Probabilistic classification of elemental abundance distributions in Nakhla and Apollo 17 lunar dust samples

摘要

Analysis of spectral and imaging data from meteoritic samples and sample return missions would benefit significantly from a systematic, quantitative statistical classification methodology and a common set of standards for data collection [McDonald and Storrie-Lombardi, 2006]. Stochastic artificial neural networks can be trained using elemental abundance distributions for the detection of macroscopic fossils [Storrie-Lombardi and Hoover, 2004] and extant microbial life [Storrie-Lombardi and Hoover, 2005]. These non-linear algorithms are particularly attractive since they can produce a Bayesian estimate of the classification accuracy of either human experts or automated, unsupervised classification algorithms. In sub-ocean and surface basalts on earth the networks can distinguish regions of biotic and abiotic alteration of basalt glass from unaltered samples using only elemental abundances as inputs [Storrie-Lombardi and Fisk, 2004b]. Recently, evidence has been presented documenting the presence of morphologic signatures in the Mars meteorite Nakhla [Fisk et al., 2004; Fisk et al., 2006] previously noted in regions of biotic alteration in sub-ocean and surface terrestrial basalts [Fisk et al., 2003; Furnes et al., 2004]. The tunneling alterations are not conclusive evidence of biotic alteration of Nakhla on Mars. However, the meteorite is well known to have experienced aqueous alteration prior to arrival on earth and is rich in carbon [Gibson et al., 2006; McKay et al., 2006]. We here present an initial application of our probabilistic classification strategy to assess elemental abundance distributions from multiple target regions in Nakhla and lunar dust samples collected by Apollo 17 astronauts. We present scanning electron microscope images and elemental abundance point distributions (C, N, O, Na2O, MgO, Al2O3, SiO2, P2O5, S, Cl, K2O, CaO, and FeO) for a series of target regions. We discuss our observations in the context of data previously presented in these meetings for extant cyanobacteria, fossil trilobites, Orgueil meteorite, and terrestrial basalt targets. These data are being added to a database that will made available to the biogeology and astrobiology communities as part of an ongoing effort to provide a quantitative probabilistic methodology for analysis of putative elemental abundance geobiological signatures.