For many regions on Mars the surface composition and its geological history has been debated in literature. Because of the limited surface coverage of in-sit measurements, either new data, or new processing methodologies are required to get a better understanding of Martian geology. This paper presents the results of a novel approach applied on the CRISM (Compact Reconnaissance Imaging Spectrometer) multispectral mapping mode dataset. So far, this dataset is under-utilized for global surface analysis. This dataset comes with common difficulties, such as noise within the spectral data, unrealistic data values and differences between orbital observations. The combination of using, spectral parameters, masking values from pre-defined thresholds and averaging the summary product values for a grid of ~5*5o have proven to be a robust method for analyzing the CRIMS mapping mode dataset. Based on an unsupervised clustering algorithm all grid pixels are classified resulting in a global map of distinct surface types. These surface types coincide with the Martian regions Northern Lowlands, Southern Highlands, Meridiani Planum, Syrtis Major, and Nili Fossae. The summary products that contributed the most on the specific surface types are defined based on a new approach, namely partial least squares discriminant analysis. These results were interpreted for (1) sulfate evaporates in Meridiani Planum, (2) hydrothermal carbonates and/or Fe-Mg phyllosilicates in Nili Fossae (3) olivine rich and potentially hydrated mineral rich Syrtis Major, and (4) the Northern lowlands with a lower mafic mineral content than the Southern highlands, exhibiting potential signs of aqueous alteration.