Rogue waves, those unexpectedly massive ridges of water that can ambush ships and beachgoers, are now more predictable thanks to a new artificial intelligence model. Mechanical engineers Thomas Breunung and Balakumar Balachandran from the University of Maryland in College Park report their findings in the July 18 issue of Scientific Reports.

These waves, which crest more than twice as high as surrounding swells, can form where converging waves amplify a single ridge or where ocean currents compress swells into powerful billows. Despite recognizing certain patterns preceding these surges, researchers had not yet developed an effective forecasting tool (SN: 6/8/15). Such a tool could be lifesaving, given that from 2011 to 2018, rogue waves were responsible for 386 deaths and the sinking of 24 ships.

Using approximately 16 million data points collected at half-hour intervals by a network of 172 ocean buoys, Breunung and Balachandran trained an AI program to identify wave patterns that typically preceded rogue waves. The program successfully predicted 75% of rogue wave occurrences at the buoys one minute in advance. When the prediction window was extended to five minutes, the accuracy was about 70%.

Remarkably, the AI program performed well even at locations where it had not been previously trained. “If you want to predict rogue waves at a new location, all you need to do is put your buoy there and you can use this program without training,” Breunung explains.

Breunung believes that by harnessing more advanced AI architectures and additional data, prediction rates could improve even further. “Maybe we can go to 4 out of 4…. We will see,” he adds.

By Impact Lab