CASTAWAY HUNTERS
When someone goes missing at sea, search and rescue plans are usually drawn up using data on weather, currents and water conditions to predict their likely trajectory. The problem is that errors can quickly accumulate until the predicted pathway is a long way adrift from what’s actually happening at sea. A new algorithm could improve the chances of locating people by predicting not their trajectory, but where they will end up. The algorithm analyses strength and direction of ocean currents, waves and surface winds, and identifies in real-time the regions of the ocean called TRAPs (TRansient Attracting Profiles), where floating objects are likely to converge.
Dr Mattia Serra, now a Schmidt Science Fellow at Harvard, developed the algorithm during his PhD with Prof George Haller at ETH Zurich. He likens TRAPs to a table on which magnets continually pop up, disappear and move. “Then throw a coin on the table,” he says. “The trajectory of the coin is very chaotic because it’s going to feel the influence of all these magnets.” The table is the ocean surface, the magnets are TRAPs, and the coin is a drifting person.
During tests, the algorithm was found to work well out in the turbulent sea off the Massachusetts coast. The team, led by MIT’s Prof Thomas Peacock, used a snapshot of local conditions to model the ocean’s behaviour and locate where the TRAPs were likely forming. Then they simulated a search and rescue mission, casting buoys and manikins into the sea, each carrying a GPS tracker. As predicted, the objects drifted towards the identified TRAPs.
Serra and his colleagues are now discussing the possibility of the US coastguard using the algorithm in search and rescue operations. The algorithm could also be used to more accurately predict the movements of oil spills.