Mathematics – Logic
Scientific paper
Dec 2010
adsabs.harvard.edu/cgi-bin/nph-data_query?bibcode=2010agufm.p51b1436r&link_type=abstract
American Geophysical Union, Fall Meeting 2010, abstract #P51B-1436
Mathematics
Logic
[3672] Mineralogy And Petrology / Planetary Mineralogy And Petrology, [5410] Planetary Sciences: Solid Surface Planets / Composition, [5464] Planetary Sciences: Solid Surface Planets / Remote Sensing, [6225] Planetary Sciences: Solar System Objects / Mars
Scientific paper
Unsupervised classification techniques were applied to TES-derived mineral abundance distributions generated by Koeppen and Hamilton [2008, JGR-Planets] to identify regions of distinct mineral assemblage on a global scale. The new mineral abundance maps of Koeppen and Hamilton [2008] were generated using an improved spectral library with a wider variety of pyroxene, olivine and feldspar solid solutions than those of Bandfield [2002, JGR-Planets]. The primary advantage of classifying mineral maps into mineral assemblage distributions is to reduce the dimensionality of the data such that the major spatial variations in lithology can be visualized and compared with geologic age, latitude, morphology, and elevation. Variations in mineral assemblage arise from a variety of geologic processes and thus are important for reconstructing geologic history. Initial work focused on abundance distributions produced at 2 pixels per degree (ppd), using nine mineral groups: olivine, feldspar, high-Ca pyroxene, low-Ca pyroxene, sulfates, quartz, high-silica phases, carbonates, and hematite. Both ISODATA and K-means classification techniques were used. The number of allowable classes imposed on the classification algorithms was iteratively increased, and the average and standard deviation of mineral abundances found for each class were compared to the global mean as well as to each other. Too few classes produce average compositions that are not distinct from the global mean, while too many classes force the separation of compositions that differ insignificantly from each other. Corresponding class distributions were compared to spectral unit maps derived by Rogers et al. [2007, JGR-Planets]. Preliminary work with 2 pixel-per-degree data shows that use of global statistics does not permit a classification scheme in which the average mineralogy for all calculated classes are outside of one standard deviation from the global mean. However, recurring classes are present despite the number of classes imposed on the algorithm. These distinctive classes were removed from the data set to permit better classification of the remaining data but also to isolate subclasses that were not separable using global statistics. Initial results are broadly consistent with spectral unit maps of Rogers et al. [2007], however new divisions in Acidalia are apparent and the highlands distributions have moderate spatial differences from previous work. Globally, units are distinguished by feldspar, low-Ca pyroxene, high-Ca pyroxene, and high-silica phase abundance. Future work will focus on classifications based on individual mineral distributions (rather than mineral groups) and 4 and 8 ppd resolution data.
Hamilton Victoria E.
Rogers David
No associations
LandOfFree
Martian Surface Composition From Multiple Datasets, Part I: Statistical Analysis of Global Mineral Distributions From MGS-TES does not yet have a rating. At this time, there are no reviews or comments for this scientific paper.
If you have personal experience with Martian Surface Composition From Multiple Datasets, Part I: Statistical Analysis of Global Mineral Distributions From MGS-TES, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Martian Surface Composition From Multiple Datasets, Part I: Statistical Analysis of Global Mineral Distributions From MGS-TES will most certainly appreciate the feedback.
Profile ID: LFWR-SCP-O-1497724