This paper used GF-5 (Gaofen-5) and EMIT (Earth Surface Mineral Dust Source Investigation) remote sensing data for mineral mapping and compared the two sets of mapping results to ensure they were accurate and reliable.

The GF-5 (Gaofen-5) satellite, launched on May 9th, 2018, is a Chinese hyperspectral observation satellite that is equipped with hyperspectral and multispectral sensors, which allow it to capture detailed images and data across many narrow spectral bands. The VNIR signal-to-noise ratio (SNR) of AHSI data is nearly 700, whereas the SNR of SWIR can reach approximately 500.

The EMIT (Earth Surface Mineral Dust Source Investigation) remote sensor that was carried by the International Space Station (ISS) on January 14th, 2022, is a hyperspectral sensor developed by the National Aeronautics and Space Administration (NASA). The EMIT provides L2A-level surface reflectance data with an extremely high signal-to-noise ratio and imaging quality. The parameters of the two satellites are listed in Table1.

In this paper, we collected GF5-AHSI data acquired on January 28th, 2019, and EMIT data acquired on February 4th, 2023, for mineral mapping. Both of the images are clear with less than 5 percent cloud cover. The GF-5 AHSI data were obtained from the GeoCloud online platform (https://geocloud.cgs.gov.cn/), and the EMIT data were obtained from the NASA Earthdata.

Table 1. Parameters of Gf-5 and EMIT hyperspectral data.

Image data processing includes several steps such as image denoising, radiometric calibration, atmospheric correction, and orthorectification.

Hyperspectral image often exhibits minor stripe noise and random noise due to variations in sensor sensitivity, operating conditions (such as radiation and temperature), and changes in the received energy. To improve the image quality of GF-5 data, it is necessary to perform data denoising, which consists of two parts: spatial denoising and spectral denoising. In this paper, the global linear stretching method and the Butterworth filtering method were used for spatial denoising and spectral denoising, respectively. The result was shown in Figure 4.

The EMIT image provides reflectance data with high quality, which means that no pre-processing is required for the data.

The mapping method used in this paper is based on the Tetracorder mineral mapping algorithm and identification rules developed by the USGS [21]. This method has been successfully applied in mineral mapping conducted by the USGS in areas such as Afghanistan and Australia. We referenced some of the optimization work done by Kaley et al. on Tetracorder and implemented the algorithm with Python programming, which supports batch processing, spectral resampling, mineral identification, calculation of absorption peak position shift, and absorption depth. These functions could help in accurately determining the compositions of minerals and aid in filtering and identifying significant mineral abundance. The mapping method can be mainly divided into the following steps: spectral continuum removal, spectral resampling, mineral spectral match, and mineral identification.

Continuum removal involves normalizing the spectrum to remove the effects of the overall reflectance level, thereby enhancing the features related to absorption bands. This step helps in standardizing the spectral signatures, making it easier to compare and analyze the absorption characteristics of different materials. Different from the common practice, this method uses “spectral interval” instead of “endpoints” for continuum removal. As a simple example shown in Figure 5, the average wavelength and reflectance values between points A and B are calculated as the left endpoint of the dolomite spectra continuum removal, and similarly, the average between points C and D for the right endpoint. The interval from A to B includes four data points, which helps to reduce the impact of noise and outliers on the continuum removal results.

This paper used the USGS digital spectral library as the source of standard mineral spectral data. Since the standard spectra obtained in the spectral library have different channels and spectral response functions from the spectra of remote sensing images, spectral resampling is required before spectra matching between the image spectra and standard spectra. A spectral convolution resampling method that uses a Gaussian function to model the spectral response function was used to resample the data by calculating the integral differences between different functions [22].

Figure 4. Data processing result of GF-5 and EMIT.

Mineral spectral match and mineral identification are based on the data that has been processed by continuum removal and spectral resampling. Each image pixel spectrum was compared to the standard spectra in the spectral library, and the coefficient of determination (R²) was calculated as the fitting parameter (F). For minerals that have multiple diagnostic features, “discriminant features” are used to distinguish and identify these minerals. For example, the absorption feature of montmorillonite is centered at 2200 nm, and it has a discriminant feature at 2340 nm. When an image spectrum of muscovite, which shows both the 2200 nm and 2340 nm absorption features, is compared to the standard spectra, the presence of the discriminant feature at 2340 nm ensures that the spectrum is not misidentified as montmorillonite, even if the fit parameter at 2200 nm is very high. The spectra of hyperspectral image pixels are often a mixture of multiple minerals. Given that the USGS standard spectral library includes standard spectra for mixed minerals, the mixed minerals in the image can be identified and classified as distinct categories. The ratio of absorption depths (Dw) between the spectral absorption features of image pixel spectra and those of standard spectra is calculated to indicate the relative abundance of minerals. This step was illustrated in Figure 6.

Figure 5. Spectra continuum removal (Left: Calculate the average of the selected endpoints for continuum removal; Right: Spectral curve after continuum removal).

Figure 6. Spectral matching and mineral identification.

The relationship between aluminum content and spectral characteristics of muscovite from the USGS spectral library is shown in Figure 7. The position of the muscovite 2200 nm absorption feature is strongly correlated with its aluminum (Al) content. Therefore, the relative aluminum content in muscovite can be indicated by the 2200 nm absorption peak position [23].

The shift in the spectral absorption feature of sericite is determined by calculating the difference in wavelengths corresponding to the positions of maximum absorption depth. The relative abundance of minerals can be obtained by calculating the absorption depth. The process was shown in Figure 8.

Figure 7. The relationship between absorption wavelength shift and Al content of sericite.

Figure 8. The calculationg method of absorption peak position shift and absorption depth.

The exchange of iron (Fe) and magnesium (Mg) often take place in chlorite, resulting in a variety of compositions. The spectral data from the USGS spectral library in Figure 9 shows that the absorption peak position of chlorite is strongly correlated with the ratio of magnesium to iron content [24]. The calculation method of the chlorite absorption feature shift is the same as that of muscovite.

Figure 9. The wavelength of the peak position in chlorite.

4.1.1. Spectral Characteristics and Distribution of Alteration Minerals

Alteration minerals such as pyrophyllite, alunite, kaolinite, montmorillonite, sericite, chlorite, and calcite were mapped in the Bijia Mountain deposit based on the spectral diagnostic features formed by different mineral functional groups and cations.

The spectral diagnostic features of these altered minerals were shown in Figure 10. Kaolinite and montmorillonite have absorption features at 2200 nm; the difference lies in the fact that montmorillonite’s absorption feature is sharper, while kaolinite has a weak feature at a slightly shorter wavelength near 2200 nm. Pyrophyllite was identified based on a sharp diagnostic absorption feature at 2165 nm and another feature at 2320 nm. Alunite has two absorption features in the Atmosphere window at 2190 nm and 2320 nm, which are flatter than the features of pyrophyllite. In addition, alunite and pyrophyllite have a secondary feature at 2210 nm within their primary absorption feature, which is crucial for the identification of the two minerals. Chlorite and calcite have absorption features at 2250 nm and 2340 nm, respectively. Notably, the spectral characteristics of chlorite and epidote are difficult to distinguish in the image, which means that the chlorite identification results in this paper may include epidote.

The distribution of alteration minerals in the Bijia Mountain mining area is shown in Figure 11. Due to the low spatial resolution of images, these minerals are often identified by mixed spectra. Both of the mapping results of EMIT and GF5 identified alunite, pyrophyllite, and pyrophyllite + kaolinite in the mining area. Alunite and pyrophyllite, visible in only a few pixels in a small area, were concentrated in the center of the mapping area, which was surrounded by a mixture of kaolinite and pyrophyllite. The propylitic zone, primarily composed of chlorite, calcite, and albite, commonly occurs at the periphery of the epithermal system. Therefore, the chlorite + calcite areas identified in this study may indicate the location of the propylitic zone. As shown in Figure 11, the altered minerals in the mining area were distributed along an EW trending that is closely related to the fault in this area. In the central region of the ore body, alunite, kaolinite, and pyrophyllite were identified. The transition zone mainly consisted of sericite and montmorillonite, while the propylitic zone was composed of calcite and chlorite.

Figure 10. The spectral features of alteration minerals in the Bijia Mountain deposit.

Figure 11. Mineral mapping results in the Bijia Mountain mine area.

To verify the remote sensing mapping results, four groups of rock samples were collected from the ore body and host rocks that had intensive alteration. Their spectra were measured by a spectrometer and compared with the standard spectra to identify the minerals (Figure 12).

The spectra analysis result showed the samples were primarily containing kaolinite, alunite, sericite, and montmorillonite, which could demonstrate the accuracy of the mineral mapping result.

4.1.2. Spectral Characteristics and Distribution of Sericite and Chlorite

Sericite is an important mineral for tracing ore-forming hydrothermal fluids. The aluminum (Al), iron (Fe), and silicon (Si) could exchange in the silica-oxygen tetrahedron of sericite, leading to the formation of various aluminum contents of sericite in different fluid conditions [25]. Due to the tetrahedral substitution of aluminum and silicon in sericite, the absorption peak position of Al-OH can shift within 2190 nm to 2228 nm, corresponding to exchange between the aluminum and silicon in sericite. For aluminum-rich sericite, the absorption feature is typically less than 2195 nm. The absorption peak position for common sericite is between 2200 nm and 2208 nm, while for low-aluminum sericite (with an Si/Al ratio > 3), the absorption peak position is between 2216 nm and 2228 nm. As shown in Figure 13, sericite was distributed along the EW trend faults on the eastern side of the mining area. Compared to the sericite in the periphery area, the absorption features of sericite in the mining area moved towards shorter wavelengths (2198 nm), which means medium aluminum (Al) content. This result indicates that there is a close relationship between the med aluminum (Al) content of sericite and the mineralized zone.

Figure 12. Measured spectra in the Bijia Mountain mine area.

Figure 13. The Al-OH absorption peak shift of sericite in the Bijia Mountain deposit.

Figure 14 shows the spatial distribution of chlorite with different magnesium contents in the mining area. With an increase in the relative magnesium (Mg) content, the absorption peak of chlorite shifts toward shorter wavelengths. The results show that most chlorite in the mapping area does not exhibit a significant tendency toward iron (Fe) or magnesium (Mg) enrichment. Chlorite with Mg-enriched characteristics is predominantly distributed along the edge of the mining area and on the eastern and western sides of the mineralized alteration zone, whereas a bit of chlorite located in the periphery of the orebody displays Fe-enriched characteristics. Overall, the mapping results suggested that the absorption features of chlorite moved towards shorter wavelengths as chlorite got closer to the ore body.

Figure 14. The absorption peak shift of chlorite in the Bijia Mountain deposit.

4.1.3. Spectral Characteristics and Distribution of Iron-Bearing Minerals

It is difficult to identify pyrite in imagery using near-infrared and short-wave infrared (NIR-SWIR) spectroscopy; the iron-bearing minerals identified in this study are primarily limonite. Due to the effect of noise in the 900 - 1100 nm near-infrared band of GF-5 data, the iron-bearing minerals are identified using EMIT data.

Original mineral pyrite in the Bijia Mountain deposit underwent oxidation on the surface, leading to the formation of limonite under relatively neutral conditions. This process results in limonitization, which serves as a significant surface alteration indicator. The relative abundance of ferric iron (Fe³⁺), derived from absorption depth analysis, indicates that limonitization mainly occurs at the center of the mining area and is spatially associated with minerals such as alunite and kaolinite (Figure 15). Minor occurrences of limonite with sericite and chlorite are found at the edge of the mining area. The mapping results indicate that the abundance and spatial distribution of limonite are closely associated with mineralization.

Figure 15. The absorption depth of iron-bearing minerals in the Bijia Mountain deposit.

Some studies indicated that high sulphur epithermal deposits are the shallow occurrence of porphyry hydrothermal systems, and they form a complete volcanic metallogenic series [26][27]. One of the typical ore-forming model theories suggests that hydrothermal fluids with gold migrate upward from the bottom at low temperature (<300˚C) and form a boiling interface. The fluids migrate through channels with sericite and pyrite in structure fractured zones, and the precious metal deposit with argillic alteration zones was formed in a funnel-shaped tectonic system above the boiling interface [28][29]. According to such research, the vertical alteration zones and minerals in the ore body profiles can indicate the migrating channels of uplifting thermal fluid, which can provide favorable evidence for epithermal deposits prospecting [30][31].

This paper analyzed spectral features on core samples of a 160-meter-deep drill, which was located on the EW-trending fault in the central region of the ore body with advanced argillic alteration on the surface (Figure 16). The alteration minerals in core samples at depths of 40 m to 160 m in the mining area were identified with a field spectrometer. The mineral identification results show that the altered minerals in the core samples mainly consist of kaolinite, sericite, chlorite, alunite, and pyrophyllite. Alunite, pyrophyllite, and kaolinite were widely distributed in samples at depths less than 100 meters, while samples in 100 to 160 meters contain sericite with minor amounts of chlorite (Figure 17). Different from the alteration zones at the surface, alteration minerals at depth lack carbonate minerals, and the chlorite is more iron-enriched, which may be related to the reducing environment and the low temperature of the fluids as they migrate to the surface. The spectral features of argillic alteration minerals and iron in core samples suggest that the drilling locations may be related to the migration channels of ore-bearing hydrothermal fluids.

Figure 16. The geological position of core samples.

Figure 17. The spectrum of core samples.

In epithermal deposits, the aluminum content of sericite is closely correlated with the alteration zone. The aluminum (Al), iron (Fe), and silicon (Si) could exchange in the silica-oxygen tetrahedron of sericite, leading to the formation of various aluminum contents of sericite in different fluid conditions.

During the early stage of mineralization, high-temperature, potassium-rich, and near-neutral magmatic fluids favor the formation of aluminum-rich sericite, which is indicative of a heat source and proximity of the mineralization center [32]. As the fluid cools and undergoes water-rock interaction with the host rocks, the pH decreases and aluminum activity declines, resulting in the formation of sericite with low aluminum content. Therefore, variations in aluminum content serve as an indicator for the transition of the hydrothermal fluid from the early mineralization stage to later-stage alteration [33].

This paper extracted the shift in the 2200 nm absorption feature of sericite to illustrate the variations in aluminum content of the sericite in the mining area.

According to Figure 18, high-aluminum sericite (absorption peak < 2196 nm) is absent in the Bijia Mountain gold deposit, where the sericite is predominantly of medium-aluminum to low-aluminum composition. Medium-aluminum sericite is distributed on the edge of the mineralized center, where it coexists with kaolinite and limonite. The aluminum content of sericite progressively decreases as one gets farther from the orebody. Medium-aluminum content sericite in the mining area indicates a low temperature, neutral to slightly acidic fluid environment, and its formation is closely associated with argillic alteration zones [34]. Variations in sericite aluminum content can delineate alteration intensity gradients and predict the location of a potential mineralization center.

During hydrothermal alteration or metamorphic processes, variations in fluid composition, oxygen fugacity, redox conditions, temperature, and pressure induce Mg-Fe isomorphic substitution in chlorite. Some studies have shown that specific element content, such as magnesium (Mg) and titanium (Ti) in propylitic alteration minerals, had a decreasing trend from the ore-forming rock body to the periphery area of the mineral deposit, making its Mg/Fe ratio a sensitive indicator of the hydrothermal mineralization environment [35]-[37].

The magnesium (Mg) content of chlorite can be described by the shift in the 2250 nm absorption peak. The variation of the chlorite 2250 nm absorption peak within the mining area is shown in Figure 19, which reflects the changes in the relative contents of magnesium (Mg) and iron (Fe) in chlorite.

Figure 18. The absorption peak of sericite in the Bijia Mountain deposit. (A) The number of pixels with a different absorption peak of sericite; (B) The percentage of various types of sericite in different alteration zones.

Figure 19. The absorption peak of chlorite in the Bijia Mountain deposit. (A) The number of pixels with a different absorption peak of chlorite; (B) The percentage of various types of chlorite in different alteration zones.

In the vertical profile, chlorite in core samples of the Bijia Mountain gold deposit exhibits iron-enriched characteristics, whereas surface chlorite is more magnesium-enriched and spatially coincides with areas containing limonite. The Mg/Fe content variation trend of chlorite and the degree of limonitization may reflect the redox conditions during different stages of mineralization, indicating the hydrothermal fluid pathways from depth toward the surface and the associated changes in mineralization conditions with depth: Under the relatively reducing conditions at depth, iron may exist in the divalent state (Fe²⁺) and participate in the formation of iron-rich chlorite. In contrast, under the oxidizing conditions at the surface, Fe²⁺ is oxidized to the trivalent state (Fe³⁺), which favors the formation of limonite and magnesium-rich chlorite.

On the surface of the mining area, iron-enriched chlorite occurs predominantly in the periphery, whereas chlorite within the central part of the mine is relatively more magnesium-enriched. A possible explanation is that chlorite from these two distinct locations formed under different geological environments. Chlorite occurring outside the mining area likely precipitated in a shallow marine sedimentary environment, with iron sourced from marine sediments. In contrast, the magnesium-rich chlorite within the mineralized alteration zone primarily formed during hydrothermal activity. This distinction means chlorite with high magnesium (Mg) content can serve as an indicator for locating the ore body.

Based on the alteration mineral assemblages in the Bijia Mountain gold deposit, the alteration zones on the surface could be identified mainly into three parts (Figure 20). The advanced argillic zone located at the center of the orebody was primarily composed of pyrophyllite, alunite, limonite, and kaolinite. The argillic zone, transition from the centre of the ore body to the edge, mainly consisted of medium aluminum content sericite, kaolinite, and montmorillonite. The propylitic zone was located at the edge of the mining area and was composed of magnesium-rich chlorite and a minor amount of calcite.

Figure 20. The alteration zone indicates mineralization in the Bijia Mountain deposit.

The hydrothermal alteration zoning observed at the surface indicates the presence of a phyllic cap composed of pyrophyllite, alunite, and kaolinite in the shallow portion of the Bijia Mountain gold deposit, marking the center of past volcanic hydrothermal activity. The advanced argillic zone is closely associated with east-west trending structures, suggesting the existence of a low-pressure alteration zone. The medium aluminum sericite occurs adjacent to the center of the orebody and constitutes a key mineral component of the transitional zone within the argillic alteration assemblage. The coexistence of chlorite and calcite at the margin of the mining area reflects a medium- to low-temperature, CO₂-rich hydrothermal environment, characteristic of typical distal alteration. The assemblage and compositional variations of alteration minerals in the Bijia Mountain mining area reflect two distinct mineralization stages—hydrothermal ore-forming and supergene oxidation, which could serve as a key indicator for locating orebodies with remote sensing data.