Supplementary Dataset for "Diagenetic history and biosignature preservation potential of fine-grained rocks at Hogwallow Flats, Jezero Crater, Mars"
Description
Supplementary Dataset for "Diagenetic history and biosignature preservation potential of fine-grained rocks at Hogwallow Flats, Jezero Crater, Mars" Data used to create Figure 15 - SuperCam infrared (IR) spectral modelling.
Files
Steps to reproduce
The modeling steps are as follows: ● Normalization of spectra: since the absolute value of the continuum depends on other parameters in addition to composition, all spectra (both experimental and laboratory) are normalized by division by their mean value. ● Selection of library endmembers in the unknown sample mixture: The endmembers are selected from families whose spectral signatures have been measured in the lab and whose composition is compatible with the results of chemical analysis of LIBS spectroscopy, thus an unsupervised approach. Within each mineral family, species were selected for their spectral signatures and to represent the diversity of measurement conditions including sample state, grain size, observation geometry, and sample purity (including other minor species). 320 endmembers in 8 mineral families were used in this work. To address photometric effects that were not modeled, three artificial spectral components were added to the mineral endmembers: a dark continuum, constant at 0.1, to adjust for the contrast of absorption bands due to the presence of phases without spectral signatures, and two continuums with positive and negative slopes (e.g., straight lines between 0 and 1 over the spectral range) to model the effect of dust scattering. ● Determination of mixing coefficients for each observation: This is a Bayesian approach where the aim is to maximize a likelihood function, which is defined as the exponential of the root mean square between the model and the data. The coefficients are assigned randomly according to a uniform distribution approach. ● Selection of endmembers whose mixing coefficient is greater than 0.01: Coefficients whose value is too low (<1 %) do not correspond to end member minerals that are significantly likely to be present and are omitted from the process. ● Statistical analysis using the Markov Chain Monte Carlo (MCMC) method: The endmembers selected in the previous step are re-injected into an MCMC sampler to calculate the posterior distributions of each of the mixture parameters. The median of these distributions corresponds to the maximum likelihood, and their width gives the sensitivity of the mixture to the presence of each endmember. Thus, a quality criterion is defined as the ratio of the median value by the width of the distribution.