The Aqueous History of Mars

The aqueous history of Mars has become contradictory subject matter, with geological and climate modelling at odds with each other. Current climate models show that the atmosphere contains only minute amounts of water, liquid water is unstable everywhere, and ice has been detected only at the North Pole (Carr, 1996). Yet, the dry river channels, alluvial fans, deltas, and valleys on Mars indicate substantial surface water and fluvial erosion during the Noachian (4.1-3.7 Ga) and parts of Hesperian (3.7-2.9 Ga) (Bishop et al., 2018; Lapotre and Lamb, 2018). Uncertainties also exist about how the water inventory has evolved. Some maintain that Mars started out with a large water inventory but lost most of it early on by escaping to space (Clifford and Hillel, 1983); others maintain that the abundant evidence of water erosion is a testament to the water’s retention on the planet (Pepin, 1994). The resulting speculation is that Mars has undergone major changes in climate. Older models speculate that Mars has experienced global, episodic changes throughout its history due to orbital and rotational motions of the planet (Ward, 1992). However, recent geologic evidence suggests that intense regional weathering due to short “warm and wet” conditions on very early Mars gave way to less intense weathering as the climate became cooler and drier (Bishop and Rampe 2016; Bishop et al. 2018). 3. Classification techniques

We will collect multi- and hyper- spectral data imagery on Earth analog and Martian soil samples. While multispectral instruments offer a handful of bands, a major limitation of multispectral sensing is that it uses average spectral information over broadband widths resulting in the loss of critical waveband information which is available in specific narrow bands. In contrast, hyperspectral sensors acquire many, very narrow, contiguous spectral bands throughout the visible, near infrared, mid-infrared, and thermal infrared portions of the electromagnetic spectrum (Govender, 2007). While high spectral resolution is required to resolve the fine spectral differences between nanominerals, sensing in many adjacent narrow wavebands can produce reflectance spectra containing considerable autocorrelation (Blackburn, 2007). The autocorrelation causes redundancy within hyperspectral data sets; therefore, it is necessary to employ appropriate techniques to characterize the main sources of spectral variability and to identify optimal wavebands that offer maximum informational content (Blackburn, 2007), otherwise known as feature selection. The goal of an optimum feature extraction method for target (mineral) detection and classification is not only to reduce the data dimensionality in order to reduce computation costs, but also to improve the efficiency of classification by extracting features that maximize the separation between the underlying classes (Cheriyadat and Bruce, 2003). Since spectral data for each material is inter- and intra-specific, careful feature extraction and classification is necessary for discriminating between the nanophase materials.