Applications of machine learning algorithms in lithological mapping of Saint Katherine Neoproterozoic rocks in the South Sinai of Egypt using hyperspectral PRISMA data.

Major Pan-African basement rock units are exposed at South Sinai, documenting a protracted geologic history from Late Mesoproterozoic to Late Neoproterozoic and forming the northeastern part of the Arabian-Nubian Shield (ANS). They form non-consanguineous diverse metamorphic and igneous rock units, which are frequently distributed throughout the entire ANS. The current study is focused on the Saint Katherine area, around Wadi Solaf-Wadi Harqus. Automatic lithological mapping has been carried out using Support Vector Machine (SVM), and Random Forest (RF) machine learning algorithms applied to hyperspectral PRISMA data with an overall accuracy (OA) of up to 91.41% and 86.64%, respectively. Six spectral signature characteristics of the widely exposed different rock units and their alteration minerals have been identified using the Halo Mineral Identifier Spectrometer Device (ASD). Six hyperspectral PRISMA hydrothermal alteration mineral indices, such as phyllic, carbonate/chlorite/epidote, clay minerals, kaolinite, ferrous silicates, and hydroxyl group, have been detected, highlighting the proposed zones for future mineral exploration in the study area. Post-collision calc-alkaline granitoids are the predominant rock varieties, represented by two magmatic series separated by NNE-SSW and NNW-SSE striking calc-alkaline dyke swarms of varied compositions (basalt, basaltic andesite, andesite, and rhyolite). The first series is dominated by granodiorite and biotite monzogranite, which is supposed to have evolved from the same magma mainly by plagioclase fractionation. The second granitic series is extremely evolved syenogranite and alkali feldspar granite. Petrography and mineral chemistry of the essential mineralogical constituents of granitoids impose constraints on their mutual relationships and evolution.