Algorithm predicts drift direction of people lost at sea

Indonesian rescuers search for the victims and wreckage of an AirAsia plane off Borneo in 2015 Keystone

A mathematical method that can speed up search-and-rescue operations at sea has been developed by scientists at the federal technology institute ETH Zurich and the Massachusetts Institute of Technology (MIT). 

Hundreds of people die at sea every year in vessel and airplane accidents, ETH Zurich noted in a statementExternal link on Tuesday. “Emergency teams have little time to rescue those in the water because the probability of finding a person alive plummets after six hours. Beyond tides and challenging weather conditions, unsteady coastal currents often make search and rescue operations exceedingly difficult.” 

An international research team led by George Haller, professor of nonlinear dynamics at ETH Zurich, has gained new insights into coastal flows. These promise to enhance the search and rescue techniques currently in use. 

Using tools from dynamical systems theory and ocean data, the team has developed an algorithm to predict where objects and people floating in water – potentially also oil spills – will drift. 

With fast-changing coastal waters, predictions today are often inaccurate owing to uncertain parameters and missing data, the statement said. “As a result, a search may be launched in the wrong location, causing a loss of precious time.” 

It’s a TRAP 

Haller’s team obtained mathematical results predicting that objects floating on the ocean’s surface should gather along a few special curves, which they call TRansient Attracting Profiles (TRAPs). 

These curves are invisible to the naked eye but can be extracted and tracked from instantaneous ocean surface current data using recent mathematical methods developed by the ETH Zurich team. 

The Swiss and US scientists tested their TRAP-based search algorithm in two separate ocean experiments off Martha’s Vineyard near the northeastern coast of the United States. Working from the same real-time data available to the Coast Guard, the team successfully identified TRAPs in the region in real time. They found that buoys and mannequins thrown into the water quickly gathered along these evolving curves. 

“Of several competing approaches tested in this project, this was the only algorithm that consistently worked in situ,” Haller said. “Our hope is that this method will become a standard part of the toolkit of coast guards everywhere.”