As the Arctic heats up, AI’s forecast is that it will release variable sea ice

From generation to generation, Arctic residents have counted seasonal sea ice, which grows and recedes throughout the year. White bears and marine mammals are based as a place to hunt and rest; The natives fish from ice holes known as pollinators, and use well-known transverse ice routes to travel from one place to another. But Arctic air and water have warmed three times faster than the rest of the planet since 1971, according to a May 2021 report. Arctic Council Report, and this warming is spreading and shrinking the ice in unpredictable ways.

Some scientists and research firms are expanding tools driven by artificial intelligence to now provide more accurate and timely predictions of the ice fragments of the Arctic Ocean and when they will be covered. AI algorithms form existing models that use physics to understand what is happening on the surface of the ocean, in a dynamic area where cold underwater currents combine strong winds to create floating ice rafts. This information is becoming increasingly valuable Members of the Arctic tribe, Commercial fishermen from places like Alaska and global shipping companies interested in taking shortcuts from open water.

Leslie Canavera, CEO of Polarctic, Lorton, a Virginia-based scientific consulting firm that has developed AI-based forecasting models, says the questionable pace of climate change means that sea ice models are becoming more accurate. That’s because they’re based on rapidly changing environmental processes.

“We don’t understand climate change and what’s happening [Arctic] system, ”says Canavera, a member of the Yupi tribe who grew up in Alaska. “We have a statistical model, but then you are looking at the average more. Then you have artificial intelligence, where you can see and learn the trends of the system. ”

Existing physics-based models contain hundreds of years of scientific records about ice conditions, current meteorological conditions, the speed and location of the polar jet stream, cloud cover, and ocean temperature. The models use this data to calculate future ice coverage. But it takes a lot of computing power to reduce numbers, and to make predictions using regular programs for a few hours or days.

Although AI also requires complex data and high initial computing power, once an algorithm is trained in the right amount of data and types, it can detect models more quickly in climate conditions than in physics-based models, according to a Thomas Anderson data. A scientist from the British Antarctic Survey, who developed an AI ice forecast called IceNet. “AI methods can run thousands of times faster, as we found in our IceNet model,” says Anderson. “And they also learn automatically. AI is no smarter. It does not replace physics-based models. I think the future is taking advantage of two sources of information. ”

Anderson and his colleagues published a new model of sea ice forecast in the magazine in August Natural Communications. IceNet uses an AI form called deep learning (It is also used to automate credit card fraud detection, operate self-driving cars, and run personal digital assistants) to train to provide a six-month forecast on each 25-square-mile network across the region, based on Arctic simulations. Climate data from 1850 to 2100 and actual observation data recorded from 1979 to 2011. Once the model was trained and given current meteorological and ocean conditions, IceNet overcame a major physics-based model to make seasonal predictions about the presence or absence of sea ice. in each square of the network, especially during the summer season, when the ice makes its annual retreat, Nature study.