Saturn's vertical and latitudinal cloud structure 1991–2004 from HST imaging in 30 filters

Abstract

We analyzed 134 images of Saturn taken by the Hubble Space Telescope between 1991 and 2004. The images cover wavelengths between 231 and 2370 nm in 30 filters. We combined some 10 million calibrated reflectivity measurements into 18,000 center-to-limb curves. We used the method of principal component analysis to find the main latitudinal and temporal variations in Saturn's atmosphere and their spectral characteristics. The first principal variation is a strong latitudinal variation of the aerosol optical depth in the upper troposphere. This structure shifts with Saturn's seasons, but the structure on small scales of latitude stays constant. The second principal variation is a variable optical depth of stratospheric aerosols. The optical depth is large at the poles and small at mid- and low latitudes with a steep gradient in-between. This structure remains essentially constant in time. The third principal variation is a variation in the tropospheric aerosol size, which has only shallow gradients with latitude, but large seasonal variations. Thus, aerosol sizes and their phase functions inferred at a particular season are not representative of Saturn's atmosphere at other seasons. Aerosols are largest in the summer and smallest in the winter. The fourth principal variation is a feature of the tropospheric aerosols with irregular latitudinal structure and fast variability, on the time scale of months. Spherical aerosols do not display the spectral characteristic of that feature. We suspect that variations in the shape of aerosols may play a role. We found a spectral feature of the imaginary index of aerosols, which darkens them near 400 nm wavelength. While we can describe Saturn's variations quite accurately, our presented model of Saturn's average atmosphere is still uncertain due to possible systematic offsets in methane absorption data and limitations of the knowledge about the shape of aerosols. In order to compare our results with those from comparable investigations, which used less than 30 filters, we fit models to spectral subsets of our data. We found very different best-fitting models, depending on the subset of filters, indicating a high sensitivity of results on the spectral sampling.