Selection of Landsat-8 Operational Land Imager (OLI) Optimal Band Combinations for Mapping Alteration Zones

Abstract

In typical alteration extraction methods, e.g., band math and principal component analysis (PCA), the bands or band combinations unitized to extract altered minerals are usually selected based on empirical models or previous rules. This results in significant differences in the alteration of mineral mapping even in the same area, thus greatly increasing the uncertainty of mineral resource prediction. In this paper, an intelligent alteration extraction approach was proposed in which an optimization algorithm, i.e., a genetic algorithm (GA), was introduced into the PCA; this approach is termed GA-PCA and is used for selecting the optimized band combinations of mineralized alterations. The proposed GA-PCA was employed to map iron oxides and hydroxyl minerals using the most commonly adopted multispectral data, i.e., Landsat-8 OLI data, at the Lalingzaohuo polymetallic deposits, China. The results showed that the spectral characteristics of GA-PCA-selected OLI band combinations in the research area were beneficial for enhancing alteration information and were more capable of suppressing the interference of vegetation information. The mapping alteration zones using the GA-PCA approach had a higher agreement with known ore spots, i.e., 25% and 33.3% in ferrous-bearing and hydroxyl-bearing deposits, compared to the classical PCA. Furthermore, two predicted targets (not shown in the classical PCA results) were precisely obtained via analyzing the GA-PCA alteration maps combined with the ore-forming geological conditions of the mine and its tectonic characteristics. This indicated that the intelligent selection of mineral alteration band combinations increased the reliability of remote sensing-based mineral exploration.

1. Introduction

The extraction of altered mineral information via multispectral remote sensing data with comprehensive analysis has become an essential method for prospecting in ore-bearing regions [1,2]. However, mineralogical alteration appears as weak anomalies in multispectral remote sensing images, often obscured by background surface features. The fundamental key to extracting alteration information and zones from remote sensing data has the following four aspects: (i) determining the unique spectral characteristics of detected mineral(s) in VNIR and SWIR regions [3,4]; (ii) eliminating the impact of vegetation, water bodies, and other interfering features [5]; (iii) performing the calculation and selection of band ratios (BR) and combinations for extracting different types of alteration information [6,7,8,9]; and (iv) utilizing alteration extraction methods for mapping alteration zones [10,11,12,13].

The multispectral scanner Landsat dataset was the first dataset utilized in the mapping of iron oxides in the 1980s [14]. According to the characteristics of the visible and near-infrared (VNIR) and shortwave infrared (SWIR) bands of Landsat data, the Landsat thematic mapper (TM) images, especially Landsat-8 data, have been widely used and are highly recommended for extracting areas of iron oxides and hydroxyl minerals that might be related to hydrothermal alteration zones of ore deposits [15,16,17]. To extract poorly exposed mineralogical alteration using Landsat-8 data, various BR and combinations are applied based on the geological background and spectral curves of minerals in the study regions. Tangestani and Moore [6] used three principal component analysis (PCA) methods to compare and analyze the porphyry copper ore alteration zones in the Meiduk area, Kerman, Iran. Timothy et al. [18] proposed a feature principal component selection method based on the spectral characteristics of typical features. Salem et al. [19] applied BR and PCA to ETM data for extracting alteration information and distinguishing between alteration zones associated with ultramafic rock and those associated with leucocratic granitic rock, in which the band ratios were b7/b5, b5/b4, and b3/b1. Son et al. [20] applied BR to extract argillic, lapillic, and sericitized information using ASTER Level 2B data. Ibrahim et al. [21] used PCA and BR to extract alteration information from pseudo-color composited and spatially filtered processed data of ETM+ and ASTER images. Zoheir et al. [22] applied BR and PCA to analyze the ASTER and Sentinel-2 datasets for lithological and hydrothermal alteration mapping to discriminate ophiolites, island arc rocks, and Fe oxides/hydroxides and carbonate alteration zones. Traorea et al. [23] integrated BR and PCA techniques with a fuzzy logic modeling method that highlighted hydrothermal alteration minerals and high perspective zones for alluvial gold exploration. Ge et al. [24] created a fused image with three band ratios (b6/b1, b6/b8A, and (b6 + b7)/b8A) for extraction of hematite + goethite, hematite + jarosite, and the mixture of iron-bearing minerals using Sentinel-2 datasets.

From the current research, mineral spectral features were a primary critical factor in extracting mineralization alteration information from multispectral remote sensing data. Mia et al. [25] applied four band ratios in Landsat-7 ETM+, i.e., the Sabins’ ratio (b5/b7, b3/b1, and b3/b5), Abram’s ratio (b5/b7, b4/b5, and b3/b1), Kauffmann ratio (b7/b4, b3/b5, and b5/b7), and Chica–Olma ratio (b5/b7, b5/b4, and b3/b1) for enhancing the alteration of iron and clay minerals. The common band ratios of ferrous iron and ferric iron (Fe³⁺/Fe²⁺ iron oxides) used in the literature were b7/b5, b3/b4, b6/5, b4/b2, and b4/b6 in Landsat-8 data [2,26]. Existing methods were based on an empirical model for band combination, followed by the use of principal component analysis and its improved hybrid form to enhance remote sensing images [27]. However, it is difficult to determine the optimal band combination due to the diversity and complexity of regional mineralization. Therefore, the subjective selection of existing band combinations leads to great variability in the mineralized alteration information extraction, which is not conducive to subsequent mineral prediction.

A genetic algorithm (GA) is an efficient heuristic search algorithm for optimal solutions, which is developed by drawing on the principles of genetics and natural selection. As an effective global optimization search algorithm, it is characterized by its generality and robustness. In recent years, it has been widely used in band selection based on multispectral [28,29,30,31] and hyperspectral remote sensing images [32,33,34,35]. The band selection method based on a genetic algorithm can solve the problem when the number of band combinations is large and difficult to traverse in the selection process for the best feature combination.

In this study, we proposed an intelligent alteration extraction approach in which the optimization algorithm, i.e., a GA, was introduced into the PCA and termed GA-PCA, and was used for selecting the optimized band combinations of mineralized alterations. The proposed GA-PCA was employed to map iron oxides and hydroxyl minerals using the most commonly adopted multispectral data, i.e., Landsat-8 OLI data in the study area of the Lalingzaohuo polymetallic deposits, China. The main contributions of this study are as follows:

• (1). A multispectral remote sensing feature set was established. Aiming to extract mineralized alteration information from multispectral remote sensing images, the spectral bands and BR were integrated by considering the regional geological background and mineral spectral characteristics.

• (2). A GA-PCA-based alteration extraction approach was presented. By introducing a genetic algorithm into the PCA, the mineralized alteration objective function and fitness function were defined to effectively select the optimal band combinations that were appropriate for alteration information extraction.

• (3). The information about iron oxides and hydroxyl mineralized alterations was effectively mapped with the selected best band combination in the study area. Furthermore, two predicted targets were precisely obtained via analyzing the alteration maps with a combination of the ore-forming geological conditions of the mine and the tectonic characteristics.

The objective of the present study was to map the alteration zones accurately and efficiently using a new alteration extraction approach via selecting the optimal band combination based on the Landsat-8 OLI satellite data of the Lalingzaohuo polymetallic deposits, China. Section 2 introduces the study area and the dataset used in this paper. In Section 3, the proposed GA-PCA approach is described in detail. The results and discussion are shown in Section 4 and Section 5. Our conclusions are presented in Section 6.

2. Study Area and Data

2.1. Geological Setting of Study Area

The study area included the western part of the Kunlun orogenic belt and the southern part of the Qaidam Basin, with a geographic location of 93°12′–93°30′E and 36°25′–36°40′N. The geotectonic unit of the study area belonged to the Qimantag ophiolite melange zone [36]. The landform of the study area was the Gobi Desert with sparse vegetation and exposed bedrocks, in which the mechanical disintegration of rocks was strong. The deep and large Kunbei faults in the south-central part of the region control the distribution of magma and rock zones and the distribution of mineral resources. The main stratigraphy is the Paleoproterozoic Baishahe formation (Pt₁b). There were porphyritic and magmatic-hydrothermally intruded sillimanite in the stratigraphy, which is a metamorphic rock formation. The Ordovician–Silurian Tanjianshan group (OSt) is a shallowly metamorphosed volcanic rock. The Devonian Yakshan group (D₁m) is a terrestrial sedimentary clastic volcanic rock. The Carboniferous Shiguaizi Formation (C1s) and Carboniferous Dagangou formation (C₁dg) are shallow marine clastic and carbonatite rocks. The Triassic Elasan formation (T₃e) is a set of terrestrial volcanics consisting of medium-basic volcanics and clastics. The diorite-type iron polymetallic deposits in the area are mainly produced in the sandy bioclastic tuffs of the Carboniferous Dagangou formation (C₁dg). The geological map of the study area is shown in Figure 1.

The mineralized localities or “spots” are listed in Table 1. Most of the known minerals in the study area are iron polymetallic deposits, such as Hongshan iron ore, Xiaozaohuo iron ore, Lalengzaohuo south iron ore, Hashatuo iron occurrences, and the iron polymetallic mines in the lower part of the Lalenggoli River. The iron ore sites are dominated by the skarn type, mainly containing magnetite, acicular iron ore, pyrite, and limonite. Molybdenum mineralization is of the skarn, porphyry, and hydrothermal types, while lead–zinc occurrences are of the hydrothermal type and copper–nickel occurrences are of the magmatic fusion type. From Table 1, one can see that two deposits were mined in the study area, including a small iron polymetallic (FM5) deposit and a large copper–nickel–cobalt (CN) deposit, and the rest of them were mineralized spots or occurrences. It should be noted that mineralized spots or occurrences are generally found during regional geological surveys or mineral prospecting and are clues for further mineral search. The mineral occurrences analyzed in this study are all already exposed. The information of these mineralized localities or “spots” was obtained from the National Mineral Properties Database published by the National Geological Library [https://www.ngac.cn/(accessed on 26 January 2023)].

2.2. Remote Sensing Data

In this study, the Landsat-8 OLI remote sensing images were used to map the alteration zones. NASA launched Landsat-8 in February 2013 as the newest mission in the Landsat series of satellites. Landsat-8 contains two sensors, an operational land imager (OLI) and a thermal infrared sensor (TIRS), which allow it to acquire high radiometric resolution (16-bit) data in the VNIR, SWIR, and TIR regions [37,38].

The OLI collects data from nine spectral bands. Seven of the nine bands are consistent with the thematic mapper (TM) and enhanced thematic mapper plus (ETM+) sensors found on earlier Landsat satellites, thus enabling compatibility with historical Landsat data while also improving measurement capabilities. Table 2 shows the iron-stained and hydroxylated minerals corresponding to the spectral characteristics of rocks and minerals, as well as their corresponding OLI bands [39,40]. Therefore, OLI data are more suitable for extracting absorption and reflection features of altered mineral spectra.

The Landsat-8 OLI scene was obtained from the Geospatial Data Cloud of China [http://www.gscloud.cn/(accessed on 18 January 2023)]. The path number and the row number were 137 and 35, respectively, and the acquired date was 17 November 2017. The image size was 928 × 856 with 6 bands (blue, green, red, NIR, SWIR1, and SWIR2). The selected OLI images of the study area had fewer clouds, a small amount of vegetation cover, and more exposed rock strata, which were suitable for remote sensing mineralization and alteration information extraction. In this study, the selected OLI images were preprocessed with radiometric calibration, atmospheric correction, and image cropping via ENVI. Then, the effects of vegetation, water, and snow were defused from the OLI images by using the normalized difference vegetation index (NDVI) [43], the modified normalized difference water index (MNDWI) [44], and the normalized difference snow index (NDSI) [45], respectively.

3. Methods

3.1. Multispectral Remote Sensing Feature Set

Let $X=\left\{x_1,x_2,\dots ,x_b\right\}$ denote the original multispectral image dataset, where $x_i$ is one of the spectral bands and b is the number of bands in the dataset X. The number of pixels in one spectral band is $N$.

$P$ is defined as the set of sums between any two bands in the dataset X:

(1)$P=\left\{x_i+x_j|x_i\in X,x_j\in X,i\ne j\right\},$

where $p$ is the number of features in set P.

$Q$ is defined as the set of band ratios between two original bands:

(2)$Q=\left\{\frac{x_i}{x_j}|x_i\in X,x_j\in X,i\ne j\right\},$

where q is the number of features in set Q.

Combining X, P, and Q constitutes the multispectral remote sensing feature set D. Figure 2 shows the structure of feature set D to the band combination Y. The number of features contained in the multispectral feature set D is d, d = b + p + q. Based on the multispectral remote sensing feature set, a binary code of length d for modeling chromosome sequences is formed. Encoding the individuals in the initial population constitutes a subset of features in the multispectral remote sensing feature set. The encoding length is the number of selected optimal features for mineralization and alteration mapping. A value of “1” in any bit of the binary code indicates that the current band selection set contains the modified feature, while a value of “0” indicates that the current band selection set does not contain the modified feature. Let $Y=\{y_1,y_2,\cdots ,y_{best}\}$ be the mineralized alteration optimal band combination, which contains the best feature bands and is selected from the multispectral remote sensing feature set D using the GA-PCA approach described below.

3.2. Mineralized Alteration Objective Function and Fitness Function

Genetic algorithms determine the goodness of chromosomes by using a fitness function [2,31,46,47,48]. In this study, the fitness function assessed whether the band combinations in a chromosome effectively captured the desired alteration information regarding the mine mineral occurrences or mineralized spots, which was defined according to Formula (3). The objective function was computed via the mean and standard deviation of the mineralized alteration information about the mineral occurrences or mineralized spots for extracting the four bands using PCA. If the obtained pixel is greater than the sum of the mean and 1.5 times the standard deviation of the principal components, it is labeled as the mineralized alteration information:

(3)$F=t\cdot \frac{1}{n}{\displaystyle \sum _{i=1}^n{G}_i}+\left(1-t\right)\cdot \left(1-U\right)$

In the above objective function, there are n mineral occurrences or mineralized spots that can be selected from the known mineralized sites list in Table 1. $G=\left\{G_1,G_2,\dots ,G_n\right\}$ is the proportion of the amount of mineralization and alteration information extracted from a predetermined area around a known mineral occurrence or mineralized spot to the study area, while $G_i$ describes the amount of alteration information extracted from the ith mine mineral occurrence or mineralized spot. The larger the value of $G_i$, the more alteration information is extracted from the ith mine mineral occurrence or mineralized spot. $U$ is the standard deviation of alteration information set G. A smaller $U$ represents more homogeneous amount of extracted alteration information within the area of all mine points. G and $U$ can be calculated using the following formulae:

(4)$G_i=\frac{c_i}{d_i}$

(5)$U=\sqrt{\frac{{\displaystyle \sum _{i=1}^n{\left(G_i-\overline{G}\right)}^2}}{n-1}}$

With this approach, the number of pixels in the preset extraction range, i.e., the detection radius r around the known mineral occurrences or mineralized spots, is used to measure the PCA extracted alteration information. Let $c=\left\{c_1,c_2,\dots ,c_n\right\}$ and $d=\left\{d_1,d_2,\dots ,d_n\right\}$ be the amount of alteration information and the number of pixels in the detection radius r, respectively.

The value of the fitness function determines the probability that the features in the multispectral remote sensing feature set D are selected and inherited into the next-generation population. In the objective function, a larger value of (1 − U) leads to a larger value of $G_i$, which means a larger F. The fitness function is defined as the maximum of the objective function F as follows:

(6)$f\left(x_i\right)=\max (F)$

3.3. Extraction of Mineralized Alteration Information from Multispectral Remote Sensing Images Based on Genetic Algorithm

By combining the established mineralized alteration objective function and fitness function, an intelligent alteration extraction approach based on the genetic algorithm (GA) and the PCA, termed GA-PCA, for selecting the optimal band combinations of mineralized alterations was proposed. Figure 3 shows the processing of the GA-PCA.

In the GA-PCA, the optimal band combinations can be iteratively searched to accurately extract mineralized alteration information. It creates new generations of candidate band combinations based on the GA, evaluates their fitness based on the defined mineralized alteration objective function and fitness function in Section 3.2, and selects the most promising band combination to carry on to the next generation. Finally, the GA-PCA converges to the selected optimal band combination. The detailed procedure is illustrated below:

• Step 1: Establish the multispectral remote sensing feature set

According to the multispectral remote sensing dataset X, datasets P, Q, and D were generated.

• Step 2: Initialize the population and set parameters

The digital encoding of multispectral remote sensing feature sets was randomly generated. Then, an initial population was produced and analyzed using a binary code. This was followed by encoding the individuals in the population and setting the coding length to the number of optimal feature combinations [28,49,50].

• Step 3: Obtain the mineralized alteration objective function

The extracted mineralization and alteration information was calculated within a predetermined area around known mining occurrences or mineralized spots using Formula (4). The standard deviation of the amount of alteration information extracted from the mining areas was calculated using Formula (5). Then, we obtained the mineralized alteration objective function using Formula (3).

• Step 4: Select the optimal feature combination for mineralization and alteration mapping.

The population initialization in step 1 was utilized to perform genetic algorithm optimization. First, the initialized populations of different feature subsets in the multispectral remote sensing feature set were processed via PCA. The fitness value of the principal components corresponding to each subset was calculated, i.e., compute the fitness function and sort the optimal individuals in this generation to participate in the next-generation populations based on the operation of the genetic operators. The crossover [51] and mutation [52,53] processes of the GA-PCA approach were performed using the standard genetic algorithm. Crossover is a genetic operator that combines genetic information from two parent individuals to produce one or more offspring. Mutation is a genetic operator that alters one or more genes in an individual with a certain probability. A variant of “0” to “1” means that one of the bands in the band combination is dropped, whereas “1” to “0” means that one of the bands in the band combination is marked. This introduces random changes in the solution space of band combinations, allowing the algorithm to explore new regions that might not be reachable through crossover alone. The fitness value of each generation was compared and the generation with the highest fitness value was selected. Then, the position of “1” in the coding chromosome was detected in this generation. The features corresponding to the i1th, i2th, …, i_bth features constituted the optimal combination of mineralization and alteration features in the optimization selection.

• Step 5: Extract mineralization and alteration information

Steps 3 and 4 were repeated until the fitness function value was no longer increasing or the maximum number of iterations was reached. The optimal combination was then used to extract the multispectral remote sensing mineralization alteration information.

4. Experiments and Results

4.1. Experimental Design

4.1.1. Comparison Strategy

To verify the effectiveness of the proposed GA-PCA approach, the physical properties of the optimal feature combination obtained by combining the spectral characteristics of rock and mineral alteration were analyzed. The classical principal component analysis (PCA) with the Crosta criterion [54,55,56] was selected to extract mineralized alteration information according to the spectral characteristics of rock and mineral alteration in the study area. We compared the proposed GA-PCA approach with the classical method and measured the degree of match between known mineral occurrences or mineralized spots and extracted mineral alterations. We conducted the experiments using MATLAB R2021a.

The primary minerals in the study area were skarn-type iron ores, and the altered minerals included hematite, pyrite, goethite, and limonite, which contained Fe³⁺ or Fe²⁺. Other altered minerals included chlorite, kaolinite, dolomite, and calcite. The spectral signatures and features of these altered minerals varied. Iron oxides containing Fe³⁺ and Fe²⁺ had corresponding mineral absorption bands within the wavelength range of 0.45 μm to 0.52 μm and 0.76 μm to 0.9 μm, i.e., absorption characteristics in the OLI bands b2, b3, and b5, whereas the OLI bands b4, b6, and b7 showed reflection features. For the BR, b4/b2, b6/b2, b6/b5, and b7/b2 enhanced the information of iron altered minerals. Typical hydroxide had corresponding mineral absorption bands in the range of 2.2 μm to 2.3 μm, i.e., the OLI band b7, while the wavelength at 1.56 μm showed strong reflection characteristics, corresponding to the OLI band b6. For the BR, b6/b7, b7/b3, and b7b5 enhanced the information of hydroxyl altered minerals. Moreover, b5/b4 reflected the effect of vegetation, and b3+b2 was used instead of b1 for broadening the vis spectrum.

According to the spectral characteristic of iron minerals in the Landsat-8 OLI images, b2, b4, b5, and b6 were selected for the PCA [16,54,55], and six OLI bands, i.e., b2, b3, b4, b5, b6, and b7, and eight band combinations, i.e., b5/b4, b4/b2, b6/b5, b6/b2, b2 + b3, b7/b2, b6/b3, and b6/b4, were selected to form the multispectral remote sensing feature set for extracting iron-stained alteration information. For hydroxyl staining alteration information extraction, the OLI bands b2, b5, b6, and b7 were selected for the PCA, and six OLI bands, i.e., b2, b3, b4, b5, b6, and b7, and five band combinations, i.e., b6/b7, b2 + b3, b5/b4, b7/b5, and b7/b3, were selected to build the multispectral remote sensing feature set.

4.1.2. Evaluation Metrics

In this study, two metrics, i.e., the accuracy and efficiency of mineralized alteration information extraction, were used to evaluate the extraction results. The extraction accuracy of mineralized alteration information is the ratio of the number of mineral occurrences or mineralized spots within the extracted alteration information to the total number of mineral occurrences or mineralized spots, which can be expressed as follows:

(7)$l=\frac{h^{\prime }}{h}$

where h′ is the number of mineral occurrences or mineralized spots within the extracted alteration information, and h is the total number of mineral occurrences or mineralized spots in the study area. A larger value of l represents a higher accuracy of the extracted alteration information.

The extraction efficiency of alteration information is calculated using the ratio of mineralized alteration area to total area as follows:

(8)$e=\left(\frac{s^{\prime }}{s}\right)\cdot l$

where s′ is the area of the mineral occurrences or mineralized spots, and s denotes the total area based on the alteration information within the study area. A larger e value indicates a higher probability of finding minerals based on the extracted alteration information. It is more beneficial in the search for mineral deposits.

4.2. Experimental Results

4.2.1. Extraction of Mineralized Alteration Information via PCA

• Iron-stained alteration information extraction via PCA

The eigenvectors, eigenvalues, and standard deviations (std) obtained for each principal component are listed in Table 3. As shown in Table 3, PC4 mainly reflected the information at b4 with a positive sign, which was opposite to b2 and b5. Thus, PC4 was used as the main component of iron staining indication.

• 2.. Hydroxyl alteration information extraction via PCA

The eigenvectors, eigenvalues, and standard deviations of each principal component are shown in Table 4. As one can see from Table 4, PC4 mainly reflects the information of b6 and b7 with opposite signs, and b5 and b7 have the same sign. Since hydroxyl-containing minerals have a strong reflection in b6 and a strong absorption in b7, PC4 is the main component of hydroxyl anomaly indication.

4.2.2. Extraction of Mineralized Alteration Information Based on the GA-PCA

The digital encoding, i.e., the binary sequence of chromosomes of the selected multispectral remote sensing feature set, was randomly generated. The initial population d was set to 100, with 50 iterations, a reproduction probability of 0.9, a crossover probability of 0.7, and a mutation probability of between 0.02 and 0.1. The encoding length was the number of optimal features for selected mineralization and alteration. In order to compare and verify with the classical PCA, we set the best encoding length to four, i.e., we selected four optimal features from the multispectral remote sensing feature set. In the experiment, a range of 10 to 150 pixels was selected for the mineralization and alteration area centered around known mineral occurrences or mineralized spots. Furthermore, considering the measurement and balance of alteration information within each mining area, different values of t were taken for analysis in the experiment, and the value range of U was between 0 and 1 in the mineralization alteration objective function.

• Iron-stained alteration information extraction based on the GA-PCA

The F1, F2, F4, and FM5 mineral occurrences or mineralized spots in the study area were utilized in the selection of the optimal feature combination for iron-stained alteration. The F3 was used to validate the results of the iron-stained alteration information extraction. The optimal feature combination of iron-stained alteration information under different parameters was obtained by selecting the parameters best, t, and k (Table 5).

As one can see from Table 5, different combinations of features were obtained with changes in the parameters t and k. When t = 0.75 and k = 1.5, the maximum fitness value (0.6848) was obtained. The optimal feature combination was b6, b5/b4, b6/b5, and b7/b2, with the principal component 1 and an optimal radius of 74. Combined with the geological background of the study area and the mineral spectral characteristics, iron-bearing minerals exhibited high reflection at b6 and b7, and b6/b5 and b7/b2 enhanced the alteration information of iron-bearing minerals. The b5/b4 band reflected the effect of vegetation.

Table 6 shows the eigenvectors, eigenvalues, and standard deviations using the GA-PCA with the combination b6, b5/b4, b6/b5, and b7/b2. From Table 6, one can see that PC2 mainly reflected the information of b6, with an eigenvector of 0.930356, and PC4 mainly reflected the information of b5/b4. PC1 mainly reflected the information of b7/b2, which was consistent with the strong reflection characteristics of iron-bearing minerals at b7/b2. The eigenvector of b5/b4 was 0.030703, which was of low value and better enhanced the information of iron-stained alteration.

• 2.. Hydroxyl staining alteration information extraction based on the GA-PCA

The S, M1, CM and the ZM, PZ, CN sites in the study area were utilized in the selection of the optimal feature combination for hydroxyl mineralized alteration. The M2 mineral occurrence was used to verify the results of the hydroxyl mineralized alteration information extraction. Table 7 lists the optimal feature combination of hydroxyl alteration information obtained under different parameters.

As shown in Table 7, when t = 0.7 and k = 1.5, the fitness value obtained was 0.72988, which was the maximum fitness value obtained under these parameters. The optimal feature combination was b3, b5/b4, b6/b7, and b7/b3 with PC 6 (PC2 was inverted), and the optimal region radius was 10. Hydroxyl minerals had high reflectance at b6 and strong feature absorption properties at b7. Therefore, b6/b7 enhanced the hydroxyl alteration information and b5/b4 also reflected the vegetation information.

Table 8 shows the eigenvectors, eigenvalues, and standard deviations obtained using the GA-PCA approach with different combinations of hydroxyl features. As one can see from Table 8, the low value of b5/b4 in PC1 had a good inhibition effect on vegetation, and PC1 mainly reflected the information of b7/b3. PC3 mainly reflected the information of b6/b7, and the eigenvector of b5/b4 was 0.513680, which cannot inhibit the vegetation information. PC4 mainly reflected the information on vegetation of b5/b4. PC2 mainly reflected the information of b3. The eigenvector of PC2 was 0.925380, and b5/b4 showed a low value, while the eigenvector of b6/b7 was negative, which was inconsistent with the fact that hydroxyl-bearing minerals had a strong reflection at b6 and a strong absorption at b7. Therefore, the inverse of PC2 can be used as the main component for extracting hydroxyl alteration information.

4.3. Analysis of Mineralized Alteration Information Extraction Results

4.3.1. Extracted Mineralized Alteration Information

In this study, the extracted information on iron staining and hydroxyl alteration was graded for anomalies based on a thresholding strategy. The mean value plus k times the standard deviation was used to classify the information into strong, medium, and weak anomaly grades, and the value of k was generally taken to be in the range between 1.5 and 3. Figure 4 and Figure 5 show the anomaly grading maps of iron staining and hydroxyl alteration extracted via the classical PCA and the proposed GA-PCA.

The information on iron-stained alteration obtained using the classical PCA was 3.915% in the study area. The information on iron-stained alteration obtained using the proposed GA-PCA was 5.855%, as shown in Table 9.

Table 10 shows that the hydroxyl alteration information obtained via the classical PCA was 5.885% in the study area, and the information on hydroxyl alteration obtained via the proposed GA-PCA was 8.914%.

As can be seen from Figure 4 and Figure 5, Table 9 and Table 10, the first-, second-, and third-level alteration information obtained using the GA-PCA was more uniformly distributed and significantly greater, and thus, the results were better than those obtained using the classical PCA.

4.3.2. Consistency with Known Mineral Occurrences or Mineralized Spots

For the PCA results shown in Figure 4a, the iron-stained alteration information around F1 and F4 was better, but the alteration information around the remaining iron ore points was not obvious. The consistency between the known mineral occurrences or mineralized spots and the extracted iron-stained alteration information is shown in Table 11.

As one can see from Table 11, the extraction accuracy of iron-stained alteration information acquired via the classical PCA was 50% and the extraction efficiency was 12.77%, while the extraction accuracy of the proposed GA-PCA was 75% and the extraction efficiency was 12.81%. The GA-PCA results showed fewer false-negative spots, only at F4. As can be seen in Figure 4b, the false-positives for the proposed GA-PCA were slightly higher than for the classical PCA, and there was more third-level iron-stained alteration information acquired (blue areas) where no mineral sites were present. Therefore, the results of iron-stained alteration information extracted based on the GA-PCA were better than the classical PCA.

In Figure 4a, only the hydroxyl alteration information around the Zn-Mo (ZM) mineral sites of Lalingzaohuo, the Cu-Ni (CN) deposits, and the lead–zinc (PZ) mineralization spot was well distributed. It can be observed from Figure 4b that the distribution of iron-stained alteration information was obvious around the Mo ore (M1) in the west of Lalingzaohuo, the Mo (Cu) occurrences (CM) in eastern Lalingzaohuo, and the Cu-Pb-Zn mineralization (PZ) and Ni-Cu deposits (CN). The consistency between the known mineral occurrences or mineralized spots and the extracted hydroxyl alteration information is shown in Table 12.

As one can see from Table 12, the extraction accuracy of hydroxyl alteration information acquired via the classical PCA was 50%, and the extraction efficiency was 8.53%. The extraction accuracy of hydroxyl alteration information by using the proposed GA-PCA was 83.3%, and the extraction efficiency was 6.36%. The GA-PCA results show fewer false-negative spots, only at ZM. As can be seen in Figure 5a, the false-positives for the classical PCA were higher than those for the GA-PCA, which extracted more third-level hydroxyl alteration information (yellow areas) where there was no mineral site. It should be noted that the hydroxyl alteration information extracted via the proposed GA-PCA was widely distributed with a high consistency, which led to a slightly lower extraction efficiency than the classical PCA.

5. Discussion

5.1. Analysis of Results between Extracted Alteration and Lithology

In this subsection, the relationship between the GA-PCA extracted mineral alteration information and the regional structure and lithology was analyzed by combining the results with the geological background of the study area.

For iron-stained alteration, the mineral occurrences F1 and F2 mainly presented a striped distribution with faults. F1 was mainly located on the grayish white-light flesh red medium-grained monzogranite (nγT₃) and in the carboniferous Shiguaizi formation (C₁s) with volcanic and clastic rocks. The iron polymetallic deposits of F2 were mainly distributed above the quaternary alluvial (Q) with bioclastic limestone and feldspar lithic sandstone. The zone with intense iron-stained alteration phenomenon was between the late Triassic monzonitic granite and carbonate strata. The mine was composed of magnetite and copper–zinc ores, and the alteration types were mainly skarnization and limonitization. The minerals were mainly magnetite and magnetic pyrite. However, F1 and F2 were located at the edge of the anomaly area. It may be due to the fact that both F1 and F2 belonged to semi-concealed ore deposits, and their surface outcrop was buried by sand. The spectral characteristics of the altered rocks were interfered by felsic minerals, although the presence of “desert varnish” and “iron oxide film” on the surface of rocks could not be entirely dismissed. It is worth noting that the iron-stained alteration information around F1 and F2 were mainly distributed in the NW–ES direction consistent with the direction of construction, which should be a reflection of the tectonic control of hydrothermal activity and the alteration of surrounding rocks. The magnetic ore of the Lalinggaoli river iron polymetallic deposit, i.e., FM5, was mainly in the shape of a pea pod and had a clear boundary with surrounding rocks. This deposit mainly occurred in the contact zone between the strata and magmatic rocks. The iron-stained alteration information extracted via the GA-PCA was uniformly distributed near the Lalinggaoli river iron polymetallic deposit.

For hydroxyl alteration, the stratigraphy of the mineral occurrence S was in the Devonian Maoniushan formation (D₃m), which had black mica quartz schist interspersed with white marble, striated mélange, and garnet rocks. The main ore-control structure was a parallel arrangement of NW faults intersecting with nearby EW faults, and the ore was mainly distributed on both sides of the faults. This was consistent with the distribution of hydroxyl alteration information extracted from the surrounding areas of the mineral occurrences or mineralized spots. The stratigraphy of the CM deposit was in the Paleoproterozoic Baishahe formation (Pt₁b), and the lithology was gray-black cloud plagioclase gneiss with the intruded rock being a medium-acidic rock body. The CM deposit belonged to a porphyry molybdenum deposit at an intersection of NW and NE faults, which was consistent with the spread of the alteration information extracted. The stronger the alteration, the more the molybdenum mineralization. The Xiari Hamu nickel–copper deposit, i.e., CN, was distributed in the gneiss section of the Paleoproterozoic Baishahe formation (Pt₁b); its pyroxene was extremely developed, which belonged to the magma melting type of deposits. The hydroxyl alteration information around the CN deposit that was extracted via the GA-PCA was mainly in the NE, NW, and EW directions, which was consistent with the direction of faults.

5.2. Extraction Efficiency Analysis of Alteration in Terms of Target Prediction

Figure 6 shows the predicted target areas based on the mineralization and alteration information extracted via the GA-PCA.As one can see from Figure 6a, two pea-pod-shaped clusters of iron-stained alteration information are distributed in the rectangular predicted area. Therefore, we predicted that there was an ore body existing in this area (the red rectangular region). To validate the accuracy of this prediction, we found that an iron ore, i.e., F3, was in this predicted area, whose information was not utilized when selecting the optimal feature combination for extracting iron-stained alteration. The stratigraphy of F3 was in the schist section of the Paleoproterozoic Baishahe formation (Pt₁b), and the marble in this area had a certain controlling effect on mineralization. The general strike around this mineral occurrence was nearly EW, with the western section deviating southward. The minerals were mainly magnetite, followed by sphalerite, chalcopyrite, and pyrite. The spatial distribution of the iron-stained alteration around the site was mainly consistent with the fault direction.

As can be seen in Figure 6b, there are NW, NE, and EW distributions of hydroxyl alteration information in the rectangular predicted area. It was consistent with the distribution of fault structures in the predicted area. A larger amount of primary and secondary hydroxyl alteration information was distributed in this area (the red rectangular region). Furthermore, the molybdenum mineral occurrence, i.e., M2, was in this predicted area, whose information was not utilized when selecting the optimal feature combination for extracting hydroxyl alteration. The stratigraphy of the M2 deposit was in the grayish white-light grayish yellow medium-grained granodiorite (γδT₃), with widespread intrusive rocks and medium-acidic rocks. The mineralization was mainly pyroxene molybdenum mineralization, which reflected that it was a porphyry-type deposit, and the alteration was mainly potassium, kaolinization, and glauconite. The tectonic structure around the mineral occurrence was the NE and NW rupture and EW reverse rupture. The development of magmatic rocks in the area provides the heat source support for mineralization, and the NW and EW trending faults provided a channel for hydrothermal fluid transportation for mineralization. The gathering of metal minerals in peripheral and intrusive rocks was very favorable for the formation of mineral deposits in this area.

6. Conclusions

This study proposes a new alteration extraction approach, i.e., GA-PCA, to select the optimal combination features for mapping mineralized alteration information. It made full use of the excellent searching ability of the GA and constructed a mineralized alteration fitness function to achieve the efficient extraction of multi-spectral remote sensing alteration information. Comparative experiments were carried out in the Lalingzaohuo polymetallic deposits, China, using the Landsat-8 OLI data. The extraction accuracy and efficiency indicated that the extracted alteration information results were better and more consistent with the known mineral occurrences or mineralized spots than the traditional method. Moreover, two predicted targets (not shown in the alteration extraction results when using the traditional method) were precisely obtained by analyzing the alteration maps combined with the ore-forming geological conditions of the mine and its tectonic characteristics. The analysis demonstrated that the intelligent selection of feature combinations can increase the reliability of mineral alteration information mapping.

However, the limited spectral resolution of Landsat-8 OLI in the VNIR and SWIR bands means that we were only able to map alteration information with rich iron oxides and clay minerals. In the future, we will endeavor to optimize band combination and selection strategy via the fusion of other types of multispectral images with broad swath width, e.g., ASTER [56], to effectively map alteration information.

Author Contributions

Conceptualization, C.Y. and H.Z.; methodology, H.J., L.D. and M.Z.; software, L.D. and H.J.; validation, L.D., M.Z. and H.Z.; formal analysis, C.Y.; investigation, L.D.; resources, L.D.; data curation, M.Z.; writing—original draft preparation, C.Y. and M.Z.; writing—review and editing, C.Y. and M.Z.; visualization, L.D. and M.Z.; supervision, C.Y. and H.Z.; funding acquisition, C.Y. and H.Z. All authors have read and agreed to the published version of the manuscript.

Data Availability Statement

The OLI images used in this paper are available at http://www.gscloud.cn/(accessed on 18 January 2023). Datasets generated or analyzed during this study are available from the corresponding author upon reasonable request.

Conflicts of Interest

The authors declare no conflicts of interest.

Footnotes

Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

Figures and Tables

Figure 1. Regional geological map of the study area (according to ref. [36]).

Figure 2. Mapping of feature set D to the band combination Y.

Figure 3. The flowchart and the processing of the proposed GA-PCA.

Figure 4. Iron-stained alteration extraction: (a,b) are the anomaly maps obtained using the classical PCA and the proposed GA-PCA, respectively.

Figure 5. Hydroxyl alteration extraction: (a,b) are the anomaly maps obtained using the classical PCA and the proposed GA-PCA, respectively.

Figure 6. Predicted target areas for mineralization alteration extracted via GA-PCA: (a) shows the iron ore mineralization target area and (b) shows the hydroxyl mineral mineralization target area.

Table 1

The types and locations of the mineral occurrences or mineralized spots in the study area are acquired from https://www.ngac.cn/(accessed on 26 January 2022) and ref. [36].

No.	Mineral Type	Symbol	Mineral Scale
1	Iron mineral occurrence	F1	Mineral occurrence
2	Iron mineral occurrence	F2	Mineral occurrence
3	Iron mine	F3	Mineral occurrence
4	Iron mineralized spot	F4	Mineralized spot
5	Iron polymetallic deposit	FM5	Small mineral deposit
6	Garnet mineralized spot	S	Mineral occurrence
7	Molybdenum deposit	M1	Mineral occurrence
8	Copper molybdenum deposit	CM	Mineral occurrence
9	Zinc–molybdenum mineral occurrence	ZM	Mineralized spot
10	Lead–zinc mineralization	PZ	Mineralized spot
11	Copper–nickel–cobalt deposit	CN	Large mineral deposit
12	Molybdenum mineral occurrence	M2	Mineral occurrence

Table 2

Absorption and reflection characteristic of minerals based on Landsat-8 OLI images (refs. [41,42]).

Band Number	Band Description	Central Wavelength	Spatial Resolution	Iron-Stained Minerals (Fe2+, Fe3+)	Hydroxylated Minerals (OH-)
1	Coastal Aerosol	0.43–0.45 µm	30	-	-
2	Blue	0.45–0.51 µm	30	Absorption	-
3	Green	0.53–0.59 µm	30	Absorption	-
4	Red	0.64–0.67 µm	30	Reflection	-
5	Near-Infrared	0.85–0.88 µm	30	Absorption	-
6	SWIR 1	1.57–1.65 µm	30	Reflection	Reflection
7	SWIR 2	2.11–2.29 µm	30	Reflection	Absorption
8	Panchromatic	0.50–0.68 µm	15	-	-
9	Cirrus	1.36–1.38 µm	30	-	-

Table 3

Eigenvectors, eigenvalues, and standard deviations of iron-stained alteration information extracted via PCA.

PC	b2	b4	b5	b6	Eigenvalues	Std.
PC1	0.252180	0.488017	0.545615	0.632889	0.025077	0.158358
PC2	−0.275537	−0.390598	−0.418844	0.772064	0.000140	0.011815
PC3	0.881357	−0.005541	−0.468924	0.057348	0.000091	0.009535
PC4	−0.289299	0.780539	−0.554062	−0.008938	0.000009	0.002994

Table 4

Eigenvectors, eigenvalues, and standard deviations of hydroxyl alteration information extracted by means of PCA.

PC	b2	b5	b6	b7	Eigenvalues	Std.
PC1	0.238781	0.514266	0.602137	0.562090	0.027869	0.166940
PC2	−0.110139	−0.753069	0.086773	0.642827	0.000243	0.015588
PC3	0.928397	−0.109779	−0.346501	0.077235	0.000097	0.009838
PC4	0.262550	−0.395432	0.714030	−0.514647	0.000020	0.004492

Table 5

The optimal feature combination and fitness of iron-stained alteration information obtained using the GA-PCA.

	k = 1.5				k = 2.0
t	r	Combinations	PC	Fitness	r	Combinations	PC	Fitness
0.60	73	b3, b4, b5/b4, b7/b2	PC6	0.68461	72	b5, b6, b7, b7/b2	PC2	0.44609
0.65	73	b3, b4, b5/b4, b7/b2	PC1	0.6562	79	b6/b5, b6/b2, b7/b2, b6/b3	PC1	0.62026
0.70	73	b3, b4, b5/b4, b7/b2	PC1	0.6562	76	b3, b4, b5/b4, b7/b2	PC1	0.6562
0.72	73	b2, b3, b5, b7/b2	PC1	0.6842	79	b5, b7, b2 + b3, b7/b2	PC1	0.62022
0.75	74	b6, b5/b4, b6/b5, b7/b2	PC1	0.6848	79	b5, b7, b2 + b3, b7/b2	PC1	0.62022
0.80	99	b2, b5, b5/b4, b4/b2	PC6	0.42827	76	b2, b6, b7, b5/b4	PC7	0.65992

Table 6

Eigenvectors, eigenvalues, and standard deviations of hydroxyl alteration information based on GA-PCA.

PC	b6	b5/b4	b6/b5	b7/b2	Eigenvalues	Std.
PC1	0.098876	0.030703	0.196784	0.974965	0.030595	0.174915
PC2	0.930356	−0.004336	−0.366067	−0.020330	0.011063	0.105179
PC3	0.332721	−0.331784	0.860476	−0.196971	0.002244	0.047376
PC4	−0.118142	−0.942846	−0.294707	0.101156	0.000367	0.019155

Table 7

The optimal feature combination and fitness of hydroxyl staining information obtained using the GA-PCA.

	k = 1.5				k = 2.0
t	r	Combinations	PC	Fitness	r	Combinations	PC	Fitness
0.60	18	b3, b5, b6/b7, b7/b5	PC1	0.61051	71	b3, b5/b4, b7/b5, b7/b3	PC6	0.63293
0.65	18	b3, b5, b6/b7, b7/b5	PC1	0.61051	65	b2, b7, b6/b7, b5/b4	PC5	0.63378
0.70	10	b3, b5/b4, b6/b7, b7/b3	PC6	0.72988	10	b7, b6/b7, b7/b5, b7/b3	PC6	0.6749
0.72	18	b3, b5, b6/b7, b7/b5	PC1	0.69287	10	b3, b5/b4, b7/b5, b7/b3	PC6	0.69322
0.75	18	b3, b5, b6/b7, b7/b5	PC1	0.69287	10	b7, b6/b7, b7/b5, b7/b3	PC6	0.71683
0.80	18	b3, b5, b6/b7, b7/b5	PC1	0.61051	10	b7, b6/b7, b7/b5, b7/b3	PC6	0.67875

Table 8

Eigenvectors, eigenvalues, and standard deviations of hydroxyl alteration information obtained using GA-PCA.

PC	b3	b5/b4	b6/b7	b7/b3	Eigenvalues	Std.
PC1	0.235071	−0.048950	0.141031	−0.960445	0.009726	0.098623
PC2	0.925380	−0.078867	−0.322334	0.183177	0.003617	0.060142
PC3	0.287904	0.513680	0.792119	0.160599	0.001243	0.035256
PC4	0.074333	−0.852946	0.498758	0.134902	0.000314	0.017722

Table 9

Percentage of iron-stained alteration information in the study area.

Method	Total	First-Level	Second-Level	Third-Level
PCA	4.490%	2.103%	1.052%	1.329%
GA-PCA	5.855%	2.581%	1.173%	2.101%

Table 10

Percentage of hydroxyl alteration information in the study area.

Method	Total	First-Level	Second-Level	Third-Level
PCA	5.885%	1.396%	1.406%	3.083%
GA-PCA	8.914%	3.586%	2.555%	2.773%

Table 11

Consistency between the known mineral occurrences or mineralized spots and the extracted iron-stained alteration information.

ID	PCA	GA-PCA
F1	Yes	Yes
F2	No	Yes
F4	Yes	No
FM5	No	Yes

Table 12

Consistency between the known mineral occurrences or mineralized spots and the extracted hydroxyl alteration information.

ID	PCA	GA-PCA
S	No	Yes
M1	No	Yes
CM	No	Yes
ZM	Yes	No
PZ	Yes	Yes
CN	Yes	Yes

References

1. Dong, P. A test study on alteration mapping using Landsat TM data and geographic information system. Methodology and Practice of Remote Sensing for Mineral Exploration; Science Press: Beijing, China, 1995; pp. 24-28.

2. Pour, A.B.; Park, T.Y.S.; Park, Y.; Hong, J.K.; Muslim, A.M.; Laufer, A.; Crispini, L.; Pradhan, B.; Zoheir, B.; Rahmani, O. et al. Landsat-8, Advanced Spaceborne Thermal Emission and Reflection Radiometer, and WorldView-3 Multispectral Satellite Imagery for Prospecting Copper-Gold Mineralization in the Northeastern Inglefield Mobile Belt (IMB), Northwest Greenland. Remote Sens.; 2019; 11, 2430. [DOI: https://dx.doi.org/10.3390/rs11202430]

3. Hunt, G.R.; Ashley, R.P. Spectra of altered rocks in the visible and near infrared. Econ. Geol.; 1979; 74, pp. 1613-1629. [DOI: https://dx.doi.org/10.2113/gsecongeo.74.7.1613]

4. Pour, A.B.; Hashim, M. The application of ASTER remote sensing data to porphyry copper and epithermal gold deposits. Ore Geol. Rev.; 2012; 44, pp. 1-9. [DOI: https://dx.doi.org/10.1016/j.oregeorev.2011.09.009]

5. Han, L.; Zhao, B.; Wu, J.J.; Wu, T.T.; Feng, M. A new method for extraction of alteration information using the Landsat 8 imagery in a heavily vegetated and sediments-covered region: A case study from Zhejiang Province, E. China. Geol. J.; 2018; 53, pp. 33-43. [DOI: https://dx.doi.org/10.1002/gj.2988]

6. Tangestani, M.H.; Moore, F. Comparison of three principal component analysis techniques to porphyry copper alteration mapping: A case study, Meiduk area, Kerman, Iran. Can. J. Remote Sens.; 2001; 27, pp. 176-182. [DOI: https://dx.doi.org/10.1080/07038992.2001.10854931]

7. Shebl, A.; Csámer, Á. Lithological, structural and hydrothermal alteration mapping utilizing remote sensing datasets: A case study around Um Salim area, Egypt. IOP Conf. Ser. Earth Environ. Sci.; 2021; 942, 012032. [DOI: https://dx.doi.org/10.1088/1755-1315/942/1/012032]

8. El-Desoky, H.M.; Shebl, A.; Abdel-Rahman, A.M.; Fahmy, W.; El-Awny, H.; El-Sherif, A.M.; El-Rahmany, M.M.; Csamer, A. Multiscale mineralogical investigations for mineral potentiality mapping of Ras El-Kharit-Wadi Khashir district, Southern Eastern Desert, Egypt. Egypt. J. Remote Sens.; 2022; 25, pp. 941-960. [DOI: https://dx.doi.org/10.1016/j.ejrs.2022.09.001]

9. Badawi, M.; Abdelatif, M.; Shebl, A.; Makroum, F.; Shalaby, A.; Nemeth, N. Mapping Structurally Controlled Alterations Sparked by Hydrothermal Activity in the Fatira-Abu Zawal Area, Eastern Desert, Egypt. Acta Geol. Sin-Engl.; 2023; 97, pp. 662-680. [DOI: https://dx.doi.org/10.1111/1755-6724.15019]

10. Fereydooni, H.; Moradzadeh, A.; Pahlavani, P.; Mojeddifar, S. Full unmixing hydrothermal alteration minerals mapping by integration of pattern recognition network and directed matched filtering algorithm. Earth Sci. Inform.; 2020; 13, pp. 417-431. [DOI: https://dx.doi.org/10.1007/s12145-019-00422-y]

11. Yousefi, M.; Tabatabaei, S.H.; Rikhtehgaran, R.; Pour, A.B.; Pradhan, B. Detection of alteration zones using the Dirichlet process Stick-Breaking model-based clustering algorithm to hyperion data: The case study of Kuh-Panj porphyry copper deposits, Southern Iran. Geocarto Int.; 2022; 37, pp. 9788-9816. [DOI: https://dx.doi.org/10.1080/10106049.2022.2025917]

12. Shirazi, A.; Hezarkhani, A.; Pour, A.B.; Shirazy, A.; Hashim, M. Neuro-Fuzzy-AHP (NFAHP) Technique for Copper Exploration Using Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) and Geological Datasets in the Sahlabad Mining Area, East Iran. Remote Sens.; 2022; 14, 5562. [DOI: https://dx.doi.org/10.3390/rs14215562]

13. Aali, A.A.; Shirazy, A.; Shirazi, A.; Pour, A.B.; Hezarkhani, A.; Maghsoudi, A.; Hashim, M.; Khakmardan, S. Fusion of Remote Sensing, Magnetometric, and Geological Data to Identify Polymetallic Mineral Potential Zones in Chakchak Region, Yazd, Iran. Remote Sens.; 2022; 14, 6018. [DOI: https://dx.doi.org/10.3390/rs14236018]

14. Goetz, A.F.; Rowan, L.C. Geologic remote sensing. Science; 1981; 211, pp. 781-791. [DOI: https://dx.doi.org/10.1126/science.211.4484.781] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/17740369]

15. Pour, A.B.; Hashim, M.; Hong, J.K.; Park, Y. Lithological and alteration mineral mapping in poorly exposed lithologies using Landsat-8 and ASTER satellite data: North-eastern Graham Land, Antarctic Peninsula. Ore Geol. Rev.; 2019; 108, pp. 112-133. [DOI: https://dx.doi.org/10.1016/j.oregeorev.2017.07.018]

16. Sekandari, M.; Masoumi, I.; Pour, A.B.; Muslim, A.M.; Rahmani, O.; Hashim, M.; Zoheir, B.; Pradhan, B.; Misra, A.; Aminpour, S.M. Application of Landsat-8, Sentinel-2, ASTER and WorldView-3 Spectral Imagery for Exploration of Carbonate-Hosted Pb-Zn Deposits in the Central Iranian Terrane (CIT). Remote Sens.; 2020; 12, 1239. [DOI: https://dx.doi.org/10.3390/rs12081239]

17. Ishagh, M.M.; Pour, A.B.; Benali, H.; Idriss, A.M.; Reyoug, S.A.S.; Muslim, A.M.; Hossain, M.S. Lithological and alteration mapping using Landsat 8 and ASTER satellite data in the Reguibat Shield (West African Craton), North of Mauritania: Implications for uranium exploration. Arab. J. Geosci.; 2021; 14, 2576. [DOI: https://dx.doi.org/10.1007/s12517-021-08846-x]

18. Kusky, T.M.; Ramadan, T.M. Structural controls on Neoproterozoic mineralization in the South Eastern Desert, Egypt: An integrated field, Landsat TM, and SIR-C/X SAR approach. J. Afr. Earth Sci.; 2002; 35, pp. 107-121. [DOI: https://dx.doi.org/10.1016/S0899-5362(02)00029-5]

19. Salem, S.M.; Arafa, S.A.; Ramadan, T.M.; El Gammal, E.S.A. Exploration of copper deposits in Wadi El Regeita area, Southern Sinai, Egypt, with contribution of remote sensing and geophysical data. Arab. J. Geosci.; 2013; 6, pp. 321-335. [DOI: https://dx.doi.org/10.1007/s12517-011-0346-z]

20. Son, Y.S.; Kang, M.K.; Yoon, W.J. Lithological and mineralogical survey of the Oyu Tolgoi region, Southeastern Gobi, Mongolia using ASTER reflectance and emissivity data. Int. J. Appl. Earth Obs.; 2014; 26, pp. 205-216. [DOI: https://dx.doi.org/10.1016/j.jag.2013.07.004]

21. Ibrahim, W.S.; Watanabe, K.; Yonezu, K. Structural and litho-tectonic controls on Neoproterozoic base metal sulfide and gold mineralization in North Hamisana shear zone, South Eastern Desert, Egypt: The integrated field, structural, Landsat 7 ETM + and ASTER data approach. Ore Geol. Rev.; 2016; 79, pp. 62-77. [DOI: https://dx.doi.org/10.1016/j.oregeorev.2016.05.012]

22. Zoheir, B.; Abd El-Wahed, M.; Pour, A.B.; Abdelnasser, A. Orogenic Gold in Transpression and Transtension Zones: Field and Remote Sensing Studies of the Barramiya-Mueilha Sector, Egypt. Remote Sens.; 2019; 11, 2122. [DOI: https://dx.doi.org/10.3390/rs11182122]

23. Traore, M.; Wambo, J.D.T.; Ndepete, C.P.; Tekin, S.; Pour, A.B.; Muslim, A.M. Lithological and alteration mineral mapping for alluvial gold exploration in the south east of Birao area, Central African Republic using Landsat-8 Operational Land Imager (OLI) data. J. Afr. Earth Sci.; 2020; 170, 103933. [DOI: https://dx.doi.org/10.1016/j.jafrearsci.2020.103933]

24. Ge, W.Y.; Cheng, Q.M.; Jing, L.H.; Wang, F.; Zhao, M.L.; Ding, H.F. Assessment of the Capability of Sentinel-2 Imagery for Iron-Bearing Minerals Mapping: A Case Study in the Cuprite Area, Nevada. Remote Sens.; 2020; 12, 3028. [DOI: https://dx.doi.org/10.3390/rs12183028]

25. Mia, M.B.; Fujimitsu, Y. Mapping hydrothermal altered mineral deposits using Landsat 7 ETM + image in and around Kuju volcano, Kyushu, Japan. J. Earth Syst. Sci.; 2012; 121, pp. 1049-1057. [DOI: https://dx.doi.org/10.1007/s12040-012-0211-9]

26. Van der Meer, F.D.; van der Werff, H.M.A.; van Ruitenbeek, F.J.A.; Hecker, C.A.; Bakker, W.H.; Noomen, M.F.; van der Meijde, M.; Carranza, E.J.M.; de Smeth, J.B.; Woldai, T. Multi- and hyperspectral geologic remote sensing: A review. Int. J. Appl. Earth Obs. Geoinf.; 2012; 14, pp. 112-128. [DOI: https://dx.doi.org/10.1016/j.jag.2011.08.002]

27. Shebl, A.; Abdellatif, M.; Badawi, M.; Dawoud, M.; Fahil, A.S. Towards better delineation of hydrothermal alterations via multi-sensor remote sensing and airborne geophysical data. Sci. Rep.; 2023; 13, 7406. [DOI: https://dx.doi.org/10.1038/s41598-023-34531-y]

28. Zhu, R.C.; Jia, Y.X.; Wang, J.F.; Xu, C.L.; Meng, Z.; Sui, S.; Zhang, Z.T.; Zhu, Y.; Zhang, L.; Wang, J. et al. Synthesized optimal design via Parallel Genetic Algorithm of multispectral metasurfaces with ultra-wideband microwave absorption, low infrared emissivity and visible transparency. Infrared Phys. Techn.; 2021; 117, 103826. [DOI: https://dx.doi.org/10.1016/j.infrared.2021.103826]

29. Xing, J.; Li, Y.Q.; Cui, S.L. The spectral-voltage calibration technology of multispectral pyrometers based on Gauss-Newton-genetic algorithm method for nonsource temperature regions. Microw. Opt. Technol. Lett.; 2023; 66, e33715. [DOI: https://dx.doi.org/10.1002/mop.33715]

30. Cao, Y.; Feng, W.; Quan, Y.; Ren, A.; Xing, M. A Novel Forest Disater Monitoring Method Based on FCM and Neighborhood Factor Genetic Algorithm Using Multispectral Data. Proceedings of the 2021 IEEE International Geoscience and Remote Sensing Symposium IGARSS; Brussels, Belgium, 11–16 July 2021; pp. 3761-3764.

31. Zhao, W.J.; Ma, H.; Zhou, C.; Zhou, C.Q.; Li, Z.L. Soil Salinity Inversion Model Based on BPNN Optimization Algorithm for UAV Multispectral Remote Sensing. IEEE J.-Stars.; 2023; 16, pp. 6038-6047. [DOI: https://dx.doi.org/10.1109/JSTARS.2023.3284019]

32. Yang, Y.; Wang, X.; Huang, M.; Zhu, Q.; Guo, Y.; Xu, L.; Zhou, Z. Hyperspectral band selection based on dual evaluation measures and improved nondominated sorting genetic algorithm. J. Appl. Remote Sens.; 2021; 15, 028504. [DOI: https://dx.doi.org/10.1117/1.JRS.15.028504]

33. Singh, P.S.; Karthikeyan, S. Enhanced classification of remotely sensed hyperspectral images through efficient band selection using autoencoders and genetic algorithm. Neural Comput. Appl.; 2022; 34, pp. 21539-21550. [DOI: https://dx.doi.org/10.1007/s00521-021-06121-4]

34. Zhao, H.S.; Bruzzone, L.; Guan, R.C.; Zhou, F.F.; Yang, C. Spectral-Spatial Genetic Algorithm-Based Unsupervised Band Selection for Hyperspectral Image Classification. IEEE Trans. Geosci. Remote; 2021; 59, pp. 9616-9632. [DOI: https://dx.doi.org/10.1109/TGRS.2020.3047223]

35. Aghaee, R.; Momeni, M.; Moallem, P. Semisupervised Band Selection From Hyperspectral Images Using Levy Flight-Based Genetic Algorithm. IEEE Geosci. Remote Sens.; 2022; 19, 6006805. [DOI: https://dx.doi.org/10.1109/LGRS.2022.3147272]

36. Chen, J.H.; Wu, H.Y.; Niu, X.L.; Niu, S.D.; Wang, Y.C.; Wang, J.H.; Wang, Z. Geology and geochronology of the Lalingzaohuo cobalt-bearing copper polymetallic skarn deposit, East Kunlun. Ore Geol. Rev.; 2023; 154, 105349. [DOI: https://dx.doi.org/10.1016/j.oregeorev.2023.105349]

37. Lin, N.; Chen, Y.L.; Lu, L.J. Mineral Potential Mapping Using a Conjugate Gradient Logistic Regression Model. Nat. Resour. Res.; 2020; 29, pp. 173-188. [DOI: https://dx.doi.org/10.1007/s11053-019-09509-1]

38. Chaves, M.E.D.; Picoli, M.C.A.; Sanches, I.D. Recent Applications of Landsat 8/OLI and Sentinel-2/MSI for Land Use and Land Cover Mapping: A Systematic Review. Remote Sens.; 2020; 12, 3062. [DOI: https://dx.doi.org/10.3390/rs12183062]

39. Adiri, Z.; El Harti, A.; Jellouli, A.; Lhissou, R.; Maacha, L.; Azmi, M.; Zouhair, M.; Bachaoui, E.M. Comparison of Landsat-8, ASTER and Sentinel 1 satellite remote sensing data in automatic lineaments extraction: A case study of Sidi Flah-Bouskour inlier, Moroccan Anti Atlas. Adv. Space Res.; 2017; 60, pp. 2355-2367. [DOI: https://dx.doi.org/10.1016/j.asr.2017.09.006]

40. Javhar, A.; Chen, X.; Bao, A.M.; Jamshed, A.; Yunus, M.; Jovid, A.; Latipa, T. Comparison of Multi-Resolution Optical Landsat-8, Sentinel-2 and Radar Sentinel-1 Data for Automatic Lineament Extraction: A Case Study of Alichur Area, SE Pamir. Remote Sens.; 2019; 11, 778. [DOI: https://dx.doi.org/10.3390/rs11070778]

41. Roy, D.P.; Wulder, M.A.; Loveland, T.R.; Woodcock, C.E.; Allen, R.G.; Anderson, M.C.; Helder, D.; Irons, J.R.; Johnson, D.M.; Kennedy, R. et al. Landsat-8: Science and product vision for terrestrial global change research. Remote Sens. Environ.; 2014; 145, pp. 154-172. [DOI: https://dx.doi.org/10.1016/j.rse.2014.02.001]

42. Said, C.; Hahou, Y.; Achmani, J.; Abdessalam, O.; Mohamed, E.A.; Larbi, B.; Assia, I.; Zineb, A.; Slimane, S.; Lahcen, O. et al. Litho-structural and hydrothermal alteration mapping for mineral prospection in the Maider basin of Morocco based on remote sensing and field investigations. Remote Sens. Appl. Soc. Environ.; 2023; 31, 100980.

43. Kriegler, F.J.; Malila, W.A.; Nalepka, R.F.; Richardson, W. Preprocessing transformations and their effect on multispectral recognition. Remote Sens. Environ.; 1969; 6, pp. 97-132.

44. Xu, H. Modification of Normalized Difference Water Index (NDWI) to Enhance Open Water Features in Remotely Sensed Imagery. Int. J. Remote Sens.; 2006; 27, pp. 3025-3033. [DOI: https://dx.doi.org/10.1080/01431160600589179]

45. Hall, D.K.; Riggs, G.A.; Salomonson, V.V. Development of methods for mapping global snow cover using moderate resolution imaging spectoradiometer data. Remote Sens. Environ.; 1995; 54, pp. 127-140. [DOI: https://dx.doi.org/10.1016/0034-4257(95)00137-P]

46. Katoch, S.; Chauhan, S.S.; Kumar, V. A review on genetic algorithm: Past, present, and future. Multimed. Tools Appl.; 2021; 80, pp. 8091-8126. [DOI: https://dx.doi.org/10.1007/s11042-020-10139-6] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/33162782]

47. Lambora, A.; Gupta, K.; Chopra, K. Genetic algorithm-A literature review. Proceedings of the 2019 International Conference on Machine Learning, Big Data, Cloud and Parallel Computing (COMITCon); Faridabad, India, 14–16 February 2019; pp. 380-384.

48. Chen, W.; Chen, Y.; Tsangaratos, P.; Ilia, I.; Wang, X. Combining Evolutionary Algorithms and Machine Learning Models in Landslide Susceptibility Assessments. Remote Sens.; 2020; 12, 3854. [DOI: https://dx.doi.org/10.3390/rs12233854]

49. Lee, J.Y.; Kim, M.S.; Kim, C.T.; Lee, J.J. Study on encoding schemes in compact genetic algorithm for the continuous numerical problems. Proceedings of the Sice Annual Conference; Takamatsu, Japan, 17–20 September 2007; pp. 2685-2690.

50. Shi, G.; Iima, H.; Sannomiya, N. A new encoding scheme for solving job shop problems by genetic algorithm. Proceedings of the 35th IEEE Conference on Decision and Control; Kobe, Japan, 13 December 1996, pp. 4395-4400.

51. Hassanat, A.; Almohammadi, K.; Alkafaween, E.; Abunawas, E.; Hammouri, A.; Prasath, V.B.S. Choosing Mutation and Crossover Ratios for Genetic Algorithms—A Review with a New Dynamic Approach. Information; 2019; 10, 390. [DOI: https://dx.doi.org/10.3390/info10120390]

52. Liu, D.H. Mathematical modeling analysis of genetic algorithms under schema theorem. J. Comput. Methods Sci.; 2019; 19, pp. S131-S137. [DOI: https://dx.doi.org/10.3233/JCM-191019]

53. Jebari, K.; Madiafi, M. Selection methods for genetic algorithms. Int. J. Emerg. Sci.; 2013; 3, pp. 333-344.

54. Sabins, F.F. Remote sensing for mineral exploration. Ore Geol. Rev.; 1999; 14, pp. 157-183. [DOI: https://dx.doi.org/10.1016/S0169-1368(99)00007-4]

55. Ali, A.; Pour, A. Lithological mapping and hydrothermal alteration using Landsat 8 data: A case study in ariab mining district, red sea hills, Sudan. Int. J. Sci. Basic Appl. Res.; 2014; 3, pp. 199-208. [DOI: https://dx.doi.org/10.14419/ijbas.v3i3.2821]

56. Zhang, T.; Yi, G.; Li, H.; Wang, Z.; Tang, J.; Zhong, K.; Li, Y.; Wang, Q.; Bie, X. Integrating Data of ASTER and Landsat-8 OLI (AO) for Hydrothermal Alteration Mineral Mapping in Duolong Porphyry Cu-Au Deposit, Tibetan Plateau, China. Remote Sens.; 2016; 8, 890. [DOI: https://dx.doi.org/10.3390/rs8110890]