The Arctic has warmed much faster than the rest of the world in recent decades and Arctic sea ice is disappearing. So far, however, it has only been possible to reliably measure sea ice thickness using satellites in winter, but not during the important melting period in summer. Scientists have now developed a method that, for the first time, can also be used to reliably interpret satellite data on sea ice thickness during the summer. This is important for ice forecasting and shipping in Arctic waters, but also for climate research and even for weather forecasts in our latitudes.
Arctic sea ice plays a crucial role not only for the North Pole region and the Arctic Ocean, but far beyond. The extent and thickness of these floating sheets of ice affect how much sunlight is reflected back into space and, in turn, how much the Earth warms. However, the thickness of the sea ice also determines how much light still reaches the ocean below, and thus influences the activity of marine algae and thus the entire food chain of the Arctic Ocean. Last but not least, the thickness of the sea ice also determines whether it can support polar bears and humans, for example, and whether icebreakers can break through it. In recent decades, however, the behavior of the Arctic sea ice has changed increasingly: Because the Arctic is warming disproportionately, more ice melts in the summer months than before and fewer and fewer areas of sea ice are thick enough to defy this summer thaw period. As a result, the amount of perennial, permanent sea ice is also decreasing.
AI helps with summer measurements problem
Radar satellites have been used since the 1980s to measure the extent and thickness of the ice in the Arctic. Since 2010, this has primarily been done by the CryoSat-2 satellite from the European Space Agency ESA, which scans and x-rays the Arctic sea ice all year round. Using the reflections of the radar waves, scientists can identify the boundary between ice and water and thus determine the ice thickness - theoretically. In practice, however, this only works reliably in the Arctic winter, from October to April, because then the ice and snow are cold and dry. "In the summer months, the satellites are blinded by pools of snow and meltwater that collect on the sea ice surface," explains lead author Jack Landy from the University of Tromsø. It is then no longer clear from the radar data where the water of the melt pond ends and the ice begins. Ironically, at the time when the sea ice is thawing the most and the greatest changes occur, the radar satellite has so far been largely blind to the thickness of the ice.
In order to change that, Landy and his team have now been looking for a way to make the satellite radar data collected in the summer usable. To do this, they used artificial intelligence to analyze the data, which had previously been difficult to evaluate, more precisely. For training, the adaptive system received radar data from Cryosat-2, in which it was known and identified which wave patterns originate from ice surfaces and which originate from free water or snow containing water. Using this data, the AI system learned to interpret and distinguish between the specific wave patterns that the human eye can hardly distinguish. With the help of the algorithm trained in this way and supported by numerical simulations, Landy and his team were able for the first time to determine the Arctic ice thickness with sufficient accuracy even during the summer months.
"Completely new insights"
The result is a dataset showing for the first time the thickness of sea ice across the Arctic and year-round. Because data from the period from 2011 to 2020 has now been evaluated, ice researchers can now estimate more precisely how strong the seasonal and interannual fluctuations are. "The year-round and Arctic-wide data give us completely new insights into the interactions between the atmosphere and the ocean," says co-author Thomas Krumpen from the Alfred Wegener Institute for Polar and Marine Research in Bremerhaven. The new satellite data on summer ice thickness can now also be used to predict ice extent and ice volume, making them much more precise than before. According to the team, the results are of great importance for shipping in the Arctic, but also for future weather and climate forecasts.
"With the new satellite data, we're finally able to make sea ice forecasts based on ice thickness, and not just for winter, but also for summer," explains Landy. “We can predict whether or not there will be ice at a given location in September by measuring the ice thickness in May. That will reduce the safety risks for ships and fishing boats.” But the forecasts for the weather in the mid-latitudes could also improve with the information now available from the Arctic.
Source: Jack Landy (University of Tromsø, Norway) et al., Nature, doi: 10.1038/s41586-022-05058-5