Scientists mark a milestone with PlanetNet as the "deep learning" approach passes its first demonstration with flying colors, mapping out a Saturn storm in great detail.
It's a new approach that's expected to significantly increase scientists' understanding of planetary atmospheres.
"PlanetNet enables us to analyze much bigger volumes of data, and this gives insights into the large-scale dynamics of Saturn," explained co-author Caitlin Griffith, a professor at the University of Arizona, in a news release from the university. "The results reveal atmospheric features that were previously undetected."
A Closer Look At Saturn
The study, published in the journal Nature Astronomy, reveals how the team of scientists used PlanetNet to map out the turbulent regions of Saturn's atmosphere.
PlanetNet uses infrared data collected by the Visible and Infrared Mapping Spectrometer instrument on NASA's Cassini spacecraft. For this new study, the researchers used the deep learning algorithm to take a closer look at a dataset of storm system observed back in February 2008.
Previous analysis of this dataset identified the rare presence of ammonia in the planet's atmosphere, taking on the form of an S-shaped cloud.
With the map produced by PlanetNet, the team discovered that this cloud is actually part of a more massive upwelling of ammonia ice clouds surrounding a dark storm. Plus the scientists spotted a similar feature around another smaller storm, which means that upwellings of ammonia ice clouds are common in Saturn.
The map also showed that there are significant differences between the center of the storms and its surrounding regions.
How PlanetNet Works
Developed by scientists from UA and University College London, PlanetNet works by first scanning the data for signs of clustering in the atmosphere's cloud structure and gas composition. Then the algorithm removes uncertainties and analyzes spectral and spatial properties.
With all these data combined, PlanetNet is able to produce a map that features the major components of Saturn's storms with unprecedented and unparalleled precision.
According to Griffith, the algorithm can easily be applied to other planets and datasets, making it an invaluable potential tool in future research.
Ingo Waldmann, lead author and deputy director of the UCL Centre for Space and Exoplanet Data, added that while Cassini and other similar missions are capable of gathering data, traditional analysis techniques have limitations, from accuracy to length of time.
"Deep learning enables pattern recognition across diverse, multiple data sets," explained Waldmann. "This gives us the potential to analyze atmospheric phenomena over large areas and from different viewing angles, and to make new associations between the shape of features and the chemical and physical properties that create them."