AI can do more than Draw the planets as bowls of soup. It’s now helping researchers get better data on climate change by teaching Earth observation satellites how to measure the thickness of Arctic ice throughout the year.
Satellites have been monitoring Earth’s icy Arctic for decades now, but the quality of these observations has long been dependent on the season. In winter, when the ice is hard, cold and dry, measurement techniques are simple and effective. But things get even more difficult in the summer, when pools of meltwater form on the icy surface. From space, these meltwater pools are highly reflective, blinding satellite instruments, or making pools look like an open ocean. Under these conditions, satellites cannot distinguish between seawater and thawing ice.
in a new way paper Published in Nature last week, the researchers describe how they were able to use artificial intelligence to overcome these limitations. Their work enabled the first record of ice thickness throughout the year from a satellite.
The satellite in question belongs to the European Space Agency CryoSat-2, launched in 2010. CryoSat-2 uses a synthetic aperture radar system that acts as an altimeter: under ideal conditions, it can measure its height above the ocean and over ice floating on the water. These comparison measurements allow the calculation of the thickness estimate. But for five months out of the year, the quality of those notes was seriously compromised.
To bridge the knowledge gap, researchers use a variety of buoys, planes and ships to monitor summer ice levels, but none of these methods provide the extensive coverage that a satellite can.
To enable CryoSat-2 to continue making useful observations from May through September, when it is usually ineffective, the team worked with previous satellite data from 2011 to 2020, and computer modeling, to teach the system how to recognize the difference between meltwater and an open ocean. This machine learning technique made it possible to obtain a complete record of ice thickness for 12 months, something that a spacecraft had not achieved before.
Perhaps applications of this new capability are of direct value to shipping, making weather pattern prediction easier, and providing guidance on when northern sea lanes are likely to be closed for the winter. It will also make those predictions available very early in the season.
In the long term, the data will be useful to climate scientists who hope to understand the processes involved in sea level changes throughout the year.
Michel Tsamados (University College London) was one of the authors of the paper. He explains, “When we use the new ice thickness data in advanced climate models, it will improve our short-term forecasts of mid-latitude weather and long-term forecasts that show the climate we will have in the future.”
Jack Landy et al. “Record of sea ice thickness throughout the year from the CryoSat-2 satellite.” temper nature.
Featured image: Melting Arctic sea ice, photographed from the Alfred Wegener IceBird Institute’s Sea Ice Survey. Credit: Alfred Wegener Institute/Esther Horvath.