Machine learning will produce fast and reliable maps of Arctic sea ice
09-12-2019

Machine learning will produce fast and reliable maps of Arctic sea ice

Machine learning will produce fast and reliable maps of Arctic sea ice. A new report from the European Space Agency (ESA) describes how experts will soon be using artificial intelligence to chart sea-ice conditions. The technology will provide fast and accurate sea-ice information to improve the efficiency and safety of marine operations in the Arctic.

As the Arctic warms at a rate that is double the global average, the ongoing changes in the region’s sea ice are extreme and often unpredictable. It is crucial that maps of sea-ice conditions and forecasts are accurate for safe navigation and improved planning, particularly in the Arctic.

For years, experts have charted sea-ice conditions manually by using satellite data, but the large expanse of the Arctic makes this a tedious and time-consuming job. Scientists have identified an urgent need for automated ice observations that can be integrated into ice forecast models.

In an effort to keep Arctic sea-ice data current and relevant, the Danish Meteorological Institute (DMI) and Technical University of Denmark have initiated the Automated Sea Ice Products (ASIP) project.

Funded by Innovation Fund Denmark, the goal of the study is to create an automatic sea-ice service that can provide detailed information in a more timely manner. 

ASIP will combine Copernicus Sentinel-1 imagery with other satellite data to resolve any uncertainties that may emerge in SAR imagery under various ocean conditions,  such as strong winds. 

According to the ESA report, ASIP uses a “convolutional neural network system” that has been trained with large datasets of ice charts to automatically generate ice maps.

“ASIP will be a great opportunity for users to have an up-to-date map of sea-ice products. We are currently working hard to get this in production and validate it with both the ice experts and the users,” said David Malmgren-Hansen, who is a machine learning expert at DTU Compute.

The research was presented by Malmgren-Hansen at this year’s Φ-week event, which was focused on Earth observation and FutureEO. 

By Chrissy Sexton, Earth.com Staff Writer

Image Credit: ESA

News coming your way
The biggest news about our planet delivered to you each day
Subscribe