G-mode classification of spectroscopic data

We present the results of the application of the G-mode method to the spectral classification of the icy satellites of the giant planets. G-mode is a multivariate statistical technique for the classification of samples depending on many variables. Here this method is tested on the infrared spectra acquired by the Cassini/VIMS instrument onboard the Cassini spacecraft. This work demonstrates the suitability of automatic spectral classification methods for the study of fair resolution spectra, such as those from VIMS. Our data set is composed by two different kinds of data: Observations of point targets (Galilean satellites data) and observations with medium spatial resolution (Phoebe data). In both situations, the G-mode classification performed well. In the first case, of a large number of subpixel observations of the Galilean satellites, through the G-mode it was possible to find statistically meaningful spectral groups of observations. In the case of Phoebe, of some spatially resolved observations, the G-mode classification of the infrared spectra of the surface led to several types, dominated by the different illumination geometry of the pixels, because, due to the irregular shape of the satellite, a proper illumination correction was not trivial to apply. Nevertheless, the decrease of the confidence level of the test as well as the re-application of the G-mode on the main type found, led to further types, whose statistical distance can be related to different chemical abundances. We plan to use the G-mode also on the data coming from ongoing and future observations of the icy Saturnian satellites.