2nd engine flameout. ROC jumps to -2100. Speed continues to drop to around 251 but then stabilizes again.

    1:37:04 Left AC bus powered up; SATCOM logs on; 7th arc. ROC -2900. Speed still stable at 259. Course still stable at 58.

    1:37:26 Begin significant course deviation. New course 69. Altitude 25000.

    1:44:45 Last transmission. Altitude 3175. ROC -4100.

    Distance from 7th arc logon to last transmission: 15 nm.

  23. Don Thompson says:

    @Joseph Coleman

    Thank you for the pointer.

    How curious. The ‘pin’, or rather a box, has now moved some 250 nautical miles north east of the previously imagined GDTAAA location.

    While there seems to be the usual hurrah from Team WA cheerleaders, this new GDTAAA derived location is twice the distance of the last from a favoured particle drift seed point. Putting further distance between two points would typically be something disheartening?

    First thoughts, I’m sure further review will be forthcoming.

  24. Victor Iannello says:

    New junk science is presented, which contradicts the previous junk science, but is still ludicrous. Anybody that agrees with this garbage is not subject competent.

    Just look at Figure 3, which says at 17:56 a signal transmitted from a WSPR station in Madison, Alabama, and received by a station in Urbana, Illinois, was measurably perturbed by MH370 flying near Malaysia. This is laughable.

    Flipping through the report, I see page after page of false claims and incorrect understandings.

    Yet, Geoffrey Thomas again promotes this nonsense with two new articles.

    These people have no shame.

  25. Joseph Coleman says:

    Seeing that flight path image in the report, even as a layman like myself, thinks WTF. It Reminds me of this Scene from family Guy With the One Guy taking notes, while ones dishing out their thoughts. The last five words spoken, sound fitting.

  26. Victor Iannello says:

    Geoffrey Thomas (@airline_ratings) blocked me on Twitter. Sounds about right. As if blocking me will somehow prevent exposing the truth.

  27. Mick Gilbert says:

    So we now have a new location 400 km north of the first one, 300 km north of Professor Pattiaratchi’s 32.5°S 96.5°E and the cheer squad want us to forget everything they’ve said over the past five years regarding Broken Ridge and hark back to the very earliest and least precise claim of ‘between 33°S and 28°S along the 7th arc’. It’s laughable. Apparently, a bit like Groucho Marx’s character, these fellows have principles but if they don’t work out, well, they have others.

    As Michael Glynn has posited, GDTAAA is so flaky that it can put the target anywhere the authors want. As I opined very early in the piece, it’s a Ouija Board for one (now two, it would appear).

    Let’s just hope that Ocean Infinity didn’t utter the terms WSPR or GDTAAA when they were negotiating with the Malaysians for a new search agreement. For those who’ve been waving their fists about the Malaysians wanting more information from OI, THIS is why. If any one thing has done more damage to the likelihood of a new search it is this GDTAAA nonsense.

  28. sk999 says:

    According to Don Thompson, “the ‘pin’ … has now moved some 250 nautical miles north east …”

    If that rate were to continue, in approximately 2 years the WSPR ‘pin’ will coincide with Mike Chillit’s ‘pin’ (based on the “perpendicular bisector method”), which is on the west side of the Zenith Plateau at latitude 22.18S. It would be an amazing convergence of … whatever.

    Mark Twain anticipated this process some time ago. “There is something fascinating about science. One gets such wholesale returns of conjecture out of such a trifling investment of fact.”

  29. Victor Iannello says:

    @Mick Gilbert: Other than in their own echo chamber, this new proposal is not being taken seriously. Even if we put the layers of technical problems aside, this new site is not consistent with the precision claimed for the previous site. That’s quite easy to see.

  30. Paul Smithson says:

    Yes, isn’t it amazing that the first estimate invested with uncanny precision had been discarded so casually?

  31. Mick Gilbert says:

    @Victor Iannello
    @Paul Smithson

    For anyone interested here’s a side-by-side of the two GDTAAA “flight paths”, most recent on the left, earlier one on the right

    Amongst things to note are the fact that the “holding pattern” has shifted some 550 kilometres to the south of where it was originally, and as Paul noted elsewhere, the pilot apparently knew where the arcs where and managed uncanny precision in flying along the 4th and 5th arcs. WT actual F?!

    Bear in mind, the first one was described by the authors as follows,

    This flight path of MH370 has been described in detail with a granularity of two minute steps and accuracy within 18 nm.

    They also stated,

    WSPR technology provides credible new evidence to help determine the crash location of MH370 and we therefore recommend a further search be authorised. The offer by Ocean Infinity to recommence an underwater search in 2023 should be accepted.

    They then went on to publish a “Confidence Analysis”, stating,

    It is thus concluded that the positions presented in the previously published papers are sound and that any associated risk is at a reasonable level.
    The crash location remains as originally defined at 33.177°S 95.300°E and is shown to be a sound conclusion.

    Of course, each one of those proclamations was greeted with rapturous applause from the fawning cheer squad.

    And now everything that went before is just jettisoned without so much as a by your leave. It’s now difficult to determine which that crew possesses less of; intelligence or character. The needle seems to be bumping EMPTY on both gauges.

  32. Victor Iannello says:

    @All: The contributors to this blog are probably the best able to recommend a new search area. If you were asked, where would you recommend? What would be the search strategy?

  33. 370Location says:

    @VictorI & @All: Thanks for inviting search strategies.

    I expect most are familiar with my acoustic analysis. Detections from at least 12 hydrophones in the IO and 40 regional seismometers point to a loud noise directly on the 7th Arc near Java. The noise was described in both Curtin reports, with correct bearing and timing. The event occurred at 01:15:18z, about 55 minutes after the last ping. I currently believe that a large section of sinking MH370 impacting the seabed best matches the unique acoustic signatures.

    The location uncertainty is as good as the seismic epicenter, about 2-4 km radius. We have very good timing of signal arrivals within fractions of a second, but the speed of sound in the local crust varies.

    That error radius is quite narrow compared to other candidate sites being considered by Ocean Infinity, and their capable AUVs could cover the broader site in a day. I have made the case publicly that searching the acoustic site first, perhaps as a test for the new equipment, would be a good hedge against a far larger commitment to months scanning an expansion of the previous search area. Even if the odds are considered low by some, should MH370 debris be found at the acoustic site, it could save many millions in expenses that would not be reimbursed.

    While Ocean Infinity has the best prospects for finding the plane in an extended search, there may still be room for another team to make an attempt at the acoustic site.

    A new acoustic strategy is to calibrate the seismic timing. Large multi-ton weights could be sunk at the candidate site, which should be detectable on the local seismometers. Knowing the exact epicenter of the reference signal would give a corrected site accuracy within about 300 meters. That error is smaller than the expected size of the debris field.

    There are several international teams with capability of diving to 3,400m to look for either visible debris or with sonar. The site is in tropical waters so there is no need to plan around Austral winter.

    Hopefully, an open invitation will give some competitive incentive to search the site first.

    The Java anomaly site is in Indonesian EEZ waters. My understanding is that Indonesia requires a permit for an underwater survey there. Indonesia may also want to get directly involved, because if MH370 is found there, they may take on treaty responsibility for recovery and also the ensuing investigation.

    Details on the acoustic detection, and how it is a good match for all the factual evidence, can be found on my website:

    Huge thanks to all for this unique forum!
    — Ed Anderson

  34. TBill says:

    In response to your post, I have been studying new flight path concepts. It was actually the WSPR controversy that prompted me to try to understand how an Active Pilot flight path may have started and ended.

    Here is what I now believe happened:
    (1) Pilot took a curved path south from ISBIX to Arc5 (eg; 180CTH or 181CMH)
    (2) After Arc5, pilot descended and slowed markedly. Rt engine was probably shut to save fuel.
    (3) Aircraft hit Arc7 in the vicinity of 32s, with fuel remaining
    (4) Arc7 I believe was deliberate action to turn off SATCOM (radio silence) followed by descent to fly away hidden under the thick cloud layer

    The distance flown/glided under the clouds is an estimated 150-300 nm depending on if efforts were taken to save fuel along the route (bleed air off, etc). I envision FL400/MRC from Arc1 to Arc5.

    Parenthetically, the N571 offset after Arc1 may have been Left not Right, or alternately it may have been a jog left over to flight path B466.

    In this scenario the crash site location cannot be determined by flight data. It is necessary to use debris drift, acoustic or other evidence, and human behavior analysis to try to figure out the pilot’s end game.

    I would personally strongly suspect that the Active Pilot’s intent was to crash in a hard-to-search area, by turning East and flying far from Arc7 along the Broken Ridge fracture zone.

  35. Joseph Coleman says:


    Only if ya wanna listen, “Ma Name is Jeff” been chatting about ya. [VI: Link shortened to remove query data.]

  36. Don Thompson says:

    @Joseph Coleman

    That ‘show‘ was gold. I’m not at all sure what genre I’d pigeonhole it with.

    If I may share a tip: the part of the URL that begins ?fbclid=. Best to omit that, and all that follows, when copying or pasting. Facebook doesn’t deserve to track your every web interaction!

  37. Victor Iannello says:

    @Joseph Coleman: I had already listened to the podcast before you submitted your comment. I actually found the interview to be entertaining, even if I don’t believe MH370 was diverted to Kazakhstan, and even if I don’t believe that Blaine Gibson is a Russian agent who planted debris.

  38. Victor Iannello says:

    @TBill: What would be your proposed search strategy?

  39. Victor Iannello says:

    @370Location: Given that the acoustic event you’ve identified is so well localized, I think you make a solid scientific case for expending a small amount of time looking there. I have said this previously.

  40. TBill says:

    Re: Search Strategy
    Of course I would have to ask search experts for advice, how to grid off the area, but I would think we are looking for debris pieces as a hint. Conceptually scanning some easy areas along the north edge of BR, maybe the level trough at the bottom. Similar to your work for 34s, someone needs to look at the direction that the currents would take debris and look downstream for debris hints.

    Before that of course I could use a collaborator or DSTG Bayesian approach to model where MH370 may have hit Arc7.

  41. Victor Iannello says:

    @All: Nobody else has a proposal for how to conduct the search? For instance, latitude range along the arc, distance from the arc, whether to rescan any previous areas?

  42. Joseph Coleman says:

    Perhaps finish off all blue shaded areas of Bayesian from original search first area including any nooks and crannies to be sure, as a lot of work went into it, be nice to finish it. Then similar Bayesian method from 34S from arc then Bayesian from 30S and so on North. Not sure how that would work out as boxes for scanning. Perhaps scan as boxes, first then scan lower probability area outside boxes after for each stage. Only suggesting this as there are a few pins along the 7th Arc from different studies over the years. And looking in as a layman its’s Gotta to be around one of them.

  43. Don Thompson says:

    I had begun to write quite a long comment in reply to @Victor Iannello’s question

    However, I’ll try to be brief.

    If one considers today’s subsea survey, search and image acquisition capabilities with those prevailing in 2014, I don’t believe one would conduct a seafloor search in the way that it was carried out in 2014.

    That is not to say anything was executed in the wrong way in 2014 thru 2017, rather that the ‘white heat of technology’ has enabled significant progress between then and now. Therefore, I believe the task would be conducted differently today. The ‘state of the art’ in 2014-2017 dictated that multiple vessels, each dragging a ‘towfish’ was the only practicable answer. While, an element of ‘the future’ solution, a Hugin AUV, was exploited in the 2014-2017 campaign it was certainly deployable only in limited seasonal ‘windows’ and not as often as might have been ideal.

    Operating a ‘towfish’ at the end of a 10km cable is a challenging task. The ATSB’s report, ‘The Operational Search for MH370’, doesn’t hide the issues encountered. These included:
    1. tow-data umbilical cable reliability – umbilical performance degraded over time during each swing;
    2. benchmarking the towfish image acquisition performance – initially deployed 3D targets were used, later, the scattered debris of a lost ship was adopted, not to ignore that two quite different image acquisition technologies were employed – real aperture side scan sonar and synthetic aperture side scan sonar;
    I’ll expand on those items with:
    2a. benchmarking image acquisition performance – capabilities weren’t evaluated across the prevailing spectrum of seafloor conditions – the benchmark targets were each located on seafloor areas favourable to discrimination;
    3. towfish stability. When winching over seafloor features so as to maintain optimal altitude the towfish platforms were not stable, not conducive to good image/target acquisition. For example, SAS towfish did not account for pitch or roll, yaw was unstable.

    For these reasons, I believe that planning for any renewed search should look afresh at an approach, first, for Zone 2 of the 2014-2017 (S32.77º, E95.47º to S35.00º, E92.82º), then Zone 1 (S35.00º, E92.82º to S39.43º E85.46º). Zone 1 is arguably a lower priority, however, a location does exist in that Zone where image acquisition is effectively absent. An approach, developed for today’s capabilities, exploiting new technology, new practices, experience of further wide-area deep ocean operations, would likely be different to that adopted in 2014-2017. Compare, contrast, expose differences – the inevitable but unavoidable compromises of 2014-2017: go explore the differences.

    I also have a minor observation for the sequence of events immediately prior to 00:19:29UTC that I believe reinforces previous rigorous analyses of the descent and suggests that the ocean impact occurred close to, not further from, the line of 7th arc.

    That is, the debris field has been missed in previous efforts.

    Finally, to be clear: prior to 2014 there had never been an attempt to survey, map, and search 120,000km² of ocean floor at the depths prevailing in the area, across the spectrum of bathymetry prevailing in the area, delivering the consistent resolution necessary to discriminate a debris field. I don’t suggest that anyone involved 2014-2017 failed to rise to the challenge, the challenge was too demanding and the tools not adequate across the spectrum of demands.

    Not sure I achieved brief. But it’s my view at this time, a summary, not all the detail.

  44. sk999 says:

    Joseph Coleman,

    Thanks for the link to the podcast interview with JW. Nice to know that he is still out there. Jeff was dismayed that no one would take his ideas seriously. That is not true. I did so — or a least tried to. My conclusions are written up in a paper titled “Routes North and other Idiocies”, which can be found here:

  45. Mick Gilbert says:

    @Victor Iannello

    Regarding a renewed search, I note that the “X marks the spot” approach has been entirely unsuccessful to date. I’d also posit that the commoditisation of underwater search by the likes of OI has changed the paradigm somewhat from the initial tow-fish days.

    If you were to first start with the largest area needed to essentially guarantee success and then pare away the impossible, improbable, unlikely and less likely areas the process might go something like this:

    The wreckage is somewhere on Earth (509.6 million km²).

    If you exclude fringe theories such as those involving planting evidence, hacking SATCOMs and shoot-down/cover ups you can eliminate as impossible or improbable essentially everywhere other than the Eastern Southern Indian Ocean (35 million km²).

    If you are satisfied with the legitimacy and accuracy of the Inmarsat data and its subsequent general interpretation then we have the seventh arc to work from. Let’s say that the extent of the 7th arc in the Eastern Southern Indian is 7,000 km.

    We can almost certainly trim the southern most 1,000 km off that based on the aircraft’s known fuel load and known performance.

    The question that arises then is how far from the 7th arc could the aircraft have travelled? The most fundamental variable here is whether the 7th arc is associated with fuel exhaustion. The likelihood of that association being true increases with latitude (eg at 39°S, the 7th arc essentially marks the fuel range whereas at say 8°S there would still be some fuel on board). This gives rise to an inverted triangular powered flight zone with its long axis centred on the arc and its apex around 39°S. I haven’t plotted it but it wouldn’t be too hard to do – it would likely run to about 4 million km².

    The next question is whether there was a controlled glide subsequent to fuel exhaustion? If so, that would extend the triangular flight zone by say 225 km, bringing us up to something like 5.6 million km².

    There has to be a probability approaching 1 that the wreckage is in that triangular area of some 5.6 million km².

    Oceanography can be used to pare away less likely parts of that. Based on the drift studies the probability that the impact was in the band 25°S – 35°S would likely be around 0.85.

    The trapezoidal, high probability zone created would likely be about 2 million km².

    That all gives rise to an area that could not presently be searched in one seasonal campaign. If we’re serious about finding the wreckage that means a multi-year program.

    Running through the centre of that zone is the swath that has been previously searched. I would be disinclined to commit any effort to re-searching the previously searched area.

    In terms of creating priorities to determine the annual programs, there’s the UGIB point and Broken Ridge towards the southern end of the zone so that might be a starting point. I’d be inclined to prioritise latitude coverage over distance from the arc.

  46. ventus45 says:

    Ed @370Location.

    Your idea of dropping weights to calibrate the seismometer data is a very good one. To that end, there should be multiple drops of hydrodynamically identical drop bodies at or near your site, i.e. they should all be the same size and shape, (either like an aerodynamic bomb as dropped by aircraft, or perhaps as a hydrodynamic cylinder like a torpedo), complete with fins, to ensure that they will not topple, or otherwise depart from a vertical trajectory (ie, they must go straight down). All should weighted differently, say in half ton increments (500kg, 1,000kg, 1,500kg, 2,000kg, 2,500kg all the way up to 10,000kg), and suitably instrumented, such that each one transmits it’s descent rate continuously, so that it’s speed, and thus it’s kinetic energy at impact with the sea floor can be determined. The drop bodies should thus be fitted with a trailing wire system (similar to that used by wire guided torpedoes) to transmit the data continuously back to the drop ship. Depending on the instrument package used, a lot of other useful data could also be obtained, eg, temperature, salinity, local sound velocity, etc.

  47. TBill says:

    Comment Re: Inmarsat Data
    In my opinion, we are not exploiting the apparent accuracy of the Inmarsat data. It was really quite some years ago that I noticed almost perfect fit to Arc5. When we make the assumption that Arc6 must also fit as passive flight, then we are possibly discounting the flight path story that the Inmarsat Arcs are trying to tell us.

    In other forums, I have been trying to say, we really do not need WSPR or spy satellite or contrails. We have a special Inmarsat satellite assigned to 9M-MRO getting the data for us. It would probably be quite hard to do better than that.

  48. 370Location says:

    @VictorI – I do recall you said that if you ever were to make a search site recommendation, that you would suggest mine as a secondary. Yours may be the highest acknowledgment of my work so far. Much appreciated.

    @ventus45 – I don’t want to complicate the possible ways of calibrating the seismic signals. I suggested multiple drops of multi-ton weights because at least two could be correlated together to extract timing well below the ambient noise.

    Using a barrage of different weights would only help if there were unknown nonlinearities in the seabed response, which I don’t think is a factor.

    As for streamlining the weights, based on the AF447 debris field I estimate that any massive object sunk from the surface will land within 100m at the 3400m seabed. Currents are not a factor in the few minutes it would take from surface to seabed.

    The only advantage I see to instrumenting a sunk calibration mass would be to get the exact seabed impact time. Temp/salinity etc have been useful for refining hydrophone arrivals, but that is coarse and this is a tighter crustal velocity calibration.

    The dropped multi-ton mass could be anything from a chunk of cement riprap to a towed mothball vessel. The heavier it is, the better the the signal to noise ratio on seismometers.

    One factor to consider is that the closest public seismometer GE.CISI (Cisomped, Java) is no longer operational. Indonesia has a closer IM.CMJI Cimerak seismometer in their private network that may have the best 2014 data and could be used for calibration. We might also recruit someone to install a Raspberry Shake amateur seismometer near the GE.CISI site to help with the calibration, since that was the best data source at the time.

    BTW, a helicopter went down in Malaysia near Ipoh yesterday. Search teams did not find it before nightfall, so I attempted to localize it with the only available seismometer. See my tweet thread at:

    The chopper was just found this morning, and it appears that the lone pilot was alive at recovery but did not survive transport to the hospital. Condolences go to his family and friends.

    My first estimates of seismic P-Wave to S-Wave delay timings appear to be off, and the multiplier 8.0 I used to calculate a search arc did not match the local crustal speeds for central Malaysia. The strong seismic detection itself may have been a local event near the seismometer that had nothing to do with the crash. Still, I believe that further investigation and refinement may reveal that seismometers can be useful for locating remote aircraft crash sites.

  49. Paul Smithson says:

    @victor, I believe you already know my search recommendation.

    BTO-based route-optimisation from Arc 1 indicates peak probability zone at arc 7 in the 39.5-40.0 zone. A waypoint defined route is obtainable very close to the optimum, namely M0.84 FL360 flight IGEBO-40S85E or IGEBO-RUNUT-40S85E. For what it’s worth, I’d bet on MEKAR-SANOB-IGEBO-40S85E but its not really material to the conclusion.

    At cruise level, this track intersects the 7th arc at 39.6S. In practice, marginally north of that if significant altitude had already been lost by 00:19:30. I’d expect the wreckage to be very close to the track angle/arc intersection and not more than 10NM from the arc defined at 25000ft.

    This area has not been searched.

  50. Don Thompson says:

    Just to add.

    I don’t expect the approach, to which I have alluded, to require a significant proportion of the original 120,000km² to be revisited.


    Is OE-FGR’s loss a useful characterisation candidate to you? Certainly a smaller, lighter, aircraft but much shallower sea (<250m).

  51. 370Location says:

    @Don Thompson:

    The last OE-FGR ADSB entry is:
    2022-09-04 17:42:22 UTC 57.6648N 21.0940E

    The two closest seismometers are Gotland UP.GOTU and GE.SLIT on the Latvian coast. Neither share data on the public seismic network. The best chance of detecting the Cessna jet crash with available data might be stacking/correlating the six amateur sensors around Stockholm.

    I don’t see the value in retrieving that data and going down that rabbit hole. Even if I were to show that the Cessna crash could be detected by seismometers as a surface impact or on the seabed as it sank, it wouldn’t change any minds about MH370 acoustics.

  52. Victor Iannello says:

    @Paul Smithson: What do you believe is the southern limit of what should be searched?

  53. Victor Iannello says:

    @TBill said: In my opinion, we are not exploiting the apparent accuracy of the Inmarsat data.

    Why do you believe the MH370 BTO data is more accurate than for the many other flights from which the uncertainty was derived?

  54. Paul Smithson says:

    @victor. 40.5S would be a generous southern max. So basically 1 degree of arc, from 39.5S to 40.5S. But my expectation is you’d be successful between 39.5 and 39.8.

  55. TBill says:

    Let me come back to you on that, but briefly one measurement that I look at is the actual (“with a ruler”) mathematical distance (ground speed) between the Arcs, eg; for 180 south. I believe the LNAV approach may noticeably sacrifice Arc2-Arc3 accuracy (it assumes those Arcs are too close) in order to bring Arc5-Arc6 into compliance. What I actually use is DrB’s CBTO spreadsheet, and also Mike Exner’s Arcs on Google Earth.

    Re: Cloud Layer below 22 South
    If the flying-under-cloud-layer scenario has merit, it might be helpful to get information from experts to better understand the cloud top heights and bottom heights. The reason I mention it, I assume most are aware that NASA was recently asked to comment on poss MH370 contrails in the area, and NASA gave some interesting data on the clouds. I am aware of 3 or 4 weather satellite images in the area (Tim Vasquez, Terra, Aqua, Himawara) but lack the expertise to work with the data very well.

    Also reflecting on Mick’s large search area, to the extent that some of the miles flown may have been low altitude, that helps reduce the potential miles flown.

    Boeing states (in SIR) 2,806 potential miles flown from 1825 at FL400, to that add +125-nm potential glide, so we are now up to 2,931 miles potential, now add 0-to-5% more fuel if fuel saving techniques were employed: that is what we are up against in my view.

  56. Victor Iannello says:

    @370Location said: I don’t see the value in retrieving that data and going down that rabbit hole. Even if I were to show that the Cessna crash could be detected by seismometers as a surface impact or on the seabed as it sank, it wouldn’t change any minds about MH370 acoustics.

    I think it would be helpful to better understand the application limits of your acoustic analysis. If either the impact of a Citation jet on the surface or seabed was detectable, that would be an important result.

  57. Victor Iannello says:

    @Don Thompson: One of my ongoing disappointments is that to this day, we still don’t know what areas within previous search limits were not searched or have low quality data due to challenging terrain or equipment issues. Those would be high on my list of areas to search.

  58. 370Location says:

    @VictorI, @Don,

    Please excuse my bad manners. I usually have better sense than to post when tired and cranky.

    I’ll look into what data might be available for the Citation, but those are portable single vertical axis hobby/school seismometers around busy Stockholm, 200km away.

    Better prospects for analysis might be for a Turbo Otter N725TH that went down off Whidbey Island near Seattle on Sep 4. There’s ADS-B available, and it looks like loss of power. A wreck has been detected by sonar, and NTSB is looking for an ROV to confirm it’s the Otter. (If it’s not, then seismic location might also be helpful to the search).

    There will be much better seismometer coverage in that area, with 3-axis data available plus nearby RasPiShake sensors within 5km for comparison.

  59. Mike Glynn says:

    For anyone interested I have done a formal analysis of the major navigational error that WSPR is susceptible to. Victor informs me that it has had some discussion previously. There are other navigation-based errors which I will address in due course. Thanks to Andy Banks for checking the principles and confirming them.

  60. Victor Iannello says:

    @Mike Glynn: Mick Gilbert has probably provided the most comments on this blog relating to inaccuracies of the Maidenhead grid locator. Here’s a link to the first of his comments I could find on the topic:

    In a comment following Mick’s I made this observation:

    @Mick Gilbert: Active hams query the database to determine the precise location of stations. It doesn’t appear there was any attempt to do this. Not that it really matters, because you would just more precisely get a wrong answer.

  61. Viking says:


    Since the crash in the Baltic sea happend nearby, I may be able to get access to professional seismic data. Do you want me to try?

  62. Victor Iannello says:

    The many issues related to WSPR tracking tend to fall into one of three broad areas:

    1) Impossibility of detecting scattered signals (physics of radar scatter, propagation loss, noise levels, Doppler discrimination, background clutter)

    2) Imprecision of localizing target (propagation paths, duration of transmissions, Maidenhead locator resolution)

    3) False statistical inference (no correlation of signal variation to target position, cherry picking of data)

    Any one of these issues is sufficient to invalidate WSPR tracking. Combined, the WSPR tracking claims are devastated, as there are layers upon layers of errors.

  63. 370Location says:

    Thanks for the offer. I can access most of the data available on the public seismic networks, but nothing close to the Citation crash site:

    What I found on closer inspection of the portable RasPiShake geophones (AM network in the above map) is that their locations are intentionally obscured by up to 1 km for privacy. It’s done to avoid EU GDPR fines and also theft.

    Much like the WSPR maidenhead grid uncertainty, it reduces the value of the data, or requires acquiring more precise info. Beamforming seismic signals from a few km around Stockholm would be difficult with 1 km uncertainty. There may also be time reporting inaccuracies of up to a second.

    I’ve obtained 1.3GB data from 170 stations within 1 degree (111km) from the Whitbey Island Otter crash +/- 2 hrs. More crunching is required to test it.

    A fourth issue category with WSPR is the false premise that signals are equally likely to take the longest great circle path around the globe rather than a shortest optimal path. Beyond the 20,000 km antipodal distance, there is no shortest path. So, the optimal path could be multipath combinations from all directions, spirals around the globe, or following the solar terminator. Granted, this might be lumped into your areas 1 and 2, but it’s a huge disconnect from reality.

    The abandonment of all the viral WSPR speculation and accuracy is telling. Tossing out unreliable data points shifted the endpoint 480 km after the author was claiming 10 km accuracy. And because the previous data released revealed the many flaws, this new endpoint allows no peer review.

  64. Victor Iannello says:

    @370Location: There may be some WSPR tracking supporters that believe whether or not it is valid, it should be advocated because it may be the new and credible evidence that Malaysia needs to resume the search. I am not in that camp. I believe it is important to find MH370’s debris field on the seabed, but using false science to justify a restarted search is dishonest and also will backfire as a strategy, as the outlandish claims will eventually be refuted, and credibility will be lost in the scientific community and ultimately with Malaysia.

  65. Viking says:


    I know a professional geologist, so I think I can get more data than you. I have just written an email to him. However, we are both busy, so you probably must wait 5-10 days, and I may have to mail it directly to you to avoid GDPR problems. Hope that is ok.

  66. Mick Gilbert says:

    @Victor Iannello

    It is fairly clear that there is an evolving attempt from some quarters to re-write history to suit their own narratives.

    As we all know, Professor Pattiaratchi’s early work apparently showed that a crash location between 28°S and 33°S could have gotten the flaperon to Réunion in the known timeframe. And, at that time 28°S – 33°S was entirely consistent with a concurrent European study which placed the crash site between 28°S and 35°S. The European study was published as a proper research paper so we understand the basics of their workings and the basis of their findings. This was all pre-2017.

    By early 2017, with more than 20 debris finds to work from, Professor Pattiaratchi had apparently refined his work to focus on 32.5°S 96.5°E as the origin point (we’ve seen no scientific paper detailing this work, just articles promoting the conclusion). And 32.5°S 96.5°E has been his “X marks the spot” ever since.

    When the Guessing, Doodling and Tracing brigade determined that 33.177°S 95.3°E was THE spot (even though they managed to mis-map it at the time), the cheer squad was happy to point out the proximity to Professor Pattiaratchi’s 32.5°S 96.5°E. Great minds agree, this must be the place, release the pigeons, start the search.

    Now we find that because the Guessing and Doodling brigade realised that they forgot to carry the 2 or some such nonsense (this despite them publishing technical and confidence (ahem) ‘papers’ confirming the accuracy of their earlier work), their X has moved some 300 kilometres north of Professor Pattiaratchi’s spot. So now, we’re all just meant to forget 32.5°S 96.5°E and revert to the now 8 year old, unrefined 28°S – 33°S band as validation that THIS is now the spot.

    Most of us following this note:

    1. The unwarranted criticism of the ATSB and GeoScience Australia from the usual suspects for not doing more about reviewing scans for the first spot these fellows landed on (despite the fact that they had mis-mapped it, and neither the ATSB nor GS could conjure up scans that didn’t exist). Again, the long suffering Australian taxpayer footed the bill for that exercise and now we find out it wasn’t warranted. Oops. No apology forthcoming for wasting the ATSB’s and GS’s time and money on that exercise, of course.

    2. The aggressive fist waving, table-banging, megaphone negotiation tactics employed by the cheer squad to get the Malaysian Government to sign up OI to immediately and unquestioningly to search the area surrounding 31°10′ S 95°18′ E. And the incandescent opprobrium directed at the Malaysians when they had the temerity to ask for more information. And then, oops, wrong spot. I don’t know if any of that crew have given any thought to what would have happened if OI had been signed up to search what was, by their own current views, THE WRONG SPOT.

    3. The evolving narrative regarding culpability for the disappearance. We’ve witnessed this crab-like scurrying from years of publicly stated views that Captain Zahaire Shah was guilty of murder/suicide, culminating recently in the headline grabbing “sensational new revelation” of a “holding pattern for 22 minutes off Indonesia” and the suggestion that the Captain may have been communicating with the Malaysian government, to now pilot disorientation brought on by hypoxia. Frankly, there’s the sensational new revelation, that of someone being in a persistent hypoxic state for seven hours.

    The whole thing is an exercise in intellectual plasticity that frankly beggars belief.

    As to any suggestion that this bunkum clears the “credible new evidence” hurdle, any vestige of credibility went up in smoke when the end point shifted 400 kilometres. We can only hope that OI had the good sense not to utter the terms WSPR or GDTAAA when they were with Malaysian government officials recently.

  67. Victor Iannello says:

    @Mick Gilbert: Well said. And I agree wholeheartedly with your advice that a new search proposal should steer well clear of the WSPR tracking fiasco, lest it be ridiculed.

  68. Victor Iannello says:

    A new article from Geoffrey Thomas with quotes from the WSPR tracking advocate on Mike Glynn’s paper on azimuth error. No mention of the imprecision of the Maidenhead grid locator. Now, the claim is the station locations are from license data,, and other similar databases.

  69. Mike Glynn says:

    I have just done a random sampling of the WSPR MK1 Technical analysis which details the LAT/LONG of the TX and RX points of the links used. Every one I looked at was at the centre of its respective grid.

    So it is a little strange to see RG claiming to have tracked down the actual positions when he has not provided any supporting evidence to that claim.

    I am working on a further critique of the track plotting techniques used in the WSPR Mark 1 track. I’d do Mk2 as well but we have no data regarding how that was done. RG has broken every rule in the book with the first analysis with regards to the techniques that should be used to make accurate position fixes.

    The concept of an “Area of Uncertainty” that accompanies any such radio fix, particularly one where the signals have travels extraordinary distances is not addressed anywhere. HF signals are apparently laser beams.

    The LORAN system which used HF signals two orders of magnitude more powerful than WSPR, contained accurate timing signals and used multiple stations for triangulation, had a maximum useable range of 1500nm. After that, the azimuth errors were too great for navigation purposes. Somehow WSPR is different?

  70. Victor Iannello says:

    Readers of this blog recognize the pattern when an informed contributor criticizes obvious errors in the WSPR tracking claims. The WSPR tracking proponent publishes a long-winded irrelevant response that does not address the core issue. Then, Geoffrey Thomas publishes portions or all of the long-winded response as if it refuted the criticism.

    These people have no shame.

  71. sk999 says:

    The WSPR transmission protocol only allows for the encoding of a 4 character Maidenhead locator in a single transmission. An option is available to send the next 2 characters in a subsequent transmission, but that is rarely done. So how is it that the WSPR database provides 6-character locators for the transmiting station most of the time? The secret is described in the following link:

    When you upload a file with spots that you have received to the database, you can enter your call sign and 6-digit locator, and the database links that locator to spots that others report of your transmissions. Kind of clever, but obviously prone to error because so much depends on hand-entered information being accurate.

  72. Mick Gilbert says:

    @Mike Glynn

    Re: “HF signals are apparently laser beams.

    Yes Mike, that is exactly one of the many ludicrous claims made regarding this rank bunkum.


    “WSPR is like a bunch of tripwires or laser beams (graphic below), but they work in every direction over the horizon to the other side of the globe,” Mr. Godfrey says.”

  73. 370Location says:

    Thanks, Steve, for that keen WSPR insight. It explains why some 4 char maidenheads I checked are ships or balloons. Any transmit-only or mobile station could be off by 50 miles, even with a 6 char code.

    @Victor, @Don, Re Otter crash acoustics:
    Looking at the arrival energy so far, the nearest seismometers show diffuse traces at around 3:06:40 PM of the reported loud bang before descent, but no obvious surface/seabed impact.

    I’m trying more advanced correlation algorithms.

  74. Victor Iannello says:

    Here’s my summary of what contributors have proposed for the new search:

    1) Search 32S-36S, with an emphasis on unsearched areas out to some glide distance from the 7th arc. Supporting evidence: Good BTO/BFO match for automated flight, fuel consumption model, drift models, satellite images. Challenges: Much of the area close to the 7th arc was already searched; requires a maneuver in addition to a turn between 18:28 and 19:41.

    2) Search 39.5S – 40S out to 10 NM from 25,000′ 7th arc. Supporting evidence: Acceptable BTO/BFO match, requires only one turn between 18:28 and 19:41, area close to 7th arc not searched. Challenges: Not supported by fuel consumption model, drift models.

    3) Search a 4 km radius around 8.36S 107.92E. Supporting evidence: Acoustic event near 7th arc. Challenges: Not supported by drift models, no debris finds near Java, no evidence linking acoustic event to MH370


  75. Mick Gilbert says:

    @Victor Iannello

    I’d be inclined include further north than 32°S in the “priority” zone, say 28°S – 36°S.

  76. Victor Iannello says:

    @Mick Gilbert: The fit to the BTO/BFO and drift model data is weaker in 28S-32S, but your recommendation is noted.

  77. 370Location says:


    OK, I was addressing your invitation about where to search and with what strategy. I could go on about the advantages and countering challenges, but I’ll keep it brief. Thanks for inviting more comments.

    Multiple drift models, especially early unbiased ones, show that drift from the Java site takes the bulk of the particles to where debris was found. A lack of evidence on nearby beaches isn’t significant. Those drift models also take debris directly away from the Java coast. Prior IG reports have excluded the Java site (with dubious probability math) and because the debris arrives too early. I do not see early arrival as a showstopper. Late modeled Java particles are still arriving Reunion in 2015, and we know that MH370 debris pieces have been found years after expected arrival.

    What links the acoustic event to MH370 is the proximity to the 7th Arc, which is at the center of that tiny 4 km error radius. It was also the strongest and sole acoustic event anywhere near the 7th Arc that hasn’t been searched. It’s beyond coincidence (and admittedly beyond the accuracy of the arcs depending on arc altitude). Odds of an anomalous event (strong on distant hydrophones, weak on local seismometers) being that close anywhere along the arc within an hour of impact are astronomical. The next nearest cataloged quake to the arc in a decade was in 2016.

    Primarily, it incorporates new evidence, and is not an estimate derived by optimizing imprecise doppler values or influenced by assumptions about FIR boundaries, unpiloted flight/heading modes, spins, dives, piloted glide range, eyewitnesses, missed searches, or masterminded conspiracies.

    Either there was a flyable path to the site, or not.
    If the noise was from sinking MH370, the cause might be revealed when debris is recovered, as with the black boxes hopefully explaining the flight path no matter where the plane is eventually found.

    Two expert barnacle analyses show that growth began in warm waters that got colder, which isn’t consistent with a cold water origin so that evidence has been largely dismissed. It does however match with drift from the tropical Java site.

    There is additional acoustic infrasound evidence for a flyby at Cocos Island airport, in the wee hours at exactly the right timing for a turn there between arcs 4 and 5.

    The site has never been searched by any method. No air surveys, surface searches, sat images, or even bathymetry. It’s just too far from where anyone expected based on the optimizable unpiloted assumption, which is now changing.

    It may not be the optimal fit to all modeled predictions from the evidence, but in reality no site ever would or could be. It simply is a plausible match to all the factual evidence.

  78. TBill says:

    My inclination would be that longer glides may have been beyond Arc7.

  79. Victor Iannello says:

    @370Location: First, I used the word “challenges” and not “show-stoppers”.

    I believe the best drift model we have is what David Griffin at CSIRO has published, both because we know exactly what hydrodynamic parameters were used to describe the debris, and also because the flaperon parameters were derived from actual experimental data on flaperon replicas. David’s model predicts that an impact around 8S would result in waves of debris arriving MANY months before the debris was found. That presents a challenge for an impact near 8S because it implies either the drift model is very wrong (unlikely considering the validation and calibration with actual drifter data), or barnacle-encrusted debris was somehow missed along the shore for many months. Not a show-stopper, but it does present a challenge that is not present for impacts much further south.

    And yes, a path that fits the BTO data, the BFO data, and automated flight after 19:41 does in my mind make it more probable than a flight where the BFO data fit is not as strong AND there are multiple maneuvers. (I remain fairly agnostic about what occurred between 18:28 and 19:41. It is only necessary to show there is at least one possibility.) Again, as you say, that does not rule out your proposed path, but it makes it less likely.

    In any event, your proposed site does have the advantage that it is very well localized, it is in a location where the weather is favorable, and it is not very far from shore. That makes it economical to search, and for that reason, I think it is worthy to consider, despite the challenges I mentioned.

  80. Brian Anderson says:

    A manoeuvre after 18:28 and before 19:41 is, in my view, entirely plausible. It need not be something elaborate; a simple orbit will work admirably, and is easily accomplished just by resetting the heading selector periodically.

    An orbit after 18:28 somewhere near the head of the Malacca Strait might also support the hypoxia possibility, with the aeroplane continuing on an automated flight heading determined by the last setting of the heading selector. The diameter of the orbit, and hence the time taken to orbit, will influence the final latitude reached.

    There is no evidence for a long glide.

  81. DrB says:

    @Mick Gilbert,

    My analyses of the CSIRO drift tracks predict the probability of the Point of Impact (POI) to be 1-2 orders of magnitude lower from 7-28S than it is circa 34S. In addition, the post-crash aerial search had good coverage but found nothing from 26.5-32S.

    Thus, the region from 28S to 32S is contraindicated rather strongly by both the negative aerial search results and by the very low probabilities (more than an order of magnitude lower) resulting from my exhaustive debris drift predictions.


    The 8S location is contraindicated by the excessive residuals in BTO and BFO occurring for any single autopilot navigation mode and with any single speed setting. Multiple maneuvers with timings between the arcs could reduce the residuals but seem highly unlikely because of their complexity and the fact that the pilot was unaware of automatic, periodic SDU communications. In addition, the probability of the MH370 debris from 8S matching the finding locations, dates, and barnacle presence/absence of verified MH370 debris is at most 1% of the probability from circa 34S.

    @Paul Smithson,

    I have been unable to find any flyable route to 39.5-40.5S which acceptably matches the BTO/BFO and is reachable with available fuel, even allowing the bleed air to be off after 19:41. The farthest South at which I have been able to demonstrate both conditions being satisfied simultaneously is 36.5S. In addition, the drift probability at 39S is significantly lower than at 34S.

  82. Paul Smithson says:

    @Dr B. Although Victor mentioned acceptable fit to BTO/BFO, I don’t claim the latter. Using your BFO calculator, I obtain a strongly trended BFO residual pattern which also exceeds the 7Hz 2-sigma in places. The BTO fit, on the other hand, is exceptionally good for a path: MEKAR-[Arc 1 crossing]-SANOB-IGEBO-40S85E, or alternatively SANOB-IGEBO-RUNUT-40S85E. In both cases the 7th arc would be met at approx. 39.6 South. Drift models are sensitive to assumptions on wind-forcing and directional offset to wind. Higher wind-forcing (say 3-4%) plus directional offset should be anticipated for highly buoyant flat composite fragments – which make up the bulk of the debris finds.

    Using your fuel model and the M0.84 tables, I also obtain a significant fuel shortfall IF we assume that the second engine went down at 00:17:30. I posit the following possible contributors to bridging the fuel shortfall (happy to hear critique):

    1. Turn APU on when first engine spools down and turn on all fuel pumps. This might succeed in retrieving residual fuel from centre tank below the scavenge auto-off point.
    2. Commence drift down descent under one engine. This uses less fuel that Max continuous thrust and also maintains higher horizontal velocity than slowing down at cruise altitude.
    3. When second engine goes down, power is maintained by APU so aircraft continues under controlled flight, in a straight line – presumably still following D/D profile for as long as the airspeed will permit. Arc 6 could have been crossed in this condition – permitting much earlier main engine fuel exhaustion.
    4. Final logon results from APU going down, then line refilled due to attitude change and restarts.
    5. Potentially additional fuel accessible below fuel=0 calibration point.
    6. Lower FF/greater endurance due to exceptionally low electrical loads and lower bleed air offtake.

  83. Mick Gilbert says:


    I hear what you are saying Bobby but we need to be mindful of the relative levels of precision that can reasonably be derived from the various methodologies. The drift science would be what? three or four orders of magnitude less precise in its predictive capabilities than the satellite science. And using the absence of detections from the search surface would be similarly as imprecise, perhaps worse, I would think.

    There needs to be a critical ranking of the utility of the various methodologies and presently there appears to be a leaning to ranking them all equally. That is why I prefer to start with the area that covers essentially all possibilities, exceedingly large though it may be, and then carve off the less likely parts of that.

  84. Victor Iannello says:

    @Mick Gilbert: Whether you start from a large area and whittle down to a smaller area, or you start with a small area and relax constraints to get a larger area, OI will only be able and willing to search a certain number of square kilometers. I don’t think we are ranking all data equally. What is subjective is how to assign the relative probabilities of various data sets and assumptions.

  85. ALSM says:

    Mick: Re: “The drift science would be what? three or four orders of magnitude less precise in its predictive capabilities than the satellite science.”

    It’s more complicated than that. Let’s break it down by 3 dimensions. The BTO data has horizontal range accuracy and precision several orders of magnitude better than debris drift analysis can provide. The BFO data provides “orthogonal accuracy and precision” (position along the 7th arc) that is roughly equal to what the debris drift analysis can provide…arguably lower. OTOH, the BFO data provides (by far) the best evidence that 370 is close to the 7th arc…probably within 20nm. The missing IFE logon circa 00:21 is hard to dismiss as a coincidence.

    Weighing all that, the area close to and along the 7th arc between S30-S36 degrees remains the most likely POI location, notwithstanding the the fact that most of this are has already been searched.

  86. Paul Smithson says:

    @ALSM, agree with your statements on drift and BFO. So why would you prefer S30-S36 (which has been searched out along the arc) vs S39.6 which provides better BTO fit, more parsimonious path solution, fully compliant with waypoint-based route at M0.840/FL360, and has not been searched?

  87. ALSM says:

    @Paul: Because the consensus debris drift analysis points to the S30-S36 range and the fuel models make S40 a stretch.

  88. Victor Iannello says:

    @ALSM: The BRAN2015 drift results that David Griffin of CSIRO has provided makes it difficult to justify searching north of 32S. Chari Pattiaratchi’s drift model results generally place the impact further north than David’s, as in the past he has recommended searching as far north as 28S, although more recently he has recommended between 32S and 33S. If this is his current recommendation, that would be more reason to not search north of 32S.

    It would be helpful to know exactly what parameters were used in Chari’s drift model to account for Stokes Drift and windage. We know exactly what David used and how he arrived at the values. If by chance both Stokes Drift and windage (which both may be expressed as a fraction of wind) were ignored, then that would explain why his results predict an impact site further north, as the virtual particles in his model would move slower than expected. (His model, however, would accurately model the drift of drogued drifters.) Chari does specifically reference what is used to model ocean currents but I haven’t seen anything about wind models, which might suggest he included ocean currents but not wind.

  89. ALSM says:

    @Victor: I agree that the Southern part of S30-S36 (S32-S36) is more likely. Given how good S37.7 once looked (ATSB, IG and others), maybe the whole range between S30-S38 should be considered still posible, within which S32-S36 is most likely. But north of S30 or south of S38 is too inconsistent with the debris drift analysis.

  90. DrB says:

    @Mick Gilbert,

    You said: “The drift science would be what? three or four orders of magnitude less precise in its predictive capabilities than the satellite science.”

    My prediction method is now (after 2 1/2 years of work) much better than that. The actual precision of the predicted POI latitude in the best case, when using all the 86,400 CSIRO drift trials and all the verified MH370 debris, and when constraining the trials to originate in the vicinity of the 7th Arc, is only about 0.11 degrees at one sigma. The absolute error of the CSIRO drift model is 30 NM, or 0.50 degrees at one sigma. Therefore, the total error is only 0.51 degrees at one sigma, or 31 NM. I also note that my predicted BTO/drift POI is about 40 NM to the northeast along the 7th Arc from the intersection of the best-fit route with the 7th Arc (which is at 34.23S). Thus, it appears that the aircraft impacted after a net left-hand turn.

    The BTO/drift POI error is close to being only one order of magnitude worse than our knowledge of the location of the 7th Arc, taking into account the BTO noise and the aircraft altitude uncertainty then. Therefore, the best drift prediction can discriminate the POI location along the 7th Arc much better than fuel models and much better than BFO fits. It locates the POI along the Arc with an accuracy that is a fraction of the maximum piloted-glide range.

    Of course, one reason the drift prediction can be so good is that we have to search only along the Arc in one dimension, not in two dimensions over the whole SIO. So, the drift prediction is only a 1-D search problem (with the BTOs providing knowledge of the other dimension). Even then, it takes many days of computing with a very fast PC. The 2-D problem is not practical to do on a PC; it needs a supercomputer, and CSIRO would have to predict at least 10 million drift tracks to start with.

    @Paul Smithson,

    You said: “So why would you prefer S30-S36 (which has been searched out along the arc) vs S39.6 which provides better BTO fit, . . . .”

    The BTO fit is 39.6S is not “better” than at circa 34S. At 39.6S the BTO residuals are too small, and the correlation coefficients don’t match their expected values as well as at 34S. The small residuals at 39.6S are inconsistent with the known BTO noise. The residuals at 34S are entirely consistent with the expected BTO noise and with their expected correlation coefficients.

    @Brian Anderson,

    You said: “There is no evidence for a long glide.”

    Roger that. On the other hand, we can’t prove (yet), that a glide did not occur. I think it is highly unlikely, because our quite detailed (and calibrated using previous flights of 9M-MRO) fuel consumption model predicts MEFE at 00:17:30 only when you include the significant fuel economy of the bleed air being off after 19:41. To me, that means it was impossible for a pilot to be alive at MEFE. No pilot means no piloted glide. No piloted glide means the POI is relatively close to the 7th Arc. Not finding the debris field in previous searches means it was “missed”. Why was it “missed”? The debris field could be in one of the small areas where there are no data (or poor-quality data), or in an area where the terrain made close inspection impossible using the equipment at hand and the programmed search pattern. I believe OI learned a valuable lesson in the ARA San Juan search. Fortunately, they realized that submarine was highly likely to be near the position of its last radio call. Therefore, even after departing the search area, they unexpectedly returned for a closer look at areas which had been initially classified as geological outcroppings. One of these turned out to be the imploded submarine debris field.

    If I were planning a search, I would look in this order:

    (1) the “difficult terrain” and the “no/poor data” areas within 2 sigma of the POI predicted by the BTO/drift analyses,
    (2) a complete search of that predicted BTO/drift zone, and
    (3) a complete search within the maximum glide range from the BTO/drift POI.

  91. Paul Smithson says:


    Re Drift model confidence. I can imagine that the theoretical predictive accuracy of drift modelling might be as precise as above. However, is it not obvious that wind-forcing (by which I mean a combination of stokes drift, surface film shear and leeway-proper) may vary substantially from levels assumed? The most buoyant debris (oil slicks, seeds) exhibit wind-forcing of about 5% of windspeed – mainly because the uppermost layer of water are moving at quite a different rate to those a few cms down. An object with 15cms draught will drift in a different fashion than one with 2cms. Moreover, you can always expect an offset of wind-induced drift wrt wind direction – you don’t require an oddly floating object like a flaperon to achieve that. Wind induced drift of even 3.5% is more than double assumed by Griffin and approaches, or even exceeds, the ocean surface current vector. For these reasons, I believe that debris drift models have a very much greater margin of uncertainty than your modelled predictions above. For a discussion of depth-dependency of the wind-related vector see here:

  92. Paul Smithson says:

    In response to the second point regarding BTO fit.

    I’m flattered that you think the BTO fit of my path model is “too good”(!). It is surely axiomatic that paths with better BTO fit are more likely than ones with worse fit if BTO error distribution is gaussian.

    It is obvious that if we allow start position, time, speed and track angle to float, it will be possible to optimise path models to be better than “truth”. By allowing the start position to vary, one can “shoot the gap” between Arc 2 and 3 to minimise errors there, while the combination of azimuth and speed profile allows you to minimise errors at arcs 5 and 6. So yes, I do understand that BTO optimisation does not lead you to truth where BTO errors are minimised. That doesn’t detract from the fact that the truth is most likely to be found close to the spectrum of lowest BTO error / best fit paths.

    If you take MEKAR-SANOB as a prior, and optimise single-turn paths models based on minimisation of BTO errors, you will find it is possible to obtain a path solution with extraordinarily low BTO errors: -7, 12, 17, -31, 11, 35 (rms error Arc 2-6 = 17.5) at FL340, M0.838, initial azimuth 191.04, intersecting the 7th arc at 39.8S.

    If we now constrain the aircraft to fly via nearest waypoints (which it must do if the solution is great circle), and you constrain constant mach options to two decimals (which is maximum resolution selectable for constant mach speed), then your BTO fit is slightly less spectacular – but nonetheless very good.

    The path that I have described previously, at constant M0.84, FL360 MEKAR-SANOB-IGEBO-40S85E produces predicted BTO residuals of: -8, 27, 30, -40, -11, 3 (rms arc2-6 = 26.2), a good deal higher than the theoretically optimsed paths. Is that “too good”? If memory serves, I believe you said that rms BTO errors should be expected to be in the region of 27 – on which basis this path seems to be right on the money.

  93. 370Location says:

    @Victor, @DrB, @All,

    I proposed a simple flyable path that matched the three acoustic detections. There are no BTO error residuals, because that defines the timing of the path. I’m not using a BFO optimization strategy, so it did not factor into my path at all. To further simplify, I assumed low altitude flight. Checking for plausible speeds from that, there were no strange excursions, and the shifts matched with winds aloft. The resulting BFO residuals also seemed acceptable, better than for some of the past priority search areas.

    The CAPTION team does use an optimization strategy. They found an optimal path up to my Cocos flyby timing with near nil residuals. I have no doubt that an optimization approach for the remaining portion could find multiple solutions with minimal errors.

    The reference to excessive turns on my path may be to the additional segment after Christmas Island toward Jakarta. That was because the speed on an NE heading between arcs 5-6 was slower than arc 6-7 on a direct path. Crossing 6-7 at a less oblique angle gave one exact solution, while sticking with a simple premise of waypoint nav. An alternative that fits the timing would be for a climb to more efficient altitude after Christmas Island toward the coastal airport, followed by a faster descent near the acoustic endpoint. I don’t know if that would have a better BFO fit, because it didn’t seem necessary to get into BFO optimization modeling. If the lack of a BFO match is considered a challenge to accepting that there is a flyable path, then perhaps that’s something I need to address.

    It might be difficult to find an expert in drift modeling who would validate that a crash site at 34S is 100x more likely than at the Java acoustic origin. Most of the reports didn’t even consider an origin that far north. The same could be concluded from the IG report for 23.6S, but we have an example of a real drifter 101703 that crossed the 7th Arc there in Mar 2014 and passed by Reunion at the right time. The CSIRO model shows more debris from Java splitting north of Madagascar, while UWA and others show an even split. All have later arrivals in 2015. I don’t think any of the models are wrong, just that the variations show the imprecision. For 34S, debris would be meandering in gyres for months (some going to Australia) before getting swept into the SEC toward Madagascar, where IG set a narrow window for it to be found as it passed Reunion. For a Java origin, the debris is swept directly toward Madagascar, but could could have meandered in an eddy anywhere along the way or near Reunion. There were also news reports that the flaperon had been seen by others at least a month earlier.

    CSIRO says they were tasked to confirm that the search being conducted was compatible with the drift modeling, and they confirmed that it was. I believe that if the priority search area had been much farther north, many studies would have confirmed that was plausible, too. Again, I haven’t undertaken the task of constructing my own drift model to optimize parameters and see where debris ends up. I’ve considered it, but it’s a huge endeavor, and the end result might only demonstrate my own vulnerability to confirmation bias as a weak link.

    I’ve strived to meet the officially stated search threshold of “credible new evidence for a specific location”. If others were to convey that my acoustic analysis candidate is credible (even if it’s not their top option), it might help towards inclusion in a new search.

  94. DrB says:

    @Paul Smithson,

    You said:

    “It is surely axiomatic that paths with better BTO fit are more likely than ones with worse fit if BTO error distribution is gaussian.”

    No, that is not true. That was exactly my point. An observed probability density which MATCHES the width of the gaussian measured on previous flights is a most likely solution, not the path which has the smallest standard error.

    You also said:
    “However, is it not obvious that wind-forcing (by which I mean a combination of stokes drift, surface film shear and leeway-proper) may vary substantially from levels assumed?”

    I agree they might, but apparently not by a large overall speed factor for the found MH370 items. I tested the sensitivity of the global POI solution to the average drift speed, and I found that the assumptions used by CSIRO for leeway, etc. were confirmed to be, within the noise, the most precise POI solution (i.e., having the narrowest Joint PDF). What I did was to scale the CSIRO-predicted tracks in the time dimension, so the predicted drift time to a given location was scaled by a percentage factor. That way I could make all or one of the debris drift 10% faster, on average, for instance, or 10% slower. I was expecting to find one or a few debris for which the predicted arriving date was better matched by the MH370 debris reports with a different drift speed. Doing this for one debris at a time failed to noticeably improve the global solution (i.e., to narrow the Joint PDF). Large drift speed error factors (25%) did produce noticeable changes in the single-debris PDF, but almost universally in the wrong direction. So, despite quite a few months of searching for a better drift speed solution, and for “outlier” debris with very different drift speeds, I was unable to improve upon the CSIRO assumptions and predicted drift tracks. Perhaps there is some selection effect in that the debris which survived years at sea and remained afloat and findable have fairly similar drift speeds. Whatever the cause, the solution is not highly sensitive to the assumed leeway anyway because each arriving wave is several months wide, and all one can really do is test which wave the debris arrives in, but the exact date within the arriving wave is only important for the several debris which arrived with live barnacles. The barnacle-free debris have an arriving date uncertainty of several months, so how could one tell the difference between a slower debris being found quickly and a faster-drifting debris having a longer finding delay after arrival? You can’t. Since almost all the MH370 debris were found barnacle-free, you can understand why weeks or even months of predicted travel time error is absorbed by the arriving time uncertainty caused by the unknown finding delay. The flaperon, the Pemba flap, and the stowage closet door had barnacles attached when found, and in these three cases the known finding date allows determination of a narrow time window for the arrival. This assists in discriminating the POI, but the other (barnacle-free) debris also play a major role in the Joint PDF. In summary, my sensitivity studies indicated the average drift speeds produced by the CSIRO-assumed parameters gave as good a global solution for the POI location as any speed adjustments for one or all of the debris.

    I have seen the depth-dependence paper. However, the small-scale variation is beyond the scope of the BRAN2015 model used by CSIRO. Also, the speed sensitivity study I described above would have identified any debris for which the effect was so large that it would have arrived in a different “arriving wave”, and I found no cases where that could be demonstrated. The deviations in drift direction when the wind is blowing are a different matter, and I would not attempt to try to model that. I will argue that, if those were randomly distributed, then the assumed wind-drift direction is correct for the ensemble of debris, and the deviations in wind-drift direction from the wind direction will have a mean near zero. So, I would expect those actual wind-drift direction offsets to be small, on average, meaning there is no significant bias in the predicted POI location, just a broader PDF, which means a more uncertain POI. Also, the directional errors are already baked into the Joint PDF.

    You also said:
    “The path that I have described previously, at constant M0.84, FL360 MEKAR-SANOB-IGEBO-40S85E produces predicted BTO residuals of: -8, 27, 30, -40, -11, 3 (rms arc2-6 = 26.2), a good deal higher than the theoretically optimsed paths. Is that “too good”? If memory serves, I believe you said that rms BTO errors should be expected to be in the region of 27 – on which basis this path seems to be right on the money.”

    Using one statistic is too simplistic. The route probability used in UGIB (2020) is summarized in Table G2. It uses 9 statistics. You have to look at all of these statistics simultaneously to evaluate the probability a given route is consistent with expectations. On that basis the route presented by UGIB has the highest probability.


    You said:
    “It might be difficult to find an expert in drift modeling who would validate that a crash site at 34S is 100x more likely than at the Java acoustic origin. Most of the reports didn’t even consider an origin that far north.”

    The CSIRO study used track origins as far north as 7S, so it includes your proposed site.

    Both cursory and extremely complex calculations of origin probability using the CSIRO drift tracks indicate much lower probability at 8S compared to 34S. The exact probability factor is immaterial. If you want to argue that 8S is possible, then you need to show an improved drift model predicting 8S is probable or, at the very least, to show why the CSIRO drift tracks are quite wrong.

  95. Victor Iannello says:

    @370Location said: The same could be concluded from the IG report for 23.6S, but we have an example of a real drifter 101703 that crossed the 7th Arc there in Mar 2014 and passed by Reunion at the right time.

    That statement is not correct, as I explained in a previous comment, which also includes a figure showing where along the 7th arc the CSIRO drift model predicts the origin on March 8, 2014, for both the drifter and the flaperon.

    An undrogued drifter and the flaperon have different hydrodynamic drift parameters. As shown in the upper pane of the figure, if we only had the final position of the drifter on July 23, 2015, we would predict that the drifter could have originated over a range of latitudes from 10S to 40S, with a cluster of possible origins very close to the actual origin of 23.6S, and with a trial (12682) that closely matched the path of the actual drifter. On the other hand, for the flaperon, origins north of around 31S are highly unlikely, as shown in the lower pane of the figure.

    The probability of the flaperon starting at 8S is a very, very low probability unless it was found many months after beaching, which is contradicted by the encrusted barnacles.

  96. Victor Iannello says:

    @370Location: I had forgotten, but in March 2021 when I was investigating the drift results regarding the flaperon and drifter 101703, I also investigated whether a reporting delay might account for an impact as far north as 8S. The results are presented in this comment and this figure.

    Assuming reporting delay (the time between beaching and reporting the find) of 60 days, there is low probability of an impact north of 24S. If we look at ALL trials which reach Reunion Island independent of timing, there were 379 trials, with only 1 trial starting north of 24S (at 20S).

    As I said back then:

    This should eliminate any concerns that I have selected a time window that is biased against an impact at northern latitudes. The possibility of an impact at northern latitudes is only allowed if the drift model is not representative of the historical winds, currents, and the hydrodynamic behavior of the flaperon.

  97. 370Location says:

    @Victor, @DrB:

    I realize that IG has put an enormous effort into refining peak probabilities based on metaanalysis of CSIRO drift modeling. @Paul, @David, and others have cautioned against attempting to derive precise results from limited and imprecise data. They have made the case more clearly than I could. CSIRO author of the studies David Griffin has said that the drift results are not precise. The work was focused around modeling for one piece of debris. Other found pieces would drift differently. @Oleksandr modeled debris with a wider variety of characteristics, and results from the Java origin are a good match for where all the debris has been found, even if discovered years late.

    I followed your link back to the Mar 2021 drift discussion, where several gave solid input. It
    s clear that the wider the window, the farther north are possible origins. Sixty days is just too narrow when we’re talking about several months of difference. My followups to your point about delayed arrival still stand:

    I’ve previously said that I don’t think the drift models are wrong, just imprecise. It’s a strawman attempt to state that I must either prove the CSIRO model wrong or show that the flaperon was beached for months. I’ve never suggested either, and described more realistic alternatives.

    This gets back to the fundamental approach of optimizing a most likely origin from very limited and imprecise data. That’s a practical way to narrow down very broad possibilities when there’s nothing else to go on. However, the seismic data is very precise and can be trivially validated from public data using well developed scientific methods. Using narrow imprecise drift interpretations to exclude the possibility that the noise was from MH370 can’t be right.

    Hypothetically, suppose that the noise near Java had been louder, seismically cataloged, and discovered to exactly match the newly realized 7th ping arc evidence. I have no doubt that there would have been immediate expert validation/refinement, aerial searches for debris, scans for ULBs at the site, bathymetry, towfish, ROV, etc. It would have certainly had an influence on how the drift and BFO optimization models were developed to narrow the search zones, not to mention conclusions about barnacles.

    Here’s a future hypothetical to consider. Suppose that MH370 is eventually found at the Java acoustic anomaly, and you are reflecting back on your analytic methods. Does that possibility change the way these probabilities are being weighed, and the claims being made?

    I always consider the possibility that I’m wrong, and whether my analysis will hold up no matter where MH370 is found. And, it must be found.

    I must say that this open comment forum has been instrumental in developing theories about where to search for MH370, and motivating that effort.

    Thanks to @All. — Ed

  98. Victor Iannello says:

    @370Location said: I must say that this open comment forum has been instrumental in developing theories about where to search for MH370, and motivating that effort.

    The challenge will be to build the political pressure to persuade Malaysia to agree to a no-find, no-fee contract. The requirement of “new and credible evidence” is a vehicle for Malaysia to do nothing by default. We all know this.

  99. DrB says:


    I have never been a member of the IG, which has not been active for many years (although some of its members still are). Victor and I have been individually collaborating to model and analyze the MH370 debris drift since March 2020. We have been assisted along the entire way by David Griffin, who has been critically reviewing our several methods and results.

    As will be made clear in our upcoming paper, your concerns about the “imprecision” of drift predictions have been overcome by our method of simultaneously analyzing all verified MH370 debris and by computing a joint probability density function (JPDF) of the Point of Impact (POI) location along the 7th Arc. We find that JPDF to be very narrow, providing a precise estimate (to within a small fraction of a degree) of the crash location. Therefore, the overall location error is dominated by the (larger) systematic model error (0.5 degrees at one sigma), which has been provided by David Griffin, rather than the (smaller) precision determined by the width of the JPDF.

    Our predicted POI based on debris drift is in close agreement (within 43 NM) with the Last Estimated Position (LEP) based on maximizing the route probability using nine statistics of the satellite data [UGIB (2020)]. I emphasize that these two methods are independent, both in terms of the data used and the prediction methodologies. It is highly unlikely that both methods would agree because, coincidentally, both had the same (large) error. Rather, it is much more likely they agree because they are both accurate within their small estimated errors.

  100. Paul Smithson says:

    Dr B, regarding your earlier commentary on the pdf insensitivity yo speed errors. It’s difficult to comment very cogently without seeing the studies in question. However, I’ll not a few things:
    1. Wind vector on a fine (daily) timescale is not tightly coupled to ocean surface current vector either in direction or magnitude. In the tradewind zone, for example, the wind is fairly steady, yet the OSC has lots of eddies. So scaling by time wouldn’t really test sensitivity to wind % assumption. You really need to decouple the wind vector, vary that systematically, and see what difference it makes.
    2. I am not talking about a 10% difference. If wind forcing really was (say) 3% of wind vector it would mostly exceed the OSC vector. My guess is that the combined vector might be 50% larger, not 10%.
    3. Sensitivity analysis w.r.t. wind has been modelled in at least one published study [i’ll ho and look it up] and demonstrated a significant influence on inferred origin.
    4. Dr Griffin laudably went to the trouble to try and validate his drift parameters experimentally. However, the lagged effect of winds on surface currents and sea state as means its unlikely to generate a reliable result from a single sea trial. I’d suggest that the drift characteristics of flat bamboo plates and other buoyant objects offer a good empirical basis to start from. Those have been characterized more thoroughly and the go downwind a good deal faster than 1.2% of windspeed.

  101. Paul Smithson says:

    Another paper with experimental demonstration of dramatically different drift characteristics by depth/draught.

    “We show that currents in the upper few centimeters of the ocean may have drastically different magnitudes and directions than the average over the upper meter, a vertical extent commonly taken as the thickness of the ocean’s surface layer.”

  102. Victor Iannello says:

    @Paul Smithson said: I’d suggest that the drift characteristics of flat bamboo plates and other buoyant objects offer a good empirical basis to start from.

    Maybe if we were studying the transport and dispersion of oil slicks. BRAN2015 was extensively calibrated against undrogued drifters, which should be much more representative of the larger debris pieces that were recovered than 2-mm thick buoyant plates.

  103. Paul Smithson says:

    @Victor. Bamboo plate drifter thickness 1.75cm, not 2mm. GDP/SVP drifter diameter 35cm, approx draft 15cm. What do you suppose was the median thickness and effective draught of the debris objects recovered?

  104. Victor Iannello says:

    @Paul Smithson: Thanks for correcting me on the bamboo plate thickness. I remembered the thickness incorrectly.

    There is a range of sizes of recovered debris from MH370. The flaperon, which is probably the most useful in discriminating impact latitude, had a thickness of around 30 cm. The flap also had a thickness of around 30 cm. Others parts with significant thickness include the flap fairing (15 cm), engine cowling (10 cm), right aileron (17 cm), and flap support cowling (15 cm). The undrogued drifter dimensions are roughly the same.

    On the other hand, the bamboo plates were used to simulate the drift and dispersion of oil slicks and small floating pieces of plastic. I don’t think that’s representative of debris that was recovered (and identified) from MH370 and used in the drift study.

  105. Paul Smithson says:

    @Victor – thanks for taking the trouble to specify the thickness of those larger items. I don’t want to belabour the point, but your list above comprises the thickest items recovered. The majority of found items were relatively thin, flat panels, like those shown in Gibson’s pictures.

    As regards the flaperon, in addition to its 1.2% stokes drift the best-fit model added 10cm/S @20% left of wind. That’s equivalent to about 1.7% of typical windspeeds (6m/s, or 12kts) – so a combined wind-forced vector approaching 3% of wind.

    I don’t know whether this was resolved in later drift modelling, but I recall that “Roy” seemed to have gone much too far/fast than the models predicted – another sign that the parameters could be awry.

    I do hold Griffin’s expertise, methods and diligence in the highest esteem, so do not aim to impugn his work. I’ll leave it there.

  106. Victor Iannello says:

    @Paul Smithson: The arrival date of “Roy” is clearly an outlier. It’s possible that strong winds may have lifted “Roy” from floating flat on the surface of the water and carried it above the water in a way that is not modeled in the drift model. Or, it somehow managed to avoid getting delayed in some of the eddies and took a more direct path.

  107. Viking says:

    Concerning the possibilities to locate unnatural events with seismic detectors:

    Within the last few hours geologists have published proof from more than 20 different seismic detectors that the leaks in the Baltic gas tubes from Russia to Germany were caused by military-grade explosions. This clearly illustrates the capability to locate violent events at sea with these detectors.

  108. Viking says:

    Here you can see the signals and hear a geologist analyze the data (click the video):

    Sorry, nothing in English yet.

  109. Victor Iannello says:

    @Viking: Thank you for your comment. As we’ve discussed before, acoustic detection was also successfully used to localize the San Juan submarine implosion.

    Beyond the technical aspects of acoustic detection, possible sabotage of the Nord Stream pipelines will have profound implications.

  110. Viking says:

    @370Location @Victor

    I think it is quite encuraging for the hunt for MH370 that seismic detectors at a distance of more than 240 km detected a roughly 100 kg TNT explosion with a good signal to noise ratio. The crash of a large airplane will release more energy that this, so I guess it can be detected up to 500 km from the impact.

    Are you able to analyze data from Christmas Island (mostly the southern detector) at later times than you did before? We can potentially look for surface waves in addition to the guided waves at deep sea, and normal earthquake waves.

    Now the hunt is beginning for those who blew up the gas tubes. According to Danish military experts it cant be done without leaving significant evidence.

  111. Viking says:

    Two of the Baltic explosions look very similar as observed from Bornholm – probably the same type of bomb was used for both tubes:

    This time in english.

    Considering the dealy, I guess a mother ship was used to transport the bombs.

  112. Viking says:

    More seismic results from the Baltic explosions are coming in quickly now (approaching 40 stations). The record distance with a clear signal is 943 km (Finland). I think it is time to stop detailed reports (from my side) here:

    However it makes me very optimistic finding signal directly from the crash of MH370. At least if it happened near Christmas Island.

  113. 370Location says:

    @Victor, @DrB, @Paul:

    Conjecturing that the Roy arrival is an early outlier caused by weird properties that don’t fit the CSIRO model seems odd, because the weirdness is only needed to explain an origin near the areas already searched. Other debris arrived within months of Roy, and all are compatible with several models of drift from a Java origin. Regardless, it seems to verify that drift interpretations from imprecise data can’t be used to rule out a Java origin. I too have the greatest respect for David Griffin, and the CSIRO work is very well done. However, as with all of the evidence, we should be careful about how the data is interpreted based on our prior assumptions.

  114. 370Location says:


    Thanks for the early tip on the Baltic explosions. I have no doubt that the seismic arrivals are a precise match for the explosions. Current news reports estimate 2-3 km uncertainty from distant seismometers. That’s about what I’ve been allowing from stations 90-400 km away. Beyond that distance, the data gets lost in the noise, so I expect the explosions were stronger than the Java anomaly.

    You mentioned a geologist who may be acquiring data on the Citation crash, who may be looking at the Baltic explosions now, too. It would be significant if they were to take an interest in analyzing the Java event. Utilizing restricted Indonesian seismometers should improve the accuracy. Geologists also have sophisticated tools for calibrating seismometers and localizing an epicenter. I’ve developed my own algorithms from first principles. I believe they’re accurate, but it would be marvelous to see an independent calculation of the epicenter.

    As for surface waves, I spent quite some time chasing a very low frequency wave that arrives AU.XMIS at around 00:23:30Z. It’s prominent on the E-W axis and appears to match the frequency of a Rayleigh wave propagating over deep ocean. I tried isolating it by ellipticity through polarization analysis, but haven’t gotten a firm enough conclusion to propose it as from an MH370 surface impact. It may be slightly early, but the narrow directionality toward the Christmas Island seismometers might imply a ditching on that heading. If so, it also opens the possibility that the unexplained Curtin hydrophone event characteristics are a good match with a reflection off the 90 East rise from the Java site. That’s all speculative so far for a surface impact, but the later 1:15:18Z anomaly event is pretty definitive.

  115. Victor Iannello says:

    @370Location: The timing and location of the arrival of “Roy” is a low probability event independent of what impact latitude is assumed. That makes it not useful for discriminating impact latitude. Given enough debris, low probability events may occur that involve phenomena not modeled by CSIRO. That’s not a reason to ignore what the model is telling us for debris where there is good latitude discrimination, such as the flaperon.

  116. Viking says:


    WOW! A 00:23:30 arrival is very close to what I was hoping for (given my crash position and a guided wave transmission in the ocean). That is extremely good news. 20-70 seconds later would be bulls eye, but I would also have to make some minor corrections on my side, so we are close.

    A surface wave from my position would arrive several hours later.

    Are you able to make your results (publically) available? By the way, wouldn’t a crash in exactly the opposite direction give pretty much the same result (because of the low compressability of water and Newtons 3rd law)?

  117. 370Location says:


    Yes, Martin, it’s a very interesting signal. It hints at a surface impact. I refined the signal plots for the AU.XMIS arrival, focusing at very low frequencies below 1Hz:

    The dominant E-W 0:23:30 signal (red) has a period of 6.66 sec = 0.15 Hz. (The surface-seabed delay for a vertical impact for 3400m depth at the Java anomaly would be 2.27s.) It turns out the 0:23:30 timing is a very good match for a T-wave arrival. I estimate timing from 19:37 should be 0:23:27Z. Besides P and S wave arrivals showing in the Autocorrelation and Phase plots, the spectrogram faintly shows what may be the leaky-Rayleigh wave arrival upward frequency dispersion arc starting just after the S-wave at 0:21:03Z, leveling off toward 0.3 Hz.

    The polarization analysis on the seismometer is 180 degrees ambiguous, but I can’t say the same for the water physics of the impact. My intuition says the shock wave would travel in the direction of impact, and water would fill the void from all directions, possibly creating an additional cavitation collapse. There are several seismometers NE of the anomaly site on the Java coast, and I recall they don’t show that low frequency wave arrival.

    The .wav file snippet is for analysis, but even at the usual 60x speedup, it’s subaudible.

    I’d be curious to hear your geologist’s impression of my plot of Signal Phase azimuth over time. I believe it’s novel in revealing weaker seismic arrivals beneath strong ones, much like the hydrophone bearing graphs.

  118. 370Location says:


    One further thought. I wonder if the brief 0.15 Hz resonance on a T-wave might be related to the oscillation of a very large cavitation bubble from the impact.

  119. Viking says:


    I am extremely happy. This looks like a real breakthrough. However, I disagree on some of the fine details of your assignments – need to discuss with my professional geology friend before commenting on these things.

    I am a little curious if you tried to look some hours later for a slow surface wave? I think we do not really need to find it, but it would still be fun.


    You can now add my end position (13.42S, 106.66E) to the proposed search list. I think the debris field will most likely be within a 10-km radius of this position.

  120. Victor Iannello says:

    @Viking: If you have acoustics results that support your proposed position, that would make a stronger case.

  121. 370Location says:


    Others have tried using a tidal buoys to look for the splash surface wave. I tried filtering that data and looking at energy changes. Mostly it’s weather driven.
    I’ve added a plot to the shared folder above.

    I don’t know that seismometers could distinguish any shift in breaking wave arrivals on a nearby beach. I did spend weeks exploring frequency or phase shifts in microseisms, which are resonances between surface and seafloor driven by surface waves. Fascinating tapestry, but nothing stood out.

  122. Viking says:


    I know from a book by one of the co-inventors of highly sensitive modern seismology equipment (R.V. Jones) that seismometers can distinguish such waves in some cases.

    I think they should be better than tidal buoys for small signals, since most of the ordinary (weather driven) wave noise will average out while a slow wave from a big event will remain coherent to some extend and therefore not average out. However, success may be dependent on the local shape of the coast and wave propagation direction.

  123. Viking says:


    I may have misunderstood your comment, or you may have forgotten the details of my original manuscript where I stated:

    This is in excellent agreement with a sound feature recorded at the nuclear
    arms listening device HA01 near Cape Leeuwin (34.892˚ South, 114.153˚ East) at 00:49:42.

    Since that original statement, I slightly changed the proposed end position, but this makes only insignificant changes in the expected arrival time, since the change was mostly perpendicular to the direction towards HA01.

  124. Victor Iannello says:

    @Viking: It would be helpful to explain why your acoustic data supports 13.4S while @370Location’s acoustic data supports 8.3S, i.e., different sensor data, different methods, different interpretation, etc. Frankly, it’s been some time since I’ve investigated the acoustic data, and I forget how you two have arrived at your conclusions.

  125. 370Location says:


    Several others have tried to claim a hydrophone amplitude spike from one of the clusters in the LANL report. (And my comments about that here have triggered legal threats in the past).

    The problem is that the H01 hydrophone is an array that can determine the direction of origin. The 00:49:42 (LANL cluster 2) arrival came from back azimuth 205 degrees, which indicates an ice event. There was no significant energy from any other direction including bearing 341.4 towards your candidate site. Here’s a detailed plot of the signal arrivals matching the LANL clusters:

    It may be possible to filter coastal seismometers for even longer periods, something more like a tsunami. I’ll look into it.

  126. Viking says:

    @Victor @370Location

    The two positions are the result of very different approaches. For my position, it is based on Inmarsat data, with some minor (but important) details from other satellites, but no input from seismic data. When one then looks for seismic data and finds something consistent with the pure satellite solution, it is a strong lead. Here the CI signal is clearly best. The HA01 signal is weak, so I am sure the direction determination is uncertain. Basically only the timing can be used, but it is not strong evidence with time only.
    On the other hand I have found a lot of other data supporting my solution. I will come back to this in a couple of blog contributions in the near future.

    For the position found by 370Location the primary data source is seismic (sorry for my simplification, but it is important to make the difference clear). Most points with seismic signals cannot be linked to MH370, because they occur at times or positions which are completely inconsistent with Inmarsat data. However, one or two may be possible to reach while still getting signals consistent with Inmarsat observations. As far as I can see the best candidate is the one found by 370Location, and it is therefore worth investigating further. However, the drawback is that it can only be reached by an active pilot flying in a somewhat unusual way.

  127. Peter Norton says:

    370Location: “Several others have tried to claim a hydrophone amplitude spike from one of the clusters in the LANL report. (And my comments about that here have triggered legal threats in the past).”

    what legal threats, why and by whom ??

  128. Victor Iannello says:

    @Peter Norton: Ed made a comment about somebody’s questionable use of the acoustic data, and that person threatened to sue Ed, me, and my web server provider (Network Solutions), and demanded that the comment be removed. You can probably guess who it was.

  129. Peter Norton says:

    @Victor Iannello: If you had said “questionable use of the WSPR data”, it would have been more obvious to me, but in this case (acoustic data), I’m afraid, I must have missed something apparently. But it’s probably none of my business, I was just very surprised, that scientific discussions could lead to such threats … I mean debate is the very essence of scientific progress. Anyway, I hope you all don’t have to worry, legally speaking. Thanks for holding out.

  130. Victor Iannello says:

    @Peter Norton: Same individual. In any event, it’s not worth re-hashing. Some people thrive on the drama. I don’t. Neither does Ed.

  131. 370Location says:


    The SNR on that LANL cluster 2 event is fairly good. I could provide more detailed plots than the one I linked to, perhaps a linear plot of signal energy vs bearing in that time frame would be clearer evidence. There were two ice events at slightly different bearings at that time. I’m confident that Curtin, LANL, myself, and others doing TDOA correlation analysis did not miss a loud H01 arrival coming from the direction of Christmas Island. Your timing is plausible, but the event came from near Antarctica. We should be able to triangulate with H08 if you need an exact origin.

    Thanks for your vote of confidence on my own candidate site. Much appreciated. I’ll keep looking for ways to support other candidates with acoustic analysis.

    @Peter, @Victor:

    Sorry to distract. The timing was before WSPR went viral, just after @DennisW passed away. (And I’m sure he would have had some choice insight.)

    The NetSol complaint was a clear attempt at shutting down this blog. I redacted the name and reposted. Since then, that name is rarely invoked. It does stifle honest scientific debate, but the name can be referenced more openly on larger sites that are less vulnerable to SLAPP suits.

    Again, thanks to @Victor for keeping a serious forum.

Leave a Reply