MH370 and Other Investigations

Malaysian investigator Aslam Khan (left), Blaine Gibson (center), and Consul Zahid Raza (right) at the Ministry of Transport in Antananarivo, Madagascar, in December 2016

On August 24, Zahid Raza, serving as the Malaysian consul to Madagascar, was killed in the Malagasy capital of Antananarivo. According to reports, he was shot multiple times while seated in the driver’s seat of his car. Mr Raza is reported to be of French-Malagasy nationality.

Last December, Reuters reported that Mr Raza assisted Blaine Gibson in transferring the custody of pieces believed to be from MH370 from Madagascar to Malaysia. At that time, six pieces were transferred. This has raised questions as to whether there was a link between those MH370 parts and Mr Raza’s death.

What makes a possible link to MH370 even more suspicious is that in the time period surrounding his death, Mr Raza was expected to visit the Malagasy Ministry of Transport, retrieve additional recovered pieces, and deliver those pieces to Malaysia. In a private communication from Blaine to me, he writes (repeated here with his permission):

On August 16 possible MH 370 debris was handed over to Madagascar authorities, and authorities in Malaysia were notified. Under the agreement between the two countries, debris is supposed to be collected by Hon. Zahid Raza, the honorary Malaysian Consul in Madagascar, and delivered by private courier to Malaysia.

On August 24 the Hon. Zahid Raza was assassinated in Antananarivo. At first we did not know if he had picked the debris up before this tragedy. We just learned that the debris is still safely in the hands of Madagascar authorities. However new arrangements must be made for the collection and delivery of debris.

Our thoughts and prayers are with the family of the Hon. Zahid Raza.

In the aftermath of Mr Raza’s death, there seems to be conflicting stories about his background. He appears to be of French-Malagasy nationality, with family living in France and in the French Reunion Island. In one report, he is referred to as the “former” consul, but other reports imply he held the title of Honorary Consul at the time of his death. There is a report linking Mr Raza to a group associated with the kidnapping of residents of Indo-Pakistani descent that are living in Madagascar. (Madagascar does not grant automatic citizenship to  those born on Malagasy soil. As a result, some Indo-Pakistani residents are from families that have lived in Madagascar for over a century.) The association of Mr Raza with the kidnappers has not been confirmed, and could be disinformation. Hopefully, the facts surrounding this will surface.

Surprisingly, the assassination of Mr Raza has been met with stony silence from both Malaysia and France, despite his ties to both countries.

I join Blaine in expressing my sincere sympathy to Mr Raza’s family.

Oliver Plunkett, CEO of Ocean Infinity

I had the opportunity to converse with Oliver Plunkett, who is the CEO of Ocean Infinity (OI). My goal was to learn more about OI’s offer to search for MH370 in the Southern Indian Ocean (SIO). Although Mr Plunkett would not disclose the details of the confidential negotiations with Malaysia, he did provide information that is helpful to understanding the general terms of OI’s proposal.

First, OI’s offer is structured such that OI assumes 100% of the economic risk for the search. OI will NOT receive any payment if the wreckage is not found. So it would appear that if the success fee that OI is proposing is less than what Malaysia would have spent in conducting the search using conventional techniques, this is an extremely attractive offer.

I learned a bit more about the recent sea trials that Ocean Infinity recently conducted in the North Atlantic. The tests demonstrated that the underwater autonomous vehicles (AUVs) could be successfully launched and recovered. Each AUV also demonstrated that it could independently scan the seabed. Mr Plunkett said he was pleased with the results so far. Further work is planned at deeper depths and over a wider range of conditions. Mr Plunkett also explained that although the unmanned surface vehicles (USVs) could not be used in the roughest of sea states, the search for MH370 could nonetheless occur over a wide range of conditions. This is because the mission is to scan the seabed and identify the wreckage rather than to generate precise maps.

I inquired about the window of opportunity for completing the negotiations with Malaysia and starting the search. (We know from previous underwater search efforts that the search season in the SIO runs approximately from December to March.) Mr Plunkett is optimistic that Malaysia and OI will reach an agreement in a time frame that allows for adequate time to prepare for a search that begins this season.

Finally, I asked whether OI had already determined the specific area to search. Mr Plunkett explained that OI intends to complement its internal resources with input from other organizations and other outside experts to help define the search area. OI has already had some interaction with the ATSB, which he believes is completely committed to finding the wreckage. I don’t expect that OI’s search area will be very different than what we have been discussing here.

Over the course of our discussion, it became apparent that Mr Plunkett was aware of the many posts and discussions that appear on this blog.

With such a favorable offer on the table from an innovative and qualified firm, I remain optimistic that the seabed search will re-start. However, for the search to begin this season, the window of opportunity to complete the negotiations is narrowing. It is imperative that Malaysia not miss this opportunity.

Update on Aug 16, 2017

In a story appearing the New Straits Times, Malaysia’s Department of Civil Aviation (DCA) reveals that the OI proposal to restart the subsea search is one of several proposals that are under review. The proposals will be brought to the attention of Australia and China for their views.

Update on Aug 17, 2017

Voice370, representing the MH370 families, released this statement today which questions the delay in re-starting the search in light of the confidence expressed by CSIRO in identifying a probable impact site.

Ocean Infinity’s technology uses multiple underwater drones and surface vessels with a single host vessel

Yesterday, a support group for MH370 families released a statement claiming that a private entity has offered to resume the seabed search for the aircraft with the understanding that it would collect a fee only if the aircraft wreckage was found. Today, through Grace Nathan, a Malaysian lawyer whose mother was a passenger on MH370, we learn that the private entity is a US-based firm called Ocean Infinity.

Readers here are already familiar with Ocean Infinity. In a recent post entitled Advanced Underwater Drones May Help Find MH370, I highlighted the innovative research at Virginia Tech in developing underwater autonomous vehicles (AUVs) that could collaboratively scan the ocean floor. In an update to the article, I stated:

I was recently in a discussion that included a well-known ocean explorer who happens to be a judge in the Ocean Discovery XPrize competition.  We were having a general discussion about searching for MH370 and ways to scan the ocean floor at high resolution, and he told us about the capabilities of Ocean Infinity. Like the team at Virginia Tech, their approach is to employ a team of AUVs. From their website:

Six HUGIN autonomous underwater vehicles (AUVs) are capable of operating in 6,000 m water depth collecting high resolution data at record breaking speeds. Our AUV fleet is accompanied by six unmanned surface vehicles (USVs) to ensure precise position and constant communication.

With multiple autonomous vehicles working simultaneously utilizing innovative technology, we are able to survey huge swaths of the seabed, quickly and with outstanding accuracy. We can operate in shallow waters but excel in extreme depths, working in dynamic environments ranging from the tropics to the Arctic ice.

Because of the size and complexity of each AUV/USV pair, the capital cost of the technology from Ocean Infinity would greatly exceed the capital cost of Virginia Tech’s technology, which uses small AUVs with innovative navigation systems. On the other hand, both approaches benefit from having a single host vessel supporting multiple underwater vehicles, which offers significant operating cost and scan rate improvements compared to the conventional towfish technology.

Ocean Infinity’s seabed exploration system is commercially available today, including underwater and surface vehicles, on-board support equipment, and the host vessel. This is an exciting possibility for conducting the search for MH370 in the near future.

I can now say that the “well-known ocean explorer” was David Mearns. At the time that I posted the article, I was not aware that Ocean Infinity had any interest in searching for MH370, although I was hoping they did. The prospect of exploiting Ocean Infinity’s technology in the near future is great news.

That means that Malaysia, Australia, and China need to make a decision: Either the tri-partite countries should provide funds to re-start the search; or, the countries should fully cooperate with Ocean Infinity and other qualified entities that are interested in re-starting the search. Any other action is unacceptable.

In a TV interview with Australia’s Studio 10,  shipwreck hunter David Mearns reveals that he would like to mount a privately-funded search for MH370.

From the interview:

“But since the search has been suspended, which I think is basically an unacceptable thing to have happened, I’ve been working with some families and some experts to see if we could actually mount a privately-funded search for the plane, because it’s inexcusable that that wreckage isn’t located, because it can be found. They just have to look in the right place.”

“It can be found. The technology is there to find it. They just need to be able to look in the right place.”

“And they’re narrowing the areas. The next search will be smaller than what’s been already done.”

“And everybody should be concerned about this, because until that black box is found, and we recover the black boxes, we don’t know what happened.”

When asked if they are looking the right place, he responded, “No, because they have  not found it. When they look in the right place, they will find it. And it can be done. I’m here to tell people that it can be done.”

“And that’s the other key thing. Technology has moved on, so now that we can search much faster than before.”

When asked if he thinks he can find MH370, he responded, “I never guarantee these things, but I believe it is definitely worth doing. There’s an area that can be searched in an efficient way, and I believe we don’t just owe it to the families. But I think, internationally, it’s an important thing to do.”

“This will be the first time a major aircraft like this has been lost without any resolution or any lessons learned about why it crashed. And that is not only unacceptable, it’s inexcusable for the authorities not to be able to continue to do something.”

I should add that David Mearns has had extensive discussions and meetings with some members of the MH370 Independent Group (IG), in which many of our analyses and findings have been shared. I hope that collaboration continues.

Global coverage of Inmarsat’s I3 network, showing the overlap region of the IOR and POR satellites

Last month, we published the complete logs for all communications that occurred on March 7 and 8, 2014, between the SATCOM unit aboard airframe 9M-MRO and the Inmarsat satellite network. (All times and days refer to UTC.) This includes communications before and during MH370 as well as the previous flight, MH371, between Beijing and Kuala Lumpur. Now that we’ve had a chance to investigate the logs for several weeks, I’ve summarized some of the findings. I’ve attempted to give proper credit to the individuals that worked on various aspects. If I have inadvertently omitted an individual, just let me know.

Observation: The log-on requests at 18:25:27 and 00:19:29 both had low carrier-to-noise-density (C/No) ratios, but normal receive power levels, indicating high noise levels. Similarly low (C/No) ratios were observed several times during MH371 under normal conditions. (Mike Exner)

Inference: The low (C/No) ratios at 18:25:27 and 00:19:29 were not likely due to abnormal aircraft maneuvers or attitudes.

Observation: When MH371 was traveling in a region of overlapping satellite coverage of the IOR and POR satellites, there were multiple automatic log-offs and log-ons with no indication of problems. (Many)

Inference: The multiple log-off and log-ons seen during MH371 are not indicative of a problem with the SATCOM.

Observation: An in-flight log-on does not produce abnormal values of BFO unless the log-on was part of power up sequence following an extended period during which the SATCOM was powered down. (Many)

Inference: This increases the likelihood that the SATCOM was unpowered for an extended period of time prior to the log-on at 18:25, and increases the likelihood that the abnormal BFOs during the log-on at 00:19 were due to an increasingly high rate of descent.

Observation: Abnormally high BTO values for a log-on request burst can be corrected with an offset of 4600 μs. (Many)

Inference: The corrected value of the BTO at 00:19:29 is 23000 – 4600 = 18400 μs, as previously suggested by Inmarsat. An adjustment to the position of the 7th arc does not seem to be warranted.

Observation: Abnormally high BTO values for a log-on acknowledge burst can be corrected with an offset of N*S, where N is an integer between 1 and 5, inclusive, and S=7812.5 μs. The value of S=7812.5 μs corresponds to the width of a slot, where a frame of 500 ms is comprised of 64 slots. (Don Thompson)

Inference: The corrected value of the BTO at 00:19:37 is 49660 – 4*7812.5 = 18410 μs, which statistically agrees with the corrected value of 18400 μs at 00:19:29. Again, an adjustment to the position of the 7th arc does not seem to be warranted.

Observation:  Maintenance messages were generated after MH371 landed in Kuala Lumpur. However, no ACARS maintenance messages were generated during the flight. (@Andrew)

Inference: If a serious condition had arisen during MH371, it would have generated an ACARS message. Therefore, no serious condition arose during MH371.

Observation: When a 2nd log-on request message occurs one second after the first, it is related to initialization of the In-flight Entertainment System (IFE), and the message does not contain information about the Flight ID. If a 2nd log-on request does not occur, it suggests the IFE was not available at that time, possibly because the IFE has not yet completed its power up sequence. (Don Thompson)

Inference: This increases the likelihood that the IFE was unpowered prior to the log-on at 18:25 and unpowered prior to the log-on at 00:19.

Observation: During a log-on sequence, the SATCOM transmits a value for the “Prev Sat ID”. If the log-on occurs after a log-off request, or after a power interruption, the previous satellite value is cleared and a value of 63 (077) is transmitted. This value was transmitted for the log-on at 18:25 and the log-on at 00:19. There may be other causes for 63 to be transmitted that did not occur during MH371. (@el-gato, Don Thompson, and Richard Godfrey)

Inference: Since no log-off request was recorded prior to the log-ons at 18:25 and 00:19, it is likely that a power interruption preceded each of these log-ons.

Observation: Fuel flow data extracted from the ACARS reports for MH371 showed that the right engine burned fuel about 3.3% faster than the left during cruise. (Mike Exner, Don Thompson, Richard Godfrey)

Inference: If there was no fuel rebalancing by a pilot, the right tank for MH370 would have run dry about 15 minutes before the left tank.

Observation (preliminary): The measured values of BTO and BFO for MH371 agree with the BTO and BFO models that were used to reconstruct the flight path for MH370. (@sk999, Richard Godfrey)

Inference: The measured values of BTO and BFO for MH370 can be used to disqualify hypothetical paths with predicted values of BTO and BFO that do not match the measured values, as the ATSB and independent investigators have assumed.

In summary, the previous flight MH371 seems to have been normal in all respects. Using the satellite data from MH371, we have a higher level of confidence that for MH370, power was interrupted to the SATCOM prior to the log-ons at 18:25 and 00:19, and also higher level of confidence that the aircraft was in an increasingly steep descent at 00:19.

Considering that the newly available data generally supports the conclusions of the official investigators, it remains a mystery as to why Malaysia withheld the data for so long, and why it chose to release the data at this time.

I hope everybody is enjoying today, the Fourth of July, including Americans celebrating Independence Day.

Inmarsat’s Mark Dickinson holding the satellite data in an interview with CNN

Since we first learned of its existence, we’ve been asking for the complete record of the communications data between MH370 and Inmarsat’s satellite network. In May 2014, Malaysia released satellite data logs, but they were incomplete: fields of data were missing, and only a small number of data records from before the flight was made available. When pressed for the complete logs, Inmarsat and Malaysia both claimed the data had to be released by the other.

We now have what we believe is the complete record of communications between airframe 9M-MRO and the Inmarsat satellite network, from March 7, 2014, at 00:51 UTC, until March 8, 2014, at 01:16 UTC. This time period includes the previous flight from Beijing to Kuala Lumpur.

The satellite data was shared with me by a relative of a Chinese passenger on MH370. The data was given to him by Malaysia Airlines with the following email text:

Please find attached the Inmarsat data, for your info. Please note that these are raw data as you have requested. The authorities agree to release the data, on condition that:

1. We will not translate the data into any meaningful information as the data is proprietary to Inmarsat. The Malaysian Investigation team does not have any experts to translate these data into any meaningful information.
2. We will not translate the data into any other language, including Mandarin.
3. These data are complete and obtained from Inmarsat. Please do not manipulate the data.

I know, by having these data, you will have more questions, but I have to say that we are providing these data to satisfy your request, but we cannot answer any questions on the data because we too, cannot understand it. Only the experts from Inmarsat can.

 Hope you understand.

 Thank you

I suspect the data will confirm some assumptions, and will raise even more questions. I hope the data can help us learn more about the disappearance.

Introduction

Boeing recently conducted end-of-flight simulations for MH370 with the assumption that there was no pilot input. The results were released in November 2016 by the ATSB as part of a report entitled MH370 – Search and Debris Examination Update. Boeing observed that in simulations where the aircraft experienced a descent rate consistent with the values and timing of the last two BFO data points, the aircraft impacted the water within 15 NM of the 7th arc. However, the details and the likelihood of the configuration that caused the high rates of descent were not discussed.

In order to better understand the conditions leading to the descent rates suggested by the final BFO values, and to estimate the distance MH370 might have traveled after crossing the 7th arc,  simulations were conducted using the PMDG 777-200LR model add-on to Microsoft Flight Simulator X (FSX). After making adjustments for differences between the PMDG 777 model and MH370, the flight characteristics were recorded under various conditions.

Modeling the End-of-Flight Using FSX

9M-MRO was a B777-200ER and the PMDG 777 model is a B777-200LR. The main differences are:

• Weight: The 777-200ER has a maximum take-off weight (MTOW) of 297.6 MT, while the 777-200LR has a MTOW of 347.5 MT, so the 777-200LR is in general a heavier aircraft. In the simulations, the zero-fuel weight (ZFW) was set to 174.4 MT to be consistent with MH370.
• Engines: 9M-MRO’s version of the 777-200ER has Trent 892-17s with 41,768 kgf (90,000 lbf) of thrust, while the PMDG 777-200LR has GE90-110B1s with 50,285 kgf (110,760 lbf) of thrust. Since only the results from the simulation after fuel exhaustion are used, the difference in engine thrust is not important.
• Wings: The 777-200LR version has a wing area about 2% greater area than the 777-200ER due to raked wingtips, which decrease the wing loading and reduce drag in cruise. The small difference in aerodynamic performance from the raked wingtips would produce little if any difference in the end-of-flight scenarios considered here, and are ignored.

In order to get realistic results from the FSX simulation, it is also important to recognize and compensate for other inaccuracies of the PMDG 777 model. In particular, it was found that the behavior of many systems when components have failed is not correct. Nonetheless, the basic aerodynamic model of the PMDG 777, when not at envelope limits such as stall conditions and transonic speeds, should be sufficiently accurate to model the flight characteristics of the B777. (Even a Level D simulator is not guaranteed to be accurate outside of the aircraft’s certified flight envelope.) The details of the flight dynamics model as incorporated into FSX are described by Yves Guillaume.

To model the flight behavior after fuel exhaustion, the simulation was conducted with the following initial conditions and programmed events:

• Initial conditions are stable flight at 220 KIAS and FL350, which is representative of flying with one engine inoperative and decelerating from the dual engine cruise speed, but still holding altitude.
• At t = 0, following events occur:
  • The fuel level is set to zero. In addition to shutting down the engines, this prevents the APU from starting.
  • The two fuel cut-off switches are set to OFF. Although it may not seem necessary to set the fuel level to zero AND employ the cut-off switches, in the PMDG model, the windmilling action of the engine shafts continues to supply electrical and hydraulic power if the cut-off switches are not employed.
  • The primary flight computers (PFC) are programmed to degrade to “secondary” control law.  Although the loss of power to the left and right transfer busses removes heat to the pitot sensors, which should automatically degrade the control law from “normal” to “secondary”, in the PMDG 777 model, the control law remains in normal mode without programming this failure. In normal mode, yaw compensation and envelope protection would be available, while they are not available in secondary law.

Implications of the BFO Values on Flight Dynamics

The last two values of BFO were 182 Hz at 00:19:29 and -2 Hz at 00:19:37. Assuming a BFO bias of 150 Hz and a nominal position at (30S, 98E), a groundspeed of 385 kn, and a track of 172°T, the corresponding descent rates are 4,000 fpm and 14,400 fpm. This represents an increase of 10,400 fpm over 8 seconds. After conducting studies of the various flight conditions that might cause these descent rates, the following observations are offered:

1. If the aircraft has perfect lateral trim, i.e., the ailerons and rudder are positioned to perfectly remove any lateral (side-to-side) asymmetry of the aircraft about its longitudinal axis, the aircraft will fly relatively straight with no pilot input. The aircraft will also develop a phugoid flight pattern consisting of a damped-sinusoid vertical speed component superimposed on the quasi-steady vertical speed component. The quasi-steady vertical speed would correspond to a descent angle of about 3°, and a glide distance of greater than 110 NM is possible. However, the amplitude of the phugoid would be smaller and the period of the phugoid would be longer than the descent rates suggested by the BFO values. The case of a straight flight with a phugoid descent is therefore not studied here as it does not match the BFO values.
2. If the aircraft is flying straight with little or no bank, the descent rates can match or exceed the descent rates suggested by the BFO values if the pilot commands a single nose-down input. The aircraft would correspondingly impact the ocean close to the 7th arc.  It is also possible that the steep descent could be arrested with a subsequent nose-up input, and the aircraft could be piloted to glide some distance (> 100 NM) from the 7th arc before the aircraft impacts the ocean.  Other than noting that with pilot input, the impact point could vary greatly in distance and direction from the crossing point of the 7th arc and still match the BFO values, this scenario is not considered here.
3. If the aircraft has lateral (side-to-side) asymmetry about the longitudinal axis that is not removed by the appropriate positioning of the ailerons and/or rudder, the plane will enter a bank with a roll rate determined by the magnitude of the lateral asymmetry. As the bank angle increases, the descent rate also increases. With sufficient lateral asymmetry, the descent rates can match or exceed the descent rates suggested by the BFO values. This case is the case studied here.

Simulation of Banked Descent and No Pilot Input

After fuel exhaustion, the control wheel was turned about 4.5 units to the left in order to match the BFO values by introducing lateral asymmetry. (With the “no-pilot” assumption, the position of the control wheel in MH370 likely stayed in the neutral position. In the simulation, the control wheel was turned to reproduce lateral asymmetry resulting from other sources.) The control wheel input induced a roll rate of about 3.6 deg/s. As the bank angle increased, the descent rate correspondingly increased. In the simulation, the increase of descent rate from 4,000 fpm to 14,400 fpm required about 9 s, while the measured BFO corresponding to these descent rates are spaced at about 8 s. The lateral asymmetry caused by the rotation of the control wheel is therefore judged to be about equal to the lateral asymmetry of MH370 after fuel exhaustion.

A video showing the view from the cockpit during the simulated descent is included below. The aircraft rolls past 180° and impacts the water at a pitch angle that is almost vertical. During the descent, the speed reaches about Mach 1.1, and the descent rate approaches 60,000 fpm (593 kn). Although these speeds are outside of the performance envelopes that can be accurately modeled by the PMDG 777 model, it demonstrates that the BFO measurements are consistent with a very high speed impact.

Another video below is an outside view of the aircraft during the descent, showing how the increasing bank leads to increasing pitch down.  At about 14 s, the deployment of the ram air turbine (RAT) can be seen and heard.

The figure below shows the trajectory of the aircraft. The position is adjusted so that the path crosses the 7th arc when the descent rate is about 4,000 fpm. There is only about 50 s between the time the descent rate reaches 4,000 fpm and the time of impact, and the impact is about 4.8 NM from the crossing of the 7th arc.

No wind is included in the simulation. However, considering that the aircraft impacts the water in about a minute and the winds were to the east, the effect of wind on the distance from the 7th arc should be small.

Possible Causes of the Lateral Asymmetry

The lateral asymmetry that induces the roll and the banked descent can be caused by a number of factors, including:

1. Geometrical asymmetries (“bend”) that would cause yaw and bank when the control surfaces are in their neutral position. Normally, this would be compensated by a pilot by adjusting rudder and aileron trim. However, the amount of trim might not exactly balance the asymmetry so there would be residual out-of-trim.
2. Asymmetric position of control surfaces caused by differences in hydraulic pressure that is supplied to the control surfaces on either side of the aircraft after fuel exhaustion.

After fuel exhaustion and without the APU operating, the main source of hydraulic pressure is from the RAT, which supplies pressure to the “center” hydraulic system. As can be seen below in the synoptic displays for the flight control surfaces and the hydraulic system, when the RAT is deployed, only the right flaperon has hydraulic pressure, and the left flaperon is “bypassed”, so that the flaperon moves freely. If the position of the right flaperon is positioned slightly down compared to the left flaperon, the result would be a roll to the left.

In fact, we have an indication that this would occur. In the Aircraft Maintenance Manual (AMM) for the B777, in the section on the Ram Air Turbine System, there is this note that was found by Don Thompson:

Training Information Point

When the RAT is extended and hydraulics off, the airplane rolls left. Two to three units of right control wheel rotation are necessary to hold the wings level.

In the simulation, 4.5 units of rotation were added to simulate the banked descent. Based on the note in the AMM, it appears that this level of wheel rotation causes asymmetric flaperon positions that are similar to what would be expected when the hydraulic pressure is supplied by the RAT.

The RAT would be deployed within seconds of the second engine flameout. However, the log-on request at 00:19:29 is believed to occur about two minutes after the second engine flameout. (One minute is required to start the APU and one minute is required for the SATCOM to request a log-on after power-up.) Therefore, there is a two-minute delay between deployment of the RAT and the roll. In fact, the pressure to the left and right hydraulic systems might decay over some period of time, as determined by the consumption of hydraulic fluid by various systems as well as the limited flow that is available as the engines windmill and the engine-driven pumps operate at reduced capacity. These factors might contribute to the two-minute delay before the aircraft entered into the banked descent.

Conclusions

The BFO values at 00:19:29 and 00:19:37 suggest that MH370 was descending at an increasingly high rate. With the assumption that there were no pilot inputs, the descent rates suggest the aircraft was in a roll as it was descending. According to Boeing documentation, when the hydraulic pressure is supplied by the ram air turbine (RAT), the aircraft banks to the left, which may have been the cause of the rolling during the descent. Using the PMDG 777 model add-on to FSX, simulations were performed for a banked descent that matched the descent rates suggested by the BFO values. In the simulation, the aircraft impacts the water at speeds around Mach 1 and with nearly vertical nose-down pitch. The distance of the impact point from the crossing of the 7th arc is less than 5 NM.

A fanciful depiction of a team of AUVs searching the sea floor as proposed by researchers at Virginia Tech

Despite an underwater search of the seabed in the Southern Indian Ocean (SIO) that covered 120,000 sq km, employed five search vessels, lasted 16 months, and cost Malaysia, Australia, and China a total of US$ 226 million, the MH370 wreckage remains elusive. The only debris from MH370 that has been found are parts that have drifted across the Indian Ocean and recovered from the shores of Eastern Africa.

Because of the need for specialized equipment capable of searching as deep as 6,000 m, operated by highly-trained crews in the exceptionally harsh conditions of the SIO, the underwater search is slow, expensive, and dangerous.

First, a bathymetric survey is undertaken to map the topography of the seabed. The bathymetric survey uses “multibeam sonar” transducers mounted on the hull of the survey vessel. By transmitting an acoustic pulse and measuring the time duration to receive an echo from the sea floor, the depth of the sea floor can be mapped with a resolution of about 100 m at a rate of 1200 sq km per day.

After the sea floor is mapped, the seabed is scanned for aircraft debris by pulling a “towfish” behind a search vessel. The towfish is lowered so that it “glides” just 100 – 150 m from the sea floor. The towfish is equipped with “side scan sonar” to search on either side of the towfish, and “multibeam sonar” to search below the towfish. This allows scanning the seabed out to a distance of 1 km to either side of the towfish at a resolution of about 70 cm and a rate of 133 sq km per day. If there is an object of interest, or the seabed is difficult to scan due to challenging topography, an autonomous underwater vehicle (referred to as an AUV or drone) can be deployed to get close to the sea floor and obtain high resolution images. For instance, the drones of the type used in the MH370 search have a resolution of about 10 cm and can scan the seabed at a rate of about 17 sq km per day.

Our estimates of the location of the crash site come mainly from two bodies of evidence: satellite data that was recorded for the brief intervals that MH370 transmitted signals to the Inmarsat communications network, and from drift models that estimate the crash site based on the timing and location of debris that has been recovered from the shores of Eastern Africa. Unfortunately, neither of these data sets is sufficiently precise to provide high confidence in the location of the wreckage.

Due to the expense of seabed searching, combined with the imprecision of using the existing data sets to estimate the location of the wreckage, some are suggesting that it is not economical to do further searching with our current technology. The argument is that further searching should be suspended until we gain additional information or insight that allows us to more precisely estimate the location, or until we develop new technology that allows us to more economically search large areas of the sea floor.

It is the promise of new technology that can more economically search large areas of the sea floor that led me to the work of Dr Dan Stilwell, a professor of electrical engineering at Virginia Tech. Dan’s team conducts research in the area of marine autonomy and robotics, and they have developed small, fast, high-performance, inexpensive AUVs for the US Navy. His research team is using their extensive inventory of technology to compete in the Ocean Discovery XPrize, which aims to accelerate innovations to improve the speed, scale, and image resolution of technologies used to explore the ocean floor.

Dan Stilwell (right) and his team with one of their AUVs

The XPrize contest will require mapping 500 sq km of ocean floor with a resolution of 5 m and at a depth of 4,000 m, and also to produce high resolution photographs of various objects on the seabed, all within 24 hours. That’s quite a challenge with existing technology. Nonetheless, the prospect of winning the US$ 7 million prize has attracted interest from 21 teams from around the world.

Dan’s approach is to use a “team” of small, low-cost AUVs to cooperatively survey and scan the ocean floor. Each AUV can travel at 4 knots for 24 hours on a single battery charge. Rather than using expensive inertial guidance systems to navigate, Dan and his team are using technology developed at the Woods Hole Oceanographic Institute, whereby all the AUVs acoustically communicate and navigate using a low-bandwidth, time-division multiple access (TDMA) network. In this approach, each AUV in the team is assigned a time slice and each AUV has a synchronized clock. Each AUV measures the time delay between transmission and receipt of pulses from each of the other AUVs, and from this and other information, the relative position of all the AUVs may be determined. One node remains at the surface, which provides an absolute GPS position reference. This approach to acoustic navigation has not previously achieved the accuracy that is required for the XPrize contest, but Dan’s team will implement a number of new tricks that they expect will provide a sufficient boost in navigational performance.

Can next generation AUV technology provide an economical way to search for MH370? Consider this: Dan estimates that it would take about four of his drones to match the scan rate of a single towfish. But there are compelling economic benefits to using a team of drones. For one, each drone is relatively inexpensive–Dan and his team can build one for about US$ 125,000. Secondly, a large team of AUVs can be deployed from a single surface vessel and crew, while a towfish requires a dedicated vessel. For instance, if a cooperative team of twelve AUVs is deployed from a single vessel, that vessel would be able to scan three times as much sea floor as a vessel deploying a towfish. As sonar sensors increase in performance and miniaturization, sea floor scanning with AUVs will become even faster and cheaper.

As we struggle to squeeze every last bit of information from the existing MH370 evidence, it may be that some of our resources are better directed to improving our ability to quickly and economically search large expanses of the sea floor. Research on autonomous vehicles like that performed at Virginia Tech by Dan and his team can help us.

Update on May 31, 2017.

I was recently in a discussion that included a well-known ocean explorer who happens to be a judge in the Ocean Discovery XPrize competition.  We were having a general discussion about searching for MH370 and ways to scan the ocean floor at high resolution, and he told us about the capabilities of Ocean Infinity. Like the team at Virginia Tech, their approach is to employ a team of AUVs. From their website:

Six HUGIN autonomous underwater vehicles (AUVs) are capable of operating in 6,000 m water depth collecting high resolution data at record breaking speeds. Our AUV fleet is accompanied by six unmanned surface vehicles (USVs) to ensure precise position and constant communication.

With multiple autonomous vehicles working simultaneously utilizing innovative technology, we are able to survey huge swaths of the seabed, quickly and with outstanding accuracy. We can operate in shallow waters but excel in extreme depths, working in dynamic environments ranging from the tropics to the Arctic ice.

Because of the size and complexity of each AUV/USV pair, the capital cost of the technology from Ocean Infinity would greatly exceed the capital cost of Virginia Tech’s technology, which uses small AUVs with innovative navigation systems. On the other hand, both approaches benefit from having a single host vessel supporting multiple underwater vehicles, which offers significant operating cost and scan rate improvements compared to the conventional towfish technology.

Ocean Infinity’s seabed exploration system is commercially available today, including underwater and surface vehicles, on-board support equipment, and the host vessel. This is an exciting possibility for conducting the search for MH370 in the near future.

Drift model for a crash site at 35S latitude. Black lines are paths for debris and arrow heads are positions on December 31, 2015. (Click on image to enlarge.)

A recent report from Australia’s Commonwealth Scientific and Industrial Research Organisation (CSIRO) was greeted with much fanfare. A previous report published in December 2016 predicted that MH370 would be found along the 7^th arc at 35S latitude. With new results in, the ATSB proclaimed that they were even more confident in their findings.

Using advanced computer models of how debris from the crash might drift across the Indian Ocean towards Africa, and comparing those results to the location and timing of debris discovered along the shores of Eastern Africa, it was possible to narrow the location to 25,000 sq km of unsearched sea bed. However, the results from this model, presented in December 2016, did not predict that one of MH370’s flaperon would arrive on the French island of La Reunion by the discovery date of July 2015. This perplexed the researchers at CSIRO, who were committed to better understand the discrepancy.

Researchers knew that because the recovered flaperon floated with a portion above the water, the drift path was more heavily influence by wind and waves than debris that floated flat on the surface. The computer model for the flaperon was therefore adjusted for the extra “leeway” by measuring the drift behavior of replica flaperons that were constructed and tested by CSIRO. However, even when the computer model for the flaperon was corrected for this extra leeway, the drift models still did not predict that the flaperon would arrive in La Reunion by July 2015. That was until the American National Transportation Safety Board (NTSB) was able to locate a spare flaperon, which was modified by Australian investigators to resemble the flaperon debris that was recovered. The drift behavior of the modified flaperon was then experimentally measured just as the behavior of the replicas was previously measured. The experimental results said that the computer model should include even more leeway, and the wind would also tend to push the flaperon about 20 degrees to the left. When these effects were included in the computer model for the flaperon, the drift models now predicted an arrival date in La Reunion that was consistent with the discovery date of July 2015, and this was released in a new report. CSIRO felt as though the last missing piece of the puzzle was found.

As reported recently in the media, these results gave CSIRO even higher confidence that its drift models are correct, and MH370 would be found along the 7^th arc at 35S latitude. Surprisingly, nowhere in those stories was it reported that at this latitude, the seabed was already searched to a distance of about 20 NM from the arc without finding the plane.

Fellow IG member and co-collaborator Richard Godfrey viewed CSIRO’s drift model with skepticism. He had already performed his own drift study using a computer model he had independently developed, which I have previously discussed and published on this blog, and he came to a much different conclusion. He argued that the timing and location of the debris recovered along the shores of Africa were not consistent with a crash at 35S latitude. Rather, a crash at 30S latitude, well north of the seabed search, was much more likely.

Rather than post Richard’s short comments on this blog, I asked him to prepare a more in-depth critique of CSIRO’s work, as certainly his comments would raise many questions. Within 24 hours, a critique arrived in my inbox, which I publish in full here. In addition to the simulated trajectories of debris, Richard also includes the effect of temperature history on barnacle growth, and comments on the effect of storms in the region. He concludes: Despite the significant contribution in refining the accuracy of the drift model, the new data is interpreted as confirming the findings of the ATSB First Principles Review. The pre-conceived idea, that “other evidence” constrains the MH370 End Point to between 32°S and 36°S is a false assumption. A MH370 End Point at 35°S does not fit the fact that the underwater search has already discounted this location to a 97% level of certainty. An MH370 End Point at around 30°S does fit the available data.

While Richard was preparing his critique, I tried to independently reconcile the differing conclusions of CSIRO and Richard. Basically, CSIRO was predicting a crash site at 35S, and Richard maintained that a search at 30S had a much higher probability of success. Fortunately, the results of CSIRO’s drift studies were made available as KMZ files that could be imported into Google Earth. Using these files, as well as the recent report and the report from December 2016, I was able to piece together some information.

The arrival of debris on the shores of Eastern Africa is highly dependent on the latitude of the crash site. In general, crash sites further to the north along the 7th arc will produce debris that arrives earlier in Africa. After traveling west across the Indian Ocean towards Africa, the debris then tends to travel south. Therefore, debris reaching Eastern Africa would beach last on the shores of South Africa.

The figure at the top of the article shows the position of debris on December 31, 2015, as predicted by the CSIRO model for a crash site along the 7^th arc at 35S latitude. The debris is assumed to have “low windage”, which is consistent with the shape of the engine cowling and flap fairing that were discovered in Mozambique and South Africa in December 2015. As can be seen in the figure, CSIRO’s model does not predict that a crash site at 35S latitude would produce debris that would beach as far south along the shores of Eastern Africa as the actual debris that was found. The results of CSIRO’s model are in this respect consistent with the findings of Richard Godfrey.

So why does CSIRO maintain that a crash site at 35S produces debris of the correct location and timing as what was found? The answer lies in a panel from Figure 3.2.1 of the report from December 2016, which is shown below. The vertical axis represents the latitude of potential crash sites along the 7^th arc, and the horizontal axis represents the time of arrival along the shores of Eastern Africa, and the color represents the associated probability, with dark blue the lowest probability and red the highest. The red and white bar shown in the figure is aligned along December 2015, which is when the first debris in Eastern Africa was found. And indeed, the colors in the figure do show that for a crash site of 35S latitude, the debris will start to reach Eastern Africa around December 2015.

Probability of reaching Eastern Africa for various crash site latitudes. (From CSIRO, Dec 2016.)

What is not shown in the figure is the timing of when debris will reach various locations along the shores of Eastern Africa. Instead, as can be seen in the title above the figure, all locations along the shores of Eastern Africa from 35S latitude to the Equator are grouped together. But we know that debris will reach locations further south along the shores of Eastern Africa last. In fact, CSIRO’s own model predicts by the end of December 2015, the “non-flaperon” debris, i.e., debris with low windage that floats relatively flat on the water, will reach the shores of Eastern Africa only between 1S and 12S latitudes. On the other hand, the debris was found in Mozambique at 24S latitude and in South Africa at 34S latitude, which is well outside of the range of latitudes predicted by CSIRO’s model.

CSIRO might argue that although a crash site of 35S doesn’t allow debris to reach South Africa by December 2015, a crash site of 30S, as suggested by Richard Godfrey, would have produced debris along the northern shores of Eastern Africa well before Blaine Gibson found the portion of the horizontal stabilizer (nicknamed “No Step”) in March of 2016.  In fact, debris might have arrived well before Mr Gibson’s discovery, and either was not found, was beached and later was again carried out to sea, was caught in offshore eddies, or was found and not reported. In these cases, a distinction should be made between the date of discovery and the date of arrival. Obviously, the arrival must always precede the discovery.

Based on the results of the drift models of both CSIRO and Richard Godfrey, recent claims about the most likely crash site of MH370 should be carefully reviewed by independent investigators.

Update 1 on April 24, 2017.

For those wishing to explore the drift model results in Google Earth, the KMZ files generated by CSIRO are available for the flaperon debris, for non-flaperon low windage debris, and non-flaperon high windage debris. The particular file I used to create the image at the top of the article is the file for low windage, non-flaperon items starting at 35S latitude. Once dragged into Google Earth, simply move the time slider and observe how the particles travel in time.

Update 2 on April 25, 2017.

The Guardian has published an article that discusses our interpretation of CSIRO’s results. “Both CSIRO and the ATSB have been contacted by Guardian Australia for their response.” No response has yet been received by reporter Elle Hunt.

Update 3 on April 26, 2017.

I received the following email from David Griffin of CSIRO:

Dear Victor,

I saw the Guardian article referring to your blog. A few comments:

1. You are correct that ‘Roy’ was found at an earlier date than the model predicted. But to be fair, the model error is ‘just’ 2 months. I consider Roy’s arrival time – before anything else upstream –  to be something that is simply too hard for any present-day model to convincingly explain. You’ve seen the paths that things take. But those paths should not be interpreted too literally. Our Dec 2016 report mentions that we do not have confidence in the model’s ability to hindcast the arrival times of individual items along the African shore. That’s why we focussed on the more-robust things that the model tells us.
2. As you correctly pointed out, a 30S crash site would, according to our model, have resulted in debris washing up on Madagascan and Tanzanian shores a full year earlier than was observed. That is a discrepancy that is hard to set aside.
3. The other factor against 30S that we find very hard to discount is that 30S is right in the middle of the zone targeted most heavily by the surface search in 2014. This is the “other evidence” that Richard overlooked. Please see Section 4 of our Dec report, and Fig 4.2 of the April report. 

Best regards

David