Mapping Saturn using deep learning

I. P. Waldmann, C. A. Griffith

Research output: Contribution to journalArticlepeer-review

Abstract

Clouds and aerosols provide a unique insight into the chemical and physical processes of gas-giant planets. Mapping and characterising the spectral features indicative of the cloud structure and composition, enables an understanding of a planet’s energy budget, chemistry and atmospheric dynamics (e.g. [1, 2, 3, 4]). Current space missions to solar-system planets produce high-quality data sets, yet the sheer amount of data obtained often prohibits detailed ‘by hand’ analyses. Current techniques mainly rely on two approaches: 1) identify the existence of spectral features by dividing fluxes of two or more spectral channels; 2) perform full radiative transfer calculations for individual spectra. The first method suffers from accuracy whilst the second from scalability to the whole planetary surface. Here we developed a deep learning algorithm, PlanetNet, able to quickly and accurately map spatial/spectral features across large, heterogeneous areas of a planet. We demonstrate PlanetNet on Saturn’s 2008 storm[5], enhancing the scope of the area previously studied. Our spectral-component maps indicate compositional and cloud variations of the vast region affected by the storm showing regions of vertical upwelling, and diminished clouds at the centre of compact storms. This analysis quickly and accurately delineates the major components of Saturn’s storm, thereby indicating regions that can be probed deeper with radiative transfer models.

Original languageEnglish (US)
JournalUnknown Journal
StatePublished - Apr 29 2019

ASJC Scopus subject areas

  • General

Fingerprint Dive into the research topics of 'Mapping Saturn using deep learning'. Together they form a unique fingerprint.

Cite this