Italian Satellite May Have Detected MH370 Floating Debris

A source has disclosed that an Italian satellite that is part of the COSMO-SkyMed constellation detected three floating objects on March 21, 2014, near where MH370 is believed to have crashed in the Southern Indian Ocean on March 8, 2014. This information was never publicly released.

The three floating objects were detected at 34.9519°S, 91.6833°E; 34.5742°S, 91.8689°E; and 34.7469°S, 92.1725°E.

COSMO-SkyMed Satellite

The detections are significant because we know that a French satellite that is part of the Pleiades constellation detected what appears to be man-made floating debris on March 23, 2014, only 35 NM from where the Italian satellite had detected floating debris two days earlier. The French Military Intelligence Service shared four proximate images from Pleiades 1A with Geoscience Australia (GA) in March 2017, which then performed detailed analyses and determined that a cluster of nine objects that are probably man-made appear in one of the images near 34.5°S, 91.3°E. Australia’s Commonwealth Scientific and Industrial Research Organisation (CSIRO) then used this position information along with advanced ocean drift models to calculate the most likely point of impact (POI) to be 35.6°S, 92.8°E.

There is no definitive proof that either satellite detected floating debris from MH370. Our source also could not definitively state that there were no other floating objects detected near the 7th arc by these two satellites. However, the source believes that if there were other objects detected, they would have been shared with the MH370 search team.

The two satellites used different physical principles for detecting floating objects. The Pleiades satellite used optical sensors to capture images in multiple bands of color to achieve a pixel size of 0.5 m x 0.5 m. On the other hand, the COSMO-SkyMed satellites use Synthetic Aperture Radar (SAR) sensors to continuously scan the earth’s surface. Unfortunately, COSMO only obtained a wide-angle, low-resolution capture of the objects. On a subsequent satellite pass, attempts to capture the objects at high resolution were not successful.

Prior to 2014, researchers had already investigated using satellite SAR data to detect floating debris. For instance, in 2011, using SAR data from the crash of Air France 447 off the coast of Brazil in 2009, researchers presented a numerical method for processing the SAR data from COSMO-SkyMed to detect floating metallic objects. (HT Don Thompson.) Likely, those who have analyzed the COSMO data from March 2014 would know if the detected objects are metallic.

To determine if the objects detected by Pleiades and COSMO-SkyMed were from a common source, we used the results of a complex drift model (BRAN2015) developed by CSIRO and shared by oceanographer David Griffin. The results include the trajectories of 86,400 virtual drifters, representative of generic debris sitting flat on the surface. The virtual drifters start along the 7th arc on March 8, 2014, between latitudes 8°S and 44°S, and the trajectories are tracked for 1000 days. Our method was to find the two virtual drifters that best match the position and timing of the detections from the the two satellites. If those two virtual drifters started from nearby locations on March 8, likely the objects detected by the satellites came from a common source.

The results from the drift analysis are shown in the figure below. The yellow circles show the path of the virtual drifter that passed closest to the COSMO objects on March 21. The red circles show the path of the virtual drifter that passed closest to the Pleiades objects on March 23. These two virtual drifters start within 3.5 NM of each other on March 8, near to 35.4°S, 92.8°E. The proximity of the starting positions is consistent with a common source for the objects detected by the two satellites. That position is about 83 NM to the southwest of where a previous study estimated that MH370 crossed the 7th arc, and within the 140 NM radius recommended to search.

Floating objects detected by two satellites place MH370 impact near
35.4°S, 92.8°S. (Click on figure to enlarge)

In order to better estimate the likelihood that these objects were from MH370, we pose the following questions:

  1. Were there other detections of floating objects along the 7th arc by Pleiades, COSMO-SkyMed, or any other satellites?
  2. Were the COSMO-SkyMed detections on March 21 determined to be metallic objects?
  3. Exactly what areas along the 7th arc were surveilled by Pleiades, COSMO-SkyMed, or any other satellites?
  4. Will Airbus (the operator of the Pleiades satellites) provide the images for each color band so that independent researchers can analyze the raw data? (HT Bobby Ulich)

273 Responses to “Italian Satellite May Have Detected MH370 Floating Debris”

  1. Don Thompson says:

    The paper referenced in the above post can be supplemented with a more detailed review of the subject. Single-Look Complex COSMO-SkyMed SAR Data
    to Observe Metallic Targets at Sea

    The COSMO-SkyMed mission comprises 4 satellites delivering a daily overpass for the area of interest for MH370 whereas Pléaides comprised only 2 satellites in 2014 with overpass on alternate days. The Synthetic Aperture Radar acquisition may provide more detail than the visible/near visible spectrum of Pléaides to discriminate small features on the ocean.

    After conducting a search through available and relevant literature, I have exchanged some correspondence with one of the paper’s authors and discussed their view of COSMO-SkyMed’s ability to discern non-metallic objects (given the study’s emphasis on ‘metallic objects’), that is, objects of G/CFRP composite construction: in reply he described that buoys have been discerned with SAR imagery and that mostly-submerged wooden structures have been discriminated in fresh water. Notably, he also described how it might be possible to discriminate a localised pattern in sea surface conditions around a floating object thus aiding observation. Good knowledge of the prevailing surface conditions is also useful.

    While the mid ocean might seem like a barren void, it is necessary to consider what other objects might be present on the ocean surface, Global Fishing Watch shows that longline tuna fishing vessels were operating in the area of interest indicated by both these COSMO-SkyMed and Pléaides acquisitions considered by Geoscience Australia. A vessel should be readily discriminated by both visible and SAR imaging platforms, whereas the longline fishing gear comprises many floats and beacon rafts over their length which can be 10s of km, possibly discriminable but less likely to be readily identifiable.

    I echo Victor’s call for ADS/Astrium GEO-Information Services (for Pléaides) and ASI (for COSMO-SkyMed) to release into the public domain all primary imagery products acquired in support of the southern Indian Ocean surface search operations during March and April 2014.

  2. Victor Iannello says:

    @Don Thompson: Thanks for the detailed comment. Whatever we can learn to increase or decrease the likelihood that the satellite detections were MH370 debris is extremely useful.

  3. CanisMagnusRufus says:

    Hi Victor, it’s odd that Geoscience Australia makes no mention of these SAR images from COSMO-SkyMed in a list of satellite assets used in the search for MH370’s location near the 7th arc.

    Satellites that have provided information in the search for MH370 include:

    – The Inmarsat communication satellites that gave us the hourly Burst Timing Offset and Burst Frequency Offset data, which, together, narrowed the search down from a large portion of the entire Indian Ocean to just a segment of the 7th arc
    – The Airbus Pleiades 1A satellite that provided the visual imagery of objects which, if they were parts of the aircraft, constitute the latest and most precise clue we have
    – NOAA and AVHRR Sea Surface Temperature imagery which provided details of ocean circulation, significantly augmenting what could be deduced from other oceanographic techniques. This is measured by the Advanced Very High Resolution Radiometer (AVHRR) sensor carried by the (US) National Oceanic and Atmospheric Administration (NOAA) Polar Operational Environmental Satellites (POES)
    – Altimetry data from satellite missions operated by the respective space agencies: Jason-2 (NASA and CNES), CryoSat2 (ESA), AltiKa (CNES and ISRO). An altimeter is an instrument carried by an earth-observation satellite that measures two quantities extremely accurately – its position in space and the distance to the surface of the ocean directly below, to estimate sea surface height. Maps of sea surface height are key to estimating sea surface currents, just like winds can be inferred from maps of atmospheric pressure.
    “If we find MH370, which we all hope to do, it will be thanks to all this satellite data, particularly the altimetry data,’’ Griffin says.

    It’s interesting that CryoSAT2 is the only satellite on this list with a SAR payload, yet only its altimeter data is mentioned as it’s contribution.

    Could it be because, until recently, SAR was the exclusive domain of governments and militaries? COSMO-SkyMed appears to be no exception. According to the first link in your article,

    October 2009: TAS (Thales Alenia Space) handed over the management of the COSMO-SkyMed ground network to the Italian Defense Ministry, an event that marks the beginning of full operational use of the system by its principal customer, the Italian Defense Ministry and the ground network provider Telespazio of Rome.

    …. and if you look at Table 4
    Support ModeSAR observation side with respect to the nadir ground track
    Configuration Right-side looking (nominal) – Comment Providing better performance of northern hemisphere coverage
    Left-side looking for limited periodsComment Use intended to satisfy ASAP priority requests

    So the Italian military is in charge of the satellite, and it’s performance is better in the northern hemisphere.
    Perhaps we can remind ourselves that the this happened during the height of the Ukraine Crisis 2014 and NATO may not have wanted to disclose the tasking of it’s satellites.

    BTW, I usually roll my eyes when drift modelling of recovered debris is mentioned, but I think it’s worth pointing out something in the GA article, which clarified my understanding, and increased my confidence in drift modelling.

    The advantage for oceanographers this time was that the drift modelling only had to be applied to a two-week window – a much shorter interval than the many months relating to all the other evidence. This allowed for much greater precision.

  4. Victor Iannello says:

    @CanisMR: I suspect that the COSMO-SkyMed data set was known to CSIRO, but protocols prevented CSIRO from referring to it. I’m told that the French and Italians share Pleiades and COSMO-SkyMed satellite data by agreement. Permission for release of that data by CSIRO might have been complicated, but I am only guessing.

  5. Victor Iannello says:

    Some of us are trying to get more information about the COSMO-SkyMed detections as well as the single band data for the Pleiades images. Also, we need to better understand where along the 7th arc there were no detections. At this point, we simply don’t know whether the objects were from MH370. However, if they were from MH370, we should be able to use the drift models to fairly precisely predict the point of impact (POI). Now that 7 years has lapsed since the satellite data was obtained, there might be less sensitivity about releasing the data.

    I’ve seen some claim that it is doubtful that the detections were objects from MH370. At this point, without additional facts, I would say those claims are baseless.

  6. DrB says:


    The fact that the Pleiades images exist of a portion of the 7th Arc only 15 days after MH370 crashed is a testament to the rapidity of the analysis of the Inmarsat satellite data, the subsequent tasking, data collection, and analysis of the Italian COSMO synthetic aperture radar data, and the subsequent tasking and image collection of the French Pleiades 1A satellite. The French analysis of the Pleiades images indicated two large objects, which seemed to be in the right place to be the same objects seen by COSMO, in the vicinity of the 7th Arc near 34.5S. At that time, the Pleiades satellite data were classified, and this also restricted promulgation to foreign entities. The classification issue was overcome for Geoscience Australia, and four snippets of the Pleiades panchromatic data, one of which showed the two objects, with the other three images presumably being included to indicate the absence of large objects in nearby areas.

    Geoscience Australia performed more extensive image analyses, and produced a report indicating a significant number of sizable and “probably man-made” objects in the NE image snippet called PHR_4. Subsequently, David Griffin and his colleagues estimated the Point of Impact (POI) which would have put the objects in the positions seen 15 days later in the Pleiades image PHR_4. This and other considerations formed the basis for the later recommendation to search for the submerged debris field circa 35.6S.

    It is important to understand three facts which affected the Pleiades image analyses by Geoscience Australia:

    1. Only images of 4 small areas (20 km by 20 km each) were supplied by the French.

    2. No information was provided on the extent of the other areas imaged, nor of any specific analysis results in those areas.

    3. Only the black-and-white images were supplied by the French. The color images were not supplied.

    The supplied images are partially obscured by clouds, the sea state was such that breaking waves are seen throughout, and sun glitter is extensive, especially so in PHR_4. All three effects hinder floating object detection.

    The panchromatic images have 0.5 m specified resolution. I have computed the pixel size in the PHR_4 panchromatic image published by Geoscience Australia to be 0.50 m, in agreement with the specification. The panchromatic image uses the entire visible band. It is a grayscale image with approximately 12-bit intensity resolution. There are also multiple color bands, covering the visible range and the near infrared, which are recorded at 2.0 m resolution. A colorized panchromatic image can also be created at 0.5 m resolution by using the color image data to set the color of the 0.5 m resolution panchromatic image. This color image would be the one most useful in searching for MH370 debris. Unfortunately, the images supplied by the French to Geoscience Australia were the black-and-white panchromatic images, not the color images.

    Presently the Pleiades satellites (there are two of them) are still operating and providing commercial imaging services. I have inquired of their developer and operator, Airbus Defence and Space, whether the 2014 SIO images are available for purchase now as unclassified material. If they can be obtained, I can determine which areas of the 7th Arc were imaged in 2014, and I can process those images looking for MH370 debris, especially flaperon-sized objects. I am interested to see if there are flaperon-sized objects in the PHR_4 image and in other images to the northeast along the 7th Arc if those images exist and can be obtained. Reliable detection and classification of floating objects significantly smaller than the flaperon and the Pemba flap are unlikely given the imaging resolution of the satellite.

  7. Don Thompson says:


    I suspect you misunderstand satellite tasking. The tasking instruction is of the form, ‘when your satellite passes over a location, as dictated by its orbit, please acquire some imagery’. Tasking does not imply ‘please have your satellite move to my requested location and acquire some imagery’.

    The configuration of the ground segment, for example the location of uplink sites, needs to be considered in the responsiveness of the ‘system’. If a particular mission has only one uplink site, tasking commands can only be sent when the satellite makes an overpass within its comms antenna footprint.

    Astrium is keen to describe how the Pléaides mission is very reactive to tasking requests as it the ‘system’ employs multiple uplink sites on French territories around the globe.

    COSMO-SkyMed, COSMO being an acronym for COnstellation of small Satellites for Mediterranean basin Observation, comprises four satellites delivering a 12h revisit time for a specific location.

    Little of the mission operations are in any way secret, ‘User Guides’ to missions such as COSMO-SkyMed are Pléaides openly available and can be understood with a modicum of knowledge.

  8. DrB says:


    One of the principal issues with processing high-resolution images, even from “small” areas, is the huge file size. For instance, each of the four “snippets” from Pleiades is 20 km by 20 km in extent. At 0.5 m resolution, the grayscale panchromatic image is 40 K X 40 K pixels. At 12 bits per pixel, each image is 2.4 TB. A color image at 8 bits per 3 colors would occupy 4.8 TB. An image of the whole 7th Arc would be several petabytes in memory.
    The large data file demands “automatic target recognition” (ATR) be implemented using a high-speed computer. Human inspection is valuable, once candidate targets have been identified by computer processing. Human inspection by a trained operator can confirm object detections, reduce false alarms due to environmental features, and classify better than most current algorithms. The human brain is quite exceptional in image feature identification, under certain conditions. For instance, we have great difficulty identifying a person if their image is upside down. So, rotational invariance of complex shapes is a weak point for us humans, but it must be strong in an ATR system.

    Any useful processing of even the (2.4 TB!) “snippets” on a PC will require disassembling the image into sub-images (or “tiles”) which can be manipulated using acceptable file sizes. Then an ATR algorithm can be run on each sub-image in a time sequence. Candidate detections are found, and a small cookie-cutter image surrounding each contact is saved and sent to a classifier algorithm. The classifier algorithms extract various image features to see if the contact is possibly a real object. Those contacts which are determined to be potential objects are saved and later reviewed by a human operator. So there are 3 steps in the typical ATR system: computer detection, computer classification, and operator confirmation/classification.

    It’s very easy to build an automatic detection algorithm which has a probability of detection of nearly 100%, assuming the objects are nearly always visible. The difficulty is in keeping the probability of detection very high while, at the same time, keeping the false alarm rate extremely low. All useful ATR systems achieve a constant false alarm rate (CFAR). That is, there is a limit to the rate at which false alarms can be tolerated without overwhelming the system or the human operator. So, an ATR designer must use adaptive thresholds for detection and classification which keep the FAR low and constant. The FAR in high-data-rate systems must be extremely low, on the order of 1 per million opportunities. I have designed, built, and operated military CFAR sensors imaging floating and submerged objects in the ocean. I have also done satellite image analyses for various 3-letter agencies in the USA. My experience has given me a good understanding of the issues associated with overhead imaging of floating debris in the ocean, and I can use some of the techniques developed in my prior military work to effectively find objects which match the larger pieces of MH370 debris (and which are considerably smaller than all the “probably man-made” objects reported by Geoscience Australia). The question is, are any of the larger recovered debris clearly visible in Pleiades images?

  9. Victor Iannello says:

    @DrB: GA identified 70 objects in the 4 images (PHR1-4), of which 12 objects were judged to be “probably man-made”, and of those 12 objects, 9 were found in image PHR4. GA claims that 2 of the 70 objects were also identified by the French Ministry of Defence, but those objects were from image PHR2, and neither object was classified as “probably man-made” by GA. (GA classified one as “possible natural” and the other as “possible man-made”.)

    This demonstrates the subjectivity in identifying and classifying the objects.

    The GA report also states that: The confidence with which we are able to state that the objects observed in the images are unnatural could be increased if we were able to also study other images (from the same instrument and satellite in a similar sea-state) where debris is not expected to be found. For this reason, examination of further images is likely to be of value.

    It is clear that GA wanted more images to improve the confidence of their identifications and classifications, but was not able to obtain the images.

    I also note that the images were received by GA on March 23, 2017, which is 3 years to the day after the images were captured. It might be that this delay was mandated. It also might mean that the restrictions would be further relaxed today now that 7 years have passed since the images were obtained.

  10. TBill says:

    The 2-week debris drift analysis is interesting.
    On the one hand, the suggested crash site is somewhat beyond Arc7, as we might expect the plane kept going south. On the other hand, if we are going to find some of that rather large floating debris in the sat photos (let’s assume some of it sinks after a while), we may need to be looking well inside of Arc7. Somewhat counter-intuitive, but maybe we have to settle for finding some drifted pieces before we find the actual crash site. Possibly adds some justification to searching inside Arc7, even though we often tend to feel the plane itself flew beyond Arc7.

  11. Don Thompson says:

    @DrB wrote “40 K X 40 K pixels. At 12 bits per pixel, each image is 2.4 TB.

    Pléaides stores the 12bit image depth samples in 2 bytes of file storage when delivered as GeoTIFF user product.

    40K px x 40K pix x 2 bytes = 3.2GBytes

    The spacecraft is provided with 600GB of onboard storage. If temporary storage on the spacecraft adopted a 3 bytes per 2 pixels format, 2.4GBytes per panchromatic spot image is (obviously) the manageable trade-off for acquisition vs downlink to the ground segment.

  12. Victor Iannello says:

    @TBill: I would advocate searching for the main debris field rather than trying to figure out where floating parts may have eventually sunk.

  13. DrB says:

    @Don Thompson,

    Thank you for correcting my multiplication error regarding Pleiades image storage space.

  14. 370Location says:


    Thanks for clarifying that COSMO SAR data is shared with the French by agreement. All the reports pointed to French Ministry spokesman revealing that the SAR data was taken on Mar 21, delivered to Malaysia and AMSA on the Mar 23.“radar+echoes”+”fuzzy”+”dimensions”

    Note comments from a Malaysian unofficial that one of the objects was about the same in size as the 22x13m object reported by the Chinese.

    I found entries for your three SAR coordinates, along with a fourth, in my GE bookmarks marked as from a French Satellite, also with an extra digit of precision. The fourth object is about 200km SSW of yours. I traced that back to reports that the coordinates were disclosed on Mar 28 2014:

    Archived here:

    It would be very interesting to examine SAR images from anywhere plausible MH370 area.

    From the study Don linked, the SAR images reveal boat wakes and other surface disruptions, very much like sunglint images but at higher resolution, and regardless of cloud cover. As I already mentioned here, I believe a 200m resolution sunglint image from Mar 11 2014 may reveal an MH370 debris field near Java.
    This image has a more accurate contrast enhancement than on my website:

    To validate, I examined a similar sunglint taken a week after the Lion Air 610 crash off Jakarta. It shows boat wakes like the SAR images, and surface striations like those shown above. It’s clear to me that evn low resolution sunglint images and SAR could both show the effects of surface disruptions from oil and smaller debris affecting surface reflectivity.

    — Ed Anderson

  15. Victor Iannello says:

    @Ed Anderson: Thank you for your comment. I was not aware that the coordinates of the three objects (plus an additional one) were previously released. The coordinates evident in that photo correspond exactly (and in the same format) to what was supplied to me. I took the liberty of converting to decimal degrees. So now we can be quite sure that what was supplied to Malaysia and AMSA by France was indeed the SAR data from COSMO-SkyMed.

    Although the satellite data provides a high level of precision, we still don’t know if the objects were from MH370. The failure of the underwater search in that vicinity of course reduces (but does eliminate) the possibility that they were.

    Regarding the imagery near Java, how do you explain that lack of any debris recovered along the relatively nearby shore?

    [Your comment appeared here with a delay because the multiple links required that I manually approve the comment.]

  16. Don Thompson says:

    On March 11th 2014, the China Meteorological Administration activated the Disasters Charter thereby requesting that the Charter’s signatories assist in collection of imagery that might be useful in the search for MH370.

    Simultaneously, products were delivered to the Malaysian Remote Sensing Agency and the China Meteorological Administration.

    Aside from the originator’s archives, it’s reasonable to expect that the MRSA and CMA retain archives of the image products distributed to them. Perhaps Malaysian and Chinese next of kin groups might assist in making requests to those agencies?

  17. CanisMagnusRufus says:

    @ Victor,
    There is a nice timeline of the debris sightings in this BBC article

    It too makes reference to SAR data released by the French on Mar 23. But the location was 930 km north of previous sightings by China and Australia.

  18. 370Location says:


    Some of the articles mentioned that the aerial search took into account the Mar 21 SAR detections (presumably with drift), but of course nothing from MH370 was sighted.

    If your source only provided three coords, that would explain why the fourth was not in your report.

    You asked, “Regarding the imagery near Java, how do you explain that lack of any debris recovered along the relatively nearby shore?”

    We did discuss the short term drift not too long ago. I had used the Nullschool maps early on to see where debris would travel from the Java Anomaly. It appeared to go generally west, with first landfall expected at Cocos Islands. I pursued that possibility with plastic debris researchers who regularly visit Cocos.

    Later, @Oleksandr did a custom plot showing that the bulk of the debris did go directly into the equatorial current towards Cocos. See the animation here:

    His data sampling was daily, but the graphical plots are every 5-6 days. The 14Mar frame shows the debris field SE of the Sunda Strait. The next 20Mar frame shows the bulk going WSW. The debris then mostly moves slowly west until mid-april, when weather appears to blow a very small slowest portion of the 50,000 particles back onto Java shores. It is passing N of Cocos on 16Apr, and some is blown S onto the islands on 20Apr. The researchers had planned a Cocos Islands survey for mid-2020 and would keep an eye out for MH370 debris. Then, Covid.

    According to their reports, nearly all of the plastic junk beaching at Cocos comes from SE Asia. Seeing the modeled forward drift, reverse drift from Cocos might generally be from around the Sundha Strait, Sumatra, and Java. If that’s the case, it could be assumed that there is a lot of local debris on Indonesian beaches. With nobody expecting MH370 debris there, it might have been overlooked.

    Sat images do show a rectangular object about 4x12ft appearing at a fish camp on the S tip of Pulau Enggano in the right time frame, but it’s probably a shack or table.

  19. CanisMagnusRufus says:

    @Ed Anderson
    According to this BBC report, search aircraft were parked in Malaysia (March 14) pending overflight approval from Indonesia to search ‘south of Java.’

  20. Miden Agan says:

    @Victor @370Location

    Debris drift analysis and arrival times contradict a crash site off the south coast Java.

    The suggested crash site of the “Java Anomaly” is at 8.36°S 107.92° E, near the south coast of Java. There would be some debris on the shoreline of Southwest Java. Many people and fishermen live there, and some debris would have been seen and found. None was.

    The following are quotations from the cited Nesterov drift analysis originating at the Javanomaly, followed by dates of actual debris finds:

    “The debris field passes Cocos Island from late April to Late May 2014.”

    No debris was ever found washed ashore in Cocos Island

    “First arrivals at Madagascar and Reunion Island are as early as August 2014 and continue through early 2015.”

    The flaperon washed ashore in La Reunion in late July, 2015. MH 370 debris was not found washed ashore in Madagascar until June 2016.

    Debris from a Javanomaly crash site would have arrived where it did about a year earlier.

    Regarding the COSMO satellite photos, if they and French Pleiades satellite photos were from 370, that would mean a crash site at about 35.5°S. MH 370 debris would probably not have arrived in La Reunion and South Africa as early as it did. Drift simulations and analysis by Prof. Chari Pattiaratchi indicate the most likely crash site is between 32°S and 33°S.

  21. Victor Iannello says:

    @Miden Agan: I don’t know who you are, but you certainly have similar opinions to Blaine Gibson, including the statement that Blaine made to Geoffrey Thomas about my most recent post.

    The results of the drift simulations from Chari Pattiaratchi differ from those from David Griffin regarding the transport speed of the virtual drifters. In general, DG’s drifters move faster, so the predicted point of impact (POI) is further south along the 7th arc. DG was kind enough to share the drift results from BRAN2015 (and BRAN2016) so we can track each virtual drifter each day, and make estimates of the location of the POI based on the timing and location of the recovery of debris. DG also shared the model inputs for both “generic” and “flaperon-like” drifters, so we know how Stokes Drift and windage are included in the model. To my knowledge, CP has not published the details of his drift model.

    Since you claim that CP’s results are more likely than DG’s, perhaps you understand the technical details of DG’s and CP’s models. If so, can you please enlighten us about the differences?

  22. DrB says:

    @Miden Agan,

    You said: “Regarding the COSMO satellite photos, if they and French Pleiades satellite photos were from 370, that would mean a crash site at about 35.5°S. MH 370 debris would probably not have arrived in La Reunion and South Africa as early as it did.”

    The 35.5 S latitude estimated by Victor near the 7th Arc is the predicted location of the COSMO/Pleiades objects on the MH370 crash date. These objects, which are all much larger than the recovered MH370 debris, may not be from MH370, as several people have noted, for a variety of reasons.

    The CSIRO drift model analysis, as reported by David Griffin and his colleagues, indicated a swath of several degrees along the 7th Arc which is generally consistent with MH370 debris reports. This swath includes 35.5 S as well as the contiguous area to the NW along the Arc.

    My current analyses, using the daily predicted CSIRO tracks for 86,400 virtual drifters, appear capable of providing a much more precise crash latitude estimate from a statistical analysis using a collection of 17 non-redundant reports of identifiable MH370 debris. My method and results are currently undergoing a series of verification tests to prove their reliability and precision. That work will be published when the verifications tests are completed, assuming success. I can say, however, that currently the MH370 Point of Impact determined by this new method is to the NW from the predicted location of the Pleiades/COSMO objects on 8 March 2014, and it is roughly midway between that point and the crash latitude estimates you attribute to Dr. Pattiaratchi

  23. 370Location says:

    You can’t be implying that a reference to “grid search” and “south of Java” means the area between the Java coast and Christmas Island was ever searched. It would have been documented. Those planes were likely heading to Perth. Taking the very long way around Indonesia probably would have required fuel stops.

    @Miden Agan,
    It’s difficult to hear such certitude in statements about drift analysis that is inherently uncertain, especially for dismissal of the acoustic event site as a search candidate.

    You can read previous discussions over debris arriving earlier than it was found. We are still finding debris 6 years later.

    Debris not being found may be because nobody was looking. Absence of evidence is not evidence of absence. For the southern 7th Arc, some debris would be expected in Australia. It was watched for, but none was confirmed.

    The Java anomaly is unique among search candidates. It is not where anyone would look based on maximizing various probabilities. It is a very specific location where there was a loud noise. The true site might be well off the bell curve on drift, debris finds, BFO, FE, searched areas, or any other parameter being optimized. All that matters for a specific site is whether it is a feasible fit to the evidence we have. So, somewhere reasonable on each bell curve.

    I’ve seen people trying to dismiss the candidate for all sorts of reasons, but usually it comes down to it being unexpected. Prominent experts has called it a red herring without considering the evidence, as if it were intentionally created to distract from the real search like some doubters making up conspiracies. I believe that if the significance of the sound were recognized early on, the site would have already been searched.

    Note that COSMO didn’t take a “photo”. It uses radar which can be presented as a monochrome image. The SAR data is unfortunately still unavailable for reanalysis, like the CTBTO infrasound and recordings from several French hydrophones in the SIO.

  24. TBill says:

    I don’t know who Miden Agan is either, but as you previously guessed, it certainly seems like it might be the same person as Alvian Aldebaran from MH370 FaceBook Veritas. In any case he raises some good discussion points, and if it is the same as Alvian he has recently started a productive thread over on Veritas FB.

  25. Victor Iannello says:

    @TBill: I don’t have a FB account, so I really can’t follow threads in FB groups.

    Just out of curiosity, does Alvian Aldebaran believe there are any possibilities other than along the 7th arc where Chari Pattiaratchi proposes, or the Maldives?

  26. TBill says:

    Not sure.
    There is a poll going on Veritas re: suggested crash locations, that Alvian started, but I am not sure he gave his personal view yet. The poll has been good discussion.

    The one take-away that surprised me was, quite a few feel the original satellite photos of the debris deep south in 44 South area were MH370. This view requires various assertions, such as cover-up of debris that may have washed up on WA beaches (West OZ), disbelief of the radar flight path turning up the Straits, and so on.

    I am not comfortable using FB either, so I just read the discussions.

  27. Mick Gilbert says:

    @Victor Iannello

    Victor, do you think that any of the work currently being undertaken will pass the ‘credible new information, which leads to a specific location‘ test?

    Even if we had open access to the military radar data I can’t see how that would be useful in leading to a more specific location.

    I’m now of the view that the analytical phase of this effort is largely, if not completely, over. You and Bobby et al have one good candidate search area. Beyond that there’s possibly an argument for a far southern search area and there’s the unsearched northern area adjacent to the seventh arc. Putting together a phased search plan is likely not a difficult task now.

    However, the chances of any of the involved governments restarting the search is so vanishingly small that I’m of the view that it can be completely discounted. Realistically, the search is only going to be restarted if some non-government source of funding can be secured. That would likely be a very deep pocketed patron or patrons who might be attracted either altruistically or for the cachet associated with having solved the most intriguing mystery in aviation, perhaps a mix of both.

    Might it be time to start talking about what securing patronage might involve?

  28. Victor Iannello says:

    @Mick Gilbert: If we can increase our confidence that the satellite detections were debris from MH370, I think that would pass the test for “credible new information”. I’m not sure we can get there, but it’s worth trying. The COSMO-SkyMed detections provide evidence that the Pleiades detections were not sun glint, but the provenance of the objects remains uncertain.

    I would likely not be involved in an effort to raise private money to fund a search, although I think all of us would assist searchers with technical matters.

  29. Victor Iannello says:

    @Mike Gilbert: Let me add that before organizing another search, I would advocate for a thorough review of sonar data in targeted areas to better understand whether it is possible the debris field was missed in areas previously searched.

  30. Mick Gilbert says:

    @Victor Ianello

    Thanks Victor. On that last point, do you know who is the custodian of the sonar data? The ATSB?

  31. Don Thompson says:

    @Mick Gilbert

    The custodian of the Phase I (bathymetry) and Phase II (seafloor search) data is Geoscience Australia. A volumuninous open repository of unprocessed products from the various instuments and sensors plus the processed image products.

  32. Victor Iannello says:

    @Mike Gilbert: And to add to what @Don Thompson said, the custodian of the data collected by Ocean Infinity is Ocean Infinity. There was an effort to place the data in the public domain through a third party. I’m not sure if that ever occurred.

    However, the issue is not only access to the sonar data. To properly analyze the data, the correct software tools for reading and displaying the large data files is required, as well as the expertise to properly analyze the data.

  33. Mick Gilbert says:

    @Victor Ianello
    @Don Thompson

    Thank you both.

  34. Mick Gilbert says:

    An open question, if I may.

    If you were gifted a 125,000 search budget, how would you allocate it?

  35. Victor Iannello says:

    @Mick Gilbert: This might seem like a non-answer, but before making an allocation, I would want to understand better the probability that the debris field was missed in areas already searched.

  36. ALSM says:

    With a budget of 125,000 km^2, you could re-search the 7th arc 100 km wide from S30 to S38. It’s in there somewhere. Using Victor’s approach, the POI might be found with a quarter of that budget.

  37. 370Location says:

    @Mick Gilbert: With a gifted micro-search of only 125 km^2, I would survey the epicenter of the seismic Java anomaly that’s precisely on the low altitude 7th Arc. I expect the error of the seismic timing to be within just a few km.

  38. 370Location says:

    For those who have image processing experience, I wonder if it might be possible to use multiple passes of surface ship multibeam bathymetry scans to construct a higher resolution seafloor map.

    The 7th Arc was first explored with 150m and also 30m resolution multibeam bathymetry from the surface, which had small gaps. A second pass would have filled in most of those holes.

    There is currently a lot of active research in AI upscaling and super-resolution reconstruction from single images, with dramatic effect. There are also attempts to utilize multiple low resolution image captures to improve the noise and resolution of a computational result. Look at Google Pixel cellphone imaging for example.

    The Olex mapping system is an interesting approach to crowdsourcing of accurate bathymetric data for contributing fishermen.

    It is clearly expensive to send submersibles to scan large areas of the seafloor.

    It occurs to me that multiple scans using the best resolution from multibeam bathymetry could be computationally combined to improve the accuracy of not only the depth from echo timing, but also the signal strength that might reveal reflectivity.

    Multiple passes from a surface ship over a candidate area might be combined into a much higher resolution map with texture of the seafloor, all without deploying an AUV. Recruiting various expeditions on other quests to also cross candidate sites (multiple times) could computationally reveal evidence for sending down a survey vessel.

  39. Paul Smithson says:

    @Mick. You dont need a budget of 125000sq km. You need <1000 to cover a narrow swathe either side of 7th arc betweent 39.5 and 40.0S.

  40. Paul Smithson says:

    Arithmetic error. Make that <10,000 sq km. But still << a tenth of that budget.

  41. Mick Gilbert says:

    @Victor Ianello

    Thank you Victor, that’s a fair answer I think. How would satisfactorily improve your understanding of whether the prior search might have missed it? An independent expert review of the all the sonar data?


    Thanks Mike. Are you saying 50 km either side of the previously searched zone from S38 up to S30?


    Thanks Ed. Always nice to see some frugality. What is the location of the epicenter of the seismic Java anomaly, just by the bye?

    @Paul Smithson

    Thank you Paul, another less is best bid, nice to see. What swathe width would you nominate?

  42. ALSM says:

    Mick: Yes, 50 km each side (100km total width) X 1250 km along the arc. S30 to S38 is about 1250 km. I’m not recommending that specific area be searched. But if you had 125,000 km^2 to burn, that’s where I would put it. If I had only half that budget, I’d pick S32 to S36 100 km wide.

  43. Victor Iannello says:

    @ALSM: +/- 50 km is +/- 27 NM. That means you propose to mostly search what was already searched, and extend outwards a bit. So, do you believe there was a greater chance that the debris field was missed rather than there was an extended glide after fuel exhaustion?

  44. ALSM says:

    No Victor, I was just rounding off the numbers to equal 125,000 km^2. I figure, worst case…it’s got to be between S30 and S38 (~1250 km). So, starting with that number, that would allow 100 km wide. My Dec 9, 2017 best estimate was that it is within +/- 22 nm (90% probability), and I see no reason to change that estimate. +/- 27 nm probably bumps up the probability to maybe 95%?

    Other than the fact that it was not found, I don’t see any evidence supporting the “long glide” theory. Meanwhile, as we have discussed many times, there are several lines of evidence pointing to the POI being close to the arc. The Jeffreys Bay analysis further supports the in-flight partial breakup theory.

  45. TBill says:

    If we look at the Figure above in Victor’s article, a hard crash on Arc7 would seemingly spread a wide debris field plume over the whole of Arc7 width, and then some. Yet we have no underwater debris hits whatsoever from 25-38 South.

    Is there an argument to support that? I suppose we could argue large pieces that came off in mid-air floated away, and small pieces after the aircraft hit water catastrophically are undetectably small. But the question has to be asked, if it hit near Arc7, shouldn’t Search 1 or 2 have seen something?

    Now then I do believe some other areas of Arc7, such as BR, the debris drift will vector easterly, away from Arc7, so that might help explain no debris hits on Arc7.

  46. ALSM says:

    TBill: Re: “…would seemingly spread a wide debris field plume over the whole of Arc7 width…”. That is not what Victor’s diagram is meant to communicate. He is not suggesting that the debris is spread over that whole circle. He suggests it is concentrated in a much smaller field ( perhaps a couple of miles?) within the circle.

  47. 370Location says:

    @Mick Gilbert asks,
    “What is the location of the epicenter of the seismic Java anomaly, just by the bye?”

    Thanks for being curious. The Java Anomaly epicenter is at: 8.36S 107.92E
    The search area might be refined by seismologists with advanced tools.
    Lots of details are on my website:

    — Ed

  48. Victor Iannello says:

    @ALSM: One of the challenges in defining the search area is to assign relative probabilities of a glide versus the probability of missing the debris field. My approach would be to only re-search areas where the detection probabilities were low, i.e., areas with challenging terrain or where there were equipment issues.

    @370Location: There are many reasons to doubt that the debris field is located at your Java anomaly. However, if the epicenter really can be located with precision, then the “value per unit area” would be high, and there is a good argument for conducting a search there.

  49. ALSM says:

    Victor: I don’t see any evidence to support a “glide” after MEFE. So I agree with your approach. If it is outside the +/-22nm band, then it is still relatively close to that band…not 50-100nm out. More likely missed in one of the harder areas to scan.

  50. Mick Gilbert says:


    Thank you, Ed. That far north would make it quite the outlier but, as Victor said, if the search area was well defined and relatively quite small it would probably get a bang-for-bucks tick.

    @Victor Ianello

    To re-search or not to re-search … Picking the “hard to get at” spots between the latitudes of interest from the initial search areas and having another look at those would seem to be the most logical approach.

    Glide versus no-glide, that has been the argument from essentially Day 1. Battlelines on that front were drawn early on and there seems to have been little movement from either camp regardless of what the trickle feed of debris suggested.

  51. Mick Gilbert says:

    @Victor Iannello et al

    In response to a request in another forum I have done a Q&D amalgamation of the various bits and pieces that I have worked on with regards to the analysis of Z’s flight sim data. It is not my best writing, being largely a handful of emails bundled together, and could do with some expanded explanations and an illustration or two.

    I’m pretty sure that I have posted most of the key points on here at various times but for the sake of good order and to address my sometimes faulty recall, here it is –

    Accompanying also Q&D diagram here –

  52. Victor Iannello says:

    @Mick Gilbert: Thanks for posting. I agree that we have already discussed (and debated) the salient points here.

  53. TBill says:

    Thank you for that write-up.
    One point I noticed, you clearly state that ATSB is confirming the SimDate assigned for these cases is 2-Feb-2014.

    The reason I mention it, there seems to be some misunderstanding by some observers. Apparently there were some additional extraneous file fragments leaked, unrelated to these files, that show a different (later February_2014) SimDate.

  54. Victor Iannello says:

    @TBill said: Apparently there were some additional extraneous file fragments leaked, unrelated to these files, that show a different (later February_2014) SimDate.

    That’s doubtful. More likely, people are confusing the log book entries for an FSX flight as extracted from MK26. The flight files to the SIO are from FS9 as extracted from MK25.

  55. Joseph Coleman says:

    @Mick Gilbert

    Hey Mick interesting report could you provide if possible the co-ordinate for number 2, from your graphic, and also not up for questions or debate do you know the approx date of when the co-ordinates as seen in the sim data, where first known whether during or after the sim analysis by whoever that may have been.


  56. Mick Gilbert says:


    G’day Bill, I concur with Victor.

    There’s repeated confusion between the data files that we’re looking at and the later February date mentioned in the RMP report. Apart from anything else, that later date is for a FSX/PMDG flight file. That’s most assuredly NOT what we’re looking at here.

    I’m of a view that that file with the later February date was created by the forensics investigators as an example for their proof of method for identifying and extracting files.

  57. Mick Gilbert says:

    @Joseph Coleman

    G’day Joseph. The location of the sim aircraft just prior to it being dragged to 10N (ie the location where the flight in the Strait of Malacca ends) was likely near 6°4’N 97°34’E, essentially off to the left (south-west) of the track from 5N to VAMPI having turned off that track prior to reaching VAMPI.

    That is based on knowing where the aircraft was, its altitude and rate of climb at 5N, the 1,740kg fuel burn (64,440kg at 5N to 62,700kg at the drag to 10N) and its altitude, attitude, orientation and the Nav Radio data from when it was relocated to 10N.

    The Nav1Active and Nav2Active were tuned to the PUT and VPL VORs respectively meaning that the simulation aircraft had to be closer to Phuket and Langkawi than to the Medan and Banda Aceh VORs.

  58. TBill says:

    @Joseph Coleman
    The other point Mick is alluding to, that I like to stress: ATSB/Malaysia/FBI has all of the available data for these runs including SimTime and SimDates. However, the complete sim data is undisclosed to the public. We only have the limited portion of the data that was leaked in 2016, and staring in Oct_2017, ATSB has given important hints about the nature of the undisclosed data, as summarized by Mick.

    So we don’t know the exact time the sim aircraft hit near VAMPI/N10, but ATSB does indeed know that. Mick has a good guess, but just saying. What I see in my sims, almost exactly as soon you hit VAMPI, the radio frequency shifts and you are ready to move the aircraft to N10 (after initiating a 20 deg turn).

  59. Joseph Coleman says:

    @Mick Gilbert

    Thanks for the Info “You Guys”.
    Interesting that Uprob waypoint has been renamed to Pakra
    Noticing a spelling mistake on calling it Oprob within the PDF file.


  60. David says:

    @ALSM. Your Aug 4th, “My Dec 9, 2017 best estimate was that it is within +/- 22 nm (90% probability), and I see no reason to change that estimate. +/- 27 nm probably bumps up the probability to maybe 95%?”

    As you will recall, in late February 2020 @DennisW introduced us to Bayesian quantitative assessment of the effects of a search. As an example he used the UGIB area A1, having drawn the conclusion that the probability of the sunken wreckage being in that was 60% prior to the search. He deliberately chose a low probability of it being discovered if there, of 80%.

    The Bayesian approach offers two formulae, one to assess the remaining probability of wreckage-being in an area searched, after an unsuccessful search, the second to assess the effect of a failed search on unsearched areas.

    The formulae are simple and elegant but subject to interpretation as to how they should be applied.
    One check as to whether a method used is sound is to add all the possibilities that emerge as to where the wreckage might be. They should of course come to 100%, within calculator accuracy.

    In March 2020 I mentioned to @DennisW that the outcome of the all the possibilities of his analysis, was 123% and offered an alternative, that meeting the 100% criterion. I believe that his error was in applying the wrong Bayes formula.

    The outcome I came to, complying with that criterion, was that the post search probability of the wreckage being in the already-searched A1 area was 16%, having been 0.6*0.8, 48% before. Based on advice from @Richard to Dennis this was 80% of A1.

    There was also a rise in the probability that it was in the unsearched 20%, which had been 0.2*0.6, 12%, to 19%.

    Thus the post-search probability that it would be in the A1 area was 16% + 19% = 35%.

    The prospect of a search being successful is the product of the probability that the wreckage is in the area to be searched and the probability of finding it if there, that having fallen from 0.6*0.8, 48% to 0.35*0.8, or 28%. However if the prospects of finding it in a new search, if there, have risen from 0.8 to the ATSB’s postulation of about 0.95 then the success prospects rise to from that 28% to 0.35*0.95, or 33%.

    At 0.95 there is only 5% room left for improvement there. If Ocean Infinity’s substantial part in the search achieved more (or less, we do not know) than the ATSB’s then that would affect that 5%. Likewise it might be argued that a new search, using such as OI’s improved AUVs, would increase the success probability, though I expect the gain to be in the speed of searching. Here I continue with the 0.95 assumption as applying to both the past search and any future.

    For interest I have applied that to the UGIB A1 area in place of Dennis’s 0.8 while at the same lifting his 60% probability of the wreckage being in that area to your estimate for your area of 90%. The method used was that of my March 2020 post and the 33% success-likelihood outcome above. It is referenced in the attachment.

    The result is that the post search probability of it being in the searched area falls from 0.8*0.9 , 72% to 11%, while that of it being in the 20% unsearched area rises from 18% to 57%, that probability total then being 68% in place of 35%.

    Summarising so far, assuming a 90% probability that the wreckage was in the UGIB area A1, with a 95% probability that if there it would be detected, after a search of 80% of it the 90% has dropped to 68%.

    The hypothetical chance of success in a new search is 0.68*0.95 = 65%, is almost double the 33% of the earlier assumptions.

    The prospect of it being somewhere else than in the UGIB search area rises from the 10% unspoken for to 32%, as one would expect, to reach the probability total that it is somewhere, of 100%.

    More background generally to the above is in the attachment.

    Now on Aug 5th you posted, “More likely missed in one of the harder areas to scan.”

    Looking into that and what difference it would make by treating the difficult-to-search hilly terrain separately, also in the attachment are the workings of two related methods of assessing that, arriving at the same answer.

    They entail making some assumptions, again just for indicative purposes. Accordingly I have assumed that the hilly terrain’s area is 1/10th of that 80% of the A1 area searched and that the probability that it would be discovered in the search, if there, was just 0.2, or 20%, vice the 0.95 , 95% elsewhere, ie the probability of it being in that area is 1/10th that of it being in the 80% area as a whole. That is because, as @DennisW’s analysis implied, the probability of the wreckage settling in any part of A1, per unit area, is taken as the same, 0.6 in his case, 0.9 in this. In other words the aircraft does not know whether beneath the sea surface the terrain is hilly or not or, if piloted, the pilot would be aware of those hills.

    The first method entails separating the search of the 9/10ths from the 1/10th, as if the latter followed the former. Otherwise that method is as employed earlier.

    The second entails the one search but with the search area overall ‘find’ probability decreasing, that of the 1/10th at 0.95 being reduced by 0.75 to 0.2. Thus the overall find probability is reduced by 0.075, from 0.95 to 0.875.

    The outcome of the first method after search conclusion is that the probability of it being in the non-hilly 9/10th of the 80% searched area drops from 0.648 (0.9*0.9*0.8), or 64.8% to 8.8%; and the 1/10th hilly area rises from 7.2% to 15.6%. That of the second method total of the 80% area’s post-search probability is the same. For both methods unsearched 20% area rises from 0.18 to 0.4864864, or 18% to 48.6%.

    The “somewhere else” rises from 10% to 27%, thence total possibilities add to 100%.

    Thus the probability of it being in area A1 after search conclusion is 8.8% + 15.6% + 48.6% or 73%, so treating that hilly area as difficult to search has raised the probability of the wreckage being in A1 from 68% to a, likewise indicative, 73%.

    Reducing the probability of finding the wreckage over that 1/10th has reduced that attributed to the total 80% searched area, the second method indicating that it would reduce from 0.95 to 0.875.

    Again assuming that a future search would have a find probability of 0.95, success probability would be 0.73*0.95, or 69½%, whereas without hilly area being considered it was 0.68*0.95 or 64½%.

    For those who think that the assumption the hilly area would occupy a tenth of the searched area is too much I have looked into the outcome were that reduced to 1/20th. The new search success probability then is 70.8%*0.95, or 67%.

    Thus allowing for a lower find probability in the hilly area makes a difference.

    Summarising overall then, were your estimate of a 90% probability of wreckage to apply to the UGIB area A1, the outcome is that a likely probability of success in a new search would be around 67%. Quite substantial.

    As I mentioned above, before the search that was 0.9*0.95 or 85½%.

    While I think the 90% estimate is hard to support for the UGIB area, or for any area after a substantial part of it has been searched unsuccessfully, the above gives at least a clue as to the upper success probability bound.

    Use of the methods in the attachment could be extended to other possibilities, particularly the first.

  61. Victor Iannello says:

    @David: My personal opinion is that the probability of the debris field in A1 is closer to 50%, and perhaps not even that high, depending on the confidence we assign to the satellite detections.

    The satellite detections are a bit like the seismic location that Ed Anderson has found. Independent of which is more likely, there are reasons to doubt both, but the location precision is high, so those are high value areas to search.

  62. David says:

    @Victor. A post search probability of 50% that the wreckage was in the A1 area would require a like pre-search probability of 80.6%, assuming there was a 95% probability that if there it would have been found.

    @DennisW’s analysis had that 80.6% somewhat lower at 60%.

    The success likelihood of a new search of A1 at your 50% and at the same find probability of the previous search would be 47.5%.

    As to the economics of high value spot checking, I imagine the search cost per unit area could decrease with time, as set up costs of specialised equipment, expertise and deployment were defrayed.

    But the opportunity costs of searching other than in a quiet commercial tasking period might constrain continuous search length.

    In which case opportunity duration might be key?

  63. Victor Iannello says:

    @David said: But the opportunity costs of searching other than in a quiet commercial tasking period might constrain continuous search length.


    At this point, I see zero interest from any party that has the resources to finance or conduct a search.

    Instead, I see advocacy from different parties that believe the next search should be conducted at a particular area along the 7th arc. With the data sets we have in hand, I think it is unlikely that any one of the parties will compel the convergence towards a consensus area to search. Most (but not all) of the parties have been quite transparent in providing analysis to support their preferred area.

    I think the satellite detections have the potential to compel at least partial convergence among various parties. Unfortunately, that hinges upon positively associating the detections with debris from MH370 with some reasonable degree of certainty. At this point, that remains elusive, and that situation might not change.

  64. David says:

    @Victor. As to motivation, the incentive to find the wreckage to improve aircraft safely has diminished with increased likelihood that there are no airworthiness implications.

    Finding it might bring closure to some of the next-of-kin while re-opening old wounds for others and there is the risk for all that a new search will again fail.

    There is no FBI interest evident in finding it, to further whatever it has in-waiting.

    Motives that remain are just challenge, curiosity and potential achievement, yet these have been enough, given time, to find a variety of lost wrecks. The Amelia Earhart mystery attracts interest still.

  65. Mick Gilbert says:

    @Victor Iannello

    To take a less conventional approach to determining a search area, might you start with the largest area needed for “guaranteed” success and then eliminate the lower probability areas.

    To that end where would you sink your fence pegs in order to define an area that would yield a 99.9 percent chance of covering the underwater debris field? In terms of ±x nm either side of the 7th arc bounded by y°S and z°S, what are your values for x, y and z?

  66. Victor Iannello says:

    @Mick Gilbert: There is no “guaranteed success” until we better understand detection probabilities. We have the luxury of time since there is no new search on the horizon. I would recommend spending time on reviewing the existing data and developing and testing methods to improve detection probabilities, if that is possible.

  67. CanisMagnusRufus says:

    “the incentive to find the wreckage to improve aircraft safely has diminished with increased likelihood that there are no airworthiness implications.”

    I know that airworthiness is not the remit of ICAO, but it doesn’t help that the person in charge of ICAO these last few years appears to have not done a good job.

  68. David says:

    @Mick Gilbert. I do not think that there can be a 99.9% success probability construct for any underwater search or anything close. I expand on some of the issues in the attached, FWIW.

    To come up with a datum of a lesser maximum success probability that could be attained with confidence would, I think, require all possibilities that it was not in that area to be listed and quantified. Thence plotting a lowering success probability against reducing area would entail gradually shedding possibilities. Yet that might be achieved in multiple ways, leading to a great complexity, with each possibility being assessed and some being interdependent.

    In short I do not see a sensible way that can be done, though I am not well versed in statistics.

    Just for one example, what would be the effect of discounting the possibility that there was a pilot?

    That would reduce the search area by possibly shortening the glide distance, though there is no certainty as to how far the aircraft could have glided unmanned or what would be the probability variation from change of piloted search width (eg from glide course). Also though it would reduce the probability there would be course or altitude changes or a ditching; but again with what probability and with what effect elsewhere?

    Then there is the well-known possible effect on fuel consumption (after a/c deselection).

    So there would be an array of answers.

  69. Mick Gilbert says:

    @Victor Iannello

    Okey doke, let’s lower the bar and simplify this a bit then. Hands up if you think that there is a better than 95 percent probability that the underwater debris field lies somewhere within a area described by 250 nautical miles either side of the 7th arc between latitudes 8°30’S – 43°S.

  70. David says:

    @CanisMagnusRufus. Thanks. Maybe the US will raise that with the UN.

  71. David says:

    @Mick Gilbert. Yes.

    To jump to the end game, I would allow 5% for the most way-out theories and unknown unknowns, lopping that off any further search prospect at the outset.

    Some of UGIB A1, the most prospective unpiloted area, which also covered short-glide piloted, has been searched, leaving the remaining 20% of that as the highest priority per unit area.
    (In passing, the UGIB hardest-to-search area was within the ATSB’s search and was searched subsequently by OI. Assuming that was thorough a research of that should await any research of A1.)

    An unsuccessful search of the rest of A1 lifts the likelihood that the wreckage is in the piloted long glide area. That is more likely to extend to less than maximum glide distance from the arc, that entailing his flying square to it by chance and not choosing to continue on the powered course.

    I would think it extremely unlikely that from very high altitude and an accelerated speed at imminent fuel exhaustion that his glide would exceed 150 air miles.

    As to arc length, 30 to 42 deg, that including the glide, the whole to be adjusted for wind integrated with descent time, the arc being that for a 45,000 ft glide start.

    The highest priority after the 20% unsearched UGIB area, would be its A2 area to 100 air nm width followed by that along the rest of that arc section to 100 air nm, extending from that searched, both sides.

    The Java anomaly and Cocos Island vicinity have weak evidence, there being no explanation of how the aircraft was connected to the noise registered at Java, no sign of survivors or any emergency communication if it ditched. Still, a point search there would be warranted as convenient on the way to or from the main site, the larger Cocos area to follow last were there time.

    But back to the main area, some Bayesian sums could be done on whether concluding the long glide search to 150 air nm or to research the already searched areas though I suspect that the former would be the better prospect.

    On that, it may be that there would be little interest at that stage in a re-search of the areas searched. It seems quite possible that the low prospects there have deterred Ocean Infinity.

    That is understandable. As I mentioned previously, using @DennisW’s being-there and find probabilities of 0.6, and 0.8 (the latter was deliberately conservative) the post-search probability of it being in the searched area was 16%, with that of the unsearched 20% rising to 19%, the total 35%.

    That 16% is unattractive. However were the ATSB’s (and quite likely OI’s) find probability of 0.95 used in place of his 0.8, then that 16% drops to 4.4%, the unsearched rising to 22%, total now 26%.

    If in a new search the search of the unsearched 20% then fails, the post search probability of the total A1 area drops to 5.6% plus 1.4%, 7% , so unattractive for a search at that point, though a failed search of the 100 air nm glide area would lift that (how much could be estimated now).

    Once either, both or neither 100 to 150 nm or re-search are concluded a search of Cocos (and Java if not done already) could be considered.

    If they and the main area’s search and re-search are concluded I estimate that would amount to the 95% it-being-there probability. Without Java and Cocos perhaps that would be 92% by my guess.

    However the search from 100 nm to 150 nm for the glide, all re-searching and Cocos could be foregone on the basis that offering a low probability per area searched.

  72. ventus45 says:

    How about tackling the problem backwards.
    Perhaps it is “hidden in plain sight” so to speak, but wider than what has thus far been searched.
    1. Convince the people doing the basic Bathymetric Mapping of the worlds oceans (international project) to do our widest possible areas as a matter of priority first. That way, we get the information for free, without it being “a MH370 search”.
    2. Then, armed with that new Bathymetry, select the “easy” to search areas (benign seafloor, no mountains / canyons etc) and convince someone to search those “easy” areas first, with whatever level of technology will give a confidence of 95% or better.
    If it is found in an easy area, all well and good, but if not, by having eliminated those easy areas with confidence, we will have raised the probability of it being in the new difficult areas, or missed in the previously searched areas.
    (donning tin hat)

  73. TBill says:

    I wish you had said +- 300 nm or 350 nm.
    If MH370 hit Arc7 at 38S, there is only glide of say 125 nm to consider.
    If MH370 hit Arc7 at 30S, now the fuel models suggest more fuel more may be available. So you need to draw the error bands based on fuel model. In other words, I am not so sure Arc7 is fuel exhaustion.

  74. Victor Iannello says:

    It’s fascinating to watch FR24 track the international flights in and out of Kabul Airport and also the helicopter flights between the US Embassy and the airport.

    Let’s all hope this rapid transition to Taliban rule occurs without injury and death.

  75. TBill says:

    @Victor @DrB @Andrew
    I was giving some further thought to the unknown of MH370 fuel density, and maybe it’s a wash. Higher density gives more power, but less gallon of fuel.

    I did read some guidance from Chevron who mentions that is a complex question, but in general higher density is better even in commercial airliners when the tank is not full. But perhaps it is not a big effect.

  76. Victor Iannello says:

    @TBill: The energy per unit mass is relatively constant. That’s why kg and lb are much better measures of range.

  77. Victor Iannello says:

    All FR24 activity around Kabul has ceased, and there are reports that the airport is under fire. Prayers for all those that are trying to evacuate.

  78. Victor Iannello says:

    If @DefenceGeek on Twitter is accurate, the air operation into and out of Kabul is massive. (I’ve seen several of these flights on FR24.)

  79. Don Thompson says:


    Political/diplomatic issues evident – Indian AF C-17s routing south around Pakistan, then north over Iran to reach Kabul. KC-135R operating as SHELL26 flying orbits at typical refuelling altitude, there’s a lot that is not visible!

    I’m surprised that some civil flights continue to overfly Afghanistan while some appear to be routing around. At this time, two UAL flights to Newark are about to cross into Afghanistan.

  80. Victor Iannello says:

    @Don Thompson: Yes, a lot of the military traffic is blocked, but some is not.

    As for civilian flights, there is an FAA NOTAM restricting flights in the Kabul FIR only below FL260.

    So far, the media has been quiet about the extent of the airlift operation, possibly because they are unaware, possibly because they’ve been asked. As far as I can tell, the information from @DefenceGeek and @vcdgf555 is accurate.

  81. Mick Gilbert says:


    Okey doke, thanks David. So how much do you think you would pare back the probably from 95 percent (or 92 percent, ex-Java/Cocos) by bringing the the search width back to ±100 nm?

    Thanks Bill. So what do you think caused the re-log on and the recorded high rate of descent if not fuel exhaustion?

  82. Victor Iannello says:

    @Don Thompson: All commercial traffic over Afghanistan has been re-routed. The Kabul FIR is under the control of the military, and no ATC services will be provided to commercial traffic.

  83. TBill says:

    It could still be active pilot managing the end of flight including power settings and APU, either accidental or on purpose causing SATCOM reboot.

  84. David says:

    @Mick Gilbert. To add to my previous, a precaution before a new search would be for the searcher to check that the calibration process for the search equipment was previously, and will be, satisfactory.

    One other feature that I have looked into quantitatively but overlooked in my earlier posts, is drift of the sinking wreckage with current, particularly the slow settling wreckage. Apart from lateral displacement of all, there is the effect on bottom spread.

    Also, @ventus45 raises the issue of bathymetric survey, which might be an added cost if his proposal were not realised, to extend search width.

    Now about your question, in brief, the probability of it being in the outer glide area before there was any search, by my estimation, which included assumptions, was 5.33%.

    However it is not that simple. For example, after unsuccessful search of all of A1 and the inner glide area, then the probability of the wreckage being in that outer area would be 30%. That is expounded below.

    For an analogy, applying the’100%-it-is-somewhere’ principle to the Bayesian equations leads to something like the figurative punching of a pillow. As a search is conducted, the pillow is dented but figuratively there is still the same total volume, the rest expanding.
    After an infinite series of unsuccessful searches of prospective areas, that unprospective 5% we have talked about would have swelled to 100% probability.

    Now getting into that, what is needed as a first step is a break up the original being-there probability of 40% ‘elsewhere’, ie outside the as-yet unsearched A1 area, that consists now of the Java plus Cocos area’s 3%, the way-out and unknown unknowns at 5% while the remaining 32% is the aggregate of the half-width extension to 100 miles and that from 100 to 150 miles.

    I divide that 32% into two by supposing (for this example) that the inner 100 miles part is currently searched to an average of 30 miles so is 70 miles wide, the outer part being 50. I suppose also that the probability tapers off linearly with width to zero at 150 miles.

    The result is a five to 1 ratio between the two. Thus I allocate 5/6ths of that 32% to the inner, 1/6th to the outer, i.e. 26.67% and 5.33% respectively.

    For simplicity in this exercise I do not separate out the A2 glide area.

    In summary then the probabilities outside the A1 area before there is any search would be, Java and Cocos (‘JCI’) 3%, the way-out and unknown unknowns (‘U’) 5%, A1 60%, the inner glide (‘IG’)area its 26.67% and the outer (‘OG’) its 5.33%.

    Post the A1 search, the 80% already done, but now adding an also-unsuccessful search of the 20%, those change to A1 7%, as posted previously, then JCI 7%; U 11.6%; IG 62%; OG 12.4%.

    Note that choice of the inner glide as next to search is obvious

    Not quite so the next, after an unsuccessful search of that IG the probabilities then become; IG 7.5%; OG 30%; A1 17%; JCI 17%; U 28.3%.

    This is the 30% I mentioned earlier that the glide area from 100 to 150 miles rises to in this example.

    The next candidate indicated is the outer glide as you can see After an unsuccessful search of that the results are; OG 2.1%; IG 10.6%; A1 23.8%; JCI 23.8%; U 39.7%.

    The above searches all suppose a find probability of 0.95, as gone into previously and all meet my simple 100% it-is-somewhere test, though being calculated individually.

    As I can now see, having these data before me, my previous post’s last two paragraphs, being before the glide search calculations, drew a wrong conclusion. There can be no elimination of the possibility that the wreckage is in a searchable area until there is 100% chance that it is in the unsearchable area, whence the search would have met the 95% probability in mind. Again, that would require the pillow’s infinity of searches.

    As you can see the next candidate after the glide areas are searched is, only then, a re-search of the A1 area – or to do the Java/Cocos island areas. The latter’s probability has risen to 23.8% from 5% and now the prospect equals an A1 re-search prospect.

    So thank you for your question.

    In conclusion, this type of calculation, easy to do, can be a useful discipline and it does provide insights.

  85. Don Thompson says:


    Traffic over recent days: C-2A Greyhound making trips between the USS Ronald Reagan, UAE’s Al Minhad AB, and the naval support base at Manasas, Bahrain. Carrier spotting via ADS-B!

  86. CanisMagnusRufus says:

    The Malaysian PM has resigned. Top contenders for replacing him include the current Foreign Minister

  87. Victor Iannello says:

    @Don Thompson: A C17 registered as A7-MAP, owned by Qatar, and operating under flight LHOB247, left Doha, Qatar, and recently landed in Kabul carrying Taliban leader Abdul Ghani Baradar, who some say will become the new president.

  88. airlandseaman says:

    The top contenders to be prime minister include Muhyiddin’s deputy Ismail Sabri Yaakob, veteran lawmaker Tengku Razaleigh Hamzah and former foreign minister Hishammuddin Hussein, all from UMNO.

  89. Victor Iannello says:

    @airlandseaman: So Malaysians might get more of H2O? I wonder if his association with MH370 will damage his chances to become PM.

  90. airlandseaman says:

    Victor: Yes. That is what they are reporting. Not a good choice for MY.

  91. Andrew says:


    According to some Malaysian media outlets, Ismail Sabri (current Deputy PM) has been endorsed by the major parties as the next PM. The King is expected to announce the appointment on Friday.

  92. Mick Gilbert says:


    Thank you for your typically considered and detailed response. With regards to the calibration of equipment, I would think that it would be reasonable to expect that was the case given the organisations involved.

    Interesting to note how re-searching has slipped down the ladder.

  93. TBill says:

    Anwar Ibrahim’s hopes fading, they say.

  94. Victor Iannello says:

    Relative to MH370, I really don’t think it matters who becomes PM of Malaysia. As far as I can tell, there is not a country in the world where there is political will to re-start a search. Nor has any private company with the requisite resources and expertise expressed a desire to re-start the search. That might change if a small area was defined that had a high chance of success. With the evidence we currently have in hand, that also seems unlikely.

    I still maintain that the best path forward is to review the existing sonar data, assess its quality, and determine if there is a reasonable chance that the debris field was missed due to incorrect classification or due to low quality or missing data. If that does not occur, it becomes difficult to make new search recommendations.

  95. TBill says:

    One question would be the status of university studies on flight path reportedly initiated by Australia a few years ago.

  96. 370Location says:

    @Victor @David and @All,

    It seems that the Java Anomaly search area is so small and just far enough from the main search focus that it keeps falling into the blind spot of our perspective. It gets briefly recognized as “high value for unit area”, but then somehow ranked by @David at 3% likelihood relative to broadening areas already searched. That pushes it back into the blind spot. Now @Victor’s conclusion that a small area must have a high chance of success before anyone with the resources would bother to look, and so declines to even recommend it. Back into the blind spot.

    The search area is small enough that a single pass of a tow fish, or deployment of an AUV with sidescan sonar could rule it out. That might be part of a military training exercise either by Indonesia, or anyone with permission to search in their coastal zone. Multibeam bathymetry could be done by any capable ship transiting the area. A private operation like Ocean Infinity could completely scan a wider area around the estimated epicenter in less than a day, vs the significant resources required to broaden the searched 7th Arc.

    So, why not recommend multiple pinpointed searches for candidates with sufficient accuracy? Wouldn’t that restarting point build interest and support for resuming a broader search if nothing shows up?

    I have provided ample consistent evidence for the Java Anomaly site. It doesn’t seem to be enough to escape the blind spot, so I continue to dig deeper with the data that I can get hold of. If there were a recommendation to search the Java site, it might release restricted infrasound and hydrophone data. There were two coastal radar installations on Christmas Island watching for refugee boats that could be examined. The Cocos infrasound validation alone could narrow the overall 7th Arc search area, as recently proposed by Michel Delarche in his five part review:

    It’s gratifying to see any review of my acoustic analysis in the media, and utilization of the Coco Island flyby. I do think the author also has a blind spot for the Java event. He is focused on optimizing BFO, and forgets that the loud noise event is the basis for heading toward Java.

  97. Victor Iannello says:

    @TBill: I’ve seen some of the results of the university studies that you are referring to. To be kind, the conclusions were less than compelling.

    @370Location: I don’t think there is a high probability that your Java anomaly is associated with MH370. However, I don’t dismiss that possibility. If I were to recommend a search plan, your hotspot would be part of it, as would other areas.

  98. David says:

    @Mick Gilbert. My thanks for your interest.

    By calibration I meant check how much the equipment could be relied upon to detect the wreckage.

    IMO that would include testing using realistic targets on a realistic bottom at a realistic depth, not sheet mild steel on a hard flat plain shallow bottom, much of the steel flat and perpendicular.
    It did not include scrap composites and alloy structures, wrecked engines or undercarriage representatives.

    Certainly wrecks, drums and wires were detected during the searches and there were informal tests at depth but I for one would like to see that ingredient of the supposed around-95% probability of detection if there, verified, before continuing. Still, more cost and effort.

    As to where that hypothetical multiple search should end logically, assuming it has got as far as I have projected, were there subsequently an unsuccessful search of Java/Cocos Island and likewise a re-search of A1, the residue of way-out possibilities and the unknown unknowns (‘U’ earlier) would now amount to 72.5%, the highest probability offering being 19.3%, for a re-search of the inner glide area.

    At this point the likelihood of the wreckage being in unknown places is almost 3 times that of being in the searchable areas. That might be a reasonable point at which to draw stumps on further searching.

    Drawing such a conclusion of where to do that in advance would allow calculation of the maximum search area contemplated.

    @Victor. Given that funds and time for a new search most likely will be limited, return for the dollar/hour would entail deciding what hot spots there are, how separated and whether to search just those. As I understand it there would be just the those of A1 and Java, though obviously for any such brief search there would need to be reasonable grounds for believing the wreckage was in those.

    Earlier I attempted to simulate a searched hot spot by allocating first 1/10th of the A1 area to that hot spot, then 1/20th. In doing so I assumed that the probability the wreckage would be found in that area if there was just 20%, vice other areas’ 95%. Also in that for the reason I stated I allocated a 90% probability of it being in the A1 area pre-search, in place of @DennisW’s 60%.

    Using my method 1, search of the 9/10ths part of the 80% area searched, unsuccessfully, resulted in the probability that it was in 1/10th rising from 7.2% to 18.7%. Post search of the 1/10th at a 20% find probability, the probability that it was there, undiscovered, dropped to 15.6%, from that prior 18.7%. That is not much of a drop: the residual 15.6% is much higher than the rest of that 80% area, which dropped from 64.8% to 8.8%, the total for a new search of all of A1 being 73%, whereas without separating out that difficult to search area it was 68%.

    Reducing the hot spot area by half, 1/20th, it becomes 70.8%.

    Bear in mind please that the outcome of these figures illustrating the quite minor effect of reducing the likelihood of finding the wreckage in the A1 hot spots to 20%, might be less still if their area is even less than a 1/20th.
    Also, in case of confusion from this plethora of probabilities please bear in mind that the search sequence gone through most recently is based on DennisW’s 60%.

    Now returning to the hot spot’s probability being treated as the same per unit area as elsewhere in A1, let’s assume that all of the probability that it was in that 80% A1 searched area, i.e. 0.6*0.8 = 0.48, was taken to emanate just from the hot spot areas, 40% remaining outside A1 still. The effect of the search of that 80% of A1, containing those hot spots, the probability of finding being reduced from 95% to say my 20%, is that its subsequent probability would be 42.5%, the 20% unsearched now 13.3%, the total being-there probability now offering being 55.8% therefore.

    Also, based on the assumption that it was in that area and was 1/20th of it, the total A1 area to be searched initially would drop to 80%/20 plus 20%, or 24%, that is by about three-quarters.

    Thus with high confidence that hot spots were in, and limited to, identified small areas, the best approach obviously would be to search those first, the unsearched 20% second and quite possibly the 80% next. The downside cost would be realised only if the need for a re-search of A1 did not arise.
    I could do another example starting from completion of the 20% should that be of interest.

    @370Location. My intuitive 3% estimate was after having looked into the Java likelihood with you and separately into that of Cocos. IMO lifting that further up the list would be on the basis that as you say it is a low effort search and in an unsearched area. The Bayes outcomes do need to be complemented by a look into their probability density, supposing that any search will be time and cost constrained.

    Even so, despite being at 3% initially, using the Bayesian formulae, Java with Cocos did get a guernsey as the probability that the wreckage was in other searchable areas decreased.

    Somewhat counter intuitively, separating it from Cocos within that initial 3% would lead to its search prospects being lower.

    About the deLanche papers, I have yet to read them but in case others have the same issue I accessed the in-French versions and translated them, being unable to access the already translated items.

  99. David says:

    @Victor. Correction please.
    In the above the first sentence of paragraph 6. of that addressed to you starts with, “Now returning…..”
    After its first comma that should have read, “…., let’s assume that all of the probability that it was in that 80% A1 searched area, i.e. 0.6*0.8 = 0.48, was taken……..”

  100. Victor Iannello says:

    @David: Corrected.

  101. David says:

    @Victor. Thanks.

  102. CanisMagnusRufus says:

    ‘Sully’ Sullenberger nominated for the post of Ambassador to ICAO in Montreal.

    He is a well recognized, public figure. Wonder why the Biden administration chose him?
    Are they expecting issues to come out of ICAO, and want a sympathetic figure like him to deal with the media?
    Or is it just to garner favours from some coveted constituency?

  103. Victor Iannello says:

    There is an interesting site that collects and presents recent and live Inmarsat ACARS and ADSC messages, among other things.

    To search for messages from a known flight or plane, use this engine. For instance, after using FR24 to identify flights operating in Afghanistan, you can use the search engine to identify messages to and from the planes.

  104. Don Thompson says:


    Thank you, interesting exchanges there. TIL about Ravens!

  105. Victor Iannello says:

    @Don Thompson: I agree. I was not aware of the extent of the security measures taken for flights with passengers. I assume that messages with references to Ravens (which can be identified by searching with the keyword “Ravens”) refer to the elite Phoenix Raven program.

  106. Don Thompson says:

    The resources going into the Kabul airlift are astonishing.

    I followed one USAF C-17, 06-0615, over the past few days. It set out from Nellis AFB (20210826T0040Z) on a first leg to Joint Base Charleston (arr 20210826T00440Z, dep 20210826T0905Z), second leg to Ramstein AFB, Germany (arr 20210826T1755Z, dep 20210826T2140Z), third leg to Al-Udeid AB, Qatar. Then departed Al-Udeid for Kabul, crossing the PK-Gulf of Oman coast 20210827T1100Z, returning over that coast at 20210827T11820Z bound for Al-Udeid. This morning it departed Al-Udeid (20210828T0640Z) bound for Sigonella, Sicily.

    Late Friday evening, stresses appeared to surface during C-17 99-0062’s ex-Kabul leg while the crew were enroute, heading toward Kuwait/Al Salem AB. Their flight plan appeared to confirm destination as Al-Udeid. The crew received four separate instructions to turn around and proceed to Al-Udeid before doing so.

  107. Victor Iannello says:

    @Don Thompson: I wonder what was going on with that C17.

    In any event, the airlift should be wrapping up.

  108. Don Thompson says:

    @Victor Iannello,

    Wrapped up now. Last C-17s departed OAKB just prior to 1930UTC. A train of KC-135Rs heading south towards the Gulf of Oman now, 2030UTC, no doubt ‘hauling’ a flight of fast jets back to Al-Udeid/other regional airbases.

  109. Victor Iannello says:

    @Don Thompson: Thanks for the update.

  110. Victor Iannello says:

    Question for the group: Is anybody aware of any ongoing research or investigations that might lead to better estimates of the POI? With the lack of new evidence, I see Twitter, FB Groups, and Reddit overrun with meaningless chatter. Bobby continues to quietly refine his drift analysis, and we are trying to obtain better satellite imagery. Is there anything else of any substance?

  111. POI says:

    From my point of view there’s still nothing else of any substance other then the previous (and recent) views of your Independent Group failed to help finding the plane when the opportunity was still there.
    Your persistent confirmation-bias the plane must have dived (uncontrolled) close to the arc contributed a lot to the efforts and the failing of the final search from OI.

    Your persistent refusal to advice them to search at least ~20 miles wider each side between ~32 and ~35S beyond the already searcehd areas has contributed to this failure a lot I’m sure.

    Your ‘Group’ advised them to go north till 25S instead. Which was totally useless and contradicted all your previous calculations and views. Driven by guys like ‘Richard Godfrey’ and Mike Exner you got caught in a whirlwind of confirmation-bias that contributed to the failure of finding MH370 big time.

    The plane is resting on the bottom of the ocean not that far beyond the searched area between ~32S and ~34S east of the 7th arc. All calculations predict this area.

    The best you supplied is mainly very detailed negative evidence brought as positive evidence. Which is very usefull if you are able to look at it this way.
    All your efforts/evindence prove the plane is not where you expected it to be. This should be a wake-up call to serious people/scientists.

    But regarding the past years I don’t expect one of you is brave enough to face their failures or let alone change their views.

  112. Don Thompson says:

    @Victor Iannello

    I now have the GFW fishing vessel data to hand for the period around acquistion of the SAR and Pléaides acquisitions. Will sift, plot & report on that presently.

  113. TBill says:

    Some words.
    I thank the IG for prodigious technical contributions to our understanding MH370. I have two special MH370 mentors or heroes, and Blaine is one, and you Victor are the other. Blaine for finding NO STEP, which inspired me to start to get involved. But it was your MH370 technical philosophy and papers that I latched onto as the closest thing to my own personal inclination. I now have “grown” to the point I have my own philosophy, which is a lonely place.

    For the last several years my analyses (though admittedly not as rigorous as IG’s) make me feel that MH370’s Arc7 crossing point may be closer to 31-32s. But obviously I could be wrong. That is the nature of MH370.

    Thank you for your work.

  114. Victor Iannello says:

    @Don Thompson: That’s great work, Don. It’s surprising that this comparison wasn’t performed previously.

    @TBill: Thank you for the kind words.

    @POI: You sound a like lot somebody else that used to continuously promote searching in that area. As far as our advising to only search close to the 7th arc, that is incorrect, as you can easily see by recent search recommendations.

  115. TBill says:

    Ge Rijn certainly had a strident BR hypothesis but I do not recall if had a hunch basis. You know my approach is more flight sims and BTO/BFO analysis. As far as my thinking, I started with your NZPG path, moved over to 180s True path to get a better BFO match, and feeling that was not the best match either, thought it may be either 38s or over to 31-32s w/ various paths but 180s Mag. Tried 38s on for size, but never grew to like the 38s option.

  116. ventus45 says:

    I have been silent for some time, just monitoring.

    What I see now, is that the level of solidarity of purpose and methodology, that has been the collective strength of this group from the beginning, is showing the obvious first signs of real strain, with growing evidence of rising frustration in many, and possibly even a sense of despair in a few. But worse, I worry that once people start playing “the blame game”, that any sense of common purpose will evaporate quickly.

    None of us want that – so stop it – immediately.

    Dennis is no longer with us, but he is probably watching from on-high, and I am sure, that if he is, he would be grossly displeased.

  117. TBill says:

    Some perspective, I think we all had great excitement and some consensus for 2018 OI search, which accomplished so much (despite the negative result). Having now accomplished searching close to Arc7 from approx 25-39 South, there is naturally going to be less consensus moving forward, especially if we cannot continue to have OI searching to rule in-or-out new ideas.

  118. TBill says:

    I just noticed your comment above. Not sure how I missed it.

    I had mis-interpreted Victor’s comment above, I took Victor’s comment personally as a criticism of my personal PoI. Now I see I was wrong and that Victor was responding to you.

    Although I probably totally agree with your current assessment of the need to search much wider from Arc7, at the time of the 2018 OI search, we had a major theory that the Captain’s simulator data pointed to an NZPG waypoint. The NZPG waypoint theory points to an Arc7 crossing something like 26-29 South depending on how one does the calcs. Therefore I was (at the time) in complete support of IG and OI searching up to 25 South, even though I personally have a very strong BR hypothesis, to be honest.

    However, I think we now know, thanks to ATSB, that Malaysia was redacting the simulator data. Thanks to Mick Gilbert’s pursuit of further ATSB guidance, we now know that the NZPG waypoint theory is not supported by simulator data (as we now understand it). I personally believe, and I have a new 2021 paper, that the sim studies may represent a 180 South CMT magnetic flight path.

    DrB in 2018 showed a number of such “magnetic” paths to 31.5 POI area. Before that, @Nederland also showed some ~31 deg South path versions that seem promising.

    However, it all boils down to Malaysia is hiding the simulator and other data that could have substantially changed the overall crash location interpretation. I my view, IG/Victor’s 2016 interpretation of the simulator was superior and extraordinarily on target. However, Victor had “one hand tied behind his back” because the complete sim data was being hidden from him and the world.

    I feel the sim data would probably have been explosive, had we gotten the data before OI’s 2018 search. As it happens, Malaysia successfully hid the “complete” simulator data long enough for the world to lose interest move on.

  119. TBill says:

    Edits above:
    (1) I actually like CMH or CTH path, but the point is a path to 31-32s region
    (2) “In” my view

  120. ventus45 says:

    You said: (my highlights) “I feel the sim data would probably have been explosive, had we gotten the data before OI’s 2018 search. As it happens, Malaysia successfully hid the “complete” simulator data long enough for the world to lose interest move on.”

    By “explosive”, are you suggesting (perhaps even obliquely) that the Malaysians “deliberately” released only some elements of the “complete” simulator data, and “deliberately” hid certain other elements of the simulator data, “specifically to ensure”, that any search area that was deduced based on that incomplete data, would certainly be wrong, and thus, any search actually conducted in that area, would certainly fail ?

  121. Victor Iannello says:

    @TBill, @ventus45: My guess is that the Malaysians did not redact the simulator data. I suspect that the Malaysian investigators and the FBI independently analyzed the data, and the amount of data that was successfully extracted was not exactly the same, although there were significant file sections that were found by both groups.

  122. George Tilton says:


    Are you making the point?: “Never ascribe to malice that which is adequately explained by incompetence.” (Napoleon Bonaparte)

    My experience (from 30 years working in three-letter agencies) is incompetence wins every time…

  123. TBill says:

    Interesting interpretation not sure, is why I hedged a little bit above. So the question for ATSB is, how long have they had the so-called “complete” data.

    I am suggesting Malaysia hid the simulator data because to my eyes, the complete data we now know from ATSB is very incriminating. I have become more convinced last 12 months, it looks like an obvious MH150 diversion plan. I now speculate FBI saw that right away, as early as Week-2, and advised MY.

    When I say explosive, think back to Jeff Wise’s original article about the sim data in the New Yorker. If, at the time, he only knew the full extent of the data, it could have been a much more persuasive article. Not to mention, that article was already 2+ years after the accident. If we expect the public to understand MH370 without creating conspiracy theories, MY needed to say what the simulator showed relatively soon. An MH150 diversion plan could be a major national security crisis, so Malaysia might have needed a couple of months to “batten down the hatches” but the public needed to know in due course.

    Malaysia was simply trying to hide the sim data which may implicate the Captain, I assert. The partial, leaked sim data we finally got in 2016 implied NZPG waypoint, however, the more full understanding we now have from ATSB suggests NZPG might not have been the strategy (no flight path in the sim data).

    The other indication NZPG is not correct, is that passive flight path to NZPG does not match BFOs (eg; Arc5) very well. That’s another hint. But in those days we were not sure how much scatter could be in the BFO data.

  124. Victor Iannello says:

    @TBill: The first mention of the simulator data potentially representing flight MH150 was in this comment of mine from April 2017. When the ATSB’s final report was published in October 2017, there was a reference to MH150. I don’t know if the ATSB and the FBI had independently observed the potential significance of MH150, or whether it was after reading my comment that this link was studied.

  125. TBill says:

    Agreed I heard it from you first.
    In hindsight, I am just saying I’d be shocked if FBI did not realize the sim data was a diversion plan…not sure if they grasped MH150 but I would think so.

  126. Victor Iannello says:

    @TBill: It’s hard to know what conclusions the FBI reached, and what additional information they may have had.

  127. TBill says:

    Bumped into an apparent 5_Jan_2017 reference of Jeddah by you from Reddit.

    Your post above clarifies that ZS flew 3-Feb to Denpasar and 4-Feb to Jeddah.

  128. Victor Iannello says:

    Here is my entire comment on Reddit on Jan 5, 2017. I don’t believe the questions posed at the end of the comment were ever answered.


    There are some interesting items to note relative to the simulator data and Zaharie Shah’s schedule. All times are local to Malaysia.

    1. The flight path and take-off fuel load extracted from the simulator data is consistent with a diverted flight between Kuala Lumpur and Jeddah, where the diversion occurs at the FIR boundary between Malaysia and Indonesia.

    2. Zaharie did not work on Feb 1 and Feb 2, 2014.

    3. On Feb 3, 2014, his start of duty was 7:50 for a 9:05 scheduled departure for Denpasar, Bali, Indonesia, on MH715. He scheduled arrival back in Kuala Lumpur on MH714 was 16:06 on the same day with an end of duty of 16:51.

    4. The Shadow Volume containing the recovered simulator data was dated Feb 3, 2014.

    5. On Feb 4, 2014, his start of duty was 14:00 for a 15:14 scheduled departure for Jeddah on MH150. His scheduled arrival back in Kuala Lumpur on MH151 was 11:12 on Feb 6, 2014, with an end of duty of 11:57.

    6. On Feb 20, 2014, Flight Simulator 2004 (FS9) was uninstalled from the MK25 drive of his simulator.

    7. On Feb 21, 2014, he piloted MH370 from Kuala Lumpur to Beijing. His scheduled arrival back in Kuala Lumpur was 15:31 on Feb 22, 2014, with end of duty of 16:16 on Feb 22, 2014.

    8. On Feb 26, 2014, he piloted MH149 from Kuala Lumpur to Melbourne, Australia. He arrived back in Kuala Lumpur on MH148 on Feb 28, 2014. Media reports say that Zaharie met with his daughter in Melbourne just before the disappearance of MH370 on March 8, 2014.

    9. On March 7, 2014, Zaharie started his duty at 23:20 for MH370 with a scheduled departure time of 00:27 on March 8, 2014.

    Some questions:

    1. Why didn’t Zaharie divert Jeddah flight MH170 [MH150] on Feb 4, 2014?

    2. Why didn’t Zaharie divert Beijing flight MH370 on Feb 21, 2014?

    3. Did Zaharie visit with his daughter during his time in Melbourne on Feb 27, 2014?

  129. TBill says:

    That was a very good post by you. Your post reminds me that we were thinking MH370 might have been the actual target flight due to the timing after the flight to Melbourne. However, forgetting that prior line of thought, I tend to feel now that FBI would have had to assume that MH150 was probably flight simmed for the obvious reason (actual intent).

  130. TBill says:

    …also reminds me, back then, I was focused on the moonless night of 8-March as the reason MH370 was chosen

  131. Victor Iannello says:

    @All: Some of you may have had trouble reaching this site in the past couple of days due to SSL certificate issues. When my SSL certificate auto-renewed, it changed configuration settings that caused the problem. (The renewal should have occurred seamlessly.) The same problem caused an interruption last year.

    I believe I now understand what caused the problem, and I was able to make the change and properly install the SSL certificate. Hopefully, the website continues to perform properly. (I won’t bore you with the details of the problem.)

    Network Solutions’ technical support was useless in helping to resolve the issue. I would strongly advise that you avoid doing any business with this company.

  132. George G says:

    Working now.

  133. Mick Gilbert says:

    @Victor Iannello

    Victor, apologies in advance if this takes us into the swamps.

    A further raft of claims are circulating about the application of WSPRnet data to “detecting” and/or “tracking” MH370’s flight into the SIO.

    The associated papers show transmitter and receiver locations where the long (essentially circumglobal) path is being utilised. Looking at the accompanying data there is only one line per Tx/Rx pair. How would anyone know whether that data refers to the short or long path?

  134. Victor Iannello says:

    @Mick Gilbert: It is not possible to use the historical WSPR data (narrow band SNR and frequency shift) to detect and precisely localize aircraft under the following combination of conditions: skywave propagation over global distances, low power, and HF band frequencies. Assuming the propagation is more likely in the long path direction when much shorter short paths exist is also problematic. I know of no knowledgeable individual that believes the WSPR data can be used in the proposed manner, including the inventor of WSPR. As an amateur radio operator, I’ve been active on the HF bands for years, and I understand that deep fading is characteristic of all skywave propagation due to multipath, changes in polarization, and the dynamic characteristics of the ionosphere. The proposal that the signal strength between two stations in the US would be influenced by aircraft activity in the Indian Ocean is preposterous.

    That said, it is futile to try to persuade people that don’t want to be persuaded.

  135. Mick Gilbert says:

    @Victor Ianello

    Thanks Victor. That succinctly covers what I’ve read in numerous other explanations of the shortcomings of trying to adopt that technology to detection/tracking at such long ranges.

    I recently took the time to do a cursory review of the dataset used for a “tracking” exercise on 8 March 2014. Of 63,400 WSPRnet links in the record, only 7,234 of them were stable. From a purely statistical perspective, trying to distill anything meaningful from that data, with its 88.6 percent anomalous link rate, would be so highly problematic as to make the outcome meaningless. It would be nigh impossible to demonstrate that a “positive” result has any statistical significance.

  136. sk999 says:


    Here’s a writeup I did on WSPR, making a quantitative calcuation of the detectability of WSPR and other signals in various situations:

    It was actually interesting to show that, yes, under ideal conditions, WSPR signals can easily be detected half way around the Earth, and even beyond (so yes, long path). I also did some calculations about radio signals (from commercial broadcast stations, not WSPR) being detected after bouncing off aircraft, again showing that the detected signals behave as we expect. Combining the two, you can show that there is no chance of being able to detect an aircraft in the SIO using WSPR signals. It’s not a matter of belief – it’s a matter of physics, backed by measurement.

    Still, people have a pernicious habit of believing in their own correctness. Blondlot continued to believe in n-rays after they had been discredited. Weber continued to believe that he had discovered gravity waves after they had been discredited. Trump still thinks he won the election. Marconi continued to think that he had detected a Morse code message at Signal Hill, thus crossing the Atlantic from Ireland to Newfoundland, even though it is generally regarded as bogus today.

    Your evaluation of the 63,400 WSPR links is generally correct. In fact, without further information, you have no way to determine if even one link is “anomalous”.

    The exercise has not been without reward. I have a better grasp of diffraction and the different domains (Fresnel v. Fraunhofer.)

  137. sk999 says:

    Minor correction – the signal Marconi allegedly detected was transmitted from Cornwall (England).

  138. Mick Gilbert says:


    Outstanding! Thanks Steve.

    I’ve been trying to muster the distant memories of my high school physics and a host of reading on WSPR to the matter of this “methodology”. Your addressing all of the key issues in one paper is very helpful. I won’t pretend to understand some of your more detailed calculations but you’ve explained it in such a way that discerning the takeaways is straightforward.

  139. Viking says:


    Great work!

    I took a quick look at your file, and I liked its deep scientific approach. I did not check for tiny errors, but everything looks really solid. Only a small number of people are willing to take the fight with Bessel functions and similar difficulties these days, so you must have spent a lot of effort and time on it. I agree completely with your conclusion.

    Now you are just missing vector diffraction (HA,HA)…

    I would also encurage you to apply the equations to the transmission power received by Inmarsat from MH370. Actually, this problem is much simpler to solve, except for the angular momentum part which needs a bit extra attention (but it can be handled without vector diffraction). However, the result is inconsistent with a direct southern route. That is a major problem. In particular since the inconsistence is not just with a southern route but with the BFO and BTO values in general.


    After the problem with your homepage it is loading extremely slowly from here. Therefore, I think there may still be some minor technical issue. Unfortunately, I am too busy with other things to help at the moment, but please take a second look at it.

  140. Victor Iannello says:

    @Viking: I don’t think there is anything I can do to address the time it takes to load the page. Sometimes it is blazing fast. Sometimes it times-out and reports a database error. I believe these are issues with Network Solutions that seem to resolve on their own. The response does seem to be better today than earlier this week.

  141. Victor Iannello says:

    @sk999: Thank you for the formal treatment of the physics of using WSPR to detect aircraft. In an ideal world, your paper would elicit one or more detailed responses from proponents that address the issues you raised.

  142. ALSM says:

    Viking: re:”However, the result is inconsistent with a direct southern route. That is a major problem. In particular since the inconsistence is not just with a southern route but with the BFO and BTO values in general.”

    What are you talking about? There were no issues with any of the Inmarsat links, out or in, up or down. All good.

  143. Arto L says:


    Rx power and C/No of Ch4 links after Arc2 are unusually high. They are higher than in the other Ch4 links recorded in Inmarsat’s data on MH370 and the previous flight of 9M-MRO, and by my understanding the signal strength should’ve decreased rather than increased as the plane went further into the Indian ocean.

    One possible explanation is temp drift of the HPA output due to cabin cooling. Perhaps Z killed himself by depressurizing the cabin between arcs 2 and 3. If that is the case, then the SDU OCXO whould also have been affected by the cooling, resulting in higher BFO. This means the actual path of MH370 may have been more to the east than BFO without temp drift indicates. Non-detection of debris during the surface search speaks against this theory though.

  144. ALSM says:

    Arto L: It is pretty clear you don’t know how the Inmarsat system works. None of that pure speculation is true. The HPA power output is continuously adjusted to maintain the eirp per spec. It does not change with cabin temp. A cold cabin temperature would have no effect on the double oven OCXO frequency. I’ve looked at all the log data on 2014-03-07&8, and there is not a single value of c/n or ber that is not normal or well understood at this point.

  145. paul smithson says:

    @ALSM: “A cold cabin temperature would have no effect on the double oven OCXO frequency.” Is that empirically known/demonstrated or presumed?

  146. ALSM says:

    Paul: It is known from test data submitted to the FCC for the type certification process. I posted it here and elsewhere years ago. I don’t have access to the charts now while on travel, but I’ll repost next week if you haven’t found the data by then. From memory, the change in TX frequency over the ambient temp range +55C to -40C was <10 Hz, and basically 0Hz over most of the range.

  147. Paul Smithson says:

    Thanks, I’d certainly be interested to see that.

  148. Paul Smithson says:

    Thanks, @ALSM.That looks like approx -25hz in measured freq for ambient temp change +20 to -30 centigrade. Does that sound about right? I presume these are the values after settling? Do we know anything about dynamic changes as new equilibrium is established?

  149. Arto L says:


    Thanks for the log analysis and the certification test graph. I remember that Thales measured the temp coefficient of the OCXO frequency to be -0.65 Hz/K, which is even higher and opposite sign than the certification test result. Anyway, it doesn’t look like temp drift of the SDU frequency would be negligible in the event of decompression.

  150. Bill Tracy says:

    @Victor @ALSM
    I assume the OCXO quartz crystal is heated and held at a constant temperature (nominal 50C range), with the possible exception of the SATCOM off period. There could be some impact on the other electronics temp, I assume we are talking about here?

  151. ALSM says:

    More guesswork? No, the crystal is not heated to 50C. The entire frequency standard circuit is enclosed in an oven and the temperature is precisely controlled at the turnover point for the crystal, approximately +75C.

    If the cabin temperature decreased to -40C for 7 hours, then the BFO bias could have changed otoo +/-20Hz. But there is no evidence to support such an assumption. It is more likely that the decompression lasted 0C during that hr. It should be noted that the OCXO warm up transient at 12:50, sitting on the ramp, was nearly identical to the transient at 18:25.

  152. ALSM says:

    Correction: “It is more likely that the decompression lasted <1hr (~17:25 to 18:xx) and never dropped below 0C during that hr."

  153. paul smithson says:

    @ALSM. I hear what you are saying there – an expectation that a period of decompression might not have lasted longer than the power-out. I’m happy to go with that as a working hypothesis, though alternative scenarios are by no means ruled out (either no depressurisation at all, or depressurised for the duration).

    Let’s say that 1hr depressurised was the case, what do we know about the cooling/reheating profile of aircraft cabin under such circumstances? Is your “never below 0” based on data from actual instances of passenger aircraft depressurised at altitude? And what do we know about the time it would take to return to normal cabin temperature if it had been down to zero and cold-soaked for a while?

    Given the calibrated temperature co-efficient of the frequency, that would give us a rough guide for some what-if scenarios regarding the magnitude and direction of frequency “error” that might be expected.

  154. TBill says:

    @ALSM @Victor
    I appreciate your clarifications.
    I also agree with your assessment of the possible cabin impact of the 1725 to 1825 off period.

    Overall, Figure F1 the UIGA path report, shows a relatively smooth BFOR trend line even extrapolating back to the MH351 prior flight.

    One comment might be, however, the UIGA/IG 34s report, if I recall is basing fuel consumption (at least in some cases) on the no bleed air case, which saves 1-2% fuel. Unclear what cabin temperature that would hold.

    I personally do not see much obvious impact of cabin temp on BFO. But I do believe BFO may have drifted a little bit (4-5 units) during the flight.

    However, the UIGA report, if I recall shows a relatively smooth BFOR trend line even extrapolating back to the MH351 prior flight.

  155. TBill says:

    CORRECTION to mine above- delete last sentence

  156. ALSM says:

    Paul: fwiiw…I believe the best assumption is that there was negligible BFO bias change after 18:25. If there was any change, there is no way to know if the change was + or -. It depends on which side of the crystal turnover temp the set point falls. That determines the tc sign. I have seen both cases. So, there is no basis to assume it is one or the other, absent test data on the specific serial number, which we do not have.

  157. CanisMagnusRufus says:

    I’m disappointed there was no reply to the post.

    during the time period when SATCOM was supposedly turned off (17:20-18:25), could the radar trace of an aircraft turning back and disappearing into the Malacca strait have been made, not by MH370, but by an advanced decoy aircraft like the MALD?

    This would explain so many details except of course the Inmarsat data from 18:25 onwards.

  158. ALSM says:

    CMR: No. Not possible. 17:52 cell phone registration proves it was 9M-MRO.

  159. ALSM says:

    The WSPR investigator posted a “rebuttal” to Steve Kent’s excellent paper on WSPR. By posting the link below, I am not in any way endorsing anything stated in the document. I’m simply making it available for others to review. Personally, I don’t see any valid defenses offered by the author…only more misguided, misleading nonsense. And it is true that Joe Taylor, K1JT, said the author’s theory is “NUTS”. I’m with Joe.


  160. Mick Gilbert says:

    In the “rebuttal” the author writes,

    In total I have performed 887 GDTAAA tests. I have verified 149 progress and position indicators of various aircraft. They cannot all be right just by chance.

    Yes, they can all be right just by chance. The author has made no attempt whatsoever to demonstrate the statistical significance of any his findings.

    Even if you accept the highly contentious point that we’re looking at data influenced by long path propagation, you only need to look at the background or base rate of anomalous links on the day that any of the “validation” exercises were conducted to see that pure happenstance, that is random chance alone, can explain what the author purports to be successful detections.

    In the case of RNZAF Orion P-3C NZ4204, of the 18,636 WSPRnet links recorded between 0036 – 0818 UTC, nearly half were anomalous. That means that on the occasion that a link crossed the path of the aircraft, whether the aircraft was causing an anomaly or not, basically came down to the toss of a coin.

    In the case of the Saudi Arabian B777 HZ-AKF, between 0022 – 0336 UTC, of the 16,433 WSPRnet links recorded 14,451 were anomalous. That is a base rate for anomalous links of nearly 88 percent. This extraordinarily high background rate of anomalous links can be seen in near total lack of stable links in the four link illustrations provided in the paper.

    With regards to the Qantas A330-200 VH-EBQ (allegedly) “blind” test, the author did not post a link to the WSPRnet data he utilised. However, you can take demonstrably wrong link data (actual T + 10 minutes) and produce a roughly 50 percent “success” rate. Moreover, in that particular exercise, there were multiple instances where the target aircraft had apparently crossed a stable link without perturbing it.

  161. Victor Iannello says:

    I’ll let others correct the author’s response to their comments. I will only make a correction to the response to the comment attributed to me.

    The author wrote that I said: “I understand that deep fading is characteristic of all skywave propagation due to multipath”.

    My full comment is this: As an amateur radio operator, I’ve been active on the HF bands for years, and I understand that deep fading is characteristic of all skywave propagation due to multipath, changes in polarization, and the dynamic characteristics of the ionosphere. The proposal that the signal strength between two stations in the US would be influenced by aircraft activity in the Indian Ocean is preposterous.

    The extract from @sk999’s paper that the author believes refutes my statement is this: “The received S/N is well above the minimum detectable signal of -30dB. To summarize, the minimum detectable received power is 200 dB fainter than the transmitted power of 1 W. There is a loss of about 140 dB between the transmitter and receiver due to free space propagation over 16,000 km. Thus, there is a margin of 60dB.

    In the rebuttal, the author conflates having sufficient margin for a long distance WSPR contact (which I agree is possible) with the ability to discern the position of an airplane using changes in signal strength of skywave propagation between two stations, which even if those aircraft-induced changes in signal strength did occur (which @sk999 refutes), would be impossible to distinguish from deep fading, which is completely normal. (Fading refers to the temporal variation in signal strength.) Any amateur radio operator that has made long distance contacts on the HF bands can attest to the S-meter bouncing around even when the conditions are good. That’s why the author finds so many of the contacts are “anomalous”, as @Mick Gilbert has emphasized.

  162. ALSM says:

    Victor: Re your reply: “…the author conflates having sufficient margin for a long distance WSPR contact (which I agree is possible) with the ability to discern the position of an airplane using changes in signal strength of skywave propagation between two stations…”

    It is a constant pattern of disinformation by conflating facts to produce a false defense. For example, in the case of the author’s trying to twist my words, he states:

    “1. Mike Exner: “Nuts”.
    Steve states: “It was actually interesting to show that, yes, under ideal conditions, WSPR signals can easily be detected half way around the Earth, and even beyond (so yes, long path).”

    Well, yes, WSPR signals can be detected by WSPR receivers at a long distance, but NOT WSPR signals that can be distinguished as being reflected off of an aircraft 1000’s of km from the TX or RCVR. This type of conflation is very typical of all the defense arguments.

  163. Paul Smithson says:

    @ALSM – thanks for your earlier reply “Paul: fwiiw…I believe the best assumption is that there was negligible BFO bias change after 18:25.”

    I don’t understand why that is your base assumption if you also allow that there may have been a one hour power-off & depressurisation. From your earlier chart, a ~20 delta in temperature should produce a frequency change in the order of 10Hz, reducing back to zero as the cabin warmed up again. What am I missing?

  164. ALSM says:

    Paul: Here is what you are missing:
    1. Assuming there was a change in ambient cabin temp (no evidence for that), it would take a 50C decrease to cause a change OTOO ~20Hz. A 20C change (say, +23C to +3C) would have negligible effect according to the FCC data. IMO, a 50C change seems unlikely in 1 hour given the thermal mass of all the equipment.
    2. Moreover, even if there was a 20 Hz change in BFO bias initially at 18:25, it would go back to 150 Hz quickly as the cabin heated back up following L Main Power restoration. I’m sure the PF would be eager to warm up.
    3. Finally, and perhaps most importantly, if there was a 20 Hz bias change, there is no way to know if it was positive or negative, so the best assumption in that case is to assume no change. To assume either a positive or negative change of 20 Hz would invite a 50/50 chance of introducing a 40 Hz error.

  165. Paul Smithson says:

    “if there was a 20 Hz bias change, there is no way to know if it was positive or negative, so the best assumption in that case is to assume no change”

    Hmmm. With your premise, the “best assumption” would seem to be +/=20Hz uncertainty rather than immutable bias.

  166. Paul Smithson says:

    Typo: +/-

  167. ALSM says:

    Paul: My point was that the FCC data cannot be taken to mean that the TC will always be the same sign. We don’t know what the sign of the TC was in the case of the specific SDU in 9M-MRO. Given that, one should not take the sign of the FCC test data as indicative of the sign in 9M-MRO. The magnitude probably is representative, but not the sign necessarily.

  168. Paul Smithson says:

    We are in glorious agreement then.

  169. ALSM says:

    Paul: I took another, closer, look at the FCC data. I had to blow up the image to get a better estimate of the small frequency change for a temp drop from +23 to +3C. And for that case, you were correct that the change could be OTOO -10 Hz. That said, I still believe, given all the assumptions needed to conclude there was any bias shift at all, it is best to assume no change after 18:28 (following the OCXO settling period). Of course, for all the other reasons we have discussed, a sdev of 7Hz or so is probably appropriate.

  170. 370Location says:


    The rebuttal to my point about misuse of long paths quotes Steve writing, ‘… it is even
    possible that the “long-path” direction gives a stronger signal than the “short-path”

    Yes, there have been rare reports of long path connections. That’s completely different from pretending that very frequent short path connections are actually taking exactly the longest possible path.

    He says the science [of long path skip] has been well researched. None of his references or anything I’ve read has been about testing paths longer than halfway around the globe.

    I published a map showing that short paths through the SIO during flight MH370 were very rare. One, maybe two “spots” might be apply. He depends on frequent connections between station pairs to look at signal changes. That *only* happens with short paths.

    I published a plot that shows an exponential decrease with distance in how often contacts occur between station pairs. By halfway around the globe, the average is down to about one contact every few days. Assuming the decrease continues for long paths, his use of a 38,500km path for a 1,500km connection might come up once every few months rather than several minutes.

    I’d appreciate getting a cite for WSPR creator Joe Taylor’s comment on its use for MH370. It’s sad to see both ham radio and MH370 advocates thinking that it’s good that this is in the news, when most expert hams comment that it’s unworkable.

    Hopefully these comments on WSPR won’t put this forum at risk of another legal threat and hosting takedown attempt.

  171. 370Location says:

    @WSPR addendum

    One more point about “long” skip paths is that in reality they include anything beyond the shortest great circle route. It’s an rf signal following a favorable path of ionization, rather than the shortest possible path. Near or beyond the antipode, there is no shortest path anyway. The sk999 paper discusses destructive interference from multiple global paths at that range. The assumption that a signal could only take one mathematically precise longest global path between two points is beyond belief.

    The rebuttal also mentions the accuracy as matching the maidenhead grid accuracy. That would be only for short paths. The shorter the path, the greater the inverse projection to a longest path magnifies the grid errors on the other side of the globe.

  172. ALSM says:

    370location: You can cite this Email from Joe Taylor to Mike Exner:

    “I was not aware of these references to WSPR data. I have not read the papers on the Godfrey blog, and I am not going to waste my time on them.
    But yes, if you think a public quote could do any good, you may use ythis one:

    I do not believe that historical data from the WSPR network can provide any information useful for aircraft tracking.

    — Joe Taylor, K1JT”

  173. 370Location says:


    Thanks for a perfectly concise and diplomatic quote, straight from the source of WSPR. Please also thank K1JT for his many contributions.

    FB OM DE KE6IZN (EX WB9KGJ) — Ed Anderson

  174. CanisMagnusRufus says:

    California Oil Pipeline leak appears to become more mysterious.
    – there are suggestions that AIS may have been hacked, and the blame was laid on MV Rotterdam Express
    – the oil pipeline may have been pulled laterally up to 3 times!
    – happened on a weekend when the Huntingdon Airshow was taking place and hundreds of leisure boats were cruising up and down the coast

    Now CNN is reporting that on the same day (Oct 2), the US Navy submarine USS Connecticut ‘bumped’ into something in the S.China Sea.

  175. George G says:


    Re: @WSPR

    On October 5, 2021 at 1:42 pm you wrote:

    “I published a plot that shows an exponential decrease with distance in how often contacts occur between station pairs. By halfway around the globe, the average is down to about one contact every few days.”

    Ed, would you please kindly provide a link ?

    1. For those contacts which did make it “halfway around the globe”, for how long were conditions amenable to successful transmission/reception ?

    2. Did you cross-correlate with transmitter power ?

  176. 370Location says:

    @George G,

    Here’s a link to a log plot of a month of Mar 2014 spots:

    Follow the Facebook link for my replies on WSPR, which has a linear plot of just Mar 7, 2014.

    The number of spots in the log graph tells us how long the connection was open, at 2 minutes per spot, though it could have been separate connections. For a month of data here are the two longest spot pairs. Each lasted only a few minutes. All are at TX power of 37 dBm = 5 watts. The 22 connections for the month beyond 19,750 km were using 5-20 watts.

    (Sorry if the formatting wraps.)

    df.loc[df[“Distance”] > 19900]

    SpotID Timestamp Rx Rxgrid SNR Freq Tx \
    283477 184636126 1393739280 ZL1SKY RF74cd -5 10.140248 EA7/DL8FCL
    284304 184636140 1393739760 ZL1SKY RF74cd -6 10.140249 EA7/DL8FCL
    285196 184636965 1393740240 ZL1SKY RF74cd -25 10.140249 EA7/DL8FCL
    285444 184637253 1393740360 ZL1SKY RF74cd -20 10.140248 EA7/DL8FCL
    285647 184637459 1393740480 ZL1SKY RF74cd -18 10.140249 EA7/DL8FCL
    2326148 186695989 1394390280 EA4BPO IN80ej -27 28.126112 ZL2IT
    2327033 186697041 1394390520 EA4BPO IN80ej -22 28.126112 ZL2IT

    Txgrid Power Drift Distance Azimuth Band Version Code
    283477 IM76al 37 -1 19944 348 10 0.8_r3058 0
    284304 IM76al 37 -1 19944 348 10 0.8_r3058 0
    285196 IM76al 37 -1 19944 348 10 0.8_r3058 0
    285444 IM76al 37 0 19944 348 10 0.8_r3058 0
    285647 IM76al 37 1 19944 348 10 0.8_r3058 0
    2326148 RF80km 37 0 19905 23 28 2.0_r1714 0
    2327033 RF80km 37 0 19905 23 28 2.0_r1714 0

  177. Victor Iannello says:


    In order to better see the trends, it might be helpful to plot the histogram with larger bins, e.g., 500 km wide.

    Also, some of us do not use Facebook, so if there is particular content you would like us to see, can you provide a different link?

  178. Victor Iannello says:

    CanisMR: Authorities should be able to determine whether the ship actually moved and had its anchor snag the pipeline, and also whether either GPS or AIS was spoofed. I’m open to whatever scenario the facts reveal.

    As for the submarine collision in the South China Sea, that should also be disclosed, and hopefully in the near future. Whatever the obstacle, it is of great interest.

  179. CanisMagnusRufus says:

    @ VictorI
    The MV Rotterdam Express has been cleared by USCG of any wrongdoing and has left port (Bloomberg).
    This incident is significant because it happened in the CONUS.

    But other countries have been experiencing these sorts of incidents related to geopolitical tensions.

    MV Wakashio incident in Mauritius was a good example.

    This may have been due to Mauritius leasing Agalega island to India. Sri Lanka may have been at the receiving end of the tit-for-tat response from India when New Diamond caught fire Sep 2020 off its east coast, or the X-Press Pearl spilled chemicals off its west coast in June 2021.

    In March 2021, Israel accused Iran of an oil spill off its coast.

    MH370 has truly opened a Pandora’s Box.

  180. Don Thompson says:


    SCS, USS Connecticut. Word round these parts: HMS Vigil.

  181. George G says:



  182. 370Location says:

    @VictorI, @WSPR

    A simple histogram was the first test, but it seemed biased by repeat connections so I segmented by station pairs. I considered an RGB intensity plot to identify the more common connections, but no point to that effort.

    Here’s the simple binning of WSPR spots per your request:

    The conclusion is still that connection frequency decreases exponentially with distance, with miniscule odds of a connection at 38,600 km. What the author has effectively done is mirrored the graph about the horizontal axis, presuming that very distant connections beyond the antipode are equally as common as the shortest ones. He then goes a step further and assumes all frequent tracking matches are long paths, in addition to always reflecting off his target.

    The link to Facebook discussion is for a public group, which can be read incognito without an account.

  183. Victor Iannello says:

    @370Location: Thanks for the reformatted histogram. I think it helps. If you have time, it would also be interesting to see it with a linear vertical axis.

    Without having a Facebook account, I can only see comments in a public group at the first level, and not the subordinate level replies. Trust me on this.

  184. 370Location says:

    @VictorI, @WSPR

    Here’s the linear WSPR histogram. Falloff is dramatic.

    The one day station pair plot in the previously linked Tweet is also linear.

  185. Victor Iannello says:

    @370Location: I think the linear histogram clearly shows that long path WSPR contacts (i.e., greater than 20 km) are rare. If a significant fraction of long path contacts were included in the short path bins (as the author suggests), you would expect more frequent contacts in the 10 – 20 km range.

  186. ALSM says:

    @370Location: Can you confirm that the peak around 7500km is the natural result of there being a concentration of WSPR stations in North America and Europe?

  187. 370Location says:


    I was curious, too. Here’s a map visualization of spot contacts in the range 6000-7000 km for Mar 2014.

    (It excludes paths crossing the Pacific for simplicity.)

  188. ALSM says:

    370location: Thanks. As expected.

  189. George G says:


    “(It excludes paths crossing the Pacific for simplicity.)”
    Why ?
    Did you run out of blank pencil ?

  190. George G says:

    … black pencil

  191. 370Location says:

    @ALSM, @WSPR, @GeorgeG,

    Also curious about which stations were getting through the skip gap between 4000-5000 km, here’s an even greater waste of black pencil:

    It’s denser because this is all 84K paths. The previous map quit short of the millions available.

    It’s a special case to draw split segments for the rare contacts between longitudes that wrap beyond the edges of the map at the antimeridian.

  192. ALSM says:

    The latest WSPR paper is out (How Can WSPR Help Find MH370?), available at the the usual location. This paper is focused on a specific “spot” between HB9CZF and VK1CH at 17:16 UTC. But there were 42 spots between these two WSPR stations on 2014-03-07. Here is a link to all the transmissions from HB9CZF to VK1CH on 2014-03-07:

    It is clear from this data set that there was nothing odd about the transmission at 17:16. There was no drift or SNR “anomaly” at 17:16. The rcv signal level was fading in and out all day, as expected for all HF transmissions over that length. And the TX was transmitting 1 to 2 Hz low all day. There is absolutely nothing in this data suggesting any effect by any aircraft.

  193. Victor Iannello says:

    @ALSM said: There is absolutely nothing in this data suggesting any effect by any aircraft.

    Very true.

  194. George G says:


    Would you mind if I use your data, should I so wish, to present part of it in a question to the WSPR author ?
    I would, of course, acknowledge you as my source.

    I had been planning a question, if I thought it appropriate, concerning the data, but your presentation (xlsx) has “beaten me to the punch”. In fact it has provided more than I had been expecting.

    Note that your snr plot does show the effect of the sun quite remarkably.

  195. George G says:


    I note your special case answer to my tongue-in-cheek question. Ta.

  196. ALSM says:

    I updated the spreadsheet link above to embed the graphics data. You can accesses it here:

    …or you can download the complete WSPR data file for 20214-03-07 (and some more analysis) here:

  197. ALSM says:

    George (and everyone else): Yes, feel free to use any WSPR files I have posted. All the data is publicly available, albeit, somewhat of a pain to download and parse to usable size spreadsheets. The charts are mine.

  198. @all
    I see that you are very busy with the WSPR, I propose you to take a short break 🙂
    Persuing with the hypothesis of a fully piloted flight, we have complete reassessed the available data based on several recent additional studies we did such as sailor Kate Tee’s testimony analysis (to whom we talked to) and on seismic data identified by Ed Anderson (@370Location) at Cocos Islands.
    It seems to me that we have found a solid possible reason for the Electricity Switch back-on justifying few other things.
    In addition, the leg NILAM to Arc2 does not includes artificial turns used to “slow” down the aircraft. It proposes a logical fly-by the sailor’s boat in a matching timing.
    This reconstructed trajectory is called CAPTION (with a N) and it matches all Inmarsat measured data and most likely by a ditching which is still under investigation.
    We are still of the opinion that the hijackers (whoever they were) had a plan that they could not execute properly.
    The report and its annexes are available on the new website :
    For those interested: have a good reading.
    PS: I will not be connected for few days. So don’t get upset by my slow reaction if needed

  199. Don Thompson says:

    Re WSPR data. Should any reader have knowledge of SQL there is a publicly available query portal at

    The team running that site have transferred the entire database into an OLAP database engine, while maintaining ongoing transfer of all new WSPR reports. They’ve added useful additional fields to the reports table (e.g. a processed lat-lon for beacon and receiver locations).

  200. DrB says:

    @ George G,
    @ ALSM,

    George C’s observation about the effect of the Sun on the observed signal strength between HB9CZF and VK1CH was insightful.

    Based on my experience, radio reception over long ranges at 14 MHz is hugely better during daylight hours than at night, because of the higher ionization of the ionosphere in sunlight. So, the ionosphere refracts better in daytime than at night, generally producing less path loss and stronger signals during daylight hours.

    ALSM plotted the SNR of the WSPR links between the Swiss and Australian stations. Local noon near the midpoint of the short-path connection is circa 18 hours UTC. The time of local noon at the midpoint of the long-path route is circa 6 hours UTC. ALSM’s plot shows the lowest-SNR signals occur at roughly 8 hours UTC, and the highest-SNR, strongest signals occur roughly at 21 hours UTC. Therefore, the variation of signal strength in ALSM’s plot with time is consistent with the expected signal strength variation during daytime/nighttime over the short-path route. This is another confirmation that many, if not most or all, of those particular connections are via the shorter route.

  201. George G says:


    you wrote: “George (and everyone else): Yes, feel free to use any WSPR files I have posted. All the data is publicly available, albeit, somewhat of a pain to download and parse to usable size spreadsheets. The charts are mine.”

    It was one of the charts I was wanting to use instead of re-inventing the wheel. See attachment.

  202. DrB says:

    @George G,

    From your time plot, the lack of any successful WSPR contacts between HB9CZF and VK1CH during the period from 08:30 to 14:00 UTC, when much of the long path was in daylight, also indicates that probably none of the successful connections were via the long path. Instead, the successful connections between 15:00 and 08:00 appear to be via the short path, and primarily when it is in daylight.

  203. George G says:


    Am I confused ?
    The transmitter is close to UTC (a bit over half an hour hour earlier).
    Hence I saw the local time at the transmitter as being close to UTC, or at least close to when viewed from the other side of the world.

    The Transmitter thus is in the early evening at 17:16 UTC, in fact very close to sunset. Successful reception in south eastern Australia had started about three hours earlier at around midnight local time (neglecting Daylight Saving). So, for the next five to six hours the receiver was not in daylight. At 17:16 UTC the receiver had about another two and a half hours before dawn.

    At 22:00 UTC, the transmitter was clearly in the dark.
    The receiver had been in Daylight for a little over two hours.
    The short path between the two, via the Asian route, would have also been the nighttime route for a large part of the successful transmission period between 14:00 UTC and 22:00 UTC.

    Hence I am confused.

  204. Don Thompson says:

    @ALSM, @370Location, et al

    In the update ‘How Can WSPR…’ the author puts focus on WSPR ‘spots’ reported for 1716UTC.

    In order to explore the data ALSM used a sample of spots bounded 20140307T0000UTC to 20140307T2359UTC while I used a sample bounded +/- 24hrs around 20140307T2120UTC (I had been using that sample related to another spot that had been given focus in another piece).

    Cyclical HF signal fading is the dominant characteristic in snr across these periods.

    These charts, A and B, present the freq and snr metadata for spot reports of WSPR transmissions from HB9CZF in band 14. The plots in Chart A show how the signal fading is reflected in the snr for the spots reported by the Australian Rx sites at ca.16400m range from HB9CZF. Chart B shows the same for spots reported by other sites 500-9500km from HB9CZF (UK, Denmark, Lithuania, US and Japan).

    Widely varying snr as is typical of HF propagation.

    Taking a pedantic approach, I should not have plotted lines through the series of reported ‘spot’ information, not least as the HB9CZF beacon operated a near regular 10min transmit schedule throughout the 48hrs that I reviewed (short periods are evident where the intervals varied to 20min, 8min, and 2min for a total of 293 transmissions in the 48hrs). However, a coarse trend is evident that correlates with day-night phases. Anything could be happening in the ‘quiet’ periods.

    Chart C and Chart D others depict the distribution of snr across the 48hr sample, Aus receivers and others, respectively. The distributions may be bi-modal, if not multi-modal. Producing reliable results from these data for an investigation into ‘micro’ propagation effects is not possible.

    Further, the drift and ‘spot’ frequency derived by the receivers shows that HB9CZF was drifting down during each transmission (92.7% receivers report -1Hz drift while demodulating HB9CZF). It is also evident that poorly calibrated local oscillators exist among the receiver population.

  205. vodkaferret says:

    @Jean-Luc Marchand.

    thanks for the new caption paper. clearly a lot of work has gone into it.

    I am not in the slightest qualified to comment on the fits to BFO /BTO or other technical aspects, but it does seem there is an incongruity there that wasn’t so much in the captio (previous) work. You state there would have been enough fuel to follow a Learmouth path at around 38,000 feet, but also go to pains to point out the lengths that were taken (in your scenario) to avoid radar detection, which in many cases involve dropping to much lower altitudes.

    So… what went wrong? You believe the PIC (people in charge, using fb your terminology) planned to avoid radar but forgot about the effects on fuel? Or something went wrong but they were completely unable to think on the fly and decided yeah, let’s just follow the original plan but at different altitudes and hope for the best?

    I can see that your aim is to match a path to some data points you have decided are most important, but the levels of skill and knowledge that are implied at different parts of the flight seem inconsistent. The path and sequence of events doesn’t seem to have any consistent rationale behind it. Personally I found the original captio more consistent and compelling.

    Thanks for the work and a interesting read anyway, as I say I’m not qualified to comment on the technical aspects but I suspect they hold up.

  206. Victor Iannello says:

    The collective analyses of WSPR data from a number of contributors here (Mike, Don, George G., Mick G.) persuasively show that there was nothing statistically “special” about the WSPR spot with the propagation path that intersected the vicinity of MH370. That means there was no detectable interaction of WSPR signals with the aircraft.

    What’s attractive about the current analyses is there is no requirement to understand the physics of propagation, nor to have the empirical knowledge of HF propagation. As such, it can be more widely understood.

    This is a classic case of cherry-picking data to support a favored theory, as many people have said.

  207. DrB says:

    @George G,

    I am the one who is confused. I was off by 12 hours.

    The links between Switzerland and Australia are shown more clearly in DonT’s Chart A. It shows two periods, from 6-8 UTC and from 14-22 UTC when more links occur. In the first period, the sun is rising in Switzerland, and it is mid-afternoon in Australia. So, there is daylight in the ionosphere over the whole (short-path) route. That first period sort of makes sense to me, because almost the whole path is in sunlight. However, from 8-10 UTC there is sunlight over the whole path but poorer SNRs. Why is that?

    In the second (and longer) period, it is mid-afternoon to nighttime in Switzerland, and late evening to near sunrise in Australia. In the second period there is sunlight at one end, but darkness at the other end. I also don’t understand why the 14 MHz band is open for a longer time from 14-22 UTC.

    Using Don’s Chart B, the link from Switzerland to Illinois (K9AN) is open from 9-11 UTC and again from 22-5 UTC. The first one occurs when it is morning in Switzerland and late night in Illinois, and the second one occurs when it is nighttime in Switzerland and late afternoon to evening in Illinois. Interestingly, near 15 UTC there is sunlight over the whole path, but poorer SNRs.

    Both paths (Switzerland to Australia and Switzerland to Illinois) show there are TWO periods each day when the SNRs are better, but neither one occurs when it is local noon at the midpoint of the path.

  208. Victor Iannello says:

    @DrB: I suspect the effects can be explained when you consider the effects of both D-layer and F-layer ionization. At 14 MHz, F-layer ionization enhances refraction, but D-layer ionization attenuates it. The F-layer also builds up faster at sunrise, and persists longer after sunset, so there is a period of time along the “grey line” where propagation is better than in full daylight.

  209. sk999 says:

    8-10 UT long path, 14-22 UT short path?

  210. DrB says:

    @ Victor Iannello,

    Your theory about the two ionized layers explains the double peak in SNR. The best overall conditions would occur when the sun is near the horizon (above or below) at both ends of the path. The difference in longitudes of Switzerland and Australia is roughly equal to the length of the day in March, so the shorter path has the sun near the horizon at both ends circa 07:00 UTC. That would explain the improved SNRs seen from 06:00 to 08:00 UTC, when it is noon at mid-path along the shorter route.

    The longer-duration band opening from 14:00 to 22:00 is not as easily explained. At the center of this period, the UTC is 18:00 and the sun has set well below the horizon in Switzerland. In Australia the sun is also well below the horizon before sunrise. However, the middle of the longer path is in full sunlight, so perhaps the D-Layer attenuation is strong over it. Over the shorter path, the D-layer attenuation is at a minimum circa 18:00 UTC. Perhaps the reduced D-Layer loss then is sufficient to allow shorter path links to occur, even though the F-Layer refraction is somewhat weakened. This would also explain the extended duration of the “midnight at mid-path” (over the shorter path).

  211. Victor Iannello says:

    @DrB: That seems reasonable. When the band is open, night time paths on 14 MHz (20 m) tend to be shorter.

  212. CanisMagnusRufus says:

    @ VictorI
    There are currently reports in the media that Israel may launch airstrikes on Iranian nuclear facilities, posssibly via Azerbaijan, to prevent Iran from acheiving nuclear weapons capabilities.
    Iran has worked with North Korea for many years to jointly research & develop ballistic missiles, as well as nuclear weapons. In an article in the National Review titled Iran Probably Already Has the Bomb by R. James Woolsey et al, the authors suggest that Iran may have already acquired/tested a nuclear weapon from the North Koreans in exchange for oil or cash.
    In Sep 2016 it was revealed that the Obama administration authorized an unmarked cargo aircraft to carry pallets of cash in euros and swiss francs issued by the Netherlands to Iran, in order to return US$400 million as part of a legal settlement of a dispute between the US & Iran going back 35 yrs. The Obama administration faced a lot of criticism for this delivery of hard cash from both political parties.

    On SEPTEMBER 21, 2016, Michael B. Mukasey, witnessing before the US Senate had this to say:
    The only conceivable purpose for this money is to finance illicit activities, because the cash is untraceable. That is
    the reason that the Iranians insisted on it, and why we agreed to it is something that I think this Committee ought to probe, because I believe there are no good reasons for having agreed to it. In addition to the considerations that I pointed out in my statement, I should point out that there is present in the world another rogue State, North Korea, that is cash-starved and that has an active nuclear program. The Iranians, of course, we know have an active missile program–ballistic missile program, a program that is not useful for any purpose other than to deliver a nuclear weapon. So the question then becomes, do they intend to continue their nuclear program, in part, with the use of this kind of money?

    In 2011, an IAEA report claimed that both Iran & North Korea were using commercial airlines & charter aircraft to transship prohibited weapons technology and equipment, particularly via China.

    On JULY 28, 2015, during a subcommittee hearing on the JCPOA in U.S. House of Representatives, congressman Brad Sherman said this:
    Iran’s going to have a lot of money; North Korea has nuclear weapons and a thirst for money. What could go wrong?…
    We have got to look at planes and ships that would connect the two. I think if there is an exchange of money for nuclear material, it is much more likely to take place on a plane. A ship gives us a chance to track it and a chance to make a decision as to whether to interdict and board. Of course, a ship is also possible. We should not be encouraging the civil aviation of Iran by selling them planes and parts. We know they are going to use those planes to take thugs to Damascus to kill people. And we hope they don’t use the planes to go pick up a nuclear weapon in North Korea.
    So we have got to see what are the opportunities to interdict either the shipment of a bomb in one direction or cash in the other. We have got to keep track of what Iran does with the $56 billion to $150 billion they get from this deal. But, finally, I think we are dependent upon China, which exercises such significant control over the most critical aspects of what the North Korean Government does. If China is willing to turn a blind eye to a cash-for-bomb situation, I don’t know if we can stop it, and we certainly–if they were willing to turn that blind eye at Beijing Airport, I know we couldn’t stop it.

    That last sentence is quite curious, hence my emphasis.

    – On JW’s website, in an article titled MH370 Scenario with a Landing At Banda Aceh (Aug 23, 2014), you speculated that there may have been large sum of cash on board MH370. FdC mentioned in her book that there was valuable cargo that was transported under heavy guard and loaded onto MH370 without going through the standard screening. Do you think it’s plausible that MH370 was carrying cash from Iran to North Korea via KL & Beijing?
    – OI carried out a search for MH370 in the Indian Ocean. They have previously been involved in attempting to locate lost treasure in the oceans. If they find it, they get to keep a cut of the proceeds from the sale of the treasure. Could this explain their interest in finding MH370?
    – Do you think the Obama administration tried to buy the silence of the Iranians re: MH370 by delivering/replacing the hard currency that the Iranians lost?

  213. ALSM says:


    Re: conspiracy theory nonsense questions one step back from “aliens from space did it”:

    “Do you think it’s plausible that MH370 was carrying cash from Iran to North Korea via KL & Beijing?” NO.

    “Could this explain their interest in finding MH370?” NO.

    “Do you think the Obama administration tried to buy the silence of the Iranians…” NO.

  214. Don Thompson says:

    @J-L Marchand

    I have downloaded and began to read your most recent proposal for the loss of MH370 and the path followed by the aircraft. Scenario 2, as described, is fraught with errors. Too many to list here. Thus, one has to ask: does the proposal have any merit whatsoever?

  215. Tim says:


    I’m sorry, but for your route it appears you’ve just picked two points, Kate Tee’s fly over and the Cocos island transit, and then just made the BTO/BFO calculations fit.

    With regards to the RAT deployed scenario, as long as the systems are working correctly the battery power should not reduce and drop voltage.

    I can’t believe there was enough ash concentration to cause any orange glow. If there was I’m sure it would have been mentioned in previous official reports.

    I don’t think that’s going to hold up to much scientific scrutiny

  216. Victor Iannello says:

    @CanisMR: I think many of us in the months after the disappearance struggled to find a reason for the diversion. Most of those reasons did not stand the test of time, including the landing at Banda Aceh and including the speculation that money was in the cargo bay, which was based on rumors that security personnel were aware.

    That scenario from July 2014 was prompted by my observation that a due south route from BEDAX to the South Pole fit the BTO and BFO values from 19:41 and later quite well. However, the flight path required some way to “lose time” between 18:28 and 19:41, because you could not connect the positions with a straight line at a reasonable airspeed. The due south flight path from BEDAX to the South Pole is where we landed in the much more recent UGIB paper, so at least part of that work survived.

  217. paul smithson says:

    I have recently refined a path model that is governed entirely by waypoints. It starts precisely at the Arc1 crossing 18:28:15 and is flown at constant M0.84 FL360. No mucking around with speeds or headings to the 4th decimal or insertion of spurious manoeuvres. The BTORs at each arc (1-7) are as follows: 0, 3, 47, 49, -29, -12, 1. The path reaches 7th arc at 39.61S (unsearched). The waypoints are: MEKAR, SANOB, IGEBO, RUNUT, 40S85E. The path model commences at the Arc1 crossing with BTO error=0 along the MEKAR-SANOB leg. kmz here for anyone interested. Hover over path points for timestamps. The path modelling tool is a variation on Barry Martin’s, adapted for FL360 with spatially and temporally interpolated wind and temperature fields for that flight level.

  218. 370Location says:


    I believe the new CAPTION paper does have merit. It needn’t be consistent with other’s conclusions, and minor errors shouldn’t be cause to dismiss the whole premise.

    Their approach is the scientific one of incorporating new evidence, and building upon it. It is part of my own acoustic findings they are using. No other researchers have yet seen fit to utilize my findings, only critique them. I will take their waypoint use of the Cocos Island flyby as validation that there is another viable and flyable path up to Cocos that matches not only BFO+BTO by waypoints, but also adopts Kate Tee’s eyewitness account.

    They do that without tight 90 degree turns to follow FIR boundaries.

    Although they proceed after Cocos airport to optimize with heavy weighting on BFO accuracy (as others do), they are free to change their endpoint (as others have) away from Christmas Island. I believe that’s part of including new evidence or refined analysis, and it sometimes takes courage to change conclusions.

    It’s not biased of me to consider their work seriously. I’ve done the same for many other endpoints, and supported them when I can. CAPTION is not excluding other possibilities to favor their own conclusions.

    I’m left wondering why this hasn’t been picked up by the media. One pattern that I’m noticing is that those who claim that they are right with the most conviction seem to get into the news, even if they are posting easily debunked clickbait. CAPTION qualifies their findings, like I do, as theories. They also propose multiple possibilities, which may have been my own downfall. Even the ATSB wanted results in black and white, saying that the plane couldn’t be at multiple crash sites. I think CAPTION is taking the right approach toward revealing the truth, regardless of whether it has less media appeal.

    — Ed Anderson

  219. TBill says:

    I agree with your sentiment, at least to some extent. Captio and Caption both explore possible alternate electric configurations at IGARI which I feel is an under-explored topic. It is also a very advanced topic requiring intimate knowledge of the B777 electronic behavior as certain circuits are depowered. As far as flight path itself, it is problematic in my view. I do agree with the active pilot (or hijacker) hypothesis but not the suggested path and destinations.

  220. Mick Gilbert says:

    @Victor Iannello

    Victor, I have to give you your dues regarding your stewardship of this forum. You may not agree with everything that is posted here but you’ve enabled a courteous exchange of ideas. You don’t ban contributors for disagreeing with you. A far cry from the cultist echo chambers run in some other circles.

  221. Paul Smithson says:

    @Mick, we were expelled for failure to exhibit the required level of awed reverence. That and daring to express an opinion.

  222. Victor Iannello says:

    @Mick, @Paul: It’s been a while since anyone was banned here. I don’t know the situation are referring to, but is seems odd that both of you were banned.

  223. Mick Gilbert says:

    @Victor Iannello

    A not so long story, Victor, relating at least in part to my preference for using primary source material, like properly constituted Annex 13 accident reports, rather than Wikipedia excerpts drawn from out-of-print books on the paranormal and unexplained phenomena to reach conclusions about aircraft accidents. Apparently under the modern day “every child’s a winner” approach to learned discourse, Wikipedia is the go-to.

    That, and the usual trouble making we get up to by using facts and reasoning when engaged in discussions. It caused a few wrinkled brows with the cradle of mediocrity cultists.

  224. David says:


    The MH370 route and crash site analysis summarised and referred to below contains a lot of new investigative work:

    The theme is that gas concentrations (methane, sulphur dioxide, carbon monoxide, dark target aerosol) and contrails/cloud cleavages could disclose a new route and site.

    The author adds how other route indicators also support these findings.

    At the crash site, after a ditching he has the flaps being retracted using APU power before an explosion (lithium battery related) blows up the aircraft and its wings, that preventing ELT reception.

    There are errors (eg APU start up would not of itself account for the last log-on; engines would not implode on sinking) and questionables (eg pilot retraction of the flaps after ditching; the equating of flaperon drift with a drogued drifter; dismissal of the 7th arc 00:19:37 BFO descent rate; ambiguity in what fuel was jettisoned, likely pilot knowledge of sea state distribution) but its conclusion and new ground warrant more than such cursory comments.

    A fundamental issue is how unambiguous the reported gas and cloud anomalies are and what else might account for them. Also, how quickly these can be expected to disperse.

    I see he is reported as saying, “…our current plan is to go ahead with our search in 2022 with or without an agreement from Malaysia.”

  225. Mick Gilbert says:


    Thank you for that link. Certainly a professionally prepared report.

    Further to your points above, a rather fanciful reading of Australia’s air defence capabilities at,

    South of POLUM, the aircraft dropped to around 7000 feet … MH370 was now within the range of the Australian over-the-horizon radar and approaching the range of the Royal Australian Air Force interceptors.

    The author is talking about a location that is over 800 nm from the nearest RAAF airfield (a “bare” base) and nearly 2,000 nm from the nearest RAAF fighter squadron.

    Separately, the author’s timeline of events has “Impact” (of sufficient force to tear the engines off) at 00:20:45 with the aircraft still afloat at 00:33:15. The ELT would have started transmitting at 00:21:15.

  226. Don Thompson says:


    “new investigative work”

    No. Just no.

    The author mistepresenta the investigation findings for the Asiana freighter crash in an attempt to infer, without any basis in fact, similarity of cause due to its, and 9M-MRO’s, consignments of Li-ion batteries.

  227. David says:

    Amplification, 5th para 1st line, the APU start-up he describes is manual.

    @Mick. Re ELT. I ‘interpreted’ his remarks to the effect that it would not have been transmitting long enough to be received by a suitable satellite before the explosion.

    @Don. My second last para refers.

  228. Mick Gilbert says:


    If the beacon was transmitting for 12 minutes it would almost certainly have been picked up by the COSPAS-SARSAT SAR signal processor package on INSAT-3A. That satellite sits in a geostationary orbit at 93.5E so it would have been well positioned to detect a distress transmission I would think.

    Beyond that, one of the COSPAS-SARSAT’s then seven low Earth orbiting SAR satellites, MetOps-A, was tracking south-south-west over Central Australia about 830 km up at the time. It should have been just above the horizon from the estimated crash site between around 00:22:30 – 00:24:30 UTC.

  229. TBill says:

    Refreshing at least to get some new thinking.
    Jet Fuel often contains sulfur from the crude oil source, not as an additive. But it is not a very high amount of sulfur eg 0.1-0.2% and most of the jet fuel was gone. Whereas the bunker fuel on ships would be very high sulfur in those days up to 4%+.

  230. 370Location says:

    @David, @Mick Gilbert, @Don Thompson,

    That MH370 report on aerosols was posted to Twitter on Sep 28. It looks like @RandRolston has recently paid a PR firm for a bit of influencing to blogs and Twitter.

    Here’s my Sept 28 response there:

    “Replying to @JeanMarcGarot and @RandyRolston:
    @RandyRolston was advised via emails in late June that the HA01 hydrophone array detects the direction of origin. All three pips were ice events near Antarctica.

    His Sea Surface Temperature anomaly existed there on Mar 7 before the crash.”
    [Followed by a link to the hydrophone analysis.]

    We had several email exchanges, and he had said he would probably drop the acoustic portion. Instead he doubled down. I think he has added the Sea Surface Temp info since his draft report.

    I had concerns about his METEOSAT7 image treatment, because the contrail on his image is much darker than on the original sat image. It shows some anomalies in the bit order, which might have been the result of a thresholding/quantization that added contrast at that luminance. Rolston has since found another satellite that covered his candidate area at a slightly different time. I have not tried accessing that data.

    I also have not retrieved the Suomi atmospheric data, so I don’t know if his other hotspots were there before the crash.

    I encouraged him to refine his research, but now I’m disappointed that he is misusing the acoustic timing because he saw someone else publish similar meta-analysis of a single amplitude spike on an early graphic while ignoring the unambiguous azimuth of origin. The original LANL authors knew the bearing was toward Antarctica, which is why that blip is not a candidate. Curtin showed it, too. Publishing the claim after it has been debunked causes me to question his methods and so his claims. They would need to be verified.

  231. TBill says:

    As a chemist, it would be highly unexpected for the SOx and CO to be caused by air crash. Unfortunately prior to 2020, the oceans had heavy air pollution due to massive amounts of SOx and CO from burning high sulfur bunker fuels in marine vessels (ships). Also gases in the atmosphere become well mixed, not sinking due to molecular weight as stated. If high molecular weight gases sunk in the atmosphere, Earth could not sustain life since the 1% argon in air would sink to the bottom of the air column, but that does not happen (except apparently at the border of the atmosphere with outer space there is some striation of the gases).

    As stated above jet fuel does not have sulfur additives, as stated, it has natural sulfur from the crude oil but very low amount of sulfur compared to ship fuel.

    I do not have any documentation to say authorities agree the Asiana freighter crash created measurable gases related to a lithium battery fire.

    Quite outlandish to suggest the flaps were retracted on the ocean surface followed by fuselage explosion due to H2/CH4 build-up. I would ask if there is any example in the history of lithium batteries of a secondary explosion due to flammable gas build-up. However, if that had happened, it would be an extremely powerful type of explosion known as a detonation, and then we really would probably have an undersea sound on hydrophones.

    Contrails remain as the only possible evidence, and that is interesting, but the important final contrail below POLUM does not look very convincing to me.

    I would like to know more about the altitude of those cloud layers below 22 South on 8 March. Tim Vasquez has historically been the contrail/weather expert for MH370.

    Anyhow I do not see immediate merit in the wild scenario, but the out-of-the-box thinking makes some possible good points, which the author squanders on science fictional type scenario.

  232. Arto L says:

    @TBill, @David:

    The contrails don’t look convincing to me either. The images shown were taken between 01:30 and 02:00 UTC. However, I don’t see anything in the 00:30 METEOSAT 7 image. The shadow visible in the later images is probably just a cloud formation.

  233. David says:

    @Mick. Thanks. So, another strike.

    @TBill. Yes, if heavy molecules were not dispersed we would be asphixiated by CO2 though I am unsure about detonation speeds being reached by the shock front in an explosion of that sort.

    With your remarks now added to others of us about his work I think it fair that they be brought to his attention in case he may wish to contribute.

    I have sent a suggestion to that effect, with Victor’s concurrence, that I hope will reach him.

    @Don Thompson, 370Location. Arto. Thus possibly he will respond here to the points you make.

  234. @vodkatferret
    «the PIC planned to avoid radar but forgot about the effects on fuel? » : 1) to our view, the diversion to the north was probably not in the plan. The surrounding traffic might have been a worry for them to be detected, thus this manoeuvre.
    2) after a sequence Power Off/Power On with a RAT regime in between, the estimation of fuel by the FMC is unreliable (temporary or permanently ? ) thus at Cocos what was the fuel estimation/prediction is unknown. If they flew this trajectory they must have had sufficient decisional elements to go that way instead of going to Christmas Island for example.

  235. @Tim
    «picked two points, Kate Tee’s fly over and the Cocos island transit, and then just made the BTO/BFO calculations fit »:
    Interesting remark. How do you think the IG built they trajectory ? a priori selecting the simplest path after Arc2 (straight line) with a simplest heading (180°) which minimises the track/heading difficulty and then select the altitude to make the BTO/BFO fit. As I demonstrated there is a one-to-one relation between Arc2 location and the track. Thus they de facto fixed Arc2 location a priori as I used Kate’s boat as a waypoint. What’s wrong with this ?

    «RAT deployed scenario, as long as the systems are working correctly the battery power should not reduce and drop voltage. »
    Considering the FCOM and the electrical system description, the principle of the EMS and the concept of electrical aircraft is to use the available power at best (via shedding normally) at any time. When a total power failure occurs, the main battery takes over expecting at least one of the electrical sources of power to start as quickly as possible (the battery capacity to power the a/c by itself is for few minutes only). When the RAT provides power it is clearly documented that no charging takes place, probably because Boeing shedding (or a hardcoded equivalent maybe) principle in this critical situation requires to use the maximum available power. Thus the battery must clearly complement the RAT. I think it would be “criminal” not to do so :-). In addition, PMDG people who created the B777 simulator worked hand in hand with Boeing. Why would have they taken care, resources and energy (and money) to produce the source code to model the behavior of the battery if it was not the reality ? If this is in the model, it is because Boeing told them so. Thus I think the RAT and the battery work together to power the aircraft. The battery is rated 46 Ah, delivering 46A during 1 hour aside to the RAT is not unrealistic.

    « not enough ash concentration to cause any orange glow. If there was I’m sure it would have been mentioned in previous official reports. » When interviewing Kate Tee, the color of the burning SiO2 is the closest to the color she could remember. In addition, the VAA (Volcanic Ashes Advisories) mentioned the maximum ceiling of the ash cloud i.e. 20000ft. If I am correct, the reports always envisaged flight altitude well above this altitude. So they were not concerned. To me this explains it.

  236. @Don Thompson

    If I may… either you said too much or you said too little 🙂 Errors ? too much simplification perhaps for the sake of being concised. To avoid this « long list » to be detailed and discussed here. I have asked Victor (thanks to him) if he would accept to forward you my email address such that we could review it off line. Would you agree ? Thanks.

  237. Victor Iannello says:

    @Jean-Luc Marchand: I don’t think you understand how the UGIB route was determined. Comments below:

    How do you think the IG built they trajectory ? a priori selecting the simplest path after Arc2 (straight line) with a simplest heading (180°) which minimises the track/heading difficulty and then select the altitude to make the BTO/BFO fit.

    Hmm…not really. The due south track after 19:41 was not selected because it was a simple “heading” (I think you mean true track). Rather, it was determined that for a route along this great circle at the indicated altitude, the BTO and BFO error criteria are statistically satisfied, including the expectation for systematic error, and also the fuel consumption. An overview of this work was presented in this blog article.

    As I demonstrated there is a one-to-one relation between Arc2 location and the track.

    The conclusions from that body of work largely repeated what I presented in four years ago in a blog post.

    Thus they de facto fixed Arc2 location a priori as I used Kate’s boat as a waypoint.

    That’s incorrect. We did not select the position at 19:41 a priori, and our selected route does not depend on Kate’s description as we do not know if she saw MH370. Rather, the selection was based on the development of an accurate fuel consumption model, and well as a statistical metric for the expected random noise inherent in the recorded satellite data. These improvements allowed the rejection of hypothetical flight paths that were previously believed to be possible.

  238. Victor Iannello says:

    @Jean-Luc Marchand said: Thus the battery must clearly complement the RAT. I think it would be “criminal” not to do so :-).

    If I understand your statement, you believe the RAT has insufficient power to meet the power demands of the critical systems, and must be supplemented by the battery. I’ve never heard that before, and it seems like a bad design because there would be a reduction in safety after a period of time. What is your source for this information?

    In addition, PMDG people who created the B777 simulator worked hand in hand with Boeing. Why would have they taken care, resources and energy (and money) to produce the source code to model the behavior of the battery if it was not the reality ?

    I can tell you from my own investigations that there are all kinds of system errors in the PMDG 777 model. PMDG’s performance guide was the FCOM, and there is no guarantee that the performance of a system is accurate if not in the FCOM. Even for items in the FCOM, there are performance errors in the simulation, especially for non-standard configurations.

  239. @Victor
    First my appologies for the “epdidermic” reaction, I was oversimplifying and reacting to the tone of some comments… 🙂 of course, you put a lot of serious analysis into your work.

    Concerning, the FCOM and PMDG, I am glad you put this on the table. I am trying to find out with B777 pilots if one went that far in his/her training sessions. It seems not. Thus nobody knows what is the reaction of the electrical system staying so long under the RAT powering. If someone here has info he/she is welcome.

    PMDG: yes they are far from perfect as you already demonstrated before 🙂 nevertheless the timing of the occurence of the “battery discharge” is a new element to look at, it seems to me and a potential avenue for deeper analysis (but we are dependent on Boeing’s good will aren’t we ?).
    We thought to bring that up because for us it is striking that no matter is the actual system behaviour at this very critical moment, nobody knows. Our view is that, to avoid to face such a (unknown) situation, the electrical power was put back on (maybe based on “wrong” behaviour of a simulator used in preparation of the coup 🙂 ).

    “you believe the RAT has insufficient power to meet the power demands of the critical systems, and must be supplemented by the battery”:
    No, it has sufficient, but the battery is sollicited nevertheless…
    from the document “Boeing B-777: Fly-By- Wire Flight Controls” (Bartley) it is written “for in-flight failures of the PMGs, the PSAs draw power from any available source” this includes the main battery via the hot battery bus as mentioned few lines before in the same paragraph.

  240. sk999 says:

    From the Continental Airlines 777 Training Manual, Chapter 24

    Section 00 (General description):

    RAT supplies 7.5 kva (i.e., 7.5 kilowatts). Battery has a capacity of 47 amp-hours at 28 volts; thus, it can supply 7.5 kilowatts for about 10 minutes.

    Section 33 (Standby Power):

    “Standby power comes from either the RAT generator OR the main battery.”

    “When all ac power is lost, the main battery supplies power to the standby loads of the dc generation system UNTIL THE RAT DEPLOYS”

    “After the RAT deploys, its generator supplies ALL of the power for the standby loads”

  241. Andrew says:

    @Jean-Luc Marchand

    RE: ‘…from the document “Boeing B-777: Fly-By- Wire Flight Controls” (Bartley) it is written “for in-flight failures of the PMGs, the PSAs draw power from any available source” this includes the main battery via the hot battery bus as mentioned few lines before in the same paragraph.’

    The PMGs mentioned in that paragraph are the power sources for the flight control power supply assemblies (PSAs). They are located on the backup generators, but they are not controlled by the backup generator switches. The PMGs will continue to power the PSAs even if the backup generator switches are selected OFF, provided the associated engine is running.

  242. 370Location says:


    Here’s a WSPR graphic plot that might be useful for determining the expected number of aircraft detections by doppler shift.

    The typical frequency drift of each transmitter can be characterized by tallying up drifts reported by all receivers that picked up its signal. Vice versa for each receiver. Stats tallied include the mean drift and variance for each Tx and Rx for a month of WSPR. For any station pair, the two means summed give an expected drift offset. Subtracting that from each detection gives a frequency shift for each “spot”.

    Slow frequency drifts for Tx and Rx are compensated by using a rolling mean that tracks each station.

    The horizontal axis is the log of summed variances (not rolling) for each station pair. This puts the most frequency stable stations on the left side of the graph, and the most unstable on the right. I’m plotting the results for Mar 2021 on the assumption that there are far more frequency stable stations this year than in 2014. The graph plots over 72 million contacts, 10x the activity in Mar 2014.

    For aircraft scatter, I would expect to see a significant doppler shift that should be especially visible on the most frequency stable stations. No doppler spots appear on the left half of the graph, so they must be extremely rare.

  243. Victor Iannello says:

    @Jean-Luc Marchan said: First my appologies for the “epdidermic” reaction, I was oversimplifying and reacting to the tone of some comments… 🙂 of course, you put a lot of serious analysis into your work.

    I was reacting to your mischaracterization of how the preferred route was selected in the UGIB paper, independent of how much work went into that analysis. There was a statistical basis for selecting the route towards the South Pole which you ignore. It was not because it was the “simplest heading”, although the route towards the South Pole is indeed simple.

    Further, based on the input from contributors here, we can safely conclude that the main battery would not discharge even if the IDGs, BU Generators, and APU were all isolated, so that should not be used to justify a scenario. The PMDG 777 model cannot be deemed reliable for non-standard configurations.

  244. @Victor
    The battery discharge modelled in our simulator was found as a possible rationale of the power switch back on. It appears not to be :-).

    But CAPTION trajectory is independent of the evoked IGARI to NILAM scenarios as they both lead to Arc1 the same way 🙂

  245. TBill says:

    @Andrew @Jean-Luc
    Re: RAT deploy question
    One question that Mike Glynn and I had discussed, is it possible to stop the RAT from deploying in the CAPTION/CAPTO power scenarios? In the B777 overhead panel there is circuit breaker entitled RAT AUTO CTL, which is presumably intended for use on the ground to prevent accidental RAT deploy.

  246. Andrew says:


    RE: “…is it possible to stop the RAT from deploying in the CAPTION/CAPTO power scenarios?”

    The RAT AUTO CTL circuit breaker on the overhead panel affects electrical power to the RAT actuator solenoid for the case where hydraulic power is lost. It does not affect RAT extension for the case where electrical power to both AC transfer buses is lost.

    Automatic RAT extension for the loss of electrical power case is controlled the ELMS (Electrical Load Management System). It is possible to prevent automatic RAT extension by pulling a circuit breaker on the standby power management panel, located in the MEC. If that circuit breaker is pulled, power cannot be provided to the RAT actuator solenoid when ELMS senses the loss of electrical power to the transfer buses.

  247. Tim says:

    @Jean-Luc @All

    As someone who supports an accident scenario and subsequent damage to the electrical system in the Left side of the E&E bay(O2 bottle rupture), I think your battery depletion idea might have merit.

    Briefly, here’s the scenario I propose to explain SATCOM loss at 17:21 and it’s reinstatement at 18:25—-At IGARI the electrical damage causes ELMS to load shed all loads off the L main bus(ref Boeing Maintenance Manual). The battery is then left supplying the hot battery bus and after 1hr it’s power is depleted.

    The only power to the P310 standby panel is from the hot battery or battery no2 bus(ref BMM). The no2 bus has already lost power due to damage to the centre TRUs. So at 18:25 the P310 panel losses power, this forces ELMS to rethink it’s load shed regime. The L main AC is repowered.

    This is a difficult failure to model, and I put this scenario out for scrutiny. I think that it goes a long way to explain how a damaged system might have behaved.

  248. TBill says:

    Thank you.

    Don’t forget you also have to account for ACARS transmissions staying off.

  249. Tim says:


    I explain the lack of ACARS due to the damage to the Audio Management Unit(AMU) and the L AIMS cabinet. Both in close proximity to the bottles.

  250. Don Thompson says:

    @Tim, @TBill,

    Tim – I understood that you posit “damage to the Audio Management Unit(AMU)” as the reason that no voice radio calls were made.

    Between the two AIMS cabinets, the hosted functions are ‘shadowed’ which each other. A function, all functions, normally executing on one cabinet will be recovered on the other (with the exception of ACMF).

    Still no recorded instance of an airliner COPV experiencing a violent ‘burst’?

  251. Victor Iannello says:

    @All: I created two videos to demonstrate deep fading of radio signals that propagate by skywave on HF bands. I used my Flex-6400 to receive two AM broadcast stations at my home in Roanoke, Virginia, and then recorded the windowed display for one minute. In the videos, you can see the strength of the AM carrier in both S units and dbm. (I verified that neither the modulation nor the received bandwidth has any effect on the indicated signal strength.) The two recordings were:

    1) Shortwave broadcast station WWCR in Nashville, TN, on 7.490 MHz at a distance of 629 km. The signal strength over the one minute varies between a minimum of -53 dbm to a maximum of -32 dbm, representing a difference of 21 db, or a power ratio of 126.

    2) Time broadcast station WWV in Fort Collins, CO, on 15.000 MHz at a distance of 2,200 km. The signal strength over the one minute varies between a minimum of -104 dbm to a maximum of -83 dbm, which coincidentally again represents a difference of 21 db, or a power ratio of 126.

    With the deep fading that is experienced over relatively short distances and short periods, it becomes evident that it would be “very challenging” to attribute small deviations in SNR to interactions with aircraft over long distances.

  252. Tim says:

    @Don @TBill,

    Re lack of COMMS

    Yes, damage to AMU, L & C VHF. Perhaps everything in the E1 rack.

  253. Victor Iannello says:

    @Tim: Let’s suppose that there was an accident that somehow caused the inoperability of VHF and HF radios, SATCOM, ACARS, and transponder. After the turnback, the plane passed near Penang International Airport in good weather, an airport that was both familiar and suitable for an emergency landing. Why did the pilot not land at Penang, but remained at cruise altitude, rounded Penang Island, and turned to the northwest to intercept N571 at VAMPI?

  254. Don Thompson says:

    Tim writes,

    Briefly, here’s the scenario I propose to explain SATCOM loss at 17:21 and it’s reinstatement at 18:25—-At IGARI the electrical damage causes ELMS to load shed all loads off the L main bus(ref Boeing Maintenance Manual). The battery is then left supplying the hot battery bus and after 1hr it’s power is depleted.

    The only power to the P310 standby panel is from the hot battery or battery no2 bus(ref BMM). The no2 bus has already lost power due to damage to the centre TRUs. So at 18:25 the P310 panel losses power, this forces ELMS to rethink it’s load shed regime. The L main AC is repowered.

    This is a difficult failure to model, and I put this scenario out for scrutiny. I think that it goes a long way to explain how a damaged system might have behaved.

    Let’s pick that apart & consider one step at a time:

    (1) “O2 bottle rupture [causing] damage to the centre TRUs [by dint of TRU C1 and C2 location on E3-2 shelf in proximity, yet separated by floor panel and bulkhead, to the O2 COPVs]”

    (2) “electrical damage causes ELMS to load shed all loads off the L main bus. The battery is then left supplying the hot battery bus and after 1hr it’s power is depleted

    SATCOM is not a system that is listed as affected by a ELMS Configuration Load Shed. However, if the L Main AC Bus could no longer be suppled by its local IDG, via the cross bus tie, or the APU then SATCOM does lose power but that is not an ELMS effect, it’s a source configuration change: the L Main AC Bus, and its subsidiaries, are simply ‘cold’. On the R Main AC side, if net demand from the galleys and other cabin utilities is not exceeding supply from its 120kVA IDG, ELMS won’t be shedding anything.

    Should C1 and C2 TRUs be unserviceable, then the Main Battery Charger is switched into TR mode, by ELMS, and it supplies the HOT BAT bus (thru to BAT and CPT FLT INST buses). The Main Battery is not discharging in that config. The Main Battery Charger is supplied by the R Main AC bus via the GND SVC bus. That C1 & C2 TRUs might be unserviceable does not imply automatic RAT deployment.

    (3) “The only power to the P310 standby panel is from the hot battery or battery no2 bus(ref BMM). The no2 bus has already lost power due to damage to the centre TRUs.

    The proposition that (1) and (2) cause effects that are resolved by invocation of the Standby Power System configuration, largely configured in P310, isn’t sound.

    The HOT BAT and BAT NO 2 buses carry 28VDC from the P310 panel. TRU C1, TRU C2, the MAIN BATTERY and the MAIN BAT CHARGER supply 28VDC to the P310 panel.

    (4) “So at 18:25 the P310 panel losses power, this forces ELMS to rethink it’s load shed regime. The L main AC is repowered.

    Incorrect: there is no Main Battery discharge, 28VDC continues to be supplied and distributed via TRU L, TRU R and the Main Battery Charger in TR mode. ELMS load shed ‘regime’ on the Main AC buses does not involve monitoring loads on the TRANSFER buses or the AC-DC TRUs. The loads ‘downstream’ of the TRANSFER buses and the AC-DC TRUs are constant, no load shedding is necessary.

    The proposed conditions for automagical restoration of L Main AC bus supply and SATCOM reboot at 18:25 are not plausible (not even close).

  255. Tim says:


    Thanks for your response.

    I’m just trying to find if there is a reason the battery could become depleted.

    1)As you say the E3-2 rack should be relatively protected from damage. However, I’m exploring options. Maybe, wiring looms to this rack were severed, or a bouncing bottle caused damage to both the E1 and E3. We know at least the ADIRU had power in the lower E2 row.

    2)The maintenance manual says ‘sequential load shed, ELMS sheds ALL loads on a main bus’ for 3 different reasons. Perhaps, I’ve misinterpreted what ‘ALL’ in this case means? Also, The Main Battery Charger is in the E3-2 rack, so is next to the centre TRUs and may have been damaged. So no battery charging.

    3)The maintenance manual says P310 panel is supplied by Hot Bat and No2 bus. Both these would be unpowered once the battery is flat.

    4)In my scenario only the R TRU remains powered…….I think the battery will discharge.

  256. Tim says:


    With severe damage to the left side of the E&E bay, the pilots faced a lot more than a COMMS fail. They would have been faced with—loss of primary flight instruments, half the electrics, pitot/static data, autopilot, autothrottle, and then a decompression which probably went unnoticed initially. Hypoxic, shortly after the IGARI turn.

  257. Tim says:


    Sorry read ‘ADIRU in the lower E3 row’

  258. Victor Iannello says:

    @Tim: Even with instrumentation and automation failures, in an emergency, the crew would have attempted a landing at Penang under visual conditions and manual flying. If there was hypoxia beginning at IGARI, the programming of the FMC to intercept a radial from VAMPI and fly along N571, as well as the later maneuvers, are not explained. The hypoxia theory also requires automated flight, as the plane continued to fly at cruise altitudes for hours.

  259. Andrew says:

    For those who might be interested in the Atlas Air 3591 accident that occurred early in 2019, the FAA has issued a Safety Alert for Operators regarding inadvertent pilot activation of go-around mode on B757/767 aircraft:

  260. Victor Iannello says:

    @Andrew: Thanks for the link. What an unfortunate sequence of events that hopefully can be avoided in the future.

  261. Don Thompson says:


    ATA 100 chapter-section reference from the Maint Manual would be useful.

  262. @VI
    Thanks for your site where we all can learn from it 🙂

    “has the RAT insufficient power to meet the power demands of the critical systems ?”
    In principle it has (except for the ADIRU cf below), but in FCOM it is said in sect 6.13 page 4 : the RAT has priority to provide hydraulic power to flight control over power to the TRUs. From which I interpret that for mechanical reasons the RAT can accept a maximum load meaning that HydraulicLoad + ElectricalLoad <= MaxLoad. So where to find the momentarily missing electrical power when needed? answer the only remaing extra source… the battery via the hot bus.
    In FSB (flight Standardization Report Rev6) Training report it is said in 125.205: the RAT and the battery alone provide the safe emergency operation of the airplane (f).
    So the duet RAT+Battery act together meaning that the battery will deplete if the RAT generator cannot suffice indicated in FCOM sec 6.6 p16 "Power for standy electrical loads is provided by the main battery when the RAT Generator loads are shed." I guess it is depending on the hydraulic needs specific to each flight.
    Recharging the battery is vaguely indicated but in view of the demand described above the battery will deplete eventually.


    Thanks Don for your detailed review. I will respond shortly directly to you.
    Concerning the ADIRU the situation seems to be that the Battery Hot bus directly powers the ADIRU and not the RAT.
    In FSB (flight Standardization Report Rev6) Training report it is stated (d) in 125.205: If DC power to the ADIRU fails it is then powered by the standy hot battery bus.
    In the FMC Honeywell (C28-3641-022 , section 4.1) document similar statement is made.
    For info one ADIRU draws roughly 1.5A i.e. ~4.5A for the 3 modules which is 10% of the possible battery current during 1 hour. It is not negigeable.

    Concerning what the battery may contribute to, hand to hand with the RAT, in FCOM:
    FCDC (Flight Control Direct Current) sect 6.6 p14: The hot battery bus provides additional backup power for the left and center PSAs only.
    So if it does under Backup powering conditions why shouldn’t it under RAT power (it would be illogical to revert in stopping the battery to contribute until it depletes). Intereseting to note that the Hot Battery Bus appears to be the primary power supply to the left PSA.

    I understand that the maintenance manual has no straight answer and to get it one has to fetch all the bits and pieces here and there to build a (almost) complete picture.

    I have discussed with B777 and others pilots: piloting the aircraft under RAT+Battery powering is relatively straigtforward (based on simulator sessions). It appears that, for them, the battery would even "die" earlier than one hour and so the ADIRU would stop well before NILAM. Because of that the navigation would be ensured thanks to the SAARU realigned time to time using the analog compass.

    Thus CAPTION Scenario2 (which cannot be dismissed for the reasons above:-) has two variants:

    1) the depletion of the battery would have been anticipated and thus the power was switched back on before this could happen
    2) the battery had depleted before NILAM already and the PIC waited to be out of Malaysian radar coverage to put the power back on. Why ? to come back to a piloting situation closer to "acceptable" after a probable voluntary depressurisation and subsequent minimalist flight conditions etc.

    Have a good day for those who read this long post 🙂

  263. Don Thompson says:

    @J-L Marchand


    You quoted the FCOM as stating “Power for standy electrical loads is provided by the main battery when the RAT Generator loads are shed.

    That is a selective quote. The MAS FCOM reads “Power for standby electrical loads is provided by the main battery during deployment of the RAT and when RAT generator loads are shed.

    The power deliverable by the RAT, and the Backup Generator Converter, is adequate to supply the loads within their distribution (excepting individual faulting loads where ELMS is capable of detecting and ‘shedding’).

    Neither of these notions of cascading electrical failures or extraordinary intervention to the electrical system stand the test to explain how the SATCOM operation was restored at 1825UTC.

  264. Victor Iannello says:

    @Jean-Luc Marchand: The source of power for the PSAs are the PMGs when the engines are running. There is no need to draw power from the hot battery bus.

  265. Andrew says:

    @Jean Luc Marchand

    RE: “…in FCOM it is said in sect 6.13 page 4 : the RAT has priority to provide hydraulic power to flight control over power to the TRUs.”

    That statement is correct, but it’s only relevant for the case where the RAT has deployed due to low pressure in all three hydraulic systems. In that scenario, the RAT is the only source of hydraulic power for the primary flight controls. In your scenario, however, the primary flight controls remain powered by the L & R hydraulic systems, which are pressurised by the engine driven hydraulic pumps. The hydraulic demand on the RAT is low, especially in cruise flight where there is little manoeuvring. It is very unlikely the RAT would need to shed electrical loads to satisfy the hydraulic demand.

    RE: “Concerning the ADIRU the situation seems to be that the Battery Hot bus directly powers the ADIRU and not the RAT.”

    The ADIRU does receive power from the Hot Battery Bus; however, with the standby power system active, the Hot Battery Bus is powered by the RAT generator, via TRU 1, the Capt Flt Instrument Bus and the Battery Bus.

    RE: “FCDC (Flight Control Direct Current) sect 6.6 p14: The hot battery bus provides additional backup power for the left and center PSAs only.”

    As I said in my previous comment, the PSAs remain powered by the PMGs on the backup generators, provided the respective engine is running. There is no demand on the battery.

  266. Tim says:

    RE hypoxic flight starting at IGARI. I don’t believe there is enough evidence the autopilot was ever re-engaged. Flight along N571 has never been proven and I have shown on simulators that prolonged flight at high cruising levels could occur with flight controls degraded to ‘secondary’.

  267. Victor Iannello says:

    @Tim said: Flight along N571 has never been proven

    That means you believe the Lido Hotel image that was presented to the NOK in Beijing was fabricated by Malaysia. Independent of that that image, the final civilian radar data around Penang aligns perfectly with VAMPI, as I documented in this blog post. That does not seem possible if the plane was flying without pilot inputs in secondary mode.

    If there was hypoxia and incapacitation followed by flight at cruise altitude in secondary mode, what caused the plane to turn gently from southwest to the northwest around Penang, and then again to the south when north of Sumatra?

  268. Tim says:

    I admit I have a problem, I don’t know how a real 777 in secondary mode, hands off, will behave under real wx conditions. I doubt even a Boeing test pilot could answer that question either.

    But what we know so far, seems to fit surprisingly well with damage to the left side of the E&E bay, followed by hypoxic flight all the way to the out of control crash 7 hours later.

  269. Victor Iannello says:

    @Tim: I think you can make an argument that in NORMAL mode, the wings would return to level after the aircraft encountered an atmospheric disturbance. In SECONDARY mode, I believe the stabilizing dihedral effect would be easily overcome by aerodynamic asymmetry (e.g., rudder out of trim) or atmospheric disturbances, and the ensuing bank would progressively increase, resulting in a steep descent.

    Do we have any empirical evidence of any plane flying long distances with no pilot inputs and with the autopilot not engaged?

  270. TBill says:

    Re: Flight Sim
    How do we simulate secondary mode in FSX/PMDG, or are you using something else?

  271. @all

    Thanks for your input 🙂

Leave a Reply