A Southern MH370 search location based on underwater implosion

Newly acquired hydrophone recordings have revealed a surprising find. Among the acoustic signals characteristic of Antarctic ice break events, a strong and clear event stands out with a unique signature and location that fits with an underwater implosion. The proposed location is found by triangulation using signals from seven hydrophones at three widely separated listening stations. The very long baselines for triangulation have the potential for high accuracy with calibration, less than the the 4 km depth of the ocean at the site.

MH370 370Location.com - Google Earth Overview
MH370 370Location.com – Google Earth Overview

The new search location at approximately S 56 E 90.5 is far South, closer to Antarctica than the 7th Arc search zone, requiring that the plane flew for another 2.5 hours. A flight bearing of 182.1 degrees would have taken it through the center of the area exhaustively searched by sonar.

The plane would have needed more fuel than it was reported to be carrying for the expected 7 hour range, on a course approximately due South along the 90th meridian. Various electronics on the plane were disabled during the flight, including the ACARS signaling gear that stopped at the 7th arc.

For background on the MH370 acoustic triangulation approach, please see the proposal prepared here for accessing the needed data.


Until June 2016, the only hydrophone data obtained for this project was through IMOS, an Australian Integrated Marine Observing System with three individual hydrophones in Western Australia listening for animal activity. They were not designed for triangulation or continuous monitoring. Each hydrophone recorded for five minutes out of every fifteen, and was placed in relatively shallow water rather than directly in the SOFAR deep sound channel where sounds can propagate across oceans. Still, several implosion candidate signals had been identified for March 8 2014.

Rottnest 05:03:02 Arrival

One of the most distinct signals arrived at the Rottnest 3315_PerthCanyon hydrophone at 5:03:02 GMT but was missed by the two other idling IMOS hydrophones and could not be triangulated without other recordings. Lacking access to CTBTO hydrophone data (and time to pursue it), this project was put on hold in late 2015.

The CTBTO has said that they examined their hydrophone data for implosion events and found none.  The only other known searches for an implosion event were conducted by Curtin University in Perth on CTBTO HA01 Cape Leeuwin hydrophone data from 0000-0800 GMT. At a threshold sensitive enough to reveal dozens of events, all but two of the bearings pointed toward Antarctic ice events, producing no plausible result. Another Curtin search of HA08 Diego Garcia data from 0000-0300 GMT tracked noisy seismic surveys that obscured other acoustic events but gave a good assessment of the HA08 bearing accuracy.

On May 31 2016, long awaited recordings from the CTBTO HA01 and HA08 hydrophone arrays were provided through an undisclosed source. Spectrograms were visually scanned for candidate events, especially for 7th Arc search area timing.

With no 7th arc event matches on HA01, the HA08S recordings from Diego Garcia were examined for 7th arc timing or matchups with HA01.  All of the candidate HA08 events also appeared to have bearings pointing toward ice events. Looking more closely, several events on both hydrophones did not have the characteristic subsonic upward sweep in frequency, the dispersion being caused by reflections between the ocean surface and shallow seafloor along the coast. Perhaps these were ice events over deeper water.

Leeuwin H01 04:59:20 Arrival
Leeuwin H01 04:59:20 Arrival

The strongest signal to arrive at HA01 for several hours of recording did not have an upward sweep. It was short and broadband without resonances, and a crescendo to a very distinct peak that fell off quickly. It arrived at 04:59:20 with a bearing of 209.5 which points toward the far Western edge of Antarctica, nearly missing it. It is not surprising that the 4:59 HA01 signal was classified as an ice event along with all the others.

Another signal arrived  HA08S Diego Garcia at 05:28:52 with the same signature, the strongest for about an hour before or after. It also had a bearing of toward Antarctica, but further East at 166.8 degrees.

H08S 05:28:52 Arrival at Diego Garcia
H08S 05:28:52 Arrival at Diego Garcia

It was only by matching the 4:59 HA01 arrival with the 5:29 HA08 signal and comparing their timing that the unique nature of the signal became evident.

If both arrivals were from the same event, they arrived too early for either to be from an ice event, regardless of the bearings. The required wave speed of sound in water from an ice event to those hydrophone locations was far too high.

This signal matching was not an automated task. With the help of a very useful geo.javawa.nl/coordcalc online map plotting tool (which uses oblate great circle calculations), iterative estimates were made for intersections from hydrophone pairs by turning their arrival timing differences into a path length difference. Two receivers define a path at varying distances that arcs away from the nearest hydrophone and becomes a distant bearing.

Estimating that path by plotting the HA01/HA08 arc by distance plus path difference found a close intersection to the independent bearing lines from both hydrophones, and the sound propagation times near S56.5 E 90.3 agreed. Comparing this location with a triangulation arc between HA01 and candidate Rottnest signal found a third close match in location and timing.

Of course the real process was more convoluted. There were several false starts and attempts to get results near the 7th arc by choosing different match candidates plus doubt over the accuracy of wave speed estimates and hydrophone placement. A 2015 370Location search area proposal on the 7th arc was nearly released but fortunately retracted when recalculation found a confirming third hydrophone. That near miss was partly on the basis of discounting the accuracy of existing hydrophone placement, but also on the premise from an early false calculation. All caution has been taken to avoid a similar mistake.

Even with the evidence at hand, doubts that this was really an ice event kept resurfacing. A search was made on the alternate hydrophone for any arrivals around the right time for ice events on the candidate bearings, but no signals came close. The same online tool was used to map the intersections of three circles centered on the hydrophone locations. With various wave speeds, only slight adjustments for path errors brought the intersections nearly to a point.

Concern arose that the positive result was an artifact of the method of choosing path length differences with a common wave speed, but trying different methods came to the same convergence point. Double checking hydrophone bearing calculations only brought the intersections closer.  The spot is consistent with the bearings from the two triads, but independent of them. The HA01 hydrophone is better calibrated, so gets more emphasis in centering the search target. With ice events excluded by arrival times, the confidence is now high that this location is a good match for an implosion event.

Time of Impact

The biggest issue with the far Southern location is the distance the plane continued to fly past the 7th arc. Other end-of-flight analyses have mostly been based on a fuel exhaustion flame-out scenario around the time of the 6th arc, causing a presumed power outage, automatic deployment of an air turbine generator, and restart of the electronics around the 7th arc.  Given the new distance from the 7th arc, an estimate was made of how long it would take the plane to fly to the Southern convergence point.

The Nov 2015 Australian Defence Science and Technology MH370 analysis book (.pdf), describes a plausible range of airspeeds from Mach 0.73 to 0.84 with a separate graph relating the satellite detected BFO frequency to airspeed toward the 7th Arc. A value from the figure slightly above Mach .80 seems plausible, and .81 conveniently matches 1000 km/hr to simplify calculations. On a consistent flight bearing of 182.1 from the equator to the target location, a distance to the 7th arc crossing of 2270 km would take 2.27 hours = 2:16:12. Adding that to the 7th arc time of 00:19:37 gives an estimated earliest impact time of 02:35:49 GMT.

To match a possible impact signal, the estimated sound travel time over 3000 km to HA01 (at 1.4754 km/s) is about 33:24 for an arrival of 3:09:01. No strong signals near the bearing of 210 stand out, but plotting a correlation map has found a short HA01 impact candidate at 03:22:21 GMT. The matching HA08 data is unfortunately missing 24:24 of recording before the expected arrival time there, with no strong signals after the gap.

It was expected that an implosion sound in the SOFAR channel could travel much farther with less distortion than an impact  at the surface. The sound channel is close to the surface at this location because the lower density ice melt floats on the more saline ocean water, forming a shallow surface layer. An impact in a shallow SOFAR path should be detectable. It is possible that with enough fuel to stay airborne that this signal event could have been an impact, but unlikely.

A visual search for other implosion events at this location did not turn up obvious candidates. It may require an algorithmic matching of spectrograms with multichannel noise reduction and beam forming methods to derive any weaker implosion events. It was anticipated that these techniques would be needed to detect any implosion events at all, and the strong signal was a big surprise.

Narrowing the Search Area

Sound Speed profile from the site to H01S – Estimate is 1.4754 km/s

The search area accuracy with the long baseline triangulation is mostly determined by the wave speed accuracy over each path. Even though hydrophone location errors get magnified at the destination, they are known to within fractions of a kilometer. Small wave speed changes over thousands of kilometers also add up rapidly. This is largest current source of error, especially near the Australian coast.

A faster wave speed for 200km along the shallower coastal shelf would effectively shorten the path along that leg by 2 seconds or 3 km. Much slower wave speeds in the less saline water have an larger counter effect. The wave speed difference is less of an issue for the narrow baseline triads with 1 km spacing, since the sounds have traveled essentially the same path variations to each hydrophone.

Averaging the sound speed profile over each path should give a best approximation, especially if live buoy data could be retrieved.  The World Ocean Atlas provides gridded salinity, temperature and depth, allowing that calculation using Ocean Data View. This has resulted in path estimates for group velocities for March of 1.4754 km/s to HA01, 1.4815 to HA08, and 1.4759 to Rottnest.

The most robust narrow search path is along the H08-H01 timing difference path. This triangulation path marked in green on the map has a very long baseline and good orientation toward the target. Large changes to the two group velocities, timing errors, and hydrophone location errors shift the path E-W only a few km.  The H08-Rottnest path is also fairly robust, differing by only 4 km near the target.

While HA01-Rottnest has a long baseline of 345km, it is not oriented well against H01 for a best accuracy. The aperture is down to 100km at the 208 degrees. The path travels along shallower coastal water and up a canyon, which affects the timing.  Since the resulting bearing and path with H01 are not as consistent as the other combinations, it gets less emphasis.

The most probable location along the robust path is where the separate triad bearings from H01 and H08 cross it. Their bearings are reportedly consistent within 0.2 to 0.5 degrees, depending on the calibration and movement of the hydrophones on their moorings. Those two bearings cross about 50 km on either side of target location. H01 has slightly better calibration, and a bearing calculation using a different algorithm crosses within 25 km of the target.

Without calibration, searching an area of several km wide but fairly long would probably require more bathymetry before sidescan sonar mapping. Fortunately, a calibration can be performed by setting off small detonations or implosions near the target site, then downloading the hydrophone results after arrival. The CTBTO hydrophones are streaming realtime data, so a corrected location could be available within hours.

It is recommended that test charges be released at different depths at a precise location. Detecting the time with a surface hydrophone at the site should be accurate within milliseconds if the depth is known.  Surface charges and different crush depth calibrations could help determine the spectral dispersion signature for implosion depth or an impact/explosion event.

Inaccuracies due to group velocity estimates and non-geodesic great circle math would be compensated by the calibration. The main remaining variable after calibration would then be seasonal temperature changes, and adjustments for turbulent changes in salinity since the 2014 event.

If the accuracy is then down to fractions of a second in arrival times, the search radius could be smaller than the debris field.  A single expedition to the site could calibrate the hydrophone paths, correct the location, and launch an ROV without needing bathymetry surveys for a sonar towfish. A bomber aircraft could also accurately drop the calibration charges along with a hydrophone buoy for transmitting the precise detonation times.

Ongoing Investigation

Rather than a static scientific paper with a bibliography, this presentation has the advantage of a quicker delivery with links to sources and ongoing revisions to the supporting data, figures, maps, and .kmz data as they are developed and refined.

It would be most beneficial if the full hydrophone and infrasound dataset could be published for analysis by anyone with an interest. With proper permission, that may be forthcoming.

The search is still on for confirming or alternative implosion events. Correlation plotting has found another HA01 implosion candidate at 04:38:37 on the same 210 bearing.

Another hydrophone triad array closer to the 7th arc near Amsterdam island has been found. It was active during the event and records continuously along with two other single hydrophones deployed further North and South. This could provide confirmation for the impact event time and help narrow the location. A request for that data is pending.

Searching by hydrophones will continue until the plane is found. The CTBTO infrasound is still unexplored territory.

Many thanks to Dr Alec Duncan at Curtin.edu.au for his patience, help, and sharing of his expert knowledge on hydroacoustics.

Sincere condolences go to the friends and families of those lost on flight MH370. This effort is for you. Please forgive any insensitive language in delivering the analysis.

— Ed Anderson

Please use Google Earth on a desktop computer to load the .kml file, which contains additional notes.

H01S cross-correlation scatterplot. Red=31vs21 Green=-23vs21 Blue=31vs23

[xyz-ihs snippet=”customize-mh370-map”