The evaluation of field-based analysis in conjunction with Sentinel-2 imagery for the identification and mapping of minerals in the Vredefort Dome

Abstract

Remote sensing has become one of the fastest emerging sciences due to its earth observation capabilities. Fields such as mineral exploration, have in recent years benefited from the help of remote sensing spectroscopy. This is because every rock type has a unique spectral response to electromagnetic radiation due to their varying crystal structures and chemical compositions, which can be picked up in satellite imagery and be used in the identification and mapping of exposed rock types. The main problem is that not all sensors are capable of accurately detecting such features, therefore it becomes important to evaluate the potential of different sensors to determine which can best identify the surface geology of any given environment. This study focuses on the surface geology of the two-billion-year-old meteoritic collision that occurred in the Vredefort Dome, South Africa. This is mainly due to the fact that the geology in this region possesses unique traits such as a superimposed rock stratigraphy that has never been analysed using remote sensing spectroscopy. Seven dominant rock types, namely, granite gneiss, basalt, quartzite, meta-ironstone, banded iron formation (BIF), metadolerite and metapelite were sampled from the Donkervleit region of the Dome and their spectral reflectances were measured using a spectroradiometer in the field. Their absorption features were also analysed before and after continuum removal was applied to the spectra using ENVI software. Results showed that there were subtle, yet distinct, spectral differences between all the rock types that occurred in the Donkervleit region of the Dome but the biggest difference between the rock types could be observed between the quartzite and basalt because of their nonoverlapping mineral compositions. The variability of spectral characteristics within rock types was attributed to differences in rock texture and grain size, and the spectral differences between the rock samples were attributed primarily to differences in mineral composition. The rock identification capabilities of Sentinel-2 imagery were also assessed and it was found that only a 79% accuracy could be achieved through this method; proving that Sentinel-2 is not that efficient at mineral identification when compared to other multispectral sensors such as Aster and Landsat which have been known to produce accuracies between 82% - 92%. Keywords: Multispectral, Sentinel-2, Spectroscopy, Vredefort Dome, Geological Mapping