Machine Learning enabled detections of Mg-spinel bearing lithology using Moon Mineralogy Mapper (M3) global coverage

Abstract

The new feldspathic rock type bearing Mg-spinel mineral has been detected at small localised areas of a few hundred meters [1,2,3,4] using Moon Mineralogy Mapper (M3) data from the Chandrayaan-1 mission [5, 6]. We have employed a one-dimensional convolutional neural network (CNN) to systematically identify and map the global distribution of this lithology by utilizing global coverage of the M3 dataset. This semi-automatic analytical framework has enabled the detection of isolated occurrences (few pixels) of Mg-Spinel lithology that would have been difficult to locate manually. The CNN model incorporates supervised classification which is validated using results from already reported locations [2, 3] and further refined by deriving spectral band parameters. This new approach of identifying a specific spectral signature using a CNN model has yielded several new Mg-spinel exposure locations which are majorly distributed at high latitudes (>±60o) and at mare-highland boundary regions, including identifications from relatively noisy data. We have also checked for repeatability using M3 coverage of different optical periods and found the results to be consistent. We demonstrate the ability of the machine learning approach in detecting Mg-spinel spectral characteristics at a global scale. The new locations identified are typically 3-4 pixels wide. We are in the process of site-specific studies mainly focused on high latitude regions from south and north. Global mapping of Mg-spinel has important implications for understanding the magmatic evolution of the Moon, especially the lunar crust. Further work integrating the data from Imaging Infrared Reflectance Spectrometer (IIRS) onboard Chandrayaan-2 [7] is underway. References: [1] Pieters, C. et al. (2011) JGR 116(4), 1-14. [2] Dhingra, D. et al (2011) GRL 38, L11201. [3] Pieters, C. et al. (2014) American Mineralogist, 99(10), 1893-1910. [4] Sun et al. (2017) EPSL 465, 48-58. [5] Pieters, C. et al. (2009) Current Science 96, 500-505. [6] Goswami, J. and Annadurai, M. (2009) Current Science 96, 486-491. [9] Mathavaraj, S. et al. (2020) Acta Astronautica 177, 286-298.