In this work, an innovative approach for remote sensing data analysis is presented. A statistical multivariate approach, applied to spectroscopic data, is able to reduce and explore the large amount of data collected during planetary missions. The multivariate statistical approach implemented is a cluster analysis method together with a criterium able to identify the natural number of clusters present in the spectral data set. An evaluation of the statistical analysis methods has been developed, implemented, and applied to analyze Mars thermal emission data. We find the statistical approach readily identifies spurious data. The resulting number of clusters provides ≥105 reduction in data volume. This allows a focusing of scientific interest onto a limited number of statistically significant groups. A comparison of the results of the statistical approach to previous expert analysis of Mars thermal emission data, for the Sinus Meridiani region where a hematite-rich area of Mars has been previously detected, is provided. We find that several of the clusters reproduce the results of the expert analyses of the Sinus Meridiani hematite distribution. The current approach has the additional advantage of eliminating the time-consuming techniques of atmospheric correction, when surface features are to be investigated.