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.
Mike
Perhaps this is a positive step towards a more scientific discussion of the flaws in using historical WSPR data to track aircraft.
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
@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.
Well, for future reference, at least now we know the sounds it makes when the wheels fall off.
@Mike Glynn said: I will be explaining why I believe this in due course.
We look forward to future comments.
@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.
@Mick Gilbert
You may also wish to review my “Limited Critique”, subject of
https://mh370.radiantphysics.com/2021/12/19/wspr-cant-find-mh370/#comment-33272
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”.
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.
@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.
Hello All, any update if ocean infinity or any other team is going to resume the search for MH370?
@Tanmay: Ocean Infinity has suggested that it might search for MH370 in early 2023 or 2024.
@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.
https://www.bbc.com/news/world-europe-62619397
@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.
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:
https://www.youtube.com/watch?v=ZDFHOW6DxdA
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).
https://uk.flightaware.com/live/flight/OEFGR/history/20220904/1302Z/LEJR/L%2057.65021%2021.12345/tracklog
@Paul Smithson: Which would of course suggest that the autopilot was set to hold a constant magnetic heading.
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.
@Victor – yes, that’s my point. Its consistent with default to HH norm at route discontinuity.
Indicative winds at altitude here:
https://classic.nullschool.net/#2022/09/04/1800Z/wind/isobaric/250hPa/orthographic=-0.18,46.34,1349/loc=19.117,56.200
@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.
@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.
@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.
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.
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.
@Victor and Co, Geoffrey’s at it again https://www.airlineratings.com/news/mh370-pilot-may-have-been-disoriented-new-study-finds/?fbclid=IwAR0I68maR27_jtLSYlCaMVczmn8-0bpBC_KtAhgXOvAiQhGTbSTiWmB9Qrc
@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.
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.
@Victor
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 https://youtu.be/uOcvmTLTtNE With the One Guy taking notes, while ones dishing out their thoughts. The last five words spoken, sound fitting.
Geoffrey Thomas (@airline_ratings) blocked me on Twitter. Sounds about right. As if blocking me will somehow prevent exposing the truth.
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.
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.”
@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.
Yes, isn’t it amazing that the first estimate invested with uncanny precision had been discarded so casually?
@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 https://www.dropbox.com/s/i8ka3hqv3scr7um/gdtaa%20takes%202%20%28l%29%20and%201%20%28r%29.png?dl=0
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.
@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?
@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: https://370Location.org
Huge thanks to all for this unique forum!
— Ed Anderson
Victor
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.
@Victor
Only if ya wanna listen, “Ma Name is Jeff” been chatting about ya.
https://143mysteries.com/ [VI: Link shortened to remove query data.]
@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!
@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.
@TBill: What would be your proposed search strategy?
@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.
@Victor
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.
@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?
@victor
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.
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.
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:
https://docs.google.com/document/d/1lZXurQKxyaD1JeMRRP3eDoXoc9cn1xIqTjPa9HsbKl0/edit#
@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.
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.
@Victor
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.
@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:
https://twitter.com/370Location/status/1569124467737239552?s=20&t=DTRF8BkDW4HImtoeVpDZxQ
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.
@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.
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.
@370Location
Is OE-FGR’s loss a useful characterisation candidate to you? Certainly a smaller, lighter, aircraft but much shallower sea (<250m).
@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. https://stationview.raspberryshake.org
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.
@Paul Smithson: What do you believe is the southern limit of what should be searched?
@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?
@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.
@Victor
@Mick
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.
@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.
@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.
@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.
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.
https://www.dropbox.com/s/zg2hizxyhxjfhfl/WSPR%20Azimuth%20Error.pdf?dl=0
@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:
https://mh370.radiantphysics.com/2021/12/19/wspr-cant-find-mh370/#comment-32836
In a comment following Mick’s I made this observation:
@Mick Gilbert: Active hams query the QRZ.com 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.
@370Location
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?
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.
@Viking:
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:
https://yepat.com/Citation-OE-FGR-crash-seismics
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.
@Victor,
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.
@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.
@370Location
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.
@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.
@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.
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, QRZ.com, and other similar databases.
https://www.airlineratings.com/news/industry-news/mh370-tracking-expert-responds-to-a-recent-paper/
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?
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.
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:
http://qrp-labs.com/images/appnotes/AN002_LT.pdf
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.
@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.
From https://www.airlineratings.com/news/mh370-pilot-made-many-turns-speed-changes-new-report-reveals/
“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.”
@sk999,
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.
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
Comments?
@Victor Iannello
I’d be inclined include further north than 32°S in the “priority” zone, say 28°S – 36°S.
@Mick Gilbert: The fit to the BTO/BFO and drift model data is weaker in 28S-32S, but your recommendation is noted.
@VictorI,
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.
Challenges:
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.
Advantages:
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.
@Victor
My inclination would be that longer glides may have been beyond Arc7.
@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.
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.
@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.
@370Location,
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.
@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.
@DrB
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.
@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.
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.
@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?
@Paul: Because the consensus debris drift analysis points to the S30-S36 range and the fuel models make S40 a stretch.
@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.
@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.
@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.
@DrB.
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:
https://www.frontiersin.org/articles/10.3389/fmars.2020.00305/full
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.
@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.
@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.
@370Location,
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.
@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.
@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.
@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:
https://mh370.radiantphysics.com/2021/02/10/preliminary-accident-report-released-for-sj182/#comment-30853
https://mh370.radiantphysics.com/2021/02/10/preliminary-accident-report-released-for-sj182/#comment-30864
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
@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.
@370Location,
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.
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.
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.”
https://agupubs.onlinelibrary.wiley.com/doi/full/10.1002/2017GL075891
@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.
@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?
@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.
@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. https://www.airlineratings.com/news/new-images-mh370-debris-revealed-wreck-hunter/
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.
@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.
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.
Here you can see the signals and hear a geologist analyze the data (click the video):
https://www.svt.se/nyheter/inrikes/svt-avslojar-tva-explosioner-intill-nord-stream
Sorry, nothing in English yet.
@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.
@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.
@370Location
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.
@All
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.
Two of the Baltic explosions look very similar as observed from Bornholm – probably the same type of bomb was used for both tubes:
https://www.reuters.com/world/europe/seismograph-spiked-twice-day-baltic-pipeline-leaks-germanys-gfz-2022-09-27/
This time in english.
Considering the dealy, I guess a mother ship was used to transport the bombs.
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:
https://www.bt.dk/samfund/se-billedet-eksplosioner-registreret-i-flere-lande
However it makes me very optimistic finding signal directly from the crash of MH370. At least if it happened near Christmas Island.
@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.
@Viking:
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.
@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.
@370Location
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)?
@Viking:
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:
https://drive.google.com/drive/folders/12ie9HrUEVjGQWC9MyYbolvuJ5bu-B4jG
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.
@Viking:
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.
@370Location
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.
@Victor
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.
@Viking: If you have acoustics results that support your proposed position, that would make a stronger case.
@Viking:
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.
@370Location
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.
@Victor
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.
@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.
@Viking:
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:
https://370location.org/2016/05/a-closer-look-at-the-lanl-report-candidate-event#aug2020update
It may be possible to filter coastal seismometers for even longer periods, something more like a tsunami. I’ll look into it.
@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.
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 ??
@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.
@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.
@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.
@Viking:
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.
Hello all.
I have not been active here for some time, not wishing to spin my wheels on the wspr issue. I am glad to see the group refocusing on actual analysis and reasoned discussion. (I have been working on putting EV charging in a large condo garage…many issues.)
I am planning to review the current state of the discussion and revisit my previous analyses. Unless something surprising pops out, I will not try to do any “new” studies. But I might see it from another perspective.
Sid
@Sid Bennett: Welcome back, Sid.
I realize this conversation has moved forward but I’d like to insert the following into the conversation re: RG’s “debunk” of my position regarding the large inherent errors that will occur when the actual positions of the WSPR TX and RX positions are not known or used.
RG replied with the statement that if the positions of the Transmitter, Receiver and aircraft do not line up then no anomaly can be present.
He has omitted the fact that he has only used Grid centerline positions for TX and RX positions, and so has simply assumed those anomalies occur on the extended centerline between the grids.
The whole point of my post was to say that this alignment of all three components does occur, but almost invariably occurs away from the centreline, which means that in all probability EVERY position he has used is subject to an azimuth error of some degree.
RG has also used a high granularity of 2-minute increments to define his track. All this does is increase the number and effect of cumulative errors present in the track.
Instead of blaming a hypoxic pilot for the chaotic track, the way to remedy this would be to ONLY use proper fixes, based on the recognized geometry limits of only using two bearing line fixes which are at least 45 Degrees from each other, and three bearing fixes, (The gold standard), that also have appropriate geometry limits. In the WSPR MK1 track RG has produced one three bearing lines that encompass a tiny arc of ~16 degrees!!
That and only use fixes every 20 to 30 minutes, otherwise you’ll end up with the dog’s breakfast presented in the latest iteration.
Thats how using fixes for position determination is taught in the real world.
@Mike Glynn: Azimuthal error is just one of the MANY reasons why the WSPR-tracking claims are false. Your frustration is understandable, especially since the supposed “debunk” never addressed the issue you tried to scientifically raise. That’s been standard operating procedure for some time now, and unlikely to change, despite Geoffrey Thomas knowing full well that he is propagating garbage.
On the other hand, you can sleep easy at night knowing that you’ve pursued the truth.
@Victor Iannello:
Hi Victor, I have spotted a strange problem:
If I click on https://mh370.radiantphysics.com/2017/04/16/atsb-denies-request-from-mh370-families-for-more-info I am redirected to https://pillsonlineservices.com/diazepam-about.html
And there is much more of this:
https://www.google.com/search?q=pills+site%253Aradiantphysics.com
(archived here: https://archive.ph/h8WUa#selection-1742.0-1782.0 )
Maybe your site has been hacked ?
Feel free to delete my comment once you have solved the problem.
Best regards, Peter
@Victor: I have looked into it again, and the google results now seem cleaner (less “infected” by the pills offers), although some of them are still there:
https://archive.ph/h8WUa/image (“Buy Modalert Uk online”).
So maybe Google has been hacked and not your site.
@Victor: I now do believe it’s a Google problem and has nothing to do with your site. The links in my 1st posting work correctly now.
But the google results for your site are still “infected” by pill offers,
see an archived version of a google search on your site here:
https://archive.ph/WufNL
sorry for misidentifying the problem in the first place (but chances that your site had been hacked seemed higher to me than an Google hack).
@Peter Norton: I have found the problem and I am working to eliminate the issue. Thank you for alerting me.
Renewed interest in cockpit video recording:
https://www.reuters.com/business/aerospace-defense/air-france-airbus-trial-13-years-after-atlantic-jet-disaster-2022-10-07/
@All: Malicious changes were made to the blog, causing the Google search engine to associate this site with online pharmacies. I believe I have removed all the troublesome content, and I have increased security. However, Google searches may continue to associate this site with the malicious sites until the Google web crawler revisits and updates its information.
If you see strange behavior, please let me know.
@Victor Iannello: Glad to hear you were able to eliminate the culprit.
Just checking: Are you sure the problem was on your end? Because the more I looked into the problem, the more I had the impression, that google search results were compromised rather than your site. Maybe google was (temporarily) hacked?
Maybe your issue is similar to one of these related issues with malicious google search result redirections:
“Clicking on a Google search result sometimes redirects to malicious site”
https://security.stackexchange.com/a/197288
“How scammers abuse Google Search’s open redirect feature”
https://nakedsecurity.sophos.com/2020/05/15/how-scammers-abuse-google-searchs-open-redirect-feature/
how Google’s search result redirection works:
https://webtrickz.com/how-to-remove-redirection-to-prevent-click-tracking-while-using-google-search
hope that helps.
@Peter Norton: I think the problem originated on my end (there was definitely malicious software) which provided bad info to Google. I expect that eventually the Google data cache will correct itself after the Google bot revisits my site, but I can’t be sure.
a question to the pilots among you:
Would it be possible to follow the curvature of the Northern Arc at a constant distance and what pilot inputs would be necessary for that ?
(Just as a theoretical question. I know that search missions don’t work that way.)
@Peter Norton: What do you mean by “Northern Arc”?
well, the northern part of the 7th Arc.
In essence what is labeled “Northern Corridor” in this image:
https://www.spacesafetymagazine.com/wp-content/uploads/2014/07/mh370-search-area.png
@Peter Norton: I know of no other way than entering a series of (lat,lon) coordinates as waypoints in a flight plan. Once the route is defined, an offset to the route (right or left) can be entered.
How does an FMS fly a DME arc ?
The DME has a (lat,lon) position, and the arc is a radius (in Nm).
So, presumably, you could set the sub-satellite point as a fake DME, and use the 7th Arc radius in Nm as the arc to fly, then vary that radius as needed to generate parallel tracks to the 7th Arc.
Would that work ?
Some approach procedures contain DME arcs that are coded into the procedure in the FMC’s nav database. As far as I’m aware, the FMC doesn’t allow pilot-defined DME arcs. As Victor mentioned, I think the only way to do it would be to enter a series of closely spaced pilot-defined waypoints.
@VictorI,ventus45,Andrew: Interesting. Thank you for your insights.
Boeing about to bleed dollars again!
“A federal judge ruled Friday that relatives of people killed in the crashes of two Boeing 737 Max planes are crime victims under federal law and should have been told about private negotiations over a settlement that spared Boeing from criminal prosecution.
The full impact of the ruling is not yet clear, however. The judge said the next step is to decide what remedies the families should get for not being told of the talks with Boeing.”
As reported by CNBC
@Barry Carlson: Thanks for that. The whole episode is a very sad one for aviation.
On Twitter, Ed Anderson referenced a new study entitled “How Barnacles Could Help Find Missing Persons Lost at Sea” was conducted at the University of New South Wales (UNSW). They studied growth rates and related oxygen isotopes to water temperature history.
From a news release about the study:
Suthers said Lepas are forensically useful for flotsam that was adrift for one to three months.
“Unfortunately for crash investigators, the new, faster Lepas growth rates suggest that the large (36 mm) Lepas found on the missing Malaysian Airline flight MH370 wreckage at Reunion Island—16 months after the aircraft was believed to have crashed in 2014—were much younger than previously realized,” he said.
“These Lepas probably settled on wreckage at least halfway across the Indian Ocean, and nowhere near the crash site.”
That’s interesting, Victor. Thanks for sharing.
@Paul Smithson: Yes, very interesting. I wonder why we hadn’t seen it before Ed’s tweet. It suggests that we should not try to relate barnacles to the crash site location.
And would offer and explanation to the warm water provenance no matter where the plane went down.
R.E. Lepas, I am now more confused than ever.
In this article from old friend Jeff Wise,
http://jeffwise.net/2015/08/11/listening-to-barnacles/comment-page-1/
he consulted Charles Griffiths of the U. of Capetown, who stated that: “… it is possible to identify this as being Lepas anserifera striata …”. As far as I can tell, this subspecies does not exist.
In Appendix 2.6A to the Safety Investigation Report, Joseph, Poupin of the French National Natural History Museum stated that: “… the specimens attached to the flaperon belong to the subspecies Lepas anatifera striata …” So it is now a subspecies (striata) of a species (anatifera) different from (anserifera) but at least does exist.
However, in the Marine biology article by Mesaglio et al. from the Unversity of New South Wales, we suddenly discover that “… their specimens are identifiable as Lepas indica …” – a completely different species – and, in fact, “Such specimens have historically been misidentified as L. anatifera striata … These misidentifications have persisted for more than 35 years, probably because Memmi’s work was only available in a Russian paper (Memmi 1982) that was difficult to access in the West at the time of publication.”
Sheesh. When experts can’t agree what they are looking at and are reduced to citing, among others, Darwin (1852) as an authority, it does not lend confidence to any analyses regarding temperatures or times of immersion of a piece of aircraft debris such as a flaperon, much less the location of either a lost mariner or MH370.
@sk999: When a piece of MH370 debris is found with lots of barnacles, it suggests that beach scavengers haven’t had the time to remove the barnacles. That’s about all we can learn.
More WSPR tracking nonsense. The authors try to make the case that there is statistical evidence of an aircraft perturbing WSPR signals. In this case, the aircraft is the Citation jet that crashed in the Baltic Sea that past September.
https://www.mh370search.com/2022/10/28/oe-fgr-case-study/
I still see no identification of WSPR signals with calculated Doppler signatures, which would be the most basic of experiments to establish the validity of WSPR tracking.
@Victor Iannello
More complete and utter nonsense. No sooner do these fellows make rules for this bunkum methodology of theirs than they then break them.
They’ve previously said,
“The issue of false alarms is very important and several measures can be taken to prevent false alarms:
1. The threshold of one standard deviation is one measure.
… ”
And here they toss that purportedly “important measure” out and go down to “0.75 standard deviation from the mean” as their threshold.
These fellows fail to grasp what an Intro to Stats student would understand; setting an anomaly threshold at ±0.75σ immediately turns nearly half (45.3 percent) of a normally distributed sample into anomalies. In other words, they are baking in a potential false positive rate of 45.3 percent, essentially turning their methodology into a coin tossing exercise. Their frankly bizarre attempt to use Receiver Operating Characteristics fails to address that matter.
And once again we are treated to the same amateur hour statistical analysis that has plagued everything else these fellows have produced. Despite stating that, “We discard candidate SNR anomalies based on a small count …” over 30 percent of their data sets have less than 10 values.
So how small is too small you would do well to ask. Well some of their data sets have just four data points. Which I guess answers the question, how small is too small? Three or fewer it would appear. It is absolutely farcical to suggest that you could reliably determine anomalies from such limited samples, even with truly randomised data. And to make matters worse, some of their samples capture data for anything up to ± a month.
And once again, these fellows show a 24 hour SNR curve that clearly illustrates that SNR tends to trend over time, essentially diurnally, and then they fail to accommodate that in their analysis.
And then, to cap it all off, despite noting on p.140 that there’s a convergence of air routes with their purported WSPR paths, they’re apparently happy to just shrug that off and claim that they’ve successfully detected the target!
This screenshot shows the other traffic in the vicinity of OE-FGR (circled) at the time the authors purport to have started “tracking” it.
https://www.dropbox.com/s/e4yfhjvqnbe6f6b/screenshot_20221029_151411_flightradar24.jpg?dl=0
It is nothing short of risible nonsense to contend that that particular target aircraft could have been detected in transmission variations between two amateur radio stations at pretty much any meaningful distance let alone half the world away.
@Mick Gilbert: Thank you for that comment.
If the authors want to prove that WSPR tracking is possible, there are scientific experiments that could be run with HF scatter off of aircraft in which continuous spectral data could be collected and compared with theory for scatter strength and Doppler shift. Once the theory was confirmed, it could then be extended to more challenging cases like the tracking of MH370 which involves signals with much lower SNRs over much long distances and in which only time averaged, single-valued SNR and frequency drift is available.
Those of us that have performed HF scatter experiments have demonstrated that the WSPR tracking claims are utterly impossible by many orders of magnitude.
The theory persists only because certain media outlets (such as Geoffrey Thomas’ Airline Ratings) are intent on propagating the myth. In the scientific community, and as the commenters on this blog understand, it is clearly fake science.
@Victor
I agree with you, but I see it as a little more politically complicated than simply Geoffrey Thomas’ fault. Believe it is safe to say some NoK support WSPR. Also those observers/investigators who agree with some of the novel features of the WSPR flight path (such as diversion towards Sumatra) find it has merit in their view.
Not to mention Malaysia’s demand that future searches must be justified, basically requiring scientific breakthroughs on the location question, which is probably at the root cause of why WSPR finds support. There is strong desire to see a future search, and for some, WSPR is one possible tool for justifying the next search.
As we “MH370-ists” all know too well, but perhaps some readers are less knowledgeable, Ocean Infinity has promised to do more searching free-of-charge to Malaysia. But Malaysia has a special rule (stated above) of no searching without a better search-area definition than we have from the Inmarsat/debris drift data.
@TBill: If WSPR supporters think that Malaysia will re-start the search by hoodwinking them with this WSPR nonsense, they are mistaken. In fact, they are providing Malaysia with ammunition to NOT search, as it is not hard for Malaysia to debunk this fake science and call it out for what it is. It is a step backwards, not forwards.
I would much rather be on the side of the truth, and persuade Malaysia (if it is possible) with facts, not fantasy.
Someone wrote: “The WSPRnet data set is noisy, but with care it is possible to extract useful information.” This statement is simply not true… not even a tiny bit. Forget the statistical rabbit hole. Fundamentally…the concept violates basic information theory by many orders of magnitude. There is no possible discernable MH370 signal burred in any WSPR data.
@ALSM said: There is no possible discernable MH370 signal burred in any WSPR data.
I agree.
Yet, the uninformed say otherwise, and label the informed commenters on this blog the “detractors” without regard to the scientific reasons we KNOW WSPR tracking is utter rubbish. What a joke.
It took something like 142 pages of data to try to establish the latest WSPR tracking analysis. How could the average “joe Bloggs” begin to make heads or tails of all that?
It reminds me of an old saying about a back street tailor [in the city of your choice], trying to sell a bolt of fabric . . .”never mind the quality, feel the width”.
The ADS-B records must help a great deal to be very selective with the data points chosen, and the scrambled jets must have been a lot closer than 19nmi to have a view in the cockpit.
Considering all the other aircraft in the air in the general area, the average “joe Bloggs” ought to conclude whole basic concept is implausible.
I may be reading this incorrectly but the dismissal of the “Null Hypothesis” here, that WSPR detections that are not on the centreline of the Tx and RX WSPR stations are not valid, appears to be some sort of strange defence that the centreline between the two respective Maidenhead Grids is the only place a valid link can propagate.
Richard claimed to be using actual TX and RX station positions but there is no evidence presented to back that up. All the claimed links I checked were simply on the grid square centres. I didn’t find any that were position anywhere off the centrelines.
A small azimuth error over 25,000 km becomes a sizeable lateral error. Combine it with “position fixes” that do not obey the required geometry, the use of a fixed groundspeed and a position taken every two minutes and you have a massive cumulative position error present in a very short space of time.
@All: Simon Maskell has provided this statement to me, which I include in full with his permission:
The analysis involves a Cessna over Europe, rather than a 777 over the Indian Ocean, and, while promising, as a result of the relatively small dataset considered, is not definitive. However, it is also clear to me that the analysis implies it is probable WSPR can detect the Cessna better than can be explained by chance. It is therefore likely that an algorithm could exist that can track the aircraft’s location by processing the detections derived from the raw WSPR data.
I’ve taken a preliminary look at the reported results from the WSPR paper. Looking at the results from the Tables 6 and 7, I note that the anomaly threshold was SD=0.75, which means that if there was no effect from the plane and SNRs are normally distributed, you’d expect 45% to be anomalous and 55% to be normal. The authors’ observed true positive rate is 26/50 = 52%, versus the expected rate of 45%, so the observations are slightly better than chance. On the other hand, the observed true negative rate is 84/133 = 63% versus the expected rate of 55%, so again observations are only slightly better than pure chance.
So the data shows that AT BEST, there is a VERY weak correlation between SNR and the position of the plane near a great circle connecting a WSPR receiver and transmitter. There’s certainly not enough of a signal (if it exists at all) for reconstructing the path of MH370.
The fact that the data even slightly beats pure chance when there is no physical basis for it means the testing methodology introduced bias, whether intended or not. For instance, the time interval for the test was around 22 minutes, yet the ADS-B data exists for 4 hours 48 minutes. How was this interval selected? What does the data say for the other 4 hours 26 minutes? What is the basis for extending the low count data sets by including data from other days? Is the observed population of SNRs normally distributed? If there truly is a measurable effect due to plane position, how is the effect of other planes near the great circle path included?
Godfrey has (finally) tried to make a proper false positive test by taking his set of WSPR links for 11 time slots and determining the rate of positive and negative detections for locations different from the actual aircraft positions. The locations he uses are the aircraft positions at times +/-2, 4, and 6 minutes different from that of a WSPR slot itself. Since the aircraft does not move all that far in 2-6 minutes, if a WSPR link is aligned with the direction of travel, it can align with the aircraft location multiple times. Is this a problem? Not really – at most, one would need to adjust the number of degrees of freedom in any error analysis. However, Godfrey takes the unjustified step of discarding the duplicates, retaining only the links that cross the true aircraft position. Since the majority of the discarded links are positive detections (21 out of 28), this means that the retained links are mainly true positives while the discarded links would mainly have been false positives had they been kept. This process introduces a bias such that the WSPR hypothesis will alway be favored, even if the null hypothesis were true. A better designed test would have avoided this issue altogether.
Regarding the Maidenhead azimuth errors, I computed the rms cross-track error for each of the 183 links in the data table. The median rms is about 2.1 nm, which is the maximum cross-track error allowed by Godfrey; the peak-peak width of the distribution is about 4 to 5 times the rms. A guess is that about half the links would fall in the 2 nm radius circle if one had precise geographic coordinates of both stations. [This analysis ignores the fact that rays do not follow precise great circles (as assumed by Godfrey) due to the spheroidal shape of the earth, introducing cross-track offsets of order 10-15 nm and thus making the entire exercise total nonsense.]
@Victor Iannello
Victor, doubtless Professor Maskell is familiar with the traditional academic point system for classifying the utility of a methodology based on its area under the ROC curve results;
.90-1.0 = excellent (A)
.80-.90 = good (B)
.70-.80 = fair (C)
.60-.70 = poor (D)
.50-.60 = fail (F)
With an overall AUROC of 58.66% the authors get an F. And let’s be clear, that’s an F under the best possible circumstances, specifically a non-blind test over a very short period of time.
We might well ask whether anyone in any position of authority on either side of the negotiations to fund a renewed search would treat a methodology with essentially no predictive capability seriously. Who could possibly say that a methodology with such manifestly poor results clears the “credible” part of the 2017 “credible new information” test?
@Mick Gilbert said: Who could possibly say that a methodology with such manifestly poor results clears the “credible” part of the 2017 “credible new information” test?
Not me. Others can speak for themselves.
@sk999: Thanks, Steve. I knew there was a small bias introduced because the physical principles say there should be no statistical correlation, no matter how small. Your explanation sounds very reasonable and is not something I had considered.
@sk999: So by my count, with a threshold of 0.75, after including the discarded data, the True Positive rate remains at 52% and the False Positive rate increases from 37% to 44%, versus a rate of 45% of detections due to chance (assuming that SNRs are normally distributed). The fact that the False Positive rate is closer to the expected rate due to chance give us more confidence in these statistical results. But now the True Positive rate and the False Positive rate move much closer together (52%-44% = 8%).
The utility of the GDTAAA plane detector just went from very weak to almost imperceptible, just as the physics would predict.
The prospect of using this detector to reconstruct the path of MH370 is once again proven to be junk science.
@WSPR:
It’s astounding that the GDTAAA proponents are still trying to squeeze stasticial significance from single cherry-picked events using mind-numbing 142 page dumps of random graphics. There are too few samples with confounding errors of other aircraft flying. Their undocumented algorithm keeps changing, avoiding critique like a game of whack-a-mole.
In Dec 2021 I presented here a simple analysis using some 13 billion data samples from millions of WSPR contacts compared with thousands of known ADS-B GPS locations from aircraft at cruise altitude.
https://mh370.radiantphysics.com/2021/07/23/italian-satellite-may-have-detected-mh370-floating-debris/#comment-32779
The authors responded with gibberish:
https://mh370.radiantphysics.com/2021/12/19/wspr-cant-find-mh370/#comment-32810
That was data from a single day, and showed no statistical shift in frequency drift related to the distance of an aircraft from the signal path. Analysis of signal strength also shows no statistical shift. If that’s not convincing enough, we have the entire history of WSPR vs ADSB with trillions of samples to analyze.
Sadly, that would be pointless. The authors won’t run a study that shows the fallacy of their method, and they won’t be convinced by any “detractor” who does it for them.
It’s truly sad to see what a boondoggle this has been. Not only the fiasco of false hopes from awful pseudoscientific manipulations that still persists, but the countless collective hours wasted by so many of us debunking the fraud.
Why is this method slightly better than pure chance?
Because n is too low ?
@Peter Norton
Peter, think of a form of bias and it is probably at play with this latest nonsense. On top of, or underlying, that (take your pick) there is, as we have seen repeatedly, just plainly delinquent scholarship.
Despite noting that, ‘The WSPRnet data set is noisy, but with care it is possible to extract useful information.‘ there is a manifest lack of care on display in the way these clowns handle pretty much ever aspect of the data gathering, analysis and reporting.
It is a bit difficult to know where exactly to start but let’s go with this one first; failing to apply their own rules on data cleansing. Despite stating on page 4 of this latest comic book that,
‘Some WSPRnet stations are mobile in vehicles, ships or balloons. The SNR and drift anomalies over a 6 hour period between mobile transmitters or mobile receivers are discounted. These mobile stations can be detected from their call signs or the fact that the Maidenhead Grid locator changes during a six hour period.‘
have a look at Table 20, specifically the tx_loc data for OZ4PAT. You will undoubtedly notice what I suspect you could train a chimpanzee to spot; the tx_loc is changing over time. It is pretty clear that the Tx is moving; along the coast of France in the Bay of Biscay, to be specific.
The OZ4PAT error is particularly egregious because it evidences their failure to apply not only their own mobile station rule but also their six character Maidenhead Grid location rule, articulated on page 3. Note that all of the tx_locs for OZ4PAT in that 17:30 UTC ±1 day ±3 hours (incorrectly labelled as just ‘±3 hours’ in the comic) are four characters. When these fellows have amended those to six character locators, purportedly ‘from the registration documents of each radio station and cross checked against amateur radio call sign databases and station web sites‘ they have placed the Tx in Denmark (grid JO55tn).
A couple of things to note here. First they have just blithely converted different four character IN grid references (Maidenhead grid IN cover the west coast of France, northern Spain and the Bay of Biscay) to a single JO grid reference (a shift of some 1,300 kilometres between the two) without so much as a ‘Mmm … that looks odd.’ Second, and this goes to their purported cross-checking, there is a disagreement between two of the three commonly used amateur radio databases (qrz.com, qrzcq.com and hamcall.net) as to the registered location for OZ4PAT. qrz.com (generally the more accurate) places it at JO65gr whereas qrzcq.com places it at JO55tn (hamcall.net has no listing). JO65gr matches the user supplied address listed on qrz.com.
For fear of stating the blindingly obvious, the OZ4PAT spots at 17:30 UTC and the 17:42 UTC should never have made it into the analysis. Both of those spots were recorded by the authors as being anomalous and both were accordingly (but erroneously) tallied as True Positives. When you remove them from the analysis we’re now back to a True Positive Rate of 50 percent. AUROC would fall even further through the FAIL range to the point of being deemed useless.
And, I suspect that there are probably a couple of other mis-mapping instances that will drag their performance down even further.
Of course, shitty work on the authors part is pretty much par for the course. The question that now comes to the fore though is how exactly is this work being ‘independently reviewed’. What level of diligence is being applied in that realm such that egregious and manifest errors of this sort are just being waved through?
@Mick Gilbert: thank you for your great demonstration, you picked some really telling examples
@Mick Gilbert: Thanks for catching that. When your claimed usefulness is only marginal, mistakes like that become critical.
From physical principles, what is claimed is impossible. Then it becomes an exercise in finding the errors or false assumptions that create the small shift from pure chance.
Also, the analysis should have been done for a longer part of the flight. Since the effect is so marginal, more data would be useful.
@Victor Iannello
Hard to believe that two authors could make an error like that. Manifestly there is no cross-checking or independent validation of their work, something that you would think dual authorship would naturally lend itself to.
As alluded to above there appears to be at least one other, similar but less egregious, error that will remove two more True Positives and one True Negative from the data. I’m working on confirming that but if correct, that will bring their TPR back to <48 percent and would pull their AUROC back to below 0.5.
Meanwhile, we're in the news! https://www.airlineratings.com/news/critics-of-wsprnet-tracking-mh370-been-answered/
Apparently neither the author nor the reviewer have seen this – https://darwin.unmc.edu/dxtests/roc3.htm
or this –
https://acutecaretesting.org/en/articles/roc-curves-what-are-they-and-how-are-they-used
or this –
https://www.researchgate.net/figure/ROC-curve-and-ROC-score-categories_fig3_322161039
@Mick Gilbert: Geoffrey Thomas fails to understand the basics of good journalism. He accepts EVERYTHING the WSPR proponent says on its face value without any understanding of what he is publishing. At least he is consistent.
And of course we have the clueless that call this “courageous reporting”.
@Victor Iannello
Exactly Victor. For instance, what sort of gibberish response do Steve’s posting is this?!
‘Steve Kent omits to state that the average number of SNR anomalous WSPRnet links at each point in the cruise of the aircraft is 4.3 and not a single link as he implies. When three or more anomalous links intersect at the aircraft position, as is the case at each point in the time frame analysed, then the cross-track error is minimal.‘
For starters, Steve said nothing that would imply there was only one link at each point in the cruise.
But more to the point, who could possibly believe that cross-track errors somehow magically eliminate each other/correct themselves at the fix?! You might think that an aviation reporter might grasp that that makes no sense.
And, of course, no rebuttal from the author would be complete without a liberal sprinkling of bullshit. Compare the statement, ‘When three or more anomalous links intersect at the aircraft position, as is the case at each point in the time frame analysed …‘ to Figure 5: GDTAAA WSPRnet links at the OE-FGR position at 17:24 UTC Local View. I’m pretty sure that 17:24 UTC is one of the points in the time frame analysed and none of the three links intersect at the aircraft position.
@Mick Gilbert: There are three “links” passing within 2 NM of the plane. I think he is calling this an intersection.
@Victor Iannello
Ah, so having soundly trashed physics the author now turns his attention to geometry.
@Mick Gilbert: As I’ve said many times before, there are three camps: the informed, the uninformed, and the WSPR-proponents. At this point, I see little movement between the camps. In this sense, Geoffrey Thomas’ articles are inconsequential. It is interesting to see which camp people fall into.
Mick Gilbert,
While your links to the “academic point system” for evaluating the significance of an AUC (area under curve) value for a ROC test are entertaining, they leave open the question of whether one can compute a more formal estimate of the significance. This question turns out to be a bit of a challenge, and most discussion of ROC and AUC make no mention of confidence intervals. The following link gives an equation that works under certain assumptions:
https://ncss-wpengine.netdna-ssl.com/wp-content/themes/ncss/pdf/Procedures/PASS/Confidence_Intervals_for_the_Area_Under_an_ROC_Curve.pdf
A more complete analysis was not done until 2004 by Cortes (which I will not link to but you can find it online.)
When I compute the difference in the “true positive” and “false positive” rates for the OE-FGR flight with a threshold of 0.75 sigma (i.e. just one point on the AUC), I get a difference of 9%, which is significant at about the 1-sigma level, i.e. it should excite no one.
@370Location
Ed, you said on:
November 1, 2022 at 4:29 am
“@WSPR: It’s astounding that”
and
“the countless collective hours wasted by so many of us debunking”
Agreed.
Doubtless to the astonishment of nobody who has been following this from a clear-minded, scientific perspective but to the annoyance of the ‘cheer-squad’, further analysis of the authors’ latest work has turned up further plainly egregious errors.
To recap, the ‘case study’ as presented purportedly showed that the methodology had a True Positive Rate of 52 percent; 26 True Positives to 24 False Positives. I have previously demonstrated that the authors failed to apply their own data cleansing rules to WSPR spots involving Tx OZ4PAT. Not only was that Tx mobile (grounds for elimination on that basis alone) but the authors managed to ‘relocate’ it from its true position in the Bay of Biscay all the way to Denmark, a distance of some 1,300 kilometres! Thus, the OZ4PAT spots at 17:30 UTC and 17:42 UTC, both recorded by the authors as being anomalous and both erroneously tallied as True Positives, should never have made it into the analysis. Removing them brings the True Positive Rate to 50 percent; 24 True Positives to 24 False Positives.
Consistent with the authors’ generally poor standards in data gathering and cleansing, there are at least three other instances where they have incorrectly located the Tx. In two of those three instances, the mislocation has occurred as a result of the authors’ attempt to refine the station’s reported four character Maidenhead grid reference to a six character reference. In both instances, in what should have been picked up by the most rudimentary of data validation checks, the authors have mis-mapped the Tx by moving it to a six character Maidenhead grid location that does not fall within the reported four character Maidenhead grid. By any objective measure that is extraordinarily sloppy work; what I would categorise as a clear cut case of a delinquency of scholarship.
In the case of Tx G8LIK, the authors have arbitrarily relocated the tx_loc from IO93 (as reported) to IO92fs, a move of some 90 kilometres to the south! On the basis that there is disagreement between the three commonly used amateur radio databases (qrz.com, qrzcq.com and hamcall.net) as to the exact location G8LIK, it is difficult to see how the authors actually have, as they claim to have done, ‘cross checked against amateur radio call sign databases and station web sites’. A modicum of investigation reveals that G8LIK is almost certainly located in IO93 as reported, most likely at IO93cn.
That, of course, means that the claimed anomalous spot between G8LIK and OZ1AAB at 17:24 UTC is nowhere near the target aircraft’s location. The erroneously tallied True Positive associated with that spot should be removed from the analysis. Doing so brings the True Positive Rate down to 48.9 percent; 23 True Positives to 24 False Positives.
Tx NY2ES is a similar story. In this instance the authors have arbitrarily relocated the tx_loc from FM29 (as reported) to FN30rx, a move of some 115 kilometres to the east! Again, there is disagreement between qrz.com, qrzcq.com and hamcall.net as to the location of the transmitter but had the authors actually checked ‘the registration documents of each radio station’ as they claimed, they would have found that the FCC registration details and the qrz.com records agree. The tx_loc is FM29dm, not FN30rx.
Therefore, the anomalous spot between NY2ES and WD4AH at 17:36 UTC does not pass near the target aircraft’s location at all. Again, the erroneously tallied True Positive associated with that spot should be removed from the analysis. This brings the True Positive Rate down to 47.8 percent; 22 True Positives to 24 False Positives.
There appear to be a few further errors relating to mis-mapping and our old mate, azimuth error, once again raises its head (as anticipated by Mike Glynn). Somewhat astoundingly, but again unsurprisingly, despite knowing that they must contend with station mapping errors of up to a couple of kilometres within a six character Maidenhead grid, the authors have elected to use spots with Tx – Rx distances as short as just 41 kilometres (I kid you not).
@Mick Gilbert: Thanks for your additional digging. I also found the G8LIK “anomaly” in which the Maidenhead grid locators used for the cross-track calculation (IO92fs), the WSPR database value (IO93), the self-assigned value in qrz.com (IO93vi), and the value derived from the qrz.com address (IO93cn) all disagree.
I am quite confident that if the test was properly conducted, the investigators would find that the presence of an aircraft near the “link” has no effect on the SNR. You are showing that whatever small predictive capability is claimed is due to errors. That makes sense.
In comment made Oct 1, 2021 (found under an article posted on Sep 9, 2021), Richard wrote:
“The maximum error resulting from the quantisation of the 6 character Maidenhead Grid references is 18 nm.”
In an article from May 4, 2021, Richard wrote:
“The GDTAAA does not fulfil the GADSS requirement of tracking an aircraft to within 6 nm …”
So why is he now able to use GDTAAA to track aircraft to within 2 nm? In the Oct 1 comment, he additionally wrote:
“I use only WSPR signals with a propagation distance of greater than 1,000 km and where the transmitter and receiver stations are not in the same 2 character Maidenhead Grid reference.”
In the OE-FGR case study, 85% of the links have distances between the transmitter and receiver stations that are less than 1,000 km. For 23% of the links, the 2 character Maidenhead Grid references are the same.
The essence of good science is reproducibility. Reproducibility includes using the same selection criteria and methods each time. It means not changing the threshold for what constitutes an “anomaly” from 0.25 to 0.75 to 1.0 sigma, depending on which set of data are being evaluated. It means not changing the criterion for a minimum distance between tx and rx from 1000 to 500 to no minimum at all. It means not changing the method for computing a standard deviation from evaluating all links at a single time to evaluating a time series for a single link. It means not changing the criterion for what constitutes a candidate progress indicator from a circle of 4 nm diameter to a box 2 nm on a side to not saying anything at all and leaving the choice in the eye of the beholder.
What is the significance of a value of 59% for the AUC in an ROC test? It is not stated. Without a proper confidence interval this value is meaningless, notwithstanding any comments to the contrary by Prof. Maskell.
Here is an analysis of 97 studies where a strong positive detection (greater than 95% confidence) was claimed. Attempts at replication failed about 2/3 of the time.
https://www.scienceintheclassroom.org/research-papers/good-science-reproducible-science
@sk999 said: The essence of good science is reproducibility.
I wrote a python script to try to reproduce his results. For consistency, I used the same Maidenhead grid locators for WSPR stations, even when those are suspected to be wrong. I also used the spherical earth model for calculating the cross-track errors, which also introduces errors.
The list I generated of WSPR “links” within 2 NM of the aircraft did not exactly match his list. Perhaps he did not properly interpolate aircraft position every two minutes on the minute as would be required by his criteria. On the other hand, in those cases where there was a match between his proximate links and mine, our calculated statistics using the +/- 3 hours of data were identical, so at least that part checked.
As an example, at 17:30, id = 4663339239, F4ASK (IN96ij) is transmitting to DF1QQ/SDR (JO32sd), where the Maidenhead grid codes are from his file. The plane at that time is at (57.32695,19.8193) over the Baltic Sea. I calculate a cross-track error of 4.0 NM, while he calculates something less than 2 NM. But it’s also interesting that IN96ij is in France, and JO32sd is in Germany, and the aircraft was over the Baltic Sea, which means he is including scatter with a path that would have to start in France, pass over Germany, impinge on the plane and then backscatter back to Germany. And somehow that scattered signal is sufficiently strong to cause the received signal to strengthen by 0.82 sigma! That’s craziness.
On a positive note, I am very impressed by how quickly the WSPR.live server returns the results of automated queries. It took much less time to find the proximate links and then calculate the statistics than I thought it would.
@sk999
Steve, doubtless the authors would try to dress up the ever shifting sands of their methodology as kaizen. Anyone who has worked with quality systems would recognise what they are doing as anything but. It’s at best cherry picking, at worst just a rambling shambles of whatever grabs them on the day.
No sooner do these fellows establish a rule for data gathering, cleansing or processing, then they trash it; as illustrated above with OZ4PAT, sometimes they do so in the space of the same paper! It is farcical.
More to the point though, it demonstrates that nothing stated in any of their work can be taken at face value. And that presents as a significant issue for any review.
Regards their AUROC score, be it their claimed 58.66% or 61.02%, what needs to be remembered is that those scores are the product of numerous reported True Positives that actually aren’t True Positives at all. The actual True Positive Rate for that particular “case study” sits sub-47 percent, not far from what you would expect when you set an anomaly threshold of ±0.75 σ. That corrected TPR would bring their true AUROC down accordingly.
Of course, a significant part of the problem is that they manifestly do not understand what their ridiculously low minimum sample size does to determining x̄. Applying the usually required confidence level of 95%, their minimum sample size of just five will generate a standard error of the mean that routinely sits at 0.85-0.9 σ. In other words, the error bars on x̄ swamp their detection threshold. That, at least in part, explains the variation between true TPR and the expected 45 percent generated by the ±0.75 σ anomaly threshold.
@Victor Iannello
@Mike Glynn
As anticipated, the authors inability to stick to their own rules regarding Tx – Rx distances again throws up azimuth error problems.
Take DL2NED – HB9TMC, where the recorded spot at 17:26 UTC is deemed to be a detection. The problem for the authors here is HB9TMC; the station’s actual location is a good 2.3 kilometres east of the JN46lj reference point. That pushes the DL2NED-HB9TMC ray path 5.6 nm to the west of the target aircraft when recording for the spot started, with the aircraft constantly moving away from the ray during the recording period.
That’s another true positive that needs to be removed from the data – TPR now back to 46.7 percent.
@George G, Andrew, Victor, ALSM.
SJ182: Sriwijaya Air 2021 final report submitted to the Indonesian parliament:
https://www.thejakartapost.com/indonesia/2022/11/04/sriwijaya-air-2021-crash-due-to-mechanical-problems-final-probe-report-says.html.
No mention of complicating conditions or A/P disengagement. It will be interesting to read what they think of the recovery actions attempted.
Earlier discussion:
https://mh370.radiantphysics.com/2021/02/10/preliminary-accident-report-released-for-sj182/
As memory joggers, possibly relevant and posted among others by @Andrew, pages 4-59, 60:
http://www.tc.faa.gov/its/worldpac/techrpt/ar03-72.pdf
As to others, @Andrew’s post:
https://mh370.radiantphysics.com/2021/02/10/preliminary-accident-report-released-for-sj182/#comment-30799
Audio alerts, again @Andrew:
https://mh370.radiantphysics.com/2021/02/10/preliminary-accident-report-released-for-sj182/#comment-30799
Also conceivably relevant, though unlikely: https://www.reuters.com/business/aerospace-defense/us-requiring-inspections-wire-failure-boeing-737-classic-planes-2021-05-14/
Presumably the final report will be listed below when released publicly:
http://knkt.dephub.go.id/knkt/ntsc_home/ntsc.htm
http://knkt.dephub.go.id/knkt/ntsc_aviation/aaic_case.htm
@David: Thanks for the update. It looks like a partial failure of the autothrottle combined with cockpit complacency leading to an incorrect recovery. All very unfortunate.
@Mick Gilbert: As we’ve said now for months, independent of all the physical and numerical errors and misconceptions, a detector based on looking for values greater than 0.75 sigma is on its face doomed to failure, as we’d expect a false positive rate of around 45%. What is referred to “anomalous” is not anomalous.
The whole statistical line of analysis is 100% bogus. WSPR data contains nowhere near the SNR or frequency drift info required to detect an aircraft in the SIO. It’s a complete waste of time to go through all the statistical analysis the authors attempt. No matter the sample size, if there is no info in the data, the stat’s are irrelevant.
BTW…Applying rules like a “0.75 sigma threshold” to power ratios (S/N) expressed in dB seems wrong to me. Shouldn’t the statistics be applied to the powers (or power ratios) ratios? Not the log of the ratios?
@airlandseaman: For me, it’s an exercise in finding errors or the biases in the analysis that may lead the uninformed to believe there is a correlation where we know there is none.
I think using SNR is fine because received power (with an assumed noise level) would favor the high power signals when calculating mean and SD because the range of powers is so high.
For those interested in looking at the details, I am sharing a file that shows the WSPR spots for which the plane is proximate to a TX-RX path during cruise. To be consistent with the WSPR-tracker, I assumed:
1. The criterion for proximity to a link starting at T0 is for the cross-track error of the plane’s position to be within 2 NM at T0.
2. I used the same Maidenhead grid assignments as the WSPR-tracker without correction.
3. I assumed the paths were great circles, i.e., the earth is spherical.
4. To calculate the z-score of the SNR, I used the population of spots that fell in a +/- 3 hr window centered on T0.
With these assumptions, I found 25 WSPR spots there were proximate to the plane during cruise. Of these, 1 spot had no other spots in the +/- 3 hour window, and so no statistics were calculated. The other 24 spots had populations of 5 or more. Using a threshold of |z| > 0.75, 13 (54%) would be true positives and 11 (46%) would be false negatives, where z is the number of standard deviations from the mean.
Of interest is the average and standard deviation of the z for all the proximate spots. If the proximate spots showed no effect due to the proximity of the plane, you’d expect a mean of 0 +/- 0.2 (for n=24) and SD = 1 since the z values are already standard values. The actual values are mean = -0.099 and SD = 1.04. That means the presence of the plane has no statistically significant effect on SNR.
Please let me know if anybody finds a fault in either my calculations or my logic.
Victor: Re my comment “…analysis is 100% bogus…”, to be clear…I was referring to the GDTAAA author’s analysis, not the excellent work you, Don, Steve, Mick and MANY others have done to show that it is bogus.
@airlandseaman: Understood. We are trying to make sense out of the senseless, and have probably spent more time than is justified.
@David,
@all,
The KNKT site server has been down every time I have checked.
Reuters:
https://www.reuters.com/world/asia-pacific/faulty-automatic-engine-throttle-system-poor-pilot-monitoring-contributed-2021-2022-11-10/
https://knkt.go.id/news/read/final-report%20of%20sjy182%20accident
@George G
Thanks for posting that link George.
Hey Victor I propose a search area recommendation at the area between 43°S and 45°S based on floating debris spotted by satellites and fact sea surface vessels were looking around that vicinity before the search was shifted further north I base my findings of the fact AMSA were looking for any other leads other the the main satellite data you have and taking into account an alternative explanation if there were any intiial errors that investigators failed to take into account with the BTO and BFO values I’m well aware it’s beyond the 7th arc and that it doesn’t have the maximum fuel endurance if it went over the Malacca Straits I’m one of the MH370 followers that question the accuracy of the alleged route shown on primary radar I’m under the assumption MH370 took a completely different route by trying to head back to Kuala Lumpur in a impending disaster and the pilots succumbed to hypoxia before descending and that it took a straight course into the Indian Ocean followed by fuel exhaustion at precisely the area the search intially focused on without making any subsequent maneuvers over the Straits if you calculate where MH370 was at IGARI all the way into the SIO then it makes sense.
@Mike R: You may or may not know that Stuart Lichty produced one of the images that located the debris in the area in the SIO that you cite. @Stuart is a commenter on this blog. You can search for his comments on this blog, although his explanation for a location of debris far from the 7th arc is not “conventional”.
Out of curiosity, why do you dismiss the satellite images of debris near the 7th arc, as described in this blog article?
https://mh370.radiantphysics.com/2021/07/23/italian-satellite-may-have-detected-mh370-floating-debris/
@Victor,
@Mike R,
Follow the trail of “breadcrumbs” …. and you will realize flight MH370 aircraft is not resting in a hangar in Kazakhstan, nor is the wreckage scattered in the ocean off the coast of Vietnam, at least not in the universe that we presently occupy.
Hopefully our civilization will be able to work through the differences we are having and move forward. The reality is that we have “neighbors” whom have shown interest in our activities and impact on the environment. Are we worthy of expanding our influence beyond our planet? Judging from our history, there is reason to be wary of our species and view us as a threat.
@Mike R
I am a little surprised that you reveal 45s as your site nomination. Believe that would suggest radar data and Inmarsat data are both wrong. Another weakness is debris drift would go to Australian coast, and would be inconsistent with actual debris findings. I do know some adherents of 45s say Australia did actually recover MH370 debris that was not properly identified. You are entitled to your opinion, but here on Victor’s blog we mostly believe the radar/Inmarsat data are largely correct, based on our experience working intimately the data for some 8 years now. In other words, the so-called “official” interpretation of the data as per ATSB/Malaysia reports and search area definition.
I’m not necessarily dismissing the satellite images at 34°S there’s no doubt the ocean is full of floating debris none of them can be conclusively confirmed that the debris field were from MH370 I cite two possibilities either MH370 was missed and the search did not detect the wreckage or MH370 could be more then hundreds of miles south from the 7th arc it’s not guaranteed that we’ll find the plane in the current search area so I’m considering an alternate explaination in the event that MH370 is not found in the current 7th arc vicinity but I’m also more then willing to try the area you recommended in one of your previous blog post back in July 2021
@Mike R,
It has been pointed out before, but bears repeating; Any debris in sat images drifted to where it was seen. Many people who propose a far southern crash site based on sat images and surface searches believe that’s where the plane crashed and where a search should be conducted. Suppose MH370 debris had been spotted in an image taken two weeks after the crash, and was confirmed a week after that by surface search.
The surface search would take place where the debris was expected to drift from the image site. Two weeks of modeled reverse drift would be needed to estimate the crash location. In general, those sat images were targeted far south because that’s where debris was expected to drift from along the 7th by the time the photos were taken. So, those who believe sat images farther south show MH370 debris are essentially agreeing with a crash site near the 7th Arc.
@370Location: Using the CSIRO drift model, there is no location on the 7th arc from which debris would drift to that location in that timeframe.
@VictorI:
OK, strike the last sentence of my comment. Detailed drift wasn’t available in the first weeks, and some imagery may have been based on FE past the 7th Arc. My main point is that reverse drift over time from an image location would be needed to determine a candidate crash site. Several vocal advocates ignore that fact.
A recent European proposal to reduce the number of pilots required for commercial air transport has generated a lot of online discussion in recent days. For those who are interested, the working paper that was presented to ICAO may be accessed via the link below.
ICAO Working Paper: AN APPROACH TO NEW OPERATIONAL CONCEPTS INVOLVING EXTENDED MINIMUM CREW OPERATIONS AND SINGLE-PILOT OPERATIONS
https://www.icao.int/Meetings/a41/Documents/WP/wp_101_en.pdf
@Andrew: The cruise portions of cargo flights seems like a sensible place to test this. Unfortunately, incidents are such rare events (and even more rare during cruise) that it might be hard to gather enough data to make an assessment of the safety.
What are your thoughts?
@Victor
It won’t be easy for the manufacturers and regulators to prove an equivalent (or higher) level of safety for single-pilot ops. I think the industry already has plenty of data on the types and frequency of in-flight events that might occur. What’s missing is data on the handling of such events by pilots operating alone for hours at a time. Given the rate of critical failures is so small, I think the regulators will need to rely on simulator studies specifically designed to collect the necessary data.
One of the major issues is pilot incapacitation, which could occur for any number of reasons. There will need to be some system to monitor the alertness and ‘well being’ of the operating pilot, and to alert the resting pilot to return to the flight deck if necessary.
Some sections of the industry are already working towards minimum crew operations in the cruise for newer generation aircraft. That work is being driven primarily by Airbus with the A350, which is already capable of autonomous automatic emergency descents, complete with TCAS manoeuvring if required. Airbus is also working on a new cockpit concept to enable single-pilot operations in future generations of aircraft (Autonomous flight:
Leveraging machine learning to enable self-piloted operations). Minimum crew operations in the cruise might well happen by the end of this decade, but I think that end-to-end single-pilot operations are a very long way off.
The airlines and manufacturers will no doubt claim this proposal ultimately enhances safety, but let’s be clear: the driving factor is cost. Even if they do manage to prove an equivalent level of safety, it will be interesting to see if there is any cost benefit once all the factors are taken into account. Moreover, will the flying public be convinced that single-pilot operations are safe?
@Victor,
@Andrew,
American Eagle (Envoy) Flight 3556
The captain became incapacitated shortly after taking off from ORD Saturday night.
Unfortunately he died later in hospital.
Unfortunate Timing, re your discussion.
Source: https://www.airlive.net/an-american-eagles-pilot-died-after-fainting-during-take-off-from-chicago-ohare-airport/amp/
@All
https://mentourpilot.com/american-eagle-e175-captain-incapacitation/
@Andrew.
The public (travelling or not) don’t, and won’t care.
If it is sold to them as a cost saving (i.e. they will think that their tickets will be cheaper) they will be all for it.
The people pushing single pilot know that, they know that will get it approved, otherwise they wouldn’t have bothered going down this path.
About the only way to stop it is via an accountant versus accountant battle.
Airline accountants don’t know how to factor safety into the balance sheets, but Insurance Company Actuaries sure do know how to factor risk into insurance premiums.
@George G
Yes, the American Eagle incident is a tragic example of what can go wrong, however infrequent. Pilot incapacitation has always been an issue and it remains one of the main justifications for having a minimum of two pilots in the cockpit.
I think any move towards single-pilot operations will need to entail much more stringent medical screening requirements for pilots that will further reduce the pool of suitable people. That screening will need to include comprehensive and ongoing assessments of mental health and possibly background checks for any factors that might affect a pilot’s mental health (eg. financial stress, personal relationships). Pilots are already subject to ongoing assessments of performance and health; I wonder how many will decide that further surveillance of their personal lives is all too intrusive?
@ventus45
I’m not sure the public doesn’t care. The commentary I’ve seen to date has been overwhelmingly negative, and not just from pilots. I guess it will depend on how well the airlines and manufacturers can convince the public that single-pilot operations are “safe”.
Re: MH370 of course rogue pilot is another concern with single pilot. This is part of the problem putting MH370 in the “greatest mystery” story category. Greatest denial story, is my book title, if I ever get around to it.
Speaking of Zaharie, his favored politician, Anwar Ibrahim, was sworn in as Malaysia’s tenth Prime Minister.
https://www.cnbc.com/2022/11/24/anwar-ibrahim-appointed-as-malaysia-prime-minister.html
@Victor
I’m of the opinion that Anwar Ibrahim is highly unlikely to exhume the subject of MH370, as instinctively this will once more raise the public awareness to his own past. A subject he has, at this time managed to bury, in conjunction with the passage of the same period of time.
@Barry Carlson: Unless he is somehow able to exploit the accident for his own political gain, I agree.
The propagation of WSPR signals in the high frequency bands is dominated by the effects of diffraction, yet the discussion up to now has been singularly devoid of any meaningful analysis of the impact of these effects. I have tried to rectify that omission in the following article:
https://drive.google.com/file/d/1RB9fSzaO2Dm9lCDyVQLgBMyPu-t1X7WC/view?usp=share_link
sk999. I read it, understood it, and agree with the conclusions. How many nails does it take to seal this coffin?
@sk999: The limits of spatial resolution due to diffraction is one of a long list of reasons why WSPR tracking of aircraft is impossible.
But WSPR tracking is not about science. It’s about making absurd claims that are repeated mindlessly by the media and other supporters. I won’t speculate here about the intentions, but it’s not hard to guess.
@airlandseaman asked “How many nails does it take to seal this coffin?”
The WSPR proponents, including the chief media promoter, Geoffrey Thomas, keep doubling down with more absurd claims. Meanwhile, those that are technically informed dismissed WSPR tracking many months ago. So, I’d say there will never be enough nails to seal the coffin.
Contributors here have done our part to enlighten anybody that is interested in the truth. That’s all we can do.
@Victor @Barry
That is very interesting re: Anwar.
I agree in theory that should be similar MH370 treatment as per Mahathir admin: denial and sweep under rug, but on the other hand, the development gives MH370 more visibility.
@All: Some of you may be aware that a small aircraft (a Mooney 20J, N201RF) crashed into a power line tower near KGAI (Montgomery County, MD) while on approach for the RNAV 14 runway. The cloud ceiling was near or below minimums for an LPV approach (269′ AGL) if equipped, and well below minimums for an LNAV approach (460 ‘ AGL).
Miraculously, the plane was slowed by the power lines and lodged into the tower about 100 ft above the ground. The pilot and passenger were rescued.
There are several good YouTube videos describing the accident. As usual, Juan Browne (@blancolirio) does a good job, but without access the ATC transmissions from Potomac Approach, he is missing some clues. Others like this one include the audio. The pilot seems to have trouble intercepting the approach course, and the controller patiently helps him.
One clue that I think has been missed so far is on the CTAF audio file. Although there is no voice audio, after ATC approved the frequency change to CTAF, there are four sets of clicks, in groups of 7,7,6, and 5. This is probably the pilot trying to turn on the runway lights, which are pilot-controlled, as typically 3, 5, and 7 clicks are used for low, medium, and high intensity, respectively.
@Victor
Re: Mooney
We are getting news coverage here in the DMV (Wash DC). To some extent there is news block-out, but one of the occupants is apparently out of the hospital, and other they are not saying, but might have been released. The was a wide area power outage at the time, but did not get over to Virginia. Pilot confirms he had another incident many years ago, but is otherwise silent, but we are hearing parts of the pilot’s calls to the 911 operator. Of course it was rainy Sunday Thanksgiving go-home day and the DCA/airports were having weather delays.
@TBill: Here is a description of his first accident:
https://www.deseret.com/1992/8/5/18998028/investigators-comb-scene-of-plane-crash-below-peak-br
That said, I’m not sure you can relate the circumstance of an accident 30 years ago to this one.
He was having problems following ATC instructions for holding a heading and intercepting the approach. Following a glide path in IMC would be much more difficult. The ceiling was near or below the minimum for an LPV approach, which would require a navigation system with WAAS GPS. If his plane was not equipped for an LPV approach, the ceiling was definitely below the minimum required.
The clicks on the CTAF for the PCL suggest to me he was having problems seeing the runway. I’m curious whether the runway lights were not operational, which could have made a bad situation even worse.
@Andrew: What are your thoughts?
@Victor
I haven’t studied the aircraft’s flight path, so the following is speculation. It might be a classic case of the ‘dive and drive’ technique gone wrong. That technique was used for many years for flying non-precision approaches without vertical guidance. It is no longer recommended because of the significant risk of CFIT or runway overrun, especially in large aircraft.
The ‘dive and drive’ technique typically has the aircraft descending as quickly as possible to the MDA once past the final fix, then flying level at MDA until one of the required visual references for the landing runway has been met. Under CFR Part 91 rules, those visual references include the threshold lights, runway end identifier lights, VASI and runway lights, amongst others. As you mentioned, the pilot might have been trying to turn on the lights or to increase their intensity in an attempt to find the runway. The following chart states the KGAI runway only has medium intensity runway lighting (MIRL):
https://aeronav.faa.gov/afd/03nov2022/ne_105_03NOV2022.pdf
Regarding the approach minima, Part 91 rules allow a pilot to fly an approach even if the ceiling is below the minimum specified for the approach. The only requirement is that the aircraft must not descend below DA/MDA unless the visibility is at or greater than the specified minimum. In this case, the chart shows the pilot only needed 1 mile visibility for either of the LPV or LNAV approaches. The METAR in the Aviation Safety Network report in Juan Browne’s video stated the visibility was 1.25 miles.
KGAI RNAV (GPS) RWY 13 Approach Chart:
https://aeronav.faa.gov/d-tpp/2211/05212R14.PDF
Aviation Safety Network Occurrence Report:
https://aviation-safety.net/wikibase/301805
@Victor
The following articles about the ‘dive and drive’ technique might be of interest:
https://www.aopa.org/news-and-media/all-news/2020/march/pilot/on-instruments-diving-and-driving
http://www.flightsafety.org/asw/nov07/asw_nov07_p13-17.pdf
@Andrew: My understanding is that even a Part 91 pilot is not allowed to fly below the DA/MDA until a visual reference of the runway is seen. Whether it is prudent for the pilot to attempt the approach under these challenging conditions is another question that will be asked, and will depend on his currency and his proficiency.
Here’s a new video from the Air Safety Institute that I think is the most complete that I’ve seen on this accident.
https://www.youtube.com/watch?v=Ynq1TdB-oo8
In the video, reference is made to the clicking heard on the CTAF, probably to turn on the lights. That suggests the runway might not have been in sight, but it’s also possible the lights were seen and the pilot was trying to increase the intensity.
I still wonder if the pilot was flying the LPV or LNAV approach. If it was an LNAV approach (which has a higher minimum), an incorrect altimeter setting could have contributed to the CFIT.
@Victor
From the 911 call, the pilot basically apologized said he just got too low altitude, no mention of mechanical issues. So Juan’s description of the tendency to go too low seems possible explanation. At one point the passenger was going to climb out, but the 911 operator/responders ordered them to stay in the aircraft until they were ready to rescue them. Very good job overall by the responders. The pilot was worried the aircraft could fall out of the perch and they’d be seriously hurt. According to WTOP radio reporter, the prior incident 30-yrs ago was basically considered pilot error in NTSB report.
@Victor
PS-
Sunday was accurately forecasted to be a very stormy day for several days before. It was even in national news, how bad a day it was going to be for those returning home from the holiday. If I was a small aircraft pilot (not with my eyes) but if so, it would have been a day that I would have avoided flying.
@TBill: There are many factors that determine whether or not a flight would be safe, including the proficiency of the pilot and the capabilities/limitations of the aircraft. I’m not aware of dangerous weather conditions such as severe turbulence, icing, or thunderstorms. A missed approach due to low ceilings and a diversion to an alternate airport could have been safely executed under the right circumstances. That’s not to say that the decision to fly was prudent. We just don’t know.
@Victor
RE: “My understanding is that even a Part 91 pilot is not allowed to fly below the DA/MDA until a visual reference of the runway is seen.”
Yes, a pilot must not descend below MDA until at least one of the required “visual references for the intended runway is distinctly visible and identifiable”. I should have made that clear in the last paragraph of my previous comment.
My point was that a pilot operating under Part 91 rules is allowed to fly an approach even if the reported ceiling is below the minimum, provided the reported visibility is at or greater than the minimum. Whether that’s a smart thing to do when flying manually is another question entirely. It is very easy to become disorientated under such conditions. At the airline where I worked, it was strongly recommended that pilots use autoland at approved airports any time the visibility was less than 1,500 m or the ceiling below 300 ft.
Re N201RF…for most of the 5 minutes prior to impact with the tower, N201RF was more or less right on the correct track for the runway KGAI RNAV (GPS) RWY 13. But 24 seconds before impact, the plane changed from a 130 degree track to 110 degrees. Why? Was this due to a sudden change in the cross wind component as the plane descended? Was it a deliberate change by the pilot? If so, why?
@Victor
RE: “Here’s a new video from the Air Safety Institute that I think is the most complete that I’ve seen on this accident.”
There’s some great analysis in that video. The last few minutes has some particularly valuable advice about ‘buying time’ and not rushing into an approach when something isn’t right, as well as not transitioning to visual flight too early in low IFR conditions.
@airlandseaman asked: But 24 seconds before impact, the plane changed from a 130 degree track to 110 degrees. Why?
One explanation is this:
Andrew said:The last few minutes has some particularly valuable advice about ‘buying time’ and not rushing into an approach when something isn’t right, as well as not transitioning to visual flight too early in low IFR conditions.
I suspect he drifted off the glidepath when he transitioned from looking inside the plane to looking outside, especially since it appears (based on the CTAF clicks) he was having problems seeing the runway.
@Victor
Have you tried plotting the aircraft’s vertical profile vs the nominal 3.0° glidepath, based on a threshold crossing height of 40 ft? I only have the data off FlightAware, but it seems to me the aircraft crossed the final fix (TIMBE) well below the 2,200 ft altitude stipulated on the RNAV (GPS) RWY14 approach chart. The aircraft remained well below the nominal glidepath until it came to grief in the powerlines.
@Andrew: I really haven’t had time to dig into the data. We also have to apply a correction to the ADS-B altitude data since what is transmitted is pressure altitude. If the altimeter setting was 29.45″, the ADS-B data would be high by about 470 ft.
Re N201RF: I put up some charts and data here: bit.ly/3B7PiA2
It looks like the PF broke out at ~630 feet MSL, then remained at that altitude until 24 seconds later when he hit the south transmission line tower, after going through the top lines west of the north tower. The lines slowed the aircraft before impact with the south tower (and one or more may have broken). The track changed 20 degrees left when the PF broke out and levelled off. The ground elevation at the towers is ~470 feet. The runway altitude is 539 feet.
@Victor
I haven’t plotted all the data, but the aircraft crossed TIMBE at about 1630 ft (2100 ft pressure altitude, corrected for QNH), JOXOX at 830 ft (1300 ft corrected) and the visual descent point at 630 ft (1100 ft corrected). Those altitudes are all well below the profile depicted on the approach chart. It seems the aircraft descended way too low, presumably in an effort to get visual, and then drove in at about 630 ft (90 ft AAL!) until it hit the power lines.
@ALSM
The course change seems to have occurred shortly after the aircraft levelled off at 630 ft. I wonder if the pilot was looking for the runway lights and became distracted by another brightly lit area to the east? There appears to be a residential area and a road (Snouffer School Rd) over in that direction.
Andrew: Breaking out and the change in course happened almost simultaneously. Not a coincidence IMO. I agree that the change was likely due to something observed visually. Like the guy on the AOPA Safety video said, stay on instruments until you are sure you have the runway. That clearly did not happen in this case.
@ALSM
To which I would add: Do NOT descend below the procedural altitudes on the chart. They are there for a reason!
@All: My email provider, Rackspace, had a major global outage today. Based on their recommendations to move to Exchange Server “until further notice”, I have doubts about the viability of the company. I’ve started to migrate to another email provider, but I may not be able to access any email that was sent to me today. If anybody needs to send me a message, please let me know here, and I will contact you using another email address.
Considering how long it is taking for Rackspace to resolve this issue, I think it might have been a ransomware attack.
@All: My email should again be working.
I was fortunate to have a local copy of my email, contacts, and calendar data. I feel sorry for those who relied on Rackspace to save their email data, and are less tech savvy.
I don’t see how Rackspace as a company survives this fiasco. Once you lose the trust of your customers, it is very challenging to recover it.
In the past year, my VOIP provider had a long outage due to a ransomware attack, my web hosting provider had a security breach which permitted insertion of malware into my website, and now my (prior) email provider is suffering an extended, global outage with no estimates as to when service will return. There was a time years ago that I ran all these servers on my own computers, but eventually I migrated to other providers to leverage their expertise and security. In retrospect, I don’t think that was a bad decision, but even the experts have security issues.
https://www.offshore-energy.biz/ocean-infinitys-first-two-high-tech-ammonia-ready-armada-ships-head-for-norway/
Re N201RF private plane crash last week (flew too low into powerlines)
https://www.yahoo.com/news/ntsb-plane-hit-tower-flew-215933497.html
@TBill
Thanks – the NTSB preliminary report is available here:
https://wtop.com/wp-content/uploads/2022/12/Report_ERA23LA071_106368_12_6_2022-2_45_33-AM.pdf
Regarding N201RF: So there was nothing found indicating a mechanical or instrument malfunction. It appears the pilot simply flew too low on the glidepath, and was below the DA/MDA (depending on whether it was an LPV or LNAV approach).
Needless to say, the pilot and passenger are very fortunate to be alive.
Re: “Needless to say, the pilot and passenger are very fortunate to be alive.”
Yes, and they are alive because they struck the high wires on the west side of the south tower first, slowing them down significantly before impact with the south tower. Being the sole survivor of a C206 high tension wire strike back in 1978, I can tell you from experience, the wires can decelerate a plane to near zero in a couple hundred feet.
correction: “…west side of the NORTH tower first…”
@airlandseaman said: Being the sole survivor of a C206 high tension wire strike back in 1978, I can tell you from experience, the wires can decelerate a plane to near zero in a couple hundred feet.
I am sure many readers here would be interested in some of the details you shared privately.
@Victor
Re: “Needless to say, the pilot and passenger are very fortunate to be alive.”
Not to mention this is 2nd incident for the lucky pilot.
@TBill: Considering the serious damage to the aircraft and the nature of the injuries to the people onboard, those two events were accidents, not incidents. The regulations do make a distinction.
@airlandseaman
The reference to a C206 stirred up some memories for me too. See your private email.
Mike, the email address I had for you [at earthlink] didn’t work.
@Brian: I sent an email to you. The EarthLink address is no longer valid.
@ALL: I’ll post the story about the C206 wire strike soon.
Reported today that another piece of MH370 wreckage looks like it has been found. Identified as the left main landing gear trunnion door, it is interesting in that its condition suggests that the gear may have been down when MH370 hit the water.
Interesting to me because the early 747’s; specifically, the 200/300/SP’s that I flew early in my career, had a step in the Emergency Descent checklist that mandated lowering the gear to get the descent going. From distant memory I recall FL280 or below being the cutoff.
It was probably to do with the residual thrust at 28,000 slowing the deceleration of the aircraft and making the depressurised aircraft spend too long above 10,000 feet, but don’t quote me.
Also, the TAS above FL280 would be a limiting factor.
What I do remember from practicing it in the sim was the awesome descent rate such an action eventually generated. Much more than just having the speed brake extended. A 777 would be the same, particularly if the person flying it didn’t care too much about the condition it was left in.
@Michael John Glynn: Lowering the landing gear increases drag, which will increase the descent rate for a given airspeed and thrust setting. But if the intent is a high speed impact, and high airspeed is desired, increasing the drag makes no sense.
In the B777, for an emergency descent, thrust is set to idle, speed to Vmo = 330 KIAS, and speed brakes raised.
For a B777, Vle = Vlo = 270 KIAS/M0.82, so the gear would need to be extended before the airspeed got very high. If I recall, there is also a maximum altitude of FL210.
Also, after fuel exhaustion, the RAT cannot power the landing gear. So gear extension has to occur while one of the engines or the APU is running.
The scenario as proposed makes no sense to me.
Comments (from experienced pilots) on the digital online edition of The Times Uk article today suggest that if you want to break up the aircraft quickly you point down at the waves never mind the landing gear. The harder you hit the water the more certain the instant destruction.
Apologies..all this in layman’s terminology
@Mike Glynn
The FCTMs for the B744 and the B777 still describe a landing gear extended rapid descent for cases where structural integrity is in doubt and the airspeed must be limited. I’m not sure I’d be in a hurry to drop the gear at high altitude and descend at Vle if structural integrity were in doubt, but it’s nevertheless a technique that’s authorised by Boeing.
The gear down rapid descent is certainly impressive in the B744, where Vle is much higher than Vlo (320 KIAS vs 270 KIAS). Not so much in the B777, where Vle = Vlo = 270 KIAS, as Victor mentioned. I remember experimenting with it in the B777 sim and we found the total elapsed time for the gear down descent at Vle was similar to that for the gear up descent at Vmo.
@Victor
There’s no maximum altitude for gear extension; you might be thinking of the FL200 limit for flap operation. Also, the gear could be extended via the alternate extension system after fuel exhaustion. That system uses a dedicated DC electric hydraulic pump powered by the hot battery bus to unlock the landing gear doors and uplocks. The landing gear then extends under its own weight.
I suppose there might be a case for extending the gear if the pilot wanted to maximise the rate of descent. I doubt that he would have been too concerned about “niceties” such as Vle.
@Andrew: Yes, I was thinking of the maximum altitude for flaps extension.
If the desire is to maximize the descent rate because of a depressurization event, for instance, then adding drag with the landing gear might make sense in special cases.
However, if the intention is to create a high energy impact, adding drag makes no sense to me. After fuel exhaustion, the flight control mode degrades from NORMAL and there is no envelope protection. Just lower the nose and maximize the airspeed. What is to be gained by lowering the landing gear?
@ All,
If the underside of 9M-MRO were painted gray, following the MAS livery paint scheme, why would one think a purported landing gear door panel from 9M-MRO would have a white painted side facing outward?
In B777 photos I have seen taken in good lighting conditions, the interior surfaces of the gear well and doors appear to be white. So, my expectation is that a landing gear door panel from 9M-MRO would be painted gray on the outer/lower side and painted white on the inner/upper side.
The new Gibson debris appears to be painted white on one side, and the other side is not now painted. Instead, it appears to be delaminated CFRP, so its original paint color is unknown. The (painted) skin on that side has been torn away. The debris photos show shrapnel entered the delaminated side and exited the white side. If the white side is actually toward the interior in 9M-MRO, then the direction of shrapnel penetration appears consistent with engine blade fragments penetrating the fuselage while this panel was in a closed or flush position.
@DrB: It is hard to argue that the underbelly of 9M-MRO was painted grey, and therefore, if the recovered debris was a landing gear panel, it should be grey and not white on the outside. That completely blows up the entire thesis of extended landing gear based on punctures from the inside to the outside of the panel.
Whether or not the punctures were from engine blade fragments, I’d say that was a possibility, but far from proven, as there would be many projectiles that might have struck the panel with a high relative velocity in the cloud of parts that formed after a high speed impact.
So we have another mess created by Geoffrey Thomas publishing a false analysis, which is then propagated by other media outlets.
I am well aware that most people here adhere to the unresponsive / ghost flight to fuel exhaustion / high speed uncontrolled dive to impact camp.
Only a few people here are in the deliberate, controlled to the end / ditch camp.
I am in the deliberate ditch camp.
If I were ditching, I would have the gear up, not down, so the panel would be closed / flush.
If the panel is actually the left landing gear strut panel, and if it was actually penetrated by small high speed / high energy fragments, the only reasonable culprits I can think of are turbine blades / fragments there of.
In a ditching, the engines impact the water first.
If the ditching was conducted with the engines powered up (perhaps considerably above idle, as in say “holding her off – as long as possible – down to the stall” – at the lowest possible airspeed), then it is almost a no brainer, that when the engine suddenly ingests tons of sea water, with the rotational energies involved (think QF A380 engine number two) that the engine would tear itself apart, almost instantly, from the inside out.
The observed damage appears to be consistent with an event of that nature.
@ventus45,
What you write, above, might have value if this article of debris were a ‘trunnion door’.
However, it’s not a fragment of a ‘trunnion door’ or any other main landing gear door.
RE new debris:
First, I want to congratulate and thank Blaine for turning up another piece of 370 debris. The BG/RG Debris Analysis dated Dec 12, 2022 contains enough information to conclude that this debris is almost certainly from 9M-MRO. However, the thread of logic leading to the conclusion that the gear doors were down is flawed in several respects.
First, no pilot would drop the gear to increase the damage. The opposite is true. As already noted, the PF (if there was one) would have accelerated to the highest speed possible, which means gear and flaps up, not down.
Second, the material, color, size, thickness, shape and marks on the debris do not appear to be consistent with any of the MLG doors (including the Trunnion door).
Third, all the gear door pieces are well inside the engines. Thus, any engine debris would hit the outside (“white side”) of the Trunnion door, not the inside as suggested in the analysis.
Forth, the 4″ cuts could have been caused by any of a few thousand other pieces of debris at any point in the breakup sequence, whether in flight or at main impact.
Fifth, statements to the effect that there is no evidence that an inflight breakup occurred are not true. In fact, the report Don and I published on the Jefferys Bay spoiler debris (https://bit.ly/3c5Wj6Z) concluded, based on all the right wing debris evidence, that in flight breakup was likely.
@airlandseaman said: First, I want to congratulate and thank Blaine for turning up another piece of 370 debris.
I agree.
But now that the debris has been recovered and we have access to some photos (albeit at a lower resolution than the originals), it’s possible for more serious investigators to correctly identify the part and determine if it has probative value. The paper authored by Blaine and Richard Godfrey failed in this respect.
Predictably, the BG/RG paper was promoted by Geoffrey Thomas, and was subsequently picked up by the mainstream media, which have performed their duty of replicating the article without any attempt to substantiate any of the technical claims.
C:\Users\julia\Downloads\IMG_20221213_113827.jpg
A two page spread in the Times today. I’ve copied a rather surprising detail that the person who handed the door to Blaine..his wife had been using it as a washing board for five years not having realised its significance.
@Julia: You supplied a link to the downloaded image stored on your computer. We need an internet link to access this.
@Victor
There’s a paywall to the internet link. All I can do is copy the whole article to you. Would you like me to do this. Everyone can then read today’s Times article.
@Victor Sadly it’s not in today’s digital edition of the Times, only my printed copy. (I subscribe to both).
Now the only thing I can do is photograph the two page spread. It’s very annoying when the Times does this. It updates the online editions throughout the day and arbitrarily deletes articles. That’s why I subscribe to both.
Is this the Times article?
https://www.thetimes.co.uk/article/flight-mh370-pilot-plane-crash-evidence-qj8shv5kk
@Julia Farrington: I don’t think it’s worth the effort to post the article, as the media stories on this topic are based upon multiple layers of false information, including:
1. Misidentification of the debris (likely not the trunnion door panel).
2. Incorrect analysis of the damage.
3. Unsupported conclusion that the landing gear was down.
4. Illogical reasoning for why the pilot might have chosen to lower the landing gear.
Eventually, others will determine the probative value of this debris, if any. That analysis is unlikely to be covered by the media.
This media stories we are seeing demonstrate that the media often propagates false information promoted by a small group of people without any serious fact checking.
@Victor: It is indeed unfortunate the way the “news” spreads with zero checking for accuracy, and fails to follow up when called out for the errors. We have seen it over and over. Here is the Times Headline:
“Flight MH370 landing gear suggests ‘criminal intent’ by crash pilot; A damaged door is the first evidence suggesting pilot deliberately downed Malaysian Airlines jet”
There is not a single piece of this Headline that is true:
1. The debris is almost certainly not from any landing gear door.
2. It is certainly not the first evidence suggesting pilot deliberately downed 9M-MRO
3. It provides no new evidence of criminal intent
Garbage
@airlandseaman: As many know, I believe a diversion by the captain is the most likely explanation for the disappearance. However, like you, I believe this new debris offers ZERO additional evidence that the pilot was responsible.
@airlandseaman
Thanks
That article was in yesterday’s Times. I sent that one to @Victor yesterday.
The elusive one today doesn’t add much to yesterday’s. I’ve photographed it and sent it to Victor’s DM. Today’s Times headline reads
” Wreckage could finally solve mystery of MH370 air crash”
I was just skeptical at the idea that the wife of the fisherman had used it as a washing or ironing board for 5 years. Allegedly it was discovered 3 years after the disappearance of MH370.
I don’t believe everything I read in the Times but am disappointed of course that this particular article is so misleading. I feel for the relatives.
@ALSM. As to where this piece was from, some speculation and observations:
The close spacing of the penetrations together with one being only partial (the report’s Fig. 27, to the right) and that not being retained (Figures 26-32) might indicate that the piece was impaled, the source then withdrawn.
That might be more likely to an object striking the projections of an engine core, such the some retained compressor blades on their discs, one short and their changing angles with compressor stages. A long shot though to be sure…
Nevertheless the reverse curvature at one edge (see the report’s Fig. 25, left photo, bottom) could be consistent with part of the engine cowl or its pylon’s fairing. I see some reverse curvature in the latter, behind the engine, in Fig. 46.
Another observation is that the right photograph of Fig 25 shows glue, apparently from a stiffener or strut, with a broken piece of an attachment adjacent, these on the exposed carbon fibre substrate, ie not being removed with separation of the skin. Odd.
While the report draws unwarranted conclusions, it does have merit in its presentation of the find, depiction of the piece and of course, as you have observed, in the find itself.
In that context it is pleasing to read on that blog that the ATSB has taken action and sent the report to Boeing for identification. Normally one would expect that to come from Malaysia.
First sentence para. 2, the above, now amended for readability:
“That might be from an object striking the projections of an engine core, such as some retained compressor blades (one being short), the angles of the penetrations changing with compressor stages; the object then separating.”
@David: Projectiles from rotating equipment tend to fly out radially. Yet the trunnion door panel is aft of the engines, and not in the flight path of projectiles. That doesn’t even consider the authors’ confusion about which side of the panel is facing the engine when the landing gear is extended.
I have worked with high speed rotating machinery for decades. There is a common saying in the industry–we stand behind our machines, but not to the side.
I believe that the FBW system on the 777 will still be working if both engines fail, as the PSA’s can be powered by any available power source including the hot battery bus and the RAT.
This would limit any attempt to push the nose down to no more than 15 degrees ND.
Andy, please chime in here if this is incorrect.
Therefore, lowering the gear to increase the rate of descent may be an option.
For info only, the FL200 limitation on the flaps is because the flaps are not required to be flight-tested above this altitude during certification. All modern airliners have this limitation.
@Mike Glynn
AC power to the transfer buses is lost if both engines fail. When that happens, a large number of sensors needed by the flight control system also lose power, which causes the flight control system to degrade to secondary mode. Envelope protection is not available in secondary mode.
Regarding the flap limitations, yes, I agree. On the B777 the flaps/slats are also inhibited in primary & secondary modes above an altitude of 20,300 ft.
Thanks Andy. With that info, and the knowledge that the SATCOM behavior is indicative of fuel exhaustion, I see no need to lower the gear to modify the angle of impact as the angle can be controlled by the available flight control inputs.
One scenario that may fit the situation is that the landing gear uplocks failed under high G forces caused by the phugoid motion during the final descent.
The precedent for this is the China Airlines Flight-006 747-SP upset, caused by a mishandled engine failure at high altitude, which sent the aircraft into a spiral dive necessitating a 5g recovery.
During this recovery, the uplocks failed in both wing landing gears which caused the gear to fall into the wing gear wells, damaging the gear doors.
The body gear actuator doors were also separated from the aircraft and tire marks were found on other pieces of the airframe.
You can read the report here: https://www.ntsb.gov/investigations/AccidentReports/Reports/AAR8603.pdf
Mike,
Discussion for implications of landing gear extension and the effect extended gear might have on descent speed has been provoked by the claim that this recently surfaced ‘Tataly-Antsiraka’ article of debris is the MLG trunnion door.
However, the article of debris is not an MLG trunnion door, nor any part of the MLG doors. I do note your reference to ‘tire marks‘ – the drag/smudge mark on the ramped side of the ‘Tataly-Antsirika’ piece could be (mis)taken for a tire mark.
A process is ongoing to correctly locate the article to its origin on the aircraft, that origin appears to be quite some way from the MLG doors. We shall take a few more days to complete that effort.
I was about to post a spoiler alert about the new debris, but Don beat me to it. The Tataly-Antsirika debris article is definitely not from any part of the gear doors. We know that for several reasons to be be explained in full shortly. In sum, the debris provides ZERO evidence indicating the gear position and thus ZERO evidence the PF was dead or alive.
The extended gear theory is fantasy based on incorrect identification of the debris, incorrect damage assessment, and incorrect understanding of the effects of lowering the gear. It’s layer upon layer of falsehoods.
When was the last time we’ve seen a theory with layers of nonsense?
I really don’t know how these people sleep at night.
Just trying to narrow down where this piece is from.
Possible topside of outboard flaps.
Reason….the smooth paint band across the panel resembles patch that touches the spoiler trailing edges. See also the paint band on topside of the recovered outboard flap.
The recovered panel is curved as the flap topside.
Don and ALSM, is that where you are looking? I’ve got a good picture if you need one.
Actually, the paint band may be evident on both the inboard flaps and the flaperon!
@Don & ALSM,
AS the found piece seems to have failed along the rivet lines. I think you might be able to exactly position this part, even work out if it’s from the left or right wing. I’m going for top of the outboard flap for the moment.
I’ll keenly await for your, no doubt thorough report….
Tim: Yes, please send your photos.
[Comments here are closed. Please continue the discussion under the new article.]