The most extensive search in aviation history has failed to find Malaysian Airlines Flight MH370 after it mysteriously diverted course on March 8, 2014. Now an engineer at a top Danish university has claimed we've been looking in the wrong place.
He has come up with a possible new flight plan that conforms with the data we have. And it looks nothing like those that guided search efforts before.
It puts MH370's probable crash site off Christmas Island, south of Jakarta. And he speculates the missing aircraft's new possible track means a hijacker could have 'bailed out'.
When it comes to finding MH370, There's never been much to go on.
Shortly after the Boeing 777 took off, its radar transponders and communications systems were all shut down. It was a sinister, deliberate act. It meant that, once out of ground-based radar range, nobody could 'see' where the aircraft was.
But whoever tried to make the aircraft invisible overlooked one thing: engine monitoring sensors that would automatically report to overhead satellites.
Searchers used these signals, along with inferred flight speeds and courses, to mark-out a broad patch of the Southern Ocean far to the southwest of Perth as the most likely place MH370 went down.
But, despite extensive surface and subsurface searches, nothing has been found.
Now, Professor Martin Kristensen, an engineer at Aarhaus University in Denmark, has published a new mathematical analysis of MH370's radar and satellite data.
We know that instead of heading north to Vietnam, flight MH370 deviated to the east. It was tracked by military radar until it passed out of range into the Andaman Sea, apparently headed towards India.
From this point, only the 'handshake' calls to the Inmarsat 3F1 satellite sitting above the Indian Ocean reveal the Boeing was still in the air.
The fractional time difference between when the 'handshake' was sent, and when it was received by Inmarsat, offered investigators a clue. That time gap revealed how far the aircraft was from the satellite at the time it was made.
But little more.
So each of the hourly calls allowed a planet-spanning circle to be plotted, with the Inmarsat at its centre. MH370 was somewhere on those circles at the time each signal was received.
The vast bulk of the circle could be eliminated through simple math: how far along each arc could the aircraft — flying at its maximum speed and within the range of the fuel it carried — have been capable of reaching in the time since its last confirmed position?
Problem is, each of the seven recorded 'handshakes' produced a broader and broader range of possibilities.
However, most of these could also be eliminated.
Those arcs extended over land and populated areas — also covered by civilian and commercial radars. MH370 hadn't been detected along any of these paths.
Which is why investigators plotted possible tracks that took the Boeing 777 out to sea.
But the 'handshakes' held one more clue.
It's called 'Doppler shift'.
It's the minuscule 'stretching' of a signal as its source moves towards or away from the receiver.
The Doppler shift reveals the relative velocity of MH370 and Inmarsat at the time each 'handshake' was made. And the satellite was in an orbit that kept it almost stationary over the Indian Ocean.
Once again, it's not a smoking gun.
There's a broad range of relative velocities that can be produced by an aircraft flying at different angles, at different speeds.
But it does offer another set of boundaries within which MH370 must have followed.
It all comes down to probabilities.
And that's where Kristensen comes in.
He built a mathematical model that takes into account all of these limiting elements.
His argument is that MH370 can only have gone down where the boundaries of speed, fuel, 'handshake' signals and doppler shift all 'overlap'.
"We find four independent solutions with the final part of the flight following a great circle," his research paper declares.
The Doppler shift reveals the relative velocity of MH370 and Inmarsat at the time each 'handshake' was made. And the satellite was in an orbit that kept it almost stationary over the Indian Ocean.
Once again, it's not a smoking gun.
There's a broad range of relative velocities that can be produced by an aircraft flying at different angles, at different speeds.
But it does offer another set of boundaries within which MH370 must have followed.
It all comes down to probabilities.
And that's where Kristensen comes in.
He built a mathematical model that takes into account all of these limiting elements.
His argument is that MH370 can only have gone down where the boundaries of speed, fuel, 'handshake' signals and doppler shift all 'overlap'.
"We find four independent solutions with the final part of the flight following a great circle," his research paper declares.
The final patch of the Earth's surface which lines up with all of MH370's flight parameters sits in the Indian Ocean to the southwest of Australia's Christmas Island.
It seems counterintuitive.
To get there, MH370 must have done an abrupt U-turn in the Bay of Bengal, and then flown along the south coast of Indonesia.
"We propose instead a new, focused search zone of 3500 km2 centred at (13.279˚ South, 106.964˚ East) with slightly elliptical shape along the 7th arc and a total length of 140km and width of 30km," Kristensen's report reads.
"The probability of finding the plane there is above 90 per cent".
Departing from his math, Kristensen points to the unexplained nature of the barnacles found on MH370s debris as conforming with a crash in the nutrient-poor but warm waters off Indonesia, before entering the colder West Australian Current.
"However the line of circumstantial evidence does not stop here where some readers may already be convinced," his report adds.
Kristensen refers to an eyewitness account from a fishing boat to the north of Bandar Aceh, at Indonesia's extreme western end, that night. It reported a westward-flying aircraft diving to a lower altitude and making a turn southward. This account was just one among many such sightings — including from the Maldives far to India's south.
But this one fits with Kristensen's plots.
After this, though, he indulges in some wild speculation.
Kristensen argues that the course and behaviour of MH370 points towards a carefully planned attempt to deliberately conceal the aircraft's course.
"There is also something special about the Christmas Island route going through the intertropical convergence zone where satellite detection and long-range radar are hampered by tropical thunderstorms, indicating intelligent planning," he writes.
"Most likely the perpetrator(s) also knew about the handshakes and deliberately directed and timed the flight to get close to the worst possible mathematical data-entanglement with satellite movement through spatial correlation, making it almost impossible to find the plane because this allows for a multitude of solutions with similar fitting quality."
Why take such drastic action?
"The only plausible explanations are that they wanted to land in Banda Aceh or abort the flight by parachute," Kristensen writes.
"Since the aeroplane did not land, the only option is parachuting.
"In order to do this they had to fly low and slow … to open a hatch and get out.
"They programmed a return to normal flying-height into the autopilot before jumping.
"Therefore the plane returned to 11km height after Bandar Aceh without a pressurised cabin (due to the leak through the open hatch) causing death for everybody on board who might still have been alive."
Residential areas in both countries have suffered from deadly strikes in the conflict.