Image of a storm in Saturn's atmosphere captured by the Cassini spacecraft wide-angle camera on March 4, 2008.

Machine learning has helped researchers to find and map ammonia-rich clouds at the centre of a storm on Saturn. Credit: NASA/JPL/Space Science Institute

Planetary science

Saturn’s secrets revealed by deep learning

PlanetNet algorithm has helped scientists to understand a storm on the giant planet.

A deep-learning method developed by planetary scientists is allowing them to map other worlds. When tested on Saturn, the algorithm uncovered details of a complex atmospheric storm.

Ingo Waldmann at University College London and Caitlin Griffith at the University of Arizona in Tucson call their algorithm PlanetNet. It starts by spotting faint patterns in an image, then searches for similar patterns in other images.

The researchers fed it pictures, taken by the Cassini spacecraft, of a storm that roiled Saturn’s atmosphere in 2008.

In one small picture, PlanetNet quickly picked out features such as ammonia ice clouds. When the scientists extended the analysis to a larger area, PlanetNet revealed that the storm covered a greater area than previously thought. The program also spotted ammonia-rich clouds upwelling around the storm’s centre. Smaller neighbouring storms also showed this upwelling.

The algorithm could help scientists to map other planets quickly and accurately, the authors say.