Mapping the hydromagnesite content of Akgöl (Burdur) sediments using remote sensing and GIS techniques

This thesis study analyzes 70 samples collected from the lake-bottom sediments of Akgöl, a hydraulically closed lake basin located in the arid zone of southwestern Türkiye, using X-Ray Fluorescence (XRF), X-Ray Diffraction (XRD), and a spectroradiometer. Within the scope of remote sensing–based modelling of the mineralogical composition of the Akgöl Basin, ASTER multispectral satellite images were integrated into the workflow, and mineral distribution surfaces for 13,293 pixels were produced by jointly applying multiple linear regression and Kriging interpolation techniques in the ArcGIS environment. At this stage, a hydromagnesite (Mg₅(CO₃)₄(OH)₂4H₂O) map of the Akgöl Basin was generated using the multiple regression method. As a second stage, in order to demonstrate that ASTER data provide more accurate results in mineral mapping, a comprehensive machine learning analysis was conducted using LandSAT-8 OLI, Sentinel-2A, and ASTER satellite images. In the study, XRD and XRF data obtained from the field were used as reference values, and Random Forest regression models were produced for all three satellite datasets to generate raster outputs of various minerals and oxides, including hydromagnesite (Mg₅(CO₃)₄(OH)₂4H₂O), dolomite (CaMg(CO₃)₂), biotite (K(Mg,Fe)3(Al,Fe)Si3O10(OH,F)2), chamosite ((Fe2+;Mg;Fe3+)5 Al(Si3Al)O10(OH;O)8), MgO, CaO, Al₂O₃, SiO₂, and Fe₂O₃. While LandSAT-8 OLI, due to its broad spectral bands, produced moderate levels of accuracy particularly for Fe-oxides and CaO, Sentinel-2A demonstrated higher discrimination power for oxides such as Fe₂O₃ and Al₂O₃ thanks to its narrower spectral bandwidth. ASTER, on the other hand, yielded more successful results in the modelling of carbonates and Mg/Fe-OH minerals such as hydromagnesite, MgO, dolomite, and chamosite due to its strong SWIR and TIR bands, when compared to the other satellite images. In light of these results, the spatial modelling outputs indicate that the mineral distributions within the basin can be reliably interpreted. The geomorphological and hydrogeochemical characteristics of Akgöl show a high degree of similarity to other endorheic evaporitic systems such as Salda and Acıgöl, and the approach used in this study demonstrates methodological transferability to similar closed basins located in arid and semi-arid environments. In this respect, the study not only contributes to constraining the regional mineral potential but also strengthens methodological depth in mineral research through the integrated use of spectroscopic, geochemical, and geostatistical datasets.