Try to imagine the conditions present at every single point in the world’s oceans, at every single second. On the surface. Just beneath the surface. Far beneath the surface. With a slow-moving current. With a fast-moving current. With an easterly current. With a northerly current. With a cool temperature. With a cooler temperature. With high salinity. With higher salinity. For every single point in the oceans, at every single second.

“It’s a huge, huge data set,” says Thomas Peacock, a professor of mechanical engineering at the Massachusetts Institute of Technology (MIT). “The ocean is so complex, you could release two floats off the side of a boat in the same location, and it’s entirely possible that they’ll end up going in different directions.” This level of complexity, of course, is a key obstacle that search-and-rescue teams work to overcome every time there’s a man-overboard situation. Rescuers do their best to determine where and when, give or take, the boater fell into the water. Search teams then run computer analyses, looking at things like prevailing winds and currents, to create probability scenarios for which direction the man overboard is likely to be headed. That’s how rescuers set up search grids and, hopefully, get the boater back alive.

Now, Peacock and his colleagues have created another tool for those rescuers to add to their search methods. Working with researchers from the Swiss Federal Institute of Technology, the Woods Hole Oceanographic Institution and Virginia Tech, the MIT team used a new algorithm to analyze ocean conditions in a different way. Instead of trying to figure out where a man overboard might be drifting while starting from his last known location on the boat, this algorithm analyzes the region around the boat to determine which parts of the ocean have conditions likely to pull the person toward them. Those spots are nicknamed Traps, which stands for transient attracting profiles.

“Imagine you had a tabletop, and there’s a few magnets moving around on the tabletop,” Peacock says. “If I throw some metal coins onto the tabletop, then where is the best place for me to go looking for those coins? With the magnets traveling around, they’re likely to pull the coins onto them. In the ocean, at any given time, there are locations on the water’s surface that are acting to very strongly pull material onto them. If you have a person in the water who was released in the general area, and you want to know the best places to look, you should go to these Traps. They act like the magnets. They draw everything from the surrounding region.”

The current mathematical technique being used for search-and-rescue operations is called Lagrangian. This new technique is called Eulerian and lets researchers do the mathematical calculations faster, as well as with more common computer hardware,
Peacock says.

“The advantage of the Traps method is that the processing is really quick,” he says. “It can be done on a laptop computer in real time, and it can be updated as soon as a new data set is available.”

Along some parts of the U.S. coast—including popular boating locations such as New England, Southeast Florida and California—that new data is coming in about every 15 minutes from sensor networks that use high-frequency radar, Peacock says. Being able to process that data faster, and with less computer firepower, can let search-and-rescue teams make different use of it.

“With the method that we have, there’s no need to go to a big computer model to find these Traps,” he says. “You can just take the data from the high-frequency radar system and run the Traps processing on it. Every 15 minutes, you can be getting an updated picture of where these Traps are in the ocean, off the coast. You could tell sailors that 15 minutes ago, these were the Traps in your region.”

To prove that the algorithm works, Peacock and other researchers did tests off the coast of Martha’s Vineyard, Massachusetts, in the summer of 2018. (Peer review of their work has delayed its release until now; it was published in late May in the journal
Nature Communications.) They brought manikins with GPS trackers onto a powerboat that was about 40 feet long, and then released the manikins into the sea. They had their mathematical predictions for where the manikins should end up based on where the data analysis told them Traps would be.

They watched the GPS trackers on their computer screens for a few hours as the manikins floated, sometimes a few miles away from the boat. Sure enough, Peacock says, “they all converged onto different regions that turned out to be Traps.”

The work was done with primary funding from the National Science Foundation’s Hazards SEES program, with additional support from the Office of Naval Research and the German National Science Foundation.

Peacock says the U.S. Coast Guard has also had eyes on the project from the start; now that the research is peer reviewed and the study is public, the Coast Guard can begin its process of deciding whether to put the tool into operational use—not just for man-overboard situations, but also for tracking things like oil spills, ocean plastics and more.

“Depending on what assets they have to go out and do the search-and-rescue, this can help with the decision-making,” Peacock says. “They may run the modeling and find a very broad search area. If they know there are a few traps in that area, they might prioritize looking there. It doesn’t replace what’s already being done, but it’s an extra thing for the Coast Guard.” 

This article originally appeared in the August 2020 issue.