Using Machine Learning for Lunar Mineralogy-I: Hyperspectral Imaging of Volcanic Samples

TL;DR

The study applies VNIR hyperspectral imaging (400-1000 nm) to Vulcano volcanic samples as terrestrial analogs for lunar minerals, focusing on olivine and pyroxene. An unsupervised clustering pipeline (K-Means, Hierarchical, GMM, Spectral Clustering) complemented by PCA and NMF assesses spectral signatures and similarities to reference olivine/pyroxene spectra, revealing a predominance of olivine in several regions. PCA aids dimensionality reduction and region-specific clustering evaluation, while NMF provides similarity heatmaps that align with lunar mineralogy expectations. The results indicate robust olivine detection with K-Means and Hierarchical methods and highlight variability in GMM performance, underscoring the value of terrestrial volcanic analogs for informing lunar mineral mapping and setting the stage for direct lunar comparisons in future work.

Abstract

This study examines the mineral composition of volcanic samples similar to lunar materials, focusing on olivine and pyroxene. Using hyperspectral imaging from 400 to 1000 nm, we created data cubes to analyze the reflectance characteristics of samples from samples from Vulcano, a volcanically active island in the Aeolian Archipelago, north of Sicily, Italy, categorizing them into nine regions of interest and analyzing spectral data for each. We applied various unsupervised clustering algorithms, including K-Means, Hierarchical Clustering, GMM, and Spectral Clustering, to classify the spectral profiles. Principal Component Analysis revealed distinct spectral signatures associated with specific minerals, facilitating precise identification. Clustering performance varied by region, with K-Means achieving the highest silhouette-score of 0.47, whereas GMM performed poorly with a score of only 0.25. Non-negative Matrix Factorization aided in identifying similarities among clusters across different methods and reference spectra for olivine and pyroxene. Hierarchical clustering emerged as the most reliable technique, achieving a 94\% similarity with the olivine spectrum in one sample, whereas GMM exhibited notable variability. Overall, the analysis indicated that both Hierarchical and K-Means methods yielded lower errors in total measurements, with K-Means demonstrating superior performance in estimated dispersion and clustering. Additionally, GMM showed a higher root mean square error compared to the other models. The RMSE analysis confirmed K-Means as the most consistent algorithm across all samples, suggesting a predominance of olivine in the Vulcano region relative to pyroxene. This predominance is likely linked to historical formation conditions similar to volcanic processes on the Moon, where olivine-rich compositions are common in ancient lava flows and impact melt rocks.

Figures (7)

• Figure S1: Overview of the hyperspectral imaging setup. (Left) The Hyperspectral Imaging System (HSI) with labeled light sources (halogen lamps) and components. (Top-right) Diagram of the experimental arrangement, clearly annotating the area camera, spectrograph, halogen lamps, transportation plate with sample, and connected computer for data acquisition. (Bottom-right) Close-up view of a sample under illumination. The white reference was obtained using a Teflon whiteboard/Spectralon reference with 99% reflectance.
• Figure S2: (Right) The image displays the RGB reflectance spectra (combining 450, 550, and 650 nm) for nine distinct regions from Vulcano, a volcanically active island in the Aeolian Archipelago, north of Sicily, Italy, covering the wavelength range of 400-1000 nm. (Left) The average spectral responses of these regions reveal variations in reflectance attributed to differences in mineral composition, particularly the amounts of olivine and pyroxene present. The percentage reported for each dashed-color spectrum indicates the proportion of olivine in the olivine-pyroxene composition template from Mandon2022. The solid-color lines represent the median spectrum for each region, labeled with specific numbers in the right panel.
• Figure S3: Hyperspectral data analysis pipeline for volcanic samples. This figure illustrates the systematic approach used to process and analyze hyperspectral images, including calibration, normalization, and the application of various clustering algorithms. The pipeline ultimately predicts the mineral composition, specifically focusing on pyroxene and olivine, enhancing our understanding of volcanic material characteristics.
• Figure S4: The visualization presents the PCA-based clustering results for volcanic hyperspectral data across nine regions (Regions 1–9), evaluated using four methods: KMeans, Hierarchical, Gaussian Mixture Model (GMM), and Spectral Clustering, with $n=4$ clusters. Each panel corresponds to a specific region and clustering technique, accompanied by Silhouette Scores to assess cluster quality (with higher scores indicating better-defined clusters). Notable findings include strong clustering performance in Region 8 (KMeans: 0.47) and Region 8 (GMM: 0.47), while weaker performance is observed in Region 1 (Spectral: 0.21) and Region 6 (Spectral: 0.27).
• Figure S5: (Top) Maps illustrating the effectiveness of four clustering algorithms—K-Means, Hierarchical, Gaussian Mixture Model (GMM), and Spectral Clustering—in grouping mineral compositions into n=4 clusters within a selected 140x190 pixel area of Volcano Region 1. (Bottom) Comparative visualization of olivine and pyroxene compositions (dashed lines, see Figure \ref{['fig:Similarity_regions_Volcano']}) across clustering models, shown by solid lines with a $\sigma$ error shaded scatters.