4.2. Band Combinations
Rocks exhibit different reflective properties depending on their chemical compositions. By utilizing these properties of rocks, ferrous (Fe+ 2 and Fe+ 3) and hydroxyl-bearing (AI-OH and Mg-OH) felsic and mafic rocks can be determined in the visible and near infrared regions of the spectrum, and siliceous, carbonated and sulfated rocks in the thermal region (Rowan & Mars, 2003).
The ASTER sensor, capable of collecting data from three different regions of the spectrum (VNIR, SWIR and TIR) enables the acquisitions of the highest spectral richness in a single image using false color composite (FCC) images created with bands covering wavelengths sensitive to various chemical and mineralogical compositions. Thus, images with rich spectral content for specific purposes can be obtained through numerous combinations of bands representing rocks with high reflective properties in spectral bands that are used as the equivalent of RGB (Bierwirth, 2002; Hewson et al., 2005; Rowan & Mars, 2003; Volesky et al., 2003). Rocks with different chemical compositions, minerals in geological terms, and the elements that constitute them, stand out in bands sensitive to the wavelengths they are susceptible to base on their absorption properties, such as the prominent presence of Fe and Mg in minerals that make up ophiolitic rocks.
In ASTER images, the commonly used 731 band combination (RGB) is generally used for distinguishing different geological features. In this study, 731 false color combinations were applied to ASTER images to attempt to identify different main rock groups (Fig. 5). Only a 2% linear stretch was applied as image enhancement to the FCC image. When the obtained image is evaluated in conjunction with the geological map of the region, ultramafic rock groups stand out in shades of blue, while volcanic rocks are reddish in the northern and northeastern parts of the study area. Carbonated rocks are depicted in light brown areas, and clastic rocks are represented by light beige areas.
In Sentinel 2A images, unlike ASTER, there are mainly visible - near-infrared (VIS-VNIR) bands with fewer shortwave infrared (SWIR) bands. In this study, the 12,8,3 (RGB) combination was used to emphasize different geological features from the Sentinel 2A image (Fig. 6). When the obtained image is evaluated in conjunction with the geological map of the region, it is generally observed that shades of blue correspond to ultramafic rocks, dark brown areas to volcanic rocks of different age groups, and lighter brown and beige areas to carbonated and clastic rocks.
4.3. Band Ratioing and Classification
Based on the VNIR-SWIR absorption features, it has been revealed that Al-OH and Fe+ 3 are dominant in felsic rock spectra, whereas Fe+ 2 and Fe, Mg-OH minerals dominate the spectra of mafic-ultramafic rocks (Hewson et al., 2005; Rowan et al., 2005). Similarly, these minerals are highlighted in mapping laterites (Andrews Deller, 2006; Bierwirth, 2002) and gossan zones (Volesky et al., 2003).
The study area, with widespread and high-grade iron ore deposits in terms of reserve, is also characterized by minimal vegetation cover, making it highly suitable for remote sensing studies.
Especially in the VNIR and SWIR regions, high absorption and reflection values are associated with Fe components, including Fe+ 2 (siderite), Fe+ 3 (hematite, goethite), iron silicates (fayalite), iron oxides (hematite, magnetite), gossan (iron cap), and laterite (Fig. 7). Mapping these iron-rich formations using ASTER and Sentinel 2A imagery involved applying band ratio processes in wavelength regions that exhibit high reflectance for iron (Table 2). The resulting spectral distributions were then compared to known Fe-bearing deposits in the field.
Table 2
Band ratio combinations for identifying Fe-bearing minerals using ASTER and Sentinel 2A data.
Mineral | ASTER | Sentinel 2A |
Ferric iron (Fe3+) (hematite, gothite) | 2/1 (Rowan & Mars, 2003a) | 4/3 |
Ferrous iron (Fe2+) (siderite) | 5/3 + 1/2 (Rowan & Mars, 2003a) | 12/8 + 3/4 |
Laterite | 4/5 (Bierwirth, 2002) | 11/12 |
Gossan | 4/2 (Volesky et al., 2003) | 11/4 |
Ferrous silicates (biot, chl, amph) | 5/4 (Cudahy et al., 2001) | 12/11 |
Ferric oxides | 4/3 (Cudahy et al., 2001) | 11/8 |
The obtained ratio images were classified using a supervised classification approach with the parallelepiped algorithm. When performing these classifications, the pixels with the highest reflectance in the relevant ratio images were defined as training data for the algorithm. Threshold values for the created classes were determined by calculating the mean basic statistical values of the ratio images. As a result of these processes, for each ratio image, pixels with the highest reflectance according to the threshold values were classified, while other pixels were not.
It was observed that the known Fe-bearing minerals, among all the classifications, mainly overlapped with classification distributions related to iron oxides, Fe2+, ferric oxides and ferrous silicates. Additionally, these formations were found to be closely associated with the COF, which approximately extends in the E-W direction through the study area, and other faults. The existing formations and the iron distributions detected through remote sensing methods were found to be related to the ophiolitic rocks in the study area and the surrounding magmatic rocks (Fig. 8).
Similar band ratio processes were also performed on Sentinel 2A images. The classification distributions obtained from the analysis of Sentinel 2A images were found to be in good agreement with the ASTER. As the ASTER ratio images, the classification distributions for Fe-bearing minerals in Sentinel 2A images showed significant consistency with Fe2+, iron silicate and iron oxide distributions, which also matched known iron formations and the COF (Fig. 9).
The classification distributions obtained from both satellite images have been compared, and while there are significant overlapping areas, the regions where iron content was detected in
the ASTER ratio images cover a range of percentages within the study area. These percentages vary from a minimum of 0.821% for Fe oxides to a maximum of 2.219% for Fe silicates (Table 3). This information highlights the distribution of Fe-bearing minerals across the study area as detected by the two satellite image datasets.
Table 3
Statistical data of classified and unclassified points within the study area based on the ratio images obtained.
Class Name | ASTER | SENTINEL 2A |
Result | Npts | Pct % | Result | Npts | Pct % |
Fe2+ | Unclassified | 4.520.053 | 98,54 | Unclassified | 9.969.286 | 96,234 |
Classified | 66.955 | 1,46 | Classified | 390.182 | 3,766 |
Fe Silicates | Unclassified | 4.485.215 | 97,781 | Unclassified | 9.79.8848 | 94,588 |
Classified | 101.793 | 2,219 | Classified | 560.620 | 5,412 |
Fe Oxides | Unclassified | 4.549.365 | 99,179 | Unclassified | 10.108.823 | 97,581 |
Classified | 37.643 | 0,821 | Classified | 250.645 | 2,419 |
Laterite | Unclassified | 4.535.014 | 98,866 | Unclassified | 10.199.019 | 98,451 |
Classified | 51.994 | 1,134 | Classified | 160.449 | 1,549 |
Fe3+ | Unclassified | 4.521.591 | 98,574 | Unclassified | 10.202.908 | 98,489 |
Classified | 65.417 | 1,426 | Classified | 156.560 | 1,511 |
Gossan | Unclassified | 4.515.206 | 98,435 | Unclassified | 9.949.941 | 96,047 |
Classified | 71.802 | 1,565 | Classified | 409.527 | 3,953 |
In Sentinel 2A images, compared to the same ratio images in ASTER, a higher proportion of the study area is classified. The distribution of classified areas is greater across the entire study area when comparing all ratio images. Laterite and Fe3+ distribution ratios, in particular, are the closest results to each other, with values of approximately 1.549%, 1.511% (Sentinel 2A) and %1,134, %1,426 (ASTER) for Fe3+. However, it's important to note that the results obtained from Sentinel 2A images are roughly double that of the results from ASTER for other classification distributions.
When comparing the ratio images of each iron formation obtained from ASTER and Sentinel 2A images, they appear to be largely similar in terms of geological formation conditions, associated lithologies and structural elements. However, due to the differences in satellite imagery and spectral resolution, there are variations in the distribution densities of the detected Fe-bearing minerals. Specifically, the ratio images for ferrous iron (Fe2+) and ferric oxides show that the Fe-bearing minerals are more associated with ophiolitic rocks and to a lesser extent, with volcanic and carbonated rocks. It is examined that the results individually, it can be noted that the distribution in ASTER images is less dense compared to Sentinel 2A. (Fig. 10).
In the results for ferrous silicates and ferric iron (Fe3+), the distributions are generally associated with clastic lithologies and to a lesser extent with carbonated lithologies. It's also observed that the distribution density in the ASTER images is lower compared to Sentinel 2A (Fig. 11).
When examining the results for laterite and gossan, their distributions appear to be associated with volcanic and plutonic rocks for laterite and to a lesser extent, volcanic rocks for gossan. Similar to other iron formations, the distribution density in the ASTER results is less compared to Sentinel 2A results for laterite and gossan (Fig. 12).