ASSESSMENT OF CARBONATE ABUNDANCE IN SANDS USING HYPERSPECTRAL INDICES

Geological remote sensing applications are ever increasing as satellite and sensor technology improves in both cost-efficiency and in increased spatial and spectral bandwidth. While multiple spectral libraries of rocks and minerals have been developed for resource exploration, there is very little literature available regarding spectroscopy of sand. Spectral signatures of sand are more difficult to characterize because they are essentially mixed reflectance comprised of their constituent materials. Hyperspectral reflectance is an emerging tool that can be employed to estimate mineral abundances in sand, and when incorporated into a spectral library, has the potential to improve understanding of geologically complex and remote sites where spectra is composed of intimate mixtures. Both smoothed and continuum-removed spectra were used in conjunction with various band ratios, including the Normalized Difference Carbonate Index (NDCI). Globally collected samples were analyzed through traditional microscopic methods and compared to spectral reflectance collected via ASD Spectroradiometer. Sands ranged from fine grained (<0.11 mm) to very coarse-grained (>2 mm) and from having no carbonate content to being completely composed of biogenic carbonate. Results show a Pearson correlation of r = 0.345 between actual carbonate content and NDCI-determined content, which performed slightly better than band ratios R₁₀₆₀/R₁₁₃₀ (r = 0.110) and R₁₆₅₀/R₂₂₁₅ (r = -0.250). Low correlations between observed abundance and band-ratio predicted abundance are potentially due to the biogenic structures in the samples rather than crystalized carbonate structure. Additionally, volcanic sand samples containing magnetite and other oxide minerals likely cause lower correlation because O₃ absorption features mimic CaCO₃ features. In this study, the sand spectral library was verified by traditional microscopy which mitigates some of the issues associated with discerning composition of remotely-sensed samples.