In which satellite models trump circumspection: the case of MH370

The strange disappearance of Malaysian Airlines flight MH340 has captivated the world, myself included. In an era of instant information, it’s sobering that an entire Boeing 777 could just vanish. I am sure I am not the only one who will be uneasy the next time I board a plane for a routine journey – especially if its itinerary lies over water. Such a tragedy could literally happen to any of us, at any time.

In the course of the last few days, the main protagonists in the search effort have been the multinational ad hoc fleet of aircraft and ships scouring the surface of the vast oceans for telltale flotsam. We have been gripped by grainy images of white, green and orange objects as the search parameters narrowed inward towards the southern corridor of the Indian Ocean some 1,500 miles southwest of Perth, Australia.

But yesterday, UK science took center stage, when the Malaysian authorities announced that the plane was shown to have crashed into the sea leaving no survivors, with the airline adding that it was beyond any reasonable doubt.

I assumed, on seeing this headline, that a bit of flotsam had finally been positively ID’d. But no – it turned out that the authorities had pinned their certainties on a cutting-edge model based on the interpolation of satellite imagery by Inmarsat and information from the UK’s Air Accidents Investigation Branch. Not just data, but a computer model – and one that apparently has never been used in this way before.

At first, I was baffled that everyone seemed to accept this verdict – even the relatives, one of whom was quoted as saying, “We accept the news of the tragedy. It is fate.” Such is the power of being blinded by science – and perhaps, by the gravitas of a high-profile press conference. This morning, however, there’s been a backlash that I find a bit more understandable, and the relatives are again engaging in protests. How rigorous is this model, and on what is it based? Is it truly solid enough to draw such a firm conclusion about something that is so very important to the victims’ relatives? From my own work with computer models, I know that they can give good indications but are seldom “beyond reasonable doubt”. With the stakes as high as they are, and the tempers as frayed, I am surprised that the authorities had so much faith in science that they felt able to draw such definitive conclusions.

About Jennifer Rohn

Scientist, novelist, rock chick
This entry was posted in Scientific method, Scientific thinking. Bookmark the permalink.

10 Responses to In which satellite models trump circumspection: the case of MH370

  1. cromercrox says:

    The plane was abducted by aliens. It’s the most parsimonious explanation.

  2. Seelye Martin says:

    I think that the methods used to track the aircraft are valid, and I assume that these methods can be replicated. It’s a beautiful job, pulling this information out of a precessing geosynchronous satellite, the time delay between broadcast of the pings and their reception, and the Doppler shift of the ping frequency associated with the satellite and aircraft motion. In contrast, the darkness in the human soul that led to this aircraft ending up south of Australia may never be understood.

  3. Mark Field says:

    Well in this case I think if it is not beyond reasonable doubt, it is getting highly unlikely that other explanation will fit the data. Yes it is a mathematical model, and it predicts one piece of data – that the plane is most probably flying on the southern arc suggested from the earlier satellite data. If you believe that analysis, then there is nowhere for the plane to land and the conclusion is very depressing.

    My understanding from the Inmarsat webpage is there is a ground station trying to communicate with the aircraft via an Inmarsat satellite, and when it doesn’t hear anything for an hour it sends a ‘log on/log off’ signal – a ping – and the aircraft responds with a short message indicating it is still logged on. This whole handshake process occurred six times after the ACARS system (the main operational comms system) became non-operational for whatever reason.

    You know the time of transmission and the time of receipt of the two signals in a handshake which gives you a distance from the satellite. Mathematically this is a sphere centered on the satellite, which intersects the earths surface in approximately a large circle. There were six completed handshakes and the distance of the plane from the satellite was getting larger.

    This is (I think) the origin of the two large arcs on which the plane was believed to be flying – if you plot the six location circles the two arcs north and south are the most probable flight paths of the plane given the available data (someone please correct me if I’m wrong here, this is my interpretation)

    The later analysis used another piece of data, the burst frequency offset, which is the difference in frequency the ground station is expecting and the one received. This allows the direction and speed of the planes travel to be estimated from the Doppler shift. What I believe Inmarsat did was to look at the burst frequency offset from six other Boeing 777 aircraft flying in the area on the day the plane went missing and looked to see which aircraft directions best fitted the signal from the missing plane over time. This suggests the aircraft was travelling on the southern arc.

    The final piece of data is that there is a confirmed handshake at 0:11 UTC and no handshake at 1:15 UTC (and some indications of a partial handshake at 0:19 UTC). Given the known ground speed of the aircraft you can then plot assumed positions. The time window where something happens is consistent with the plane running out of fuel. It is also the reason why the search has now become concentrated at the southern end of the arc, approximately 1500 miles from Perth.

    So it is not certain, the frequency analysis could be flawed, we are also assuming Inmarsat knows for certain this data is from the plane in question (and Inmarsat have earlier handshakes when the ACARS system was on, so this seems reasonably certain) and the time data is correct. However, this is the simplest explanation of the flight path that fits. We also know that no unidentified aircraft entered into the airspaces of India, Russia or China or the US bases in Afghanistan, all of which have good radar coverage.

  4. Thanks for the very interesting explanations. I am not doubting the science per se – its beauty or its likelihood – and personally I don’t believe that the plane is anywhere else but “full fathoms five” in the general area indicated. My point was if you read the quotes issued by the Malaysian authorities, it states the outcome as certain. They even go as far to state that there are no survivors. I don’t see how analysis of pings and handshakes can tell us whether a water landing was attempted nor whether life rafts were deployed. It’s the sort of statement that would rile up any peer reviewer.

  5. Mark Field says:

    I agree entirely, and I apologize for going off on a tangent from your original blog post. The analysis just suggests a place to look, nothing further. We may never know what actually happened, and I really hope I’m wrong about that.

  6. No apologies necessary – I love hearing about the science.

    But wouldn’t it be great if people in general, and politicians in particular, took other modeling (e.g. climate change) this seriously?

  7. Laurence Cox says:

    I just want to add one comment about landing on the sea. While theoretically possible, there are two factors that make it much more difficult in this particular case. First, the location being in the southern ocean means that the waves have a long distance to build up in amplitude from their interaction with winds. This means that the sea-state is likely to be anything other than calm. Secondly, the aeroplane is assumed to have run out of fuel. Although the aircraft is described as “fly-by-wire”, this means that the control signals to the hydraulic actuators are electrical rather than mechanical. The control surfaces are moved by the hydraulic actuators and a complete loss of engine power will a short time later result in a complete loss of hydraulic pressure. At this point, even a fully conscious pilot would have no way of controlling the aircraft’s attitude.

  8. I have no doubt it’s difficult. Whether it’s impossible is another question.

    But yes, all important points.

  9. aeon says:

    Jennifer, I fully agree on your notion, except for one: it would not ‘rile up any peer reviewer’. What I read in published papers is often terribly oversold. Not speaking of the methodological questions reviewers fail to spot. However, I also lost years (literally) on peer reviewers who wanted me to use different, more powerful methods. Sorry to say, but scientific bandwagoning is terribly common. Current topics and methods are asked for, even if your data clearly is not suited for the analysis in question, and has clearly only remotely connections to the latest funding fad.

    But I don’t want to derail the discussion. They have to show the data, and specify the type of model in detail – afterwards, we can talk again about certainty.

    And, btw, I think, parsimony is out of the question if people are asking what happened to their relatives. Unequivocal physical evidence, not p values, are needed.

  10. Thanks for your comment, aeon. It’s possible modeling papers attract a different sort of referee attitude. But if any of my (cell biology) papers stated that my experiments “prove absolutely/beyond reasonable doubt” that such-and-such is true, I’d get reamed. In my field, we are all about “strongly suggest”/”support a model whereby” and so forth.

    “Parsimony is out of the question if people are asking what happened to their relatives. Unequivocal physical evidence, not p values, are needed.” You said this a hundred times better than I could have.