Finding MH370 is going to be a bitch

The aircraft has gone down in an area which is the undersea equivalent of the eastern slopes of the Rockies, well before anyone mapped them. Add to that a search area of thousands of square kilometres in about an isolated a spot as you can imagine, a search zone interpolated from satellite pings and you can see that it’s going to be tough.

Hey, this sounds like a job for International Rescue Bayesian search theory! Slightly more seriously such searches have a long and honourable tradition in undersea recoveries, having been used by the USN to find two H-bombs that fell into the sea of Palomares in Spain, and the wreck of the USS Scorpion. More recently Bayesian search was used to bring the search for the AF447 flight data recorders to a successful conclusion after four fruitless prior search phases.

A (very) basic description of the the techniques is that you estimate two factors that contribute to the probability of finding a lost item in a given location, the prior probability that the object is in a given location (we’re looking in the right area) and the probability of locating the object given that it is in the location your searching (difficulty of search). You then develop a probability map for each factor over the search area, multiply together and voila! You have yourself a combined (Bayesian) probability map of how likely it is that we’ll find the aircraft in a particular search area. Combining these factors is very important when the terrain over which your searching can vary so markedly, as it does in the MH370 search area.  The other critical aspect of the method is to progressively update the probability estimates as you go, for example if you searched an area and found nothing then you’d reduce the likelihood in that area, but not eliminate it completely, while increasing the likelihood in other areas using Bayes theorem.

Logically the most optimal way to expend scarce resources is then to scan along the path from highest probability search area to lowest. What this cmeans in practice is that search resources may be better placed to search the abyssal plane up track, even though it is less likely to contain the aircraft, because it’s a hell of a lot easier to search there than in the mountain ranges that lie at the end of track. This is is also where you inevitably run into conflict with peoples biases and prejudices about where the ‘best’ place to search is, with the strength of opinion usually being inversely proportional to the available facts in these situations as well.

The two great advantages of the technique are that on the one hand all information available is used coherently and on the other that it is possible to generate a cost estimate for achieving a certain success likelihood before you start searching, which is kind of useful when the search is going to cost tens of millions of dollars. Of course we are dealing with the Australian Government here so I predict, in our island bastion of anti-science, that we’ll get to see the squandering of lots of public dollars before they call in the statisticians. 🙂