“These waves are at least twice as tall as the surrounding waves. They’re unpredictable. They can come from unexpected directions, often against prevailing winds and swells. You need a whole cocktail of factors to come together wind, swell, current. It’s a chaotic, nonlinear wave interaction, where one wave will suddenly gather energy from others and explode in size.” That is how scientists, referenced in recent news reports, characterise the rare and potentially catastrophic rogue wave a phenomenon once relegated to maritime legend, now in the limelight courtesy of an array of AI-driven buoys off the coast of British Columbia.
In February of 2020, a buoy located close to Tofino, Vancouver Island, detected a rogue wave standing at 17.5 meters, which was more than double the sea state nearby. “For context, that’s about six stories,” said Scott Beatty, CEO of MarineLabs Data Systems, who oversaw the confirmation of this historic event. Background waves were only six meters tall, which made the spike all the more remarkable. The crew put the data through a stringent examination, checking the measurements three times before establishing what is currently the highest rogue wave ever seen in Canadian waters.
Differing from the fairly predictable giants found at sites such as Cortes Bank, these open-ocean giants are the result of sophisticated, nonlinear processes. The physics behind their creation is as chaotic as the waves themselves. Scholars have come to identify wind-wave interaction, nonlinear wave interactions, and modulational instability as major ingredients. When wind power is transferred to the ocean surface, it lays the foundation for the exponential growth of waves, but the actual drama is worked out within the domain of nonlinear physics. The nonlinear Schrödinger equation (NLS) describes how a minor perturbation can build up, via modulational instability, into a rogue wave through a process whereby the envelope of a wave packet develops chaotically, sometimes ending in a calamitous “blowup” of amplitude, as in recent theory (focus nonlinear Schrödinger equation).
Ocean topography and ocean currents make the situation more complex. Currents have the ability to refract and concentrate wave energy, and seamounts and ridges beneath the ocean serve as lenses to concentrate wave energy in hotspots where rogue waves have a high chance of bursting (ocean topography and wave focusing). The non-deterministic dance of these elements gives rise to rogue waves unexpectedly, threatening ships, offshore platforms, and coastal structures.
In an effort to demystify and predict these phenomena, technology has come to the forefront. The use of AI-supported oceanographic buoys is a milestone in real-time detection. These buoys, which are supported by accelerometers and sophisticated signal processing, continually monitor sea surface height and send information back to shore for analysis. Machine learning algorithms specifically long short-term memory (LSTM) neural networks, are employed to process the raw data, sampled at frequencies up to 2.56 Hz, and search through billions of data points to identify the characteristic fingerprints of rogue waves (forecasting freak waves from buoy measurements).
More recent research using the Coastal Data Information Program’s large buoy network has proven that these AI algorithms are able to identify three of every four rogue waves with as much as five minutes’ warning. The neural networks are outfitted with large datasets, learning to identify faint pre-rogue-wave signals against the statistical noise of normal sea states. “About 3000 rogue waves are correctly predicted by the neural network, which is about 77% out of all rogue-wave samples,” wrote authors of a 2024 Scientific Reports paper. That performance continues to hold when extrapolated to buoys far from the training data, indicating some universality of the physics underlying the phenomenon that the AI is picking up.
But it is a long way from being solved. The unpredictability of rogue waves has its basis in the randomness of phases of waves in the ocean and the limitations of existing models, which are rarely phase-resolving. Although the AI method is strong on classification assessing if a rogue wave is on the horizon it does not yet forecast the exact height or timing beyond a short horizon. Additionally, the most extreme and rarest ones are out of reach, with around one in four rogue waves not being detected.
As the array of AI-equipped buoys grows and data are collected, the possibility of operational rogue wave forecasting systems is brought into clearer view. For marine engineers, the stakes are high: enhanced safety procedures, responsive infrastructure planning, and real-time decision aid for offshore operations. For researchers, the data-driven strategy offers new possibilities for exploring the nonlinear ocean’s dynamics. And for big-wave surfers, the prospect of finding unsuspected surf areas concealed in the open sea’s vast, turbulent space is still a seductive, if risky, frontier.
