According to the US National Oceanic and Atmospheric Administration (NOAA), rogue waves are twice as large as nearby waves and defy normal sea conditions.
Unusually huge waves in comparison to preceding and subsequent waves are known as rogue waves, and they can endanger human life, ships, and offshore and coastal infrastructure. Regretfully, up until today, there has been no way to predict rogue waves.
The University of Maryland mathematicians Thomas Breunung and Balakumar Balachandran developed an artificial intelligence computer to identify wave patterns that preceded rogue waves, up to five minutes in advance, using billions of data points gathered by a network of 172 ocean buoys.
Deadly Unforeseen
Sea state, as used in oceanography, describes the state of a vast body of water’s surface at a specific place and time. Wave height is the basis for the World Meteorological Organization’s (WMO) sea condition classification, which ranges from 0 (no waves) to 9 (waves more than 14 m).
The US National Oceanic and Atmospheric Administration (NOAA) states that rogue waves are twice the size of neighboring waves and contradict the normal sea condition. These waves frequently arise when swells, which are caused by far-off weather systems rather than local winds, combine to create a single, larger wave. They could potentially originate from swells that are compressed by ocean currents into powerful billows.
Although rogue waves are not completely unforeseen, as scientists have long understood, there have been fatal repercussions due to the lack of a real-time forecasting technique. At least 386 people were murdered and 24 ships sunk by rogue waves in the World Ocean between 2011 and 2018, according to oceanographer E G Didenkulova (“Catalogue of rogue waves occurred in the World Ocean from 2011 to 2018 reported by mass media sources”, 2020).
The promise of artificial intelligence
Breunung and Balachandran took into account 20-minute samples that were captured by buoys in the ocean. When rogue waves (as defined by NOAA) are present in a sample, the recording stops at that moment. To train the AI algorithm, the rogue wave samples were then compared to all other samples (where rogue waves did not occur).
From then on, the program was able to forecast approximately 75% of rogue waves one minute ahead of time. About 73% of rogue waves may be anticipated five minutes ahead of time. Most importantly, the researchers showed that the tool could be used to forecast the appearance of rogue waves close to two buoys at very different depths that weren’t included in the data sets that the AI was trained on, indicating that the program might have predictive powers that could be applied to a variety of situations.
Physical parameters like ocean depths, wind speeds, and buoy placements might be added to the AI’s forecasts to increase its accuracy and forecast lead time, according to the researchers. Breunung suggested that future near-perfect forecasts might result from utilizing more data and stronger AI frameworks.
