Satellites can now measure the thickness of Arctic sea ice all year round. Previously, satellites have had trouble determining the state of the floes in summer months due to the presence of surface meltwater.
Now, using deep learning techniques, scientists have pushed past these limitations to get reliable observations all year round.
Apart from the obvious advantage to ships for navigation, there are significant benefits to climate and weather forecasting.
Having an improved understanding of the melting process in the summer months when floes are being reduced, in area and thickness, should improve the output from computer models.
“We need to tighten those predictions so we’re a lot more confident about what’s going to happen and when – and how the climate feedbacks will accelerate as a consequence.” said Jack Landy from UiT the Arctic University of Norway.
The extent of Arctic sea-ice cover has been declining for the entire period that satellites have been monitoring it, which is more than four decades – a reduction of an average rate of 13% per decade.
However, it is only since 2011 that satellites have been able to consistently measure its thickness.
The European Space Agency’s Cryosat-2 mission carries a radar to measure the difference in height between the top of the marine ice and the top of the water in the crack’s that separate the floes.
This method works well in winter months, but in summer, when the snow on top of the ice, and the ice itself, starts to melt, pooling water interferes with the radar. Scientists can not measure if the echo signal that returns to Cryosat is coming from the open ocean or from the surface of a meltpond sitting on the ice.
May through September, which is the key melt season, has been a blind spot for the satellite.
To solve the problem, researchers used an artificial intelligence technique in which an algorithm was able to learn and identify reliable observations from a vast library of synthetic radar signals.