AI could make sea ice warnings faster, cheaper, and universally available.
Seagoing vessels in the polar area need to keep tabs on sea ice and the usual source is ice warning data from The Norwegian Meteorological Institute and similar centers. Unfortunately, this is a very expensive process that needs supercomputers and satellites. AI could change that, according to Sindre Markus Fritzner, a doctoral research fellow at UiT The Arctic University of Norway. (PHYS.ORG)
Because the calculations are so enormous, the work can be done only at The Norwegian Meteorological Institute.
“If you are on a vessel in The Barents Sea, you are dependent on being connected to a network to download the warnings from The Norwegian Meteorological Institute,” explains Fritzner.
However, AI algos can recognize and predict sea ice patterns – and this could be a game-changer for shipping and cruise lines operating in ice-infested waters.
Using AI on sea ice
Currently, satellite observation data on ice cover, and other information such as ice thickness and snow depth, are fed into dynamic computer models. This is very voluminous data and only supercomputers can process it – a limited and expensive proposition.
“Dynamic models are physical models and require a lot of data to be processed,” says Fritzner. “If you are going to make warnings about future events, you need to use a supercomputer.”
However, through the use of machine learning, computer algorithms can be trained to recognize data and predict how the ice is likely to form, say, over the next week.
Fritzner loaded in the data for one week, and then another set for the following week for a long enough period to train the algo.
“Thus, it is the coherence in the development between these weeks that the machines learn itself, and in this way it can predict how it evolves,” Fritzner says.
However, a fully trained AI algo can run the calculations for the ice warnings on a regular laptop.
A game-changer for ice warnings
Equipped with the relevant program and the AI algos, a vessel can compute the forward-looking ice warnings using onboard computers or laptops
This would be a huge boon for shipping that operates close to the ice zone, as also for cruise traffic.
Fritzner’s models still need refinement as well as more data points. But the research looks promising and he does not doubt its potential.
Related Story: Satellites and AI Could Together Predict Wildfires Accurately