Reşadiye bentonite deposits, which play a significant role in Türkiye's bentonite production, are situated in Central Anatolia. Geochemical, mineralogical, and remote sensing data have been integrated to map the spatial distribution of clay minerals in the bentonite deposits and argillic areas. It is hypothesized that the bentonite samples occurred by the in-situ diagenetic alteration of rhyolite-dacite, trachyte, and andesite/basaltic andesitic composition pyroclastic rocks (ash-flow tuff). Biotite, clinoptilolite, calcite, dolomite, K-feldspar, opal-CT, quartz, and clay minerals are detected in most bentonite samples. The clay patterns determined in the bentonite samples in the X-ray diffraction (XRD) diagrams were 12.3–12.6 Å and were interpreted as being rich in Na-smectites. Mineral mapping in these deposits is essential for mining operations since the high-grade bentonite deposits can be affected by the other clay, gang, and ore minerals they contain in addition to the smectite. The sample spectra measurements matched montmorillonite and kaolin/smectite spectra. This study tests support vector machine (SVM) and artificial neural network (ANN) machine learning and MTMF subpixel algorithms in lithology and mineral mapping in Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) satellite data. It combines the power of subpixel unmixing algorithms to determine the distribution of clay and high-grade bentonites in argillic areas discriminated by machine learning. The results showed that the SVM algorithm can map better than ANN for argillic areas. Additionally, the distribution of high-grade bentonite and kaolin/smectite bearing sites in the study area is discriminated by the mixture-tuned matched filtered (MTMF) spectral classification method. As a result, this study shows that remote sensing studies can be utilized for the exploration and monitoring of high-grade bentonite sites during and/or post-mining operations.
Read full abstract