SeaHelp News

Predicting freak waves: AI should be able to predict cavemen in the future

Predicting monster waves, also known as freak waves or rogue waves (sailors call them cavemen), was previously considered impossible. This could now change thanks to special field measurements and artificial intelligence

At the beginning of January 1995, a low pressure system with a large pressure gradient forms in southern Sweden. The resulting storm sweeps southwards across the North Sea and gathers speed. In the late morning hours, it reaches hurricane force.

Around 86 nautical miles south-west of the southern tip of Norway is the oil drilling platform Draupner E. It is equipped with several state-of-the-art measuring instruments for determining the wave height. At midday, the average wave height measured at the platform reaches around twelve metres and remains at this level until shortly after 4 pm.

At this point, all the measuring instruments on Draupner E suddenly fail: one wave is much higher than all the others. It reaches 25.6 metres – more than twice the average wave height. Damage to the platform at this height serves as the final proof of the exceptional wave: since then, this monster wave – which has thus been verifiably proven for the first time – has also been known as a so-called Draupner wave.

Monster waves – scientists call them extreme waves – are rare, they are not expected and they are extremely steep. That’s what makes them so dangerous. Sailors call these types of lone giant waves caventsmänner – derived from the Latin word cavere, to take care. In English, the terms freak wave (crazy wave) and rogue wave (rogue wave) are commonly used.

Monster waves are a major hazard to shipping – now it may soon be possible to predict them

The problem: monster waves pose a considerable danger to ships, offshore infrastructure and other maritime equipment, and of course also to sports skippers on long voyages – impressively described, for example, in the Film Der Sturm (Originaltitel: The Perfect Storm), a 2000 disaster film by Wolfgang Petersen, whose plot is based on a true story from 1991.

Reliable monster wave forecasts could minimise this risk for operations at sea. But these have been in short supply until now; reliable monster wave forecasts have simply not been available.

This could change in the future. In a article by Thomas Breunung and Balakumar Balachandran, published on 18 July this year on nature.com, the authors attempt to demonstrate how monster waves could be predicted from field measurements.

An extensive buoy dataset consisting of billions of waves is used to parameterise neural networks, the scientists write in their article. This network is then “trained to distinguish waves before an extreme wave from waves that are not followed by an extreme wave”.

 

AI should be able to predict rogue waves in the future.
© Sandra Geiger | Adobe Stock

 

With the study approach, three out of four monster waves can be predicted one minute in advance

With this approach, three out of four monster waves could be correctly predicted one minute in advance, the authors promise; if the warning time is extended to five minutes, the proportion of accurate predictions is reduced to “seven out of ten monster waves”.

How does the system work? According to the study by Breunung and Balachandran, impending monster waves are – put simply – predicted using buoy data. The publicly available buoy data is searched for monster waves and thousands of 30-minute time windows containing a monster wave are extracted.

Ocean waves prior to the monster wave event are then extracted. These measurements are paired with recordings of the same length without monster waves. A special network is then used to distinguish between the two classes, namely waves that precede a monster wave and waves that are not immediately followed by a monster wave. This network is then used to predict monster waves.

Finally, it is shown that the “performance of the neural network trained on balanced data sets with an equal number of monster wave samples and non-monster wave samples is transferable to a real ocean”, where a “much higher percentage of non-monster wave samples is observed”.

As a result, monster waves could be “largely predictable with a warning time of a few minutes in the future”, the authors claim in their study. And: as a monster wave is only one example of an extreme event, it is also conceivable that the findings from this study could also be used to predict the occurrence of other extreme events, for example “forest fires, seismic activity and possibly even the climate“.

SeaHelp Service
Für tagesaktuelle Kraftstoffpreise
bitte hier klicken!
SeaHelp Service

Push Service & Newsletter

Marina MitanAdvertising

SeaHelp Neueste Artikel

Advertising
SeaHelp

Coronavirus Current

[ulc id="30297" taxs="576" posts_per_page="-1" order_by="title" order="ASC" cols="1" layout_style="minimal"]
SeaHelp News

Related Posts