This page's content is no longer actively maintained, but the material has been kept on-line for historical purposes.
The page may contain broken links or outdated information, and parts may not function in current web browsers.

An Objective Classification of Saturn Cloud Features

The datasets here were produced by k-means clustering of Narrow and Wide Angle Camera images of Saturn's atmosphere acquired by the Imaging Science Subsystem (ISS) instrument on the Cassini Saturn Orbiter spacecraft. The concept behind k-means clustering is to objectively identify regions in Saturn's atmosphere that are spectrally similar and thus identify specific cloud types. The approach we use follows that originally adopted by the International Satellite Cloud Climatology Project (ISCCP) to classify "weather states" on Earth using histograms of cloud optical thickness and cloud top pressure derived from visible and window infrared images on operational weather satellites.

Figure 1: Six panels showing clustering patterns

Figure 1: View large PNG or PDF.

For Saturn, we apply the clustering technique to ISS images at two wavelengths: a continuum wavelength (750 nm; filter CB2) that primarily views the tops of Saturn's ammonia clouds, but with some obscuration by its upper troposphere haze, and a methane band wavelength (727 nm; filter MT2) that is restricted to higher altitudes by methane absorption and senses the upper troposphere haze. The images were acquired in 2007, 2008 and 2014 and cover the 18°-60°N region in late southern summer. Histograms of binned MT2 vs. CB2 I/Fs over 1°×1° regions, normalized to remove viewing geometry effects and large-scale brightness patterns, are the input to the clustering algorithm.

The first figure shows the clusters derived from the algorithm. Each one is indicative of specific cloud types as defined by preferential values of CB2 I/F (dark to bright indicating small to large optical thickness) and MT2 I/F (dark to bright indicating low-altitude to high-altitude tops). Combined with the observed cloud morphology in the locations corresponding to each cluster, plus knowledge of the wind field, the clusters are tentatively interpreted as follows:

  • SC1: Deep convective clouds
  • SC2, SC3: Nimbostratus and altostratus/altocumulus due to baroclinic instability
  • SC4: Thin cirrus
  • SC5: Clear or nearly clear subsidence regions associated with nearby convection
  • SC6: Shallow cumulus clouds

The second figure below shows the cluster locations (red boxes) superimposed on CB2 (left) and MT2 (right) images of Saturn's great storm of 2010-2011.

Figure 2: Cluster locations superimposed on Saturn imagery

Figure 2: View large PNG


The datasets contain cluster number, latitude, and longitude for each cluster identification, for each of the days analyzed, plus a fifth dataset for Saturn's great storm of 2010-2011.

The datasets are all gzipped plain text.


The ISS instrument characteristics are described in detail in:

Porco, C.C., R.A. West, S. Squyres, A. McEwen, P. Thomas, C.D. Murray, A. Del Genio, A.P. Ingersoll, T.V. Johnson, G. Neukum, J. Veverka, L. Dones, A. Brahic, J.A. Burns, V. Haemmerle, B. Knowles, D. Dawson, T. Roatsch, K. Beurle, and W. Owen, 2004: Cassini imaging science: Instrument characteristics and capabilities and anticipated scientific investigations at Saturn. Space Sci. Rev., 115, 363-497, doi:10.1007/s11214-004-1456-7.

A full description of the clustering approach as applied to Saturn and interpretations of the clusters can be found in:

Del Genio, A.D., and J.M. Barbara, 2015: An objective classification of Saturn cloud features from Cassini ISS images. Icarus, in press, doi:10.1016/j.icarus.2016.02.011..

An overall review of the dynamics of Saturn's atmosphere as revealed by Cassini can be found in:

Del Genio, A.D., R.K. Achterberg, K.H. Baines, F.M. Flasar, P.L. Read, A. Sánchez-Lavega, and A.P. Showman, 2009: Saturn atmospheric structure and dynamics. In Saturn from Cassini-Huygens. M.K. Dougherty, L.W. Esposito, and S.M. Krimigis, Eds. Springer-Verlag, pp. 113-159, doi:10.1007/978-1-4020-9217-6_6.

PDF documents require the free Adobe Reader or compatible viewing software to be viewed.


This research was led by Dr. Anthony Del Genio. Please address any questions you may have about the data to Mr. John Barbara.