Comparison of Different Algorithms to Map Hydrothermal Alteration Zones Using ASTER Remote Sensing Data for Polymetallic Vein-Type Ore Exploration : Toroud – Chahshirin Magmatic Belt ( TCMB ) , North Iran

Polymetallic vein-type ores are important sources of precious metal and a principal type of orebody for various base-metals. In this research, Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) remote sensing data were used for mapping hydrothermal alteration zones associated with epithermal polymetallic vein-type mineralization in the Toroud–Chahshirin Magmatic Belt (TCMB), North of Iran. The TCMB is the largest known goldfield and base metals province in the central-north of Iran. Propylitic, phyllic, argillic, and advanced argillic alteration and silicification zones are typically associated with Au-Cu, Ag, and/or Pb-Zn mineralization in the TCMB. Specialized image processing techniques, namely Selective Principal Component Analysis (SPCA), Band Ratio Matrix Transformation (BRMT), Spectral Angle Mapper (SAM) and Mixture Tuned Matched Filtering (MTMF) were implemented and compared to map hydrothermal alteration minerals at the pixel and sub-pixel levels. Subtle differences between altered and non-altered rocks and hydrothermal alteration mineral assemblages were detected and mapped in the study area. The SPCA and BRMT spectral transformation algorithms discriminated the propylitic, phyllic, argillic and advanced argillic alteration and silicification zones as well as lithological units. The SAM and MTMF spectral mapping algorithms detected spectrally dominated mineral groups such as muscovite/montmorillonite/illite, hematite/jarosite, and chlorite/epidote/calcite mineral assemblages, systematically. Comprehensive fieldwork and laboratory analysis, including X-ray diffraction (XRD), petrographic study, and spectroscopy were conducted in the study area for verifying the remote sensing outputs. Results indicate several high potential zones of epithermal polymetallic vein-type mineralization in the northeastern and southwestern parts of the study area, which can be considered for future systematic exploration programs. The approach used in this Remote Sens. 2019, 11, 495; doi:10.3390/rs11050495 www.mdpi.com/journal/remotesensing Remote Sens. 2019, 11, 495 2 of 24 research has great implications for the exploration of epithermal polymetallic vein-type mineralization in other base metals provinces in Iran and semi-arid regions around the world.


Introduction
Since the Bronze Age, polymetallic vein-type ores have been important sources of precious metal and established a main type of deposit for various base-metals [1][2][3][4][5].Polymetallic vein-type ore deposits precipitated in the geological structures such as faults, fractures, brecciated rocks, and porous layers, where the pressure, temperature, and several other chemical factors are suitable for the precipitation [1,6].Moreover, during the ore mineral precipitation processes, hydrothermal fluids react with the mineral constituents of lithological units they are passing and produce hydrothermal alteration zones with distinctive mineral assemblages [7].The presence of intrusive rocks and hydrothermal alteration zones associated with polymetallic vein-type deposits provide an important guide for exploring this type of ore mineralization especially by the application of advanced satellite remote sensing data [8,9].
The Toroud-Chahshirin Magmatic Belt (TCMB) is located in Semnan province, central-north Iran (Figure 1A) and contains numerous occurrences of epithermal polymetallic vein-type mineralization (Figure 1B).The TCMB is the largest known goldfield and base metals province in central-north Iran [28,29].Hydrothermally altered zones are reportedly associated with polymetallic vein-type mineralization in this belt [29][30][31][32][33]. Propylitic, phyllic, argillic, and advanced argillic alteration and silicification occur generally with Au-Cu, Ag, and/or Pb-Zn mineralization.Therefore, mapping and identification of hydrothermal alteration mineral assemblages using ASTER satellite remote sensing data in the TCMB can be considered as a cost-effective and applicable tool for targeting and prospecting epithermal polymetallic vein-type mineralization.In this research, the Moaleman region of the TCMB was selected (Figure 1B).This region has a high potential for epithermal polymetallic vein-type mineralization, particularly anomal Cu-Au values associated with altered dacite and dacite-andesite and volcaniclastics rocks [30,34,35].Since no report on a comprehensive remote sensing investigation is available for epithermal Cu-Au exploration in the Moaleman region of the TCMB, results of an ASTER remote sensing mapping are necessary for future systematic exploration projects.The main objectives of this study are (i) to detect hydrothermal alteration mineral zones and assemblages using VNIR and SWIR spectral bands of ASTER data by application of specialized image processing techniques, including Selective Principal Component Analysis (SPCA) [36,37], Band Ratio Matrix Transformation (BRMT) [16], Spectral Angle Mapper (SAM) [38], and Mixture Tuned Matched Filtering (MTMF) [39][40][41]; (ii) to compare the results derived from SPCA and BRMT transformation algorithms and SAM and MTMF spectral mapping algorithms to map alteration minerals at the pixel and sub-pixel levels and (iii) to prospect high potential zones of epithermal Cu-Au mineralization for future systematic exploration programs in the study area.

Geology of the TCMB and Mineralization
The TCMB is situated in the central to eastern Alborz Magmatic Belt (AMB) and lies in the northern part of the Central Iran Structural Zone (CISZ) (Figure 1A).It is restricted between the E-W trending Toroud fault in the south and the E-W trending Anjilou fault in the north [29].This magmatic arc has a complex tectonic, magmatic, and stratigraphy history [29,30].Alavi et al. [43] and Alavi [44] and proposed that Torud-Chahshiran range and adjacent volcanic rocks are displaced to Eocene magmatism in the CISZ to the south.The Magmatic arc contains mainly of igneous rocks of Tertiary age, while there are also scattered outcrops of metamorphosed Paleozoic and Mesozoic rocks [45].Most of the magmatic activities in the TCMB occurred in the middle to late Eocene and have been divided into three stages from oldest to youngest, including (i) explosive volcanic activity represented by rhyolite to rhyodacite tuffs and locally andesitic lava flows, with subordinate marls, tuffaceous marlstones, and sandstones; (ii) lava flows and pyroclastic rocks of andesite, trachyandesite, and basaltic andesite composition; and (ii) subordinate dacitic-rhyodacitic rocks and hypabyssal intrusive rocks [46].

Pre-processing of the Remote Sensing Data
A cloud-free level 1B ASTER in hierarchical data format (HDF) that was acquired on 25 March 2003 was used for remote spectral analyses in this study.This image was pre-georeferenced to UTM zone 40 North projections with using the WGS-84 data.The SWIR bands re-sampled to the spatial resolution of VNIR so that all pixels of nine bands (VNIR+SWIR) with 15*15 m 2 pixel size were stacked.The level 1B data product measures radiance at the sensor without atmospheric corrections [18].Therefore, atmospheric correction is necessary before image processing analysis for converting radiance-calibrated data to apparent reflectance.The Log-residual atmospheric correction technique was applied to the ASTER image in this study.The Log Residuals calibration is capable to remove atmospheric transmittance, topographic effects, solar irradiance, and albedo effects [51,52].It produces a pseudo reflectance image, which is highly applicable for detecting absorption features related to alteration minerals.Additionally, Crosstalk correction was applied to the ASTER dataset [53].We have performed this correction by Cross-Talk correction software that is available from www.gds.aster.ersdac.or.jp.The ENVI (Environment for Visualizing Images, http://www.exelisvis.com)version 5.2 and ArcGIS version 10.3 software (Esri, Redlands, CA, USA) packages were used to process the remote sensing datasets.

Image Processing Algorithms
The main target of the specialized image processing techniques adopted in this study was to apply image processing techniques that are capable of mapping hydrothermal alteration minerals at the pixel and sub-pixel levels using VNIR+SWIR spectral bands of ASTER data.Therefore, subtle differences between altered and non-altered rocks and hydrothermal mineral assemblages could be feasible by implementing specialized image processing techniques as follows.

Principal Component Analysis (PCA)
The PCA is a multivariate statistical technique that used to reduce the dimensionally of input data and reduces the additional frequency among the data, as a result, the possibility of useful data loss is minimized.This information in terms of quantity is a very small part of the overall information content available in original bands.Spectral bands are selected that contain absorption and reflection features of alteration minerals.In this way, a new image (PC) is generated on the axes with the new coordinate system [36,37].The resulting component is more interpretable than are the original images.A PC image contains the unique contribution of eigenvector loadings (magnitude and sign) for absorption and reflection bands of alteration mineral or mineral group is able to enhance the mineral or mineral group.If the loading is positive in the reflective band of a mineral the image tone will be bright, and if it is negative, the image tone will be dark for the enhanced target mineral [37].The PCA technique has been applied to multispectral remote sensing data such as ASTER for highlighting spectral responses related to specific hydrothermal alteration minerals associated with porphyry copper mineralization [54][55][56][57][58][59].
In this research, the Selective Principal Component Analysis (SPCA) [60], known also as Directed Principal Component Analysis (DPCA) [61], was applied on VNIR+SWIR bands for mapping the specific alteration zones associated with polymetallic vein-type mineralization in the study area.The basic difference between the PCA and SPCA is that in the SPCA only a subgroup of bands is selected depending on the aims that plan to be achieved.In this study, according to the known ASTER band indices for hydrothermal alteration mineral mapping [62][63][64][65], some subsystems (specific bands) were selected for SPCA analysis.Bands 1, 2, and 4 were selected for mapping iron oxides/hydroxide minerals (Table 1A).Bands 4, 5, and 6 were designated for argillic alteration mapping (Table 1B).Bands 5, 6, and 7 were nominated to specify phyllic zone (Table 1C) and bands 7, 8, and 9 were used to map propylitic alteration zones.

Band Ratio Matrix Transformation (BRMT)
The Band Ratio Matrix Transformation (BRMT) is a semiautomatic lithological-mineralogical mapping technique, which is proposed and established for sedimentary strata discrimination [16].This analytical method extracts key spectral characteristics using VNIR and SWIR spectral bands of ASTER.Although BRMT methodology is proposed for sedimentary rocks discrimination, the effectiveness of this technique for mapping hydrothermal alteration zones in igneous rocks background is promising.The robustness of BRMT arises from a combination of statistical factors, including variance percent (V%), positive and negative correlation averages (±r), correlation averages (±rk), and contribution percent (C%).It maps key spectral characteristics using RGB color composites and rule classifier of the band ratio (BR) and band transform (BT) bands.In this study, the BRMT method was applied to VNIR+SWIR bands of ASTER data for detailed mapping of hydrothermal alteration zones in the study area.Table 2 shows positive and negative correlation and contribution percent of the band ratios n1-n36 used for BRMT transformation.Table 3 shows eigenvalues, variance percent (%V), positive and negative correlation averages (+rk > 0.1 and −rk < 0.1), and correlation percent of BT1 to BT10, which hold maximum spectral properties extracted from the image.The selected BT bands containing specific spectral properties (negative and positive contribution percent >3%) are listed in Table 4.  4. The selected band transform bands (BTs) contain specific spectral properties (negative and positive contribution percent > 3%).

Contribution > 3% BT
The MTMF is a partial unmixing, hybrid method based on the combination of well-known signal processing methodologies and liner mixture theory [39][40][41].MTMF consist of two phases, an MF calculation for abundance estimation and a mixture tuning calculation for the identification and rejection of false positives .In this study, the endmembers were extracted from the image using the Minimum Noise Fraction (MNF), Pixel Purity Index (PPI), and n-dimensional visualization techniques [69,70].The MNF method was accomplished to separate noise from the data.The PPI detects the highest purity pixels in the image.The n-dimensional visualization was used to locate, identify, and cluster the purest pixels and the most extreme spectral responses (endmembers) in the VNIR+SWIR ASTER dataset.The threshold of 2.5 was applied to PPI.The output of PPI represents as bright pixels (more purity) and dark pixels (less spectral purity).After applying the n-dimensional visualization method, 10 n-D classes (endmembers) were extracted, which indicate distinctive absorption features related to alteration minerals.Figure 2 shows the extracted endmember spectra (n-D classes) for the study area.Comparison of the absorption characteristics of the extracted endmembers (n-D classes) with the USGS spectral library is considered for identification of alteration minerals.

Spectral Angle Mapper (SAM)
The SAM is a classification method that classifies the pixels based on similar spectral properties with reference spectra [38].It uses endmember spectra that can be extracted directly from the image or spectral library.This process of calculating the spectral angle between the reference spectra and image determines the similarity of these two groups.The outputs of the algorithm are angles between zero and one.The large angle shows less similarity and the smaller angle more similarity [71,72].In this analysis, the SAM was applied to VNIR+SWIR ASTER bands with default value 0.1 (radians) as a threshold, and the spectra for running the algorithm were obtained from the USGS library as reference spectra [73].Figure 3 shows end-member minerals extracted from the USGS library, including hematite, jarosite, montmorillonite, illite, muscovite, and chlorite.

Fieldwork Data, Laboratory Analysis and Verification
For collecting field data, 10 zones showing the high spatial distribution of hydrothermal alteration minerals in the Moaleman region were selected.Global positioning system (GPS) survey was acquired using a GPS (Garmin eTrix 10, Nanjing Sifang Mapping Equipment Ltd., Jiangsu, China) with an average accuracy of 5m to find the exact location of the selected zones in the ASTER scene.Field photos were taking from alteration zones.Seventy rock samples were collected from the alteration zones and lithological units for laboratory analysis, including thin section preparation, X-ray diffraction (XRD) analysis, and analytical spectral devices (ASD) spectroscopy.The XRD analysis was applied using Advance-D8 XRD Bruker model at Central Laboratory of Damghan University, Damghan, Iran.The exposure time of powder samples (1 g) was about one hour by a monochromatic ray in the wavelength of 5.4 Å and ranging angles between 5 to 70 degrees.The step of diffraction was set as 0.2 degrees to guarantee detection of clay minerals which are detectable in low angles.The copper anode is used to generate X-rays with a voltage of 25 kilovolts (kV).The spectra of the representative samples from altered zones were measured using a FieldSpec3®spectroradiometer operating in the 0.35-2.5 µm spectral range at the Department of Ecology, Institute of Science and High Technology and the Environmental Sciences Graduate University of Advanced Technology, Kerman, Iran.The fore-optics were at a small distance from the surface under observation.Spectralon of Labsphere which is made of polytetrafluoroethylene (PTFE) and sintered halon G-80 was used as a white reference panel.About 10 measurements were performed per spot.Moreover, the Kappa coefficient was calculated using a Matlab code developed by Askari et al. [74] for SAM and MTMF results for accuracy assessment (Table 5A,B).

Alteration Mapping Results Derived from ASTER Data
Considering statistical results derived from SPCA (Table 1A-D), it evident that hydrothermally altered rocks were mapped in the SPC images that contain a unique contribution of components (eigenvectors) related to spectral characteristics of the alteration minerals.Table 1A shows eigenvector values for mapping iron oxide/hydroxide minerals in the study area.Analyzing the eigenvector loadings shows that the SPC2 contains a strong to moderate contribution of band 1 (−0.42) and band 2 (−0.39) with negative signs and a strong contribution of band 4 (0.81) with a positive sign.Iron oxide minerals have absorption features in bands 1 and 2 and reflectance features in band 4 of ASTER, respectively [54,62,75].Therefore, the SPC2 is able to enhance oxide/hydroxide minerals as bright pixels due to opposite signs of the eigenvector loadings in the absorption bands (negative signs in bands 1 and 2) and reflection band (positive sign in band 4) (Figure 4A).
Looking at the eigenvector loadings in Table 1B for mapping argillic alteration indicates that the SPC3 has a strong contribution of band 4 (−0.61) and band 5 (−0.73) with negative signs and a strong contribution of band 6 (0.67) with a positive sign, respectively.Kaolinite and alunite are main constituents of argillic alteration that normally exhibit Al-OH absorption features in bands 5 and 6 of ASTER [63].Al-OH minerals show maximum reflectance features in band 4 of ASTER that covers the spectral region of 1.6 µm [64].Thus, the argillic alteration zone appears as dark pixels in the SPC3 image because of a negative sign in band 4 (reflection band).Dark pixels in the SPC3 image were inverted to bright pixels by multiplication to −1 (Figure 4B).
Table 1C shows the eigenvector loadings for mapping phyllic alteration zone.The SPC3 shows strong eigenvector loadings for band 5 (−0.69) and band 6 (0.71) with opposite signs, while band 7 (0.02) has a very small contribution in the SPC3.The phyllic zone composed of illite/muscovite (sericite) produces an intense Al-OH absorption feature at band 6 of ASTER [52].Phyllic alteration zone in the study area manifests in bright pixels in the SPC3 image (Figure 4C).Considering the eigenvector loadings in Table 1D for identification of propylitic alteration zone in the study area, the SPC3 contains strong loadings of band 8 (−0.81) and moderate contribution of bands 7 (0.35) and 9 (0.45) with opposite signs, respectively.The propylitic alteration zone consisting of epidote, chlorite, and calcite display strong absorption features in band 8 of ASTER [64].For that reason, the SPC3 image depicts propylitic alteration zone as bright pixels in the study area (Figure 4D).
Propylitic, phyllic, argillic, and advanced argillic alteration zones were reported as dominated alteration zones with epithermal polymetallic vein-type mineralization in the study area [30].The surface distribution pattern of iron oxide/hydroxide minerals, argillic alteration zone, phyllic zone, and propylitic zone is almost similar and mostly concentrated in the central and southwestern parts of the study area (see Figure 4A-D).However, argillic zone and iron oxide/hydroxide minerals show more similar spatial distribution and strong surface abundances compare to phyllic and propylitic zones.Accordingly, a Red-Green-Blue (RGB) color composite was assigned to the SPC3 of argillic alteration, SPC3 of phyllic alteration, and SPC3 of propylitic alteration images for providing a false color-based classification image of the detected pixels. Figure 5 shows the resultant image for the study area.Argillic alteration zone appears in red and yellow colors, it is evident that the red zone can be considered as the advanced argillic and yellow zone is a combination of Argillic and phyllic alteration zones (Figure 5).The phyllic alteration zone is represented in green and cyan colors.The mixture of phyllic and propylitic alteration zones depict as cyan color.Propylitic alteration zone manifests as blue color (Figure 5).Comparison with the geological map of the study area (see Figure 1B), most of the ore mineralizations are concentrated in argillic and advanced argillic alteration zones (see Figure 5), which are associated with dacite and dicite-andesite, spilitic basalt, trachyandesite basalt, and quartz trachyandesite lithological units.Several advanced argillic (red mixed with yellow pixels) zones are observable in the southwestern and northeastern parts of the study area, which could be considered as prospective zones.Considering of statistical results calculated for the BRMT algorithm (Tables 2-4), the eigenvalues, Vi%, positive and negative correlation averages (+r k > 0.1 and −r k < 0.1) for the BT1, BT2, and BT3 are considerable (Table 3).High number of eigenvalues, Vi%, and +r k % and −r k % were estimated in the BT1, BT2, and BT3.The BT1 contain the highest eigenvalue (0.123) Vi% (83.24), +r k % (20.7), and −r k % (29.48).The BT2 shows eigenvalue of 0.00928, Vi% of 6.28, +r k % of 12.70, and −r k % of 13.36.The BT3 has an eigenvalue of 0.00643, Vi% of 4.35, +r k % of 8.51, and −r k % of 12.08 (Table 3).It shows that these BTs contain most of the spectral information that was extracted by the BRMT algorithm from the image.Therefore, the BTs were used for producing RGB false color composite to reveal the most spectrally dominated hydrothermal alteration zones and lithological units in the study area (Figure 6).With reference to SPCA results, it is discernable that argillic, phyllic, and advanced argillic alteration zones are most spectrally dominated alteration zones in the study area, which appear as magenta color in Figure 6.These alteration zones are typically concentrated in the central, southwestern, and northeastern parts of the study area, which contain mineralogically interesting zones for ore exploration.The analysis of the negative and positive contribution percent > 3% for BTs (see Tables 2 and 4) indicates that the BTs contain a negative contribution holding more specific spectral properties attributed to lithological units and hydrothermal alteration zones.Therefore, the BTs containing a negative contribution > 3% (Table 4), such as n10, n5, n19, n20, n6, n7, n22, n2, n17, n13, and n4, were used for running the rule image classifier.Figure 7 shows the resultant BRMT classification map for the study area.Eight classes (C1 to C8) were identified.The class C1 (red pixels) can be considered a moderate propylitic alteration zone that is combined with unaltered volcaniclastic rocks.This class (C1) mainly covers the eastern and northeastern parts of the study area.The class C2 (blue pixels) includes the advanced propylitic alteration zone, which is generally concentrated in the central and southwestern part of the study area (Figure 7).The advanced argillic alteration zone is depicted in class 3 (green pixels), which covers typically central and western parts and many other small exposures in the whole of the study area.Class C4 (cyan pixels) is a combination of argillic, iron oxide, and propylitic mineral assemblages and sedimentary rocks that weathered and transferred to the alluvial fan.Class 5 (brown pixels) is an accumulation of iron oxide minerals that are within the highly altered parts of argillic and phyllic zones.Sandstone and alluvium can be considered in class 6 (light yellow pixels), while class 7 (magenta pixels) might be an admixture of some weathered rocks of classes 1, 2, and 3. Class 8 (mustard pixels) represents unknown units that may consist of some weathered and transferred sedimentary rocks.By using the geological map of the study area (see Figure 1) as a reference, it is obvious that the most of the reported ore mineralizations in the study area are located in the interior of class 3 (advanced argillic alteration zone) of the BRMT classification map (Figure 7).Thus, some perspective zones (green pixels) could be considered in the southwestern and northeastern sectors.Within the VNIR+SWIR interval, for identification of the n-D classes (#1 to #10) derived from the n-Dimensional visualization analysis technique (see Figure 3), diagnostic absorption features related to electronic process intensities and crystal field transitions due to Fe2+, Fe3+, and REE transitional metals in the VNIR [76,77] and vibrational overtones of the fundamental Al-OH/Fe-OH/Mg-OH and CO3 in the SWIR [78,79] should be taken into consideration.Generally, dominant spectral patterns correspond to the most abundant minerals or spectrally dominant mineral groups.Subordinate spectral patterns produce spectral variability due to spectrally less active or less abundant phases in the rock [80].Therefore, the wavelength shift of the distinctive absorption features directly or inversely depends on the abundance of spectrally active mineral groups within the rocks.Comparison of the absorption characteristics of the n-D classes extracted for the study area with the USGS spectral library indicate some spectrally distinctive mineralogical phases (see Figure 2).The n-D class #1 does not contain any distinctive absorption features related to alteration minerals and could be considered as an unaltered/unknown class.The n-D class #2 represents diagnostic absorption features related to Fe3+ transitional metals, coinciding with bands 2 and 3 of ASTER.Thus, it contains iron oxide/hydroxide absorption features, which are attributable to hematite.There are no recognizable absorption features in the n-D class #3, which can be classified as unaltered/unknown.
Analyzing the n-D class #4 and n-D class #5 indicates spectral signatures attributed to jarosite, which correspond with bands 2, 3, and 7 of ASTER (see Figure 2).In fact, band 7 of ASTER is able to detect Fe-OH absorption features caused by jarosite and/or Fe-muscovite [81].So, the n-D class #4 might contain spectral signatures of other mineral groups with subordinate spectral patterns.The n-D class #6 represents chlorite due to major Mg, Fe-OH absorption properties in band 8 of ASTER (2.30-2.360µm).A major Al-OH absorption feature positioned in band 6 of ASTER is obvious in the n-D class #7, which reflects the spectral signatures of illite (see Figures 2 and 3).The n-D class #8 might be considered as mixed spectral signatures of illite, muscovite, and montmorillonite.The n-D class #9 is characterized by a strong absorption feature centered at 2.20 µm (coinciding with band 6 of ASTER), which is attributable to muscovite/montmorillonite (see Figures 2 and 3).The n-D class #10 exhibits mixed spectral signatures of hematite and jarosite (see Figure 2).
Fraction images of end-members (the n-D classes #1 to #10) resulting from MTMF analysis appear as a series of greyscale rule images (one for each extracted end-member) for the study area.High digital Number (DN) values (bright pixels) in the rule image represents the subpixel abundance of the end-member mineral in each pixel and map its location.The pseudo-color ramp of greyscale rule images was generated to illustrate high fractional abundance (high DN value pixels) of end-members (the n-D classes) in the study area (Figure 8).It helps to distinguish the contrast between subpixel targets and surrounding areas.This contrast expresses the fractional abundance of the target mineral present in the rule image.It should be noted here that unaltered/unknown class (the n-D class #1 and n-D class #3) was omitted during the production of Figure 8.
Figure 8 shows the pseudo-color ramp of the n-D class fraction images derived from the MTMF algorithm for the study area.Considering the fractional abundance of detected endmember minerals, muscovite, montmorillonite, and illite spectrally governed the study area, while hematite, jarosite, and chlorite have less contribution in total mixed spectral characteristics.Spatial distribution of the minerals with similar spectral features such as muscovite and montmorillonite (absorption features near 2.20 µm) and hematite and jarosite (absorption features near 0.48 µm to 0.85 µm) is comparable.It derives from the fact that ASTER multispectral signatures contain some limitations for detecting subtle differences between analogous absorption characteristics especially when mixture occurs.However, the ASTER VNIR and SWIR bands are sufficiently positioned to detect spectral feature differences between important key minerals [82].Referencing geological map of the study area (see Figure 1), muscovite/montmorillonite/illite mineral assemblages are typically concentrated in the central, southwestern, and northeastern parts associated with dacitic and andesitic units.Nevertheless, hematite/jarosite/chlorite mineral assemblages are present in low surface abundance (Figure 8).The high concentration of iron oxide/hydroxide minerals is noticeable in the western and southwestern part of the study area associated with trachyandesite basalt units.In this part, the fractional abundance of muscovite/montmorillonite/illite is not high (Figure 8).Accordingly, the southwestern and northeastern parts of the study area contain a number of mineralogically interesting zones for ore mineralizations and holding high potential zones for future systematic exploration program.
SAM classification technique was implemented for mapping the spatial distribution of prevalent minerals such as hematite, jarosite, montmorillonite, illite, muscovite, and chlorite in the alteration zones.Hematite contains absorption features in bands 1 and 3 of ASTER.Jarosite shows absorption characteristics in bands 1, 3, and 7 of ASTER.Montmorillonite displays weak absorption features in band 5 and strong absorption features in band 6 of ASTER.Illite exhibits distinctive absorption properties in bands 5 and 6 of ASTER.Muscovite has absorption features in bands 1, 2, and 6 of ASTER.Chlorite contains absorption characteristics in bands 1, 2, and 3 and diagnostic absorption features in band 8 of ASTER (see Figure 3).Therefore, these spectral absorption signatures of alteration minerals in the VNIR+SWIR bands of ASTER can be used for detecting subtle differences between alteration minerals by running the SAM algorithm.Figure 9 shows the SAM classification map for the study area.Detailed spatial distribution of the selected minerals was mapped within the alteration zones.Montmorillonite, muscovite, and illite are the most dominated minerals in the argillic, phyllic, and advanced argillic alteration zones.However, hematite and jarosite demonstrate moderate surface distribution and chlorite has very low abundance in the argillic, phyllic, and advanced argillic alteration zones.High concentration of chlorite is mapped only in the propylitic alteration zone that is associated with hematite and jarosite.Several concentrations of muscovite and illite are mapped in the southwestern and northeastern parts of the study area, which previously deliberated as high potential zones for ore mineralizations (Figure 9).

Fieldwork, Laboratory Analysis and Verification Results
Comprehensive geological fieldworks were carried out in the study area especially in the detected hydrothermal alteration zones.A number of prospects and mineralogically interesting zones were visited.The precise location of the alteration zones was recorded using GPS survey.The lithological units and alteration zones were checked and samples were collected.Part of the hand specimen was split off for a thin section and the rest was crushed to a grain size of less than 2 mm for XRD analysis.In particular, the study area contains a significant concentration of advanced argillic, argillic, and phyllic alteration zones (Figure 10A-C).However, hematite-rich altered oxidized zones and propylitic alteration zones cover also large parts of the study area (Figure 10D-F).Typically, hematite-rich altered oxidized zones and phyllic alteration zones show the close spatial relationship in many parts of the study area (See Figure 10D).
Petrographic studies were carried out on thin sections of the collected rock samples.Thin section observations indicate the transformation of primary silicate minerals (feldspars) to secondary altered minerals (sericite, clay minerals, calcite and epidote) (Figure 11A-F).Plagioclase is typically replaced by sericite, clay minerals, calcite, epidote, and quartz in the most of alteration zones.Veins and subhedral grains of opaque minerals are more observable in the thin sections of advanced argillic, argillic, and phyllic alteration zones (see Figure 11A-D).In the propylitic zone, the original minerals are fully replcaed by secondary minerals (calcite and epidote) (see Figure 11E,F).Minerals identified in the collected rock samples from hydrothermal alteration zones using XRD analysis include montmorillonite, illite, goethite, hematite, muscovite, albite, orthoclase and quartz in advanced argillic and argillic zones (Figure 12A,B); muscovite, illite, hematite, magnetite, albite, epidote, calcite, montmorillonite, and quartz in the phyllic zone (Figure 12C,D); epidote, calcite, chlorite, albite, anorthite, and quartz in the propylitic zone (Figure 12E,F).The ASD spectroscopy is sensitive to detect the presence of alteration minerals with strong absorption features in the mineralogically interesting zones.Figure 13 shows the average reflectance spectra of phyllic, gossan (hematite-rich altered oxidized zone), argillic and propylitic rock samples.The reflectance spectra from phyllic samples show three prominent absorption features near 1.40 µm, 1.90 µm, and 2.20 µm, due to vibrational overtone and combination tones involving OH-stretching modes [83,84].The absorption features near 1.40 µm and 1.90 µm in the phyllic samples can be attributed to OH stretches occurring at about 1.4 µm and the combination of the H-O-H bend with OH stretches near 1.90 µm [78].The feature near 2.20 µm is due to a combination of the OH-stretching fundamental with Al-OH bending mode [84,85].These spectral characteristics exhibit similarities to the spectra of muscovite (dominant absorption features located around 2.20 µm), which is a main alteration mineral in the phyllic zone.
Strong absorption features near 0.50 µm and 1.0 µm in the gossan samples are due to electronic transitions in iron ions (ferric and ferrous ions) [79,83].Moreover, the feature near 2.27 µm is attributed to a combination of OH stretch and Fe-OH bend [78].The two strong absorption features at about 0.50 µm and 1.0 µm are normally considered for hematite and goethite in the VNIR regions [86].Jarosite has a diagnostic absorption feature at 2.27 µm [78].The OH and H2O vibrational bands near 1.40 µm and 1.90 µm are also commonly seen in iron oxides/hydroxides spectra [86].Thus, hematite, goethite, and jarosite are major alteration mineral constituents in the gossan zone.The reflectance spectra of argillic samples display two overall absorption features at about 1.40 µm and 1.90 µm (due to the OH and H2O vibrational bands) and also consist of maximum absorption near 2.17 µm and 2.20 µm (Figure 13).Montmorillonite, kaolinite, illite, and alunite show spectral characteristics consisting of a major absorption feature at around 2.20 µm associated with a secondary feature between 2.16 and 2.18 µm, which are related to stretching vibration of the inner and outer hydroxyl groups (Al-OH bending mode) [85,87].The spectra from Propylitic samples show also characteristics absorption features near 0.50 µm and 1.0 µm (due to ferric and ferrous iron ions) and the OH and H2O vibrational bands (about 1.40 µm and 1.90 µm) (Figure 13).The absorption feature near 2.35 µm in the samples is due to a combination of OH-stretching fundamental with the Mg-OH bending mode, and the feature near 2.50 µm can be attributed to combination and overtone bands of CO3 fundamentals, respectively [78,83].Chlorite and epidote show distinctive absorption features at around 2.35 µm [85].Carbonates (calcite, aragonite, and dolomite) have diagnostic absorption features between 2.30 and 2.50 µm [88].Comparing the ASD spectroscopy results with XRD analysis indicates most diagnostic spectral features in the phyllic zone derived from muscovite, in the argillic zone resulting from montmorillonite and illite, and in the propylitic zone associated with chlorite, epidote, and calcite.Considering the XRD and ASD analysis, iron oxide/hydroxide minerals are associated with the alteration mineral assemblages in advanced argillic and argillic, phyllic, and propylitic alteration zones.Table 5A,B shows the accuracy assessment results for the MTMF and SAM methods based on GPS survey collected during fieldwork.Analysis of the statistical factors indicate that the overall accuracy and Kappa Coefficient for MTMF and SAM are 95.7 and 0.93 (see Table 5A) and 84.4 and 0.78 (see Table 5B), respectively.The assessment emphasizes that both MTMF and SAM methods provide accurate mapping results in the study area.However, the MTMF method was capable of providing more accurate results for mapping the surface distribution of hydrothermal alteration minerals.

Conclusions
ASTER remote sensing data processing provides maps of surface alteration mineralogy for the Moaleman region of the TCMB, which illustrate several high potential zones of polymetallic vein-type mineralization.The image processing algorithms implemented in this analysis, including the SPCA, BRMT, SAM, and MTMF provided a consistent way to identify and map hydrothermal alteration zones in the study area.The SPCA, BRMT spectral transformation algorithms used in this study were capable of mapping the surface distribution of hydrothermally altered rocks and lithological units.The surface distribution pattern of iron oxide/hydroxide minerals, argillic alteration zone, phyllic zone, and propylitic zone was mapped in the study area using a unique contribution of eigenvector loading in the SPC3.Iron oxide/hydroxide minerals and propylitic zones showed similar spatial distribution, while strong analogous surface distribution patterns were more obvious for argillic, phyllic, and iron oxide/hydroxide minerals zones.The BT1, BT2, and BT3 hold most of the spectral information in the image, which was extracted by the BRMT algorithm.The most spectrally dominated hydrothermal alteration zones, including argillic, phyllic, and advanced argillic alteration zones were revealed.Additionally, the BTs contain a negative contribution > 3% providing comprehensive information as eight different lithological/alteration/weathered classes (C1 to C8) for the study area.Therefore, more alteration/lithological information can be obtained from the BRMT algorithm compare to the SPCA.
Ten endmembers (n-D classes) were extracted using n-dimensional visualization method, which indicates distinctive absorption features related to alteration minerals in the study area.Spectrally distinctive mineralogical phases were identified, such as hematite, jarosite, chlorite, illite, muscovite, and montmorillonite, which may have some spectral signatures of subordinate mineralogical phases.Fraction images of end-members derived from the MTMF algorithm showed the similar sub-pixel distribution for minerals contain analogous spectral features, which can be attributed to some limitations of ASTER data for detecting subtle differences between equivalent absorption characteristics especially when mixture occurs.Detailed spatial distribution of prevalent minerals including hematite, jarosite, montmorillonite, illite, muscovite, and chlorite in the alteration zones was mapped using the SAM algorithm.Results indicate that montmorillonite, muscovite, and illite are the most dominated minerals in the argillic, phyllic, and advanced argillic alteration zones.However, hematite and jarosite demonstrate moderate surface distribution and chlorite has very low abundance in these alteration zones.A high concentration of chlorite was mapped only in the propylitic alteration zone, which was also associated with hematite and jarosite.The MTMF results confirmed the presence of alteration minerals and their spatial distribution at the subpixel level in the study area, while SAM mapped subtle differences between alteration minerals in the alteration zones.The accuracy assessment results show the MTMF method is proficient to be responsible for more accurate outcomes for mapping the surface distribution of hydrothermal alteration minerals.Hydrothermal alteration zones and minerals that mapped using ASTER data processing demonstrate good correspondence with the results of field survey, petrography, XRD analysis, and spectral measurements acquired by ASD spectrometer.A number of high potential zones of epithermal polymetallic vein-type mineralization were identified in the study area, particularly in the northeastern and southwestern sectors, which can be considered for future systematic exploration programs.The approach developed in this study can be used for the exploration of epithermal polymetallic vein-type mineralization in other base metals provinces in semi-arid regions around the world.

Figure 2 .
Figure 2. The endmember mineral extracted from the ASTER dataset using of n-Dimensional visualization method.The reference spectra were resampled to the response function of ASTER bands.The ASTER band center positions are also shown.

Figure 3 .
Figure 3. Reference endmember spectra of hematite, jarosite, montmorillonite, illite, muscovite, and chlorite obtained from the USGS spectral library for mapping alteration zones in the study area.The reference spectra were resampled to the response function of ASTER bands.The ASTER band center positions are also shown.Table 5. (A) The accuracy assessment for the Mixture Tuned Matched Filtering (MTMF) method based on GPS survey collected during the field study.(B) The accuracy assessment for the Spectral Angle Mapper (SAM) method based on GPS survey collected during the field study.

Figure 4 .
Figure 4. SPC images derived from SPCA analysis.(A) SPC2 image showing oxide/hydroxide minerals as bright pixels; (B) SPC3 image showing the argillic alteration zone as bright pixels; (C) SPC3 image showing the phyllic alteration zone as bright pixels; and (D) SPC3 image showing the propylitic alteration zone as bright pixels.

Figure 5 .
Figure 5. RGB false color composite of the SPC3 of argillic alteration (R), SPC3 of phyllic alteration (G), and SPC3 of propylitic alteration (B) images covering the study area.

Figure 6 .
Figure 6.RGB false color composite of the BT1 (R), BT2 (G), and BT3 (B) derived from the BRMT algorithm for the study area.

Figure 7 .
Figure 7. BRMT classification map for the study area.

Figure 9 .
Figure 9. Spectral Angle Mapper (SAM) classification map for the study area.

Figure 10 .
Figure 10.Field photographs of the hydrothermal alteration zones in the study area.(A) A panoramic view of argillic alteration zones; (B) a view of the advanced argillic alteration zone; (C) a regional view of the phyllic alteration zone; (D) a regional view of hematite-rich altered oxidized zones in association with phyllic alteration zones; (E) a view of the propylitic zone; (F) a close up of a specimen from the propylitic alteration zone.

Figure 12 .
Figure 12. Results of XRD analysis shows minerals of representative samples collected from (A) and (B) advanced argillic and argillic zones; (C,D) phyllic zone; and (E,F) propylitic zone.

Table 1 .
Eigenvector loadings matrix calculated using Selective Principal Component Analysis (SPCA)

Table 2 .
Positive and negative correlation (±r) and contribution percent (C%) of the band ratios n1-n36 used for Band Ratio Matrix Transformation (BRMT).