MH370 - drift analysis

David Griffin

28 February 2016 update: 'No step' item found on a sandbank off Mozambique

An item which appears to be a wing fragment, with NO STEP printed on it, has been found by Blaine Gibson on a sandbank in the Mozambique Channel. The item has been handed in to authorities for investigation. If it can be identified as being from MH370, this brings to two the number of items potentially usable for oceanographic back-tracking.

Unlike the flaperon found on La Reunion, however, this item is not heavily encrusted with sea life, so it has probably spent a significant length of time either weathering in the sun and/or washing back and forth in the sand at this or some other location. The time at sea is therefore possibly much less than the 716 days that have elapsed since 14 March 2014, and the path taken may have been two or more distinct segments.

The tracks of Global Drifter Program drifters arriving in the Mozambique Channel (during 1985-2015) were overwhelmingly from the east, as shown at right. Two were close to the ATSB search area 700d prior to being in the Mozambique Channel, suggesting that the 'no step' item could very well have followed a similar path. But other drifters arrived near Mozambique from very different points along the 7th arc, showing that those, too, could also be regarded as possible origins of the item.

As with the flaperon, therefore, we conclude that while the location of this finding does not cast doubt on the ATSB's choice of search area (based on the Inmarsat handshakes), it can not provide particularly strong support for it either, because the trajectories of drifting items are so chaotic.

Trajectories of Global Drifter Program drifters

2 October 2015 update: Composite trajectories

Few drifters complete journeys long enough to be directly relevant to the task at hand, especially since we are only interested in the subset of the drifters' trajectories that are before or after specific places, and with or without the drogue attached. We have therefore elected to create composite trajectories by adding shorter trajectories together head-to-tail. There are gaps in time and space involved, but we think that the benefits of the vastly greater number of composite trajectories more than compensates for the damage done by these gaps.

Our method is simple: at the end of a partial trajectory, we do a search for all trajectories that meet distance-away, time-of-year and drogue on/off criteria. At the ends of all those trajectories, another set of searches is done. Drifter data density is quite uneven, potentially biasing the results, so we under- or over-sample the available trajectories appropriately (by time-shifting in the latter case). We have limited the number of partial trajectories to three or four for the results shown here, so they are either 538/3 or 538/4 days long. This results in a large number of trajectories. Using the trajectories back from La Reunion, we can estimate the possible location of the flaperon 508 days earlier (the time between 8 March 2014 and 29 July 2015) by mapping the density of points 508days (+/- 30days, center-weighted) back along all the trajectories.

One way to rationalise this approach is to consider that the best way to estimate where a drifter might go, having come to a certain point, is to look at the paths of other drifters that have passed near that point. This is, after all, what we were doing by looking at single-drifter trajectories, anyway. So now, as well as single-drifter trajectories, we have very many more three- or four-drifter trajectories. We have done this for a selection of La Reunion-backward and drogued and un-drogued trajectories [list].

The image above shows our estimate of the likelihood function of where the flaperon was on 8 March 2014. The 7th arc is added for reference, with the 'high priority' search zone shown in bold. The sea-floor search zone is clearly within the zone of likely origins. The southern end of the search zone is close to the southern flank of the likelihood function, while the northern end of the search zone is towards the eastern flank, but at slightly higher likelihood than the southern half. But why is there a broad maximum NW of the arc, e.g. near 35S, 75E - 30S, 90E? Because this is where drifting buoys are evidently concentrated, moving at lower average velocity and therefore spending more time. We have verified that this is not an artefact of where drifters are deployed or lose their drogues. There is a (leaky) mechanism that concentrates them here in the centre of the gyre. Our back-tracking inevitably concludes that the most likely prior location of the flaperon was where it probably spent a lot of time, regardless of exactly where it entered the water nearby. This assumes that the flaperon drifted exactly like an undrogued drifter, which is probably not the case, being a very different shape. But it probably drifted much more like an undrogued drifter than a drogued one (in which case we would conclude that it entered the water much farther north).

The end points of forward-tracks originating near the northern half of the high-priority seafloor search zone are distributed fairly evenly around La Reunion. A similar result is obtained for the southern half, reinforcing our earlier conclusion that if the plane had crashed where it is thought to have, that Reunion would be one of the most likely places where debris would be found.

8 September 2015.


Analysis of the trajectories of satellite-tracked drifting buoys deployed in the Indian Ocean over the last 30 years confirms our earlier conclusion based on computer modelling that the MH370 flaperon found on La Reunion in July 2015 is consistent with MH370 having crashed near the 39°S-32°S segment of the 7th arc on 8 March 2014. With just one piece of MH370 found, however, the buoy data, like the computer modelling, can not significantly refine the ATSB's sea-floor search area - it just increases our confidence that the flight path analysis underpinning the choice of sea-floor search area is not wrong.


Our 5 August news item described how the 29 July 2015 discovery of a flaperon (now confirmed as being from MH370) on La Reunion island did not cast doubt on the MH370 sea-floor search area chosen by the Australian Transport Safety Bureau based on analysis of the satellite ping data (see the ATSB Fact Sheet). This conclusion was based on ocean modelling. Here, we discuss additional information that lends further support to our 5 August conclusion.

The Global Drifter Program is a highly-valued legacy of the World Ocean Circulation Experiment. The satellite-tracked drifting buoys ('drifters') measure ocean surface temperature for calibration of satellites, and atmospheric pressure for improving the accuracy of weather forecasts. The trajectories of the drifters tells us about ocean surface currents.

The term 'surface current' however, needs discussion before we connect it with the drift of the flaperon.

The flaperon is buoyant, flat, and only a few meters long. Depending on exactly how it floated, it would have drifted at the velocity of the water averaged over the top 0.5-2m. This is certainly close enough to the ocean surface for the effect of waves to be important, so the Stokes Drift would have contributed to its drift. If the flaperon had any non-negligible freeboard (projection into the air) then it would also catch the wind and 'sail' slowly through the water at some small fraction of the wind speed. This down-wind velocity is often referred to as 'leeway' and expressed as a percentage of the wind speed. We do not have an estimate of the leeway factor of the flaperon, so our modelling work was done for a range of plausible leeway factors. As an aside, we note that winds and waves are usually strongly correlated so empirical leeway factors will include the Stokes Drift if the influence of waves is not deliberately isolated.

Global Drifter Program drifters are fitted with 10m-long sea-anchors (or 'drogues') centred at about 15m depth so that, by design, they do not drift downwind like buoyant items such as the flaperon. There are several good reasons for this design, both scientific and practical. Fortunately for our present purpose, however, the drifters lose their sea-anchors after some time (Lumpkin et al. 2013), making them much more relevant to the question we now face with the flaperon than they were designed to be. Their leeway is possibly slightly higher than the flaperon's, but we think that the error associated with this is smaller than the errors of any other source of information available. We refer below to these as undrogued drifters. Trajectories of the drogued drifters are also shown to highlight the importance of the difference, and for comparison with models that have not included the effects of winds and waves, such as the recent GEOMAR study.


We have extracted from the 1985-2015 database the trajectories of all undrogued drifters relevant to MH370, i.e. drifters that were near the present MH370 sea floor search area (the high-probability segment of the Inmarsat 7th arc between 39°S and 32°S) in February-April, or near La Reunion in April-July, in any of the last 30 years.

Forward drift from 39°S-32°S near 7th arc
Seventy four undrogued drifters (right, click to expand) passed through an area surrounding the present sea-floor search area for MH370 and reported data for at least 200 days after that. A substantial fraction of them went near La Reunion within 500 days. These are drifters that transited the blue rectangle within 6 weeks of 8 March. Relaxing that criterion, we see that trajectories at other times of the year were not very different. One drifter actually beached on La Reunion.

Water temperature and barnacles
The drifting buoys record the surface temperate of the ocean. These data, along with the general northward then westward drift post-crash, show that the flaperon probably entered water warmer than 18°C within a month or two of the crash, so barnacle nauplii may have started settling and growing on the flaperon for most of the voyage.

Forward drift from other segments of the 7th arc
We have also looked at the trajectories of drifters passing though segments of the 7th arc either south of 39°S or north of 32°S. Drifters passing through the southern region have an increased tendency to go east and a lower chance of going to La Reunion. Conversely, drifters passing through the northern region mostly went west across the Indian Ocean but passed north of La Reunion, several beaching on Madagascar or Africa in less than 300d. The flaperon finding is therefore too late as well as too far south to be consistent with a crash site north of the present sea-floor search region.

Comparison with model results
The undrogued drifter trajectories differ from our model trajectories in that the real drifters passed by La Reunion principally north of La Reunion, while our model drifters were principally south of La Reunion. This is precisely the sort of model error we had in mind when we said taking model errors into account in our 5 August news item.

Trajectories leading to La Reunion
Turning the question around, the origins of undrogued drifters that passed close to La Reunion any time in 1985-2015 were principally in the southern Indian Ocean, including regions near the supposed MH370 crash site, as shown at right. Selecting only those that arrived near La Reunion in April-July does not greatly alter the picture. What does change the picture, as mentioned above, is selecting for tracks of drogued drifters that (ultimately, possibly with the drogue off, like this one) go near La Reunion. These drifters are more likely to have originated in the tropical Indian Ocean. As discussed above, however, the flaperon floated close to the ocean surface, so the tracks of drogued drifters are much less directly relevant than those of the undrogued drifters.


Taking the modelling and drifter observations together, we stand by our earlier conclusions: the finding of the flaperon is not a reason to doubt the present choice of sea-floor search area. And with only one piece of MH370 found, the presence of ocean eddies makes it essentially impossible to refine the sea-floor search area with any confidence. The flaperon finding does, however, support the flight-path analysis conclusion that the 39°S-32°S segment of the 7th arc is indeed the highest-priority search region for MH370.

Graphics Archive

Global Drifter Program

  1. Southern Indian Ocean trajectories
  2. Indian Ocean - Tasmania trajectories
  3. Individual trajectories of drifters that were near the 39°S-32°S segment of the 7th arc in Feb-April of 1985-2015. The track is blue if the drogue is attached, green after the Global Drifter Program deem it to have detached, at which point its leeward drift and exposure to Stokes drift becomes more like that of a flaperon in the water.
  4. To see all the undrogued drifters on-screen in Google Earth at the same time, we've time-shifted them all so it looks like any drifter that entered the 39°S-32°S rectangle at any Feb-May time in 1985-2015 did so on 8 March 2014. Operate the time slider to see the dispersal, as a function of elapsed time, of the drifters (bearing in mind that they were actually in the water in different years): [kmz]
  5. As above, but relaxing the Feb-May selection criterion: [kmz]
  6. Individual trajectories of drifters that were near the 39°S-32°S segment of the 7th arc at any time in 1985-2015.
  7. Individual trajectories of drifters that were near La Reunion Island in April-July of 1985-2015. [kmz]
  8. At right: tracks of some 1976 drifting buoys deployed by Dr George Cresswell of CSIRO. It is unclear how long the parachute drogues remained attached. One went from near the Abrolhos Islands (Jan 1976) to Tanzania (Dec 1976).

Model simulations using splashpoints on the 39°S to 32°S segment of the '7th arc'

Data sources: currents and winds for 2014-2015 from the Bureau of Meteorology, Stokes drift for 2010 from NOAA WaveWatch 3.
  1. Leeway 0, 1, 2.8%: [Indian Ocean] [2014-15 fli] [2014-15 mp4] [2014-15 ogg] [2014-15 avi]
  2. Leeway 0, 1, 2%: [Indian Ocean] [2014-15 fli] [2014-15 mp4] [2014-15 ogg] [2014-15 avi] [Indian Ocean - backtrack from Reunion]
  3. Leeway 1.2, 1.5, 1.8%: [Indian Ocean] [2014-15 fli] [2014-15 mp4] [2014-15 ogg] [2014-15 avi] [Indian Ocean - backtrack from Reunion] [2014-15 fli] [2014-15 mp4]
  4. Leeway 1.5: [forward] [kmz] [New York Times render]
    [backward] [kmz]

© Copyright 2015, CSIRO Australia
Use of this web site and information available from it is subject to our
Legal Notice and Disclaimer