Using hyperspectral imagery for identification of pyrophyllite-muscovite intergrowths and alunite in the shallow epithermal environment of the Yerington porphyry copper district

Hydrothermal mineral deposits are the primary source of many mineral commodities of global importance. Since hydrothermal alteration minerals associated with the formation of these mineral deposits are active in the visible and infrared range, the analysis of spectral absorption features can be used to identify the mineralogy associated with different alteration events. Some hydrothermal events are responsible for the occurrence of mineral commodities, while other events create hydrothermal alteration unrelated to the introduction of base and precious metals. Therefore, it is crucial to develop a mineral exploration strategy to rapidly identify and map the indicator minerals linked to a mineralising event. The separation of minerals of different alteration events which are spectrally active in the same overlapping range of the spectrum, is the challenge addressed in this study. High spatial resolution airborne and laboratory-based hyperspectral images are combined to detect and visualise textures of muscovite replacing pyrophyllite in the shortwave infrared (SWIR) imaging spectroscopy survey over the Buckskin Range, the volcanic-hosted lithocap part of the Yerington porphyry district, Nevada (USA). Spectral wavelength maps in different SWIR ranges are used to map the hydrothermal alteration mineralogy at both laboratory (26 μm) and airborne (1 m) scales. The airborne spectral data define outward zoning from alunite ± pyrophyllite to muscovite characterized by variable wavelength positions of its Al-OH absorption feature. The wavelength range of 1650–1850 nm is used to differentiate zones of pyrophyllite predominance over alunite within the inner domain. The laboratory data improves the characterisation of the hydrothermal alteration mineralogy, which includes alunite, pyrophyllite, muscovite, dickite, chlorite, topaz and zunyite. The textural relationship of muscovite replacing pyrophyllite is addressed through the development of a novel spectral index, the pyrophyllite-muscovite index (PMI). The characterisation of the intergrowths of pyrophyllite and muscovite at the laboratory scale is based on two aspects: (1) the definition of pervasive versus veinletcontrolled textures and (2) a subtle shift detection in the wavelength position of the Al-OH absorption feature of muscovite from 2189 to 2195 nm. The combination of the spatial patterns with the textural relationship of the pyrophyllite-muscovite association allows the identification of areas which contain the muscovite replacement of pyrophyllite. The recognition of a late muscovite replacement of pyrophyllite suggests that advanced argillic alteration reflecting intense acid leaching is followed by late near-neutral pH magmatic-hydrothermal fluids, adding K and potentially other alkali elements and metals in the epithermal environment. As a result of this study, we document the hydrothermal muscovite-pyrophyllite intergrowth relationships in the study area, thus contributing to an improved understanding of the lithocap epithermal system and a better assessment of its exploration potential for Au, Ag and Cu mineralisation. * Corresponding author. E-mail addresses: b.virgilioportela@utwente.nl (B. Portela), seppm@oregonstate.edu (M.D. Sepp), f.j.a.vanruitenbeek@utwente.nl (F.J.A. van Ruitenbeek), c.a. hecker@utwente.nl (C. Hecker), John.Dilles@oregonstate.edu (J.H. Dilles).

Hydrothermal mineral deposits are the primary source of many mineral commodities of global importance. Since hydrothermal alteration minerals associated with the formation of these mineral deposits are active in the visible and infrared range, the analysis of spectral absorption features can be used to identify the mineralogy associated with different alteration events. Some hydrothermal events are responsible for the occurrence of mineral commodities, while other events create hydrothermal alteration unrelated to the introduction of base and precious metals. Therefore, it is crucial to develop a mineral exploration strategy to rapidly identify and map the indicator minerals linked to a mineralising event. The separation of minerals of different alteration events which are spectrally active in the same overlapping range of the spectrum, is the challenge addressed in this study. High spatial resolution airborne and laboratory-based hyperspectral images are combined to detect and visualise textures of muscovite replacing pyrophyllite in the shortwave infrared (SWIR) imaging spectroscopy survey over the Buckskin Range, the volcanic-hosted lithocap part of the Yerington porphyry district, Nevada (USA).
Spectral wavelength maps in different SWIR ranges are used to map the hydrothermal alteration mineralogy at both laboratory (26 µm) and airborne (1 m) scales. The airborne spectral data define outward zoning from alunite ± pyrophyllite to muscovite characterized by variable wavelength positions of its Al-OH absorption feature. The wavelength range of 1650-1850 nm is used to differentiate zones of pyrophyllite predominance over alunite within the inner domain. The laboratory data improves the characterisation of the hydrothermal alteration mineralogy, which includes alunite, pyrophyllite, muscovite, dickite, chlorite, topaz and zunyite. The textural relationship of muscovite replacing pyrophyllite is addressed through the development of a novel spectral index, the pyrophyllite-muscovite index (PMI). The characterisation of the intergrowths of pyrophyllite and muscovite at the laboratory scale is based on two aspects: (1) the definition of pervasive versus veinletcontrolled textures and (2) a subtle shift detection in the wavelength position of the Al-OH absorption feature of muscovite from 2189 to 2195 nm. The combination of the spatial patterns with the textural relationship of the pyrophyllite-muscovite association allows the identification of areas which contain the muscovite replacement of pyrophyllite. The recognition of a late muscovite replacement of pyrophyllite suggests that advanced argillic alteration reflecting intense acid leaching is followed by late near-neutral pH magmatic-hydrothermal fluids, adding K + and potentially other alkali elements and metals in the epithermal environment. As a result of this study, we document the hydrothermal muscovite-pyrophyllite intergrowth relationships in the study area, thus contributing to an improved understanding of the lithocap epithermal system and a better assessment of its exploration potential for Au, Ag and Cu mineralisation.

Introduction
Epithermal mineral deposits are essential to the mineral commodities industry. Typically formed at shallow depths (<1.5 km), the deposits are a source of Au-Ag ± Cu (Pirajno, 2009;Simmons et al., 2005). Traditionally, these deposits can be classified into low-, intermediateand, high-sulphidation based on the sulphidation state of the hypogene sulphide assemblage (Hedenquist et al., 2000;Sillitoe and Hedenquist, 2003;Simmons et al., 2005). Alternatively, wall-rock alteration mineral assemblages can be used to divide epithermal deposits into geothermal and magmatic-hydrothermal types. The latter is associated with quartz + alunite ± aluminosilicate clays (i.e. pyrophyllite, dickite and kaolinite) assemblages which contain Au-Ag ± Cu ores (Simmons et al., 2005). These hydrothermal alteration minerals are a product of hydrolysis reactions by low-pH magmatic-hydrothermal fluids sourced from hydrous intermediate composition magmas at depth. Zones of strong magmatic-hydrothermal up-flow experience high water to rock ratios, resulting in intense acid leaching of the wall-rock, leaving only residual quartz. As these magmatic-hydrothermal fluids migrate laterally from up-flow zones, interaction with wall rocks neutralise the low-pH fluids, allowing for successively higher-pH stable mineral assemblages to form, from quartz-alunite to alunite + aluminosilicate clays, and finally only aluminosilicate clays. Identification of mineralogic zonation in these systems is useful in locating fluid up-flow zones which can be mineralised with Au-Ag ± Cu ores (Hedenquist et al., 2000;Sillitoe and Hedenquist, 2003;Simmons et al., 2005). In porphyry copper systems, advanced argillic alteration forms from the near-surface to about 1.5 km depth (e.g., Dilles and John, 2020) and has been termed the lithocap environment that lies above or slightly offset over the top parts of porphyry copper systems (Sillitoe, 1995(Sillitoe, , 2010. Many hydrothermal alteration minerals are spectrally active in the visible and infrared range (300-3000 nm) (Clark et al., 1990). By being spectrally active, these minerals absorb incoming (sun) light at specific wavelengths, while the remaining wavelengths are reflected. The absorbed energy creates the so-called "spectral absorption features", which can be diagnostic and used to identify minerals or mineral groups (Hecker et al., 2019c). In various scales, spectral methods have been used for mineralogic identification (van der Meer et al., 2012). Regional overviews of mineralogic occurrences have been generated with multispectral data (few and broad wavelength bands) from satellite-borne instruments such as Landsat TM (Thematic Mapper) (Fraser et al., 1997) and ASTER (Advanced Spaceborne Thermal Emission and Reflectance Radiometer) (Hewson et al., 2005). Detailed regional and local mineralogic analysis overview have been acquired with hyperspectral data (many narrow and contiguous wavelength bands) from airborne sensors in visible, infrared and thermal ranges (Cudahy et al., 2001;Swayze et al., 2014). Even though airborne data are an important tool to address surface mineralogy mapping, the mixture of minerals per pixel, which ranges from > 20 m to 1 m diameter, can be an issue linked to the scale of the data (van der Meer et al., 2012). To address the scale limitations of hyperspectral airborne data and increase the chance of analysing pixels containing single minerals, hyperspectral proximal sensing (high spectral and spatial resolution) have been used on samples and drill cores (Baissa et al., 2011;van Ruitenbeek et al., 2019).
Infrared spectroscopy can detect different minerals produced by hydrothermal alteration as well as some compositional characteristics of the mineral, and these can work as a proxy for hydrothermal fluid chemical composition and pH gradient (i.e., Halley et al., 2015). These different mineral assemblages and compositions can be produced by temporally different hydrothermal alteration events. Some events are responsible for the occurrence of mineral commodities, while other events create mineral alteration without any economic concentrations of economic minerals or resources. Successfully identifying and mapping the alteration minerals associated with a mineralising event is crucial for a better definition of target areas for follow-up mineral exploration studies and may lead to a more objective, effective and cost-efficient mineral exploration strategy. However, the separation of minerals of different alteration events which are spectrally active in the same overlapping wavelength range is a challenge (Bedini et al., 2009;Khashgerel et al., 2009;Lipske and Dilles, 2000;Sepp et al., 2019;Watanabe and Hedenquist, 2001).
In this study, however, we investigate spectrally active alteration minerals which occur together but are potentially associated with several different hydrothermal alteration events which is very common in long-lived porphyry copper and epithermal systems (e.g., Longo et al., 2010;Seedorff et al., 2005;Sillitoe, 2010). We tested our approach on the Buckskin Range at the Yerington district, Nevada (USA), a highsulphidation epithermal system, where Lipske (2002) provided evidence that pyrophyllite from an early-stage of advanced argillic alteration is locally replaced by muscovite during a later-stage sericitic alteration. We used laboratory-based hyperspectral images (26 µm pixel spacing) to develop a workflow to detect and visualise replacement textures of pyrophyllite-muscovite in SWIR imaging spectroscopy. High spatial resolution airborne hyperspectral imagery (1 m pixel spacing) provided average spectral responses which were comparable to the laboratory measurements. Hence, replacement textures in airborne hyperspectral imagery could be interpreted from the laboratory study. We characterised the textural relationship of the pyrophyllite and muscovite obtained from a novel spectral index designed for the laboratory data to reconstruct the hydrothermal alteration events of the area. Detailed ground-based geological and hydrothermal alteration maps, as well as mineralogical and chemical investigations, are presented in more detail in companion manuscripts (Sepp et al., 2019, in prep.).

Geologic setting
The Yerington area of western Nevada (USA) is a mining district that includes a variety of ore deposits such as porphyry Cu (Mo), Cu skarns, and both large Cu-Au-Fe oxide deposits and small lodes . The Buckskin Range, located in the western part of the Yerington District, is defined by two volcanic sequences, namely Artesia Lake and Fulstone Spring Volcanics (Dilles and Wright, 1988;Lipske and Dilles, 2000;Proffett andDilles, 1984, 2008). The Buckskin Range hosts several lithocaps or shallow advanced argillic and sericitic alteration zones which are the discharge areas for magmatic-hydrothermal fluids that produced porphyry Cu(Mo) deposits at greater depth (Lipske and Dilles, 2000). For this study, we focused on a single lithocap, named Alunite Hill by Lyon and Honey (1990), which is located in the central Buckskin range (central study area, Fig. 1).
The hydrothermally altered rocks of the study area are hosted by the Jurassic Artesia Lake Volcanics, which in the Buckskin Range consist of an 800-2000 m thick sequence of sub-aerial andesitic lava flows and breccias, volcaniclastic sandstones and conglomerates, and thin dacitic ignimbrites (Proffett and Dilles, 2008). Due to intense hydrothermal alteration, the primary rock textures were variably and partly destroyed and the primary mineralogy is extensively overprinted in zones of advanced argillic and sericitic alteration. Advanced argillic alteration zones range from ~5 m to > 100 m width and occur along with both near-vertical fractures or faults and gently dipping zones paralleling the volcanic stratigraphy. In the central part of the Buckskin Range there are three principal assemblages: (1) quartz-alunite-pyrophyllite; (2) quartzpyrophyllite-sericite; and (3) quartz-alunite-sericite (Lipske, 2002;Lipske and Dilles, 2000;Sepp et al., 2019). Sericitic alteration envelops the advanced argillic alteration in zones that are meters to tens of meters in width and is characterised by sericite (muscovite)-quartz-pyrite assemblage and a more widely distributed sericitechloritequartz ± pyrite ± hematite zones containing some stable albite and K-feldspar (Lipske and Dilles, 2000) (Fig. 1A, B).
Four aspects make this area highly suitable as a test site to define the distribution of pyrophyllite-muscovite and the proportions of the two minerals. First, the desert exposures have sparse vegetation and limited alluvial or regolith cover. Secondly, the rock exposures in the Yerington area represent a cross-section of the Middle Jurassic Yerington batholith magmatic and hydrothermal system, from paleodepths of 2-7 km in the plutonic and porphyry systems to < 2 km paleodepth of shallow sections in the volcanic environment of the Buckskin Range . These rock exposures and all pre-Miocene rocks are exposed in crosssection and have been tilted 60 • -90 • towards the west as the result of block-rotation associated with a series of east-dipping Miocene to Recent normal faults (Dilles and Einaudi, 1992;Dilles et al., 2000;Proffett, 1977). Thirdly, the pyrophyllite and muscovite paragenesis formed in the lithocap as part of the Yerington porphyry Cu (Mo) system but it developed as the result of multiple successive hydrothermal alteration events by fluids of slightly different composition. Fourthly, the volume of ground-based mapping, petrology, mineralogy, petrology, chemical and geochronology studies conducted at Yerington (e.g. (Proffett, 1977;Proffett and Dilles, 1984;Dilles and Wright, 1988;Dilles and Einaudi, 1992;Dilles et al., 2000;Lipske and Dilles, 2000;Longo et al., 2010;Sepp et al., 2019) makes this area highly suitable to calibrate airborne hyperspectral data.
The geological setting of the Buckskin Range epithermal or lithocap environment is well established. In the Buckskin Range there are three main exposures in the southern, central (Alunite Hill, area of this study) and northern parts, which have been studied respectively by Hudson and Oriel (1979), Hudson (1983), Lipske (2002), Lipske and Dilles (2000), Lyon and Honey (1990), Riedell et al. (2020), Rubin (1991), Runyon et al. (2015), respectively. In the central Buckskin Range, about 800-1000 m of Artesia Lake Volcanics are exposed. The shallowest exposure being the western part of the Range. Throughout these exposures, advanced argillic alteration is represented by coarse-grained pyrophyllite, fine-grained alunite and variable amounts of muscovite. Whereas dickite is only rarely present in shallowest exposures, other hypogene kaolinite polymorphs are absent or sparse, and illite is very rare (Lipske, 2002). Based on the general stability of these minerals and assemblages, coarse pyrophyllite formed at 280-360 • C and muscovite formed at ≥ 300 • C (Henley and Adams, 1992;Seedorff et al., 2005). The estimated pressure of the formation, assuming boiling low-density fluids is, therefore, >40-50 bars or > 500 m depth (Dilles and John, 2020). Lipske (2002) and Sepp et al. (2019) employed electron microprobe BSE (backscattered electron) imaging and QEMSCAN® (Quantitative Evaluation of Materials by Scanning Electron) mineral mapping to investigate coarse-grained pyrophyllite with sheaf and radiating structures that are intergrown along the basal sheet silicate planes (0 0 1) by muscovite and locally overgrown by muscovite. Based on these textures, the pyrophyllite formed during an early-stage hydrothermal event, while muscovite precipitated during a subsequent hydrothermal overprint. Further geological details are presented in Sepp et al. (in prep). Such a result is consistent with the theory of epithermal advanced argillic alteration, which has been proposed to be due to condensation of SO 2 -rich aqueous vapour into shallow groundwater and generation of acidic conditions by abundant H 2 SO 4 and H 2 S (Bethke et al., 2005;Henley and McNabb, 1978;Sillitoe, 2010;Simmons et al., 2005). In contrast, later muscovite was formed by less acidic aqueous fluids that were condensed fluids and relatively sulphur-poor when they ascended upward into the lithocap. Dilles et al. (2000) and Lipske and Dilles (2000) proposed that such a temporal evolution of magmatichydrothermal fluid likely reflects the crystallization of the source magma and derivation of early-stage fluids from shallower magma depths compared to late-stage fluids (~4 km vs 7 km). This temporal evolution is also recorded by the cooling-related change from potassic to sericitic alteration assemblages observed in most porphyry copper deposits (Gustafson and Hunt, 1975;Seedorff et al., 2005).

Material and methods
Surface mineralogy and the spatial distribution of the pyrophyllitemuscovite association are investigated using both airborne and laboratory data through the interpretation of the reflectance spectra and by mapping the wavelength position and depth of absorption features using the Minimum Wavelength Mapper.
A novel spectral index is used to highlight the relative proportion of the two minerals at the laboratory scale as well as their textural relationships. The textural relationships are compared with the scanning electron microscope-backscattered electron (SEM-BSE) images and merged with the information derived from the airborne wavelength maps and hydrothermal alteration mapping of the area. The methodology is explained in detail in the following sections.

Field data
This data set consists of 62 rock samples of the central study area, collected during field mapping of the study area between 2017 and 2018. These rock samples are cut to prepare thin sections for petrographic analysis and thin sections rocks slabs (billets) for the laboratorybased hyperspectral imagery. Petrographic data consists of field observations confirmed by using a FEI QUANTA 600F scanning electron microscope at Oregon State University (USA). Phyllosilicate composition is confirmed using spot a EDAX Si (Li) energy dispersive spectrometer (EDS).

Laboratory hyperspectral data
High-resolution hyperspectral images of the sample subset are acquired with the SisuCHEMA hyperspectral imager at Faculty ITC -University of Twente (The Netherlands), a high-speed push-broom system developed by Spectral Imaging Ltd., Finland (Specim, 2015). The SWIR setup consists of a Specim SWIR-LVDS-100-N25E camera and OLESMacro lens acquiring images with 384 pixels per line and a spatial resolution of 26 µm. It operates in a spectral range of 930-2540 nm, with a spectral resolution (FWHM) of 10 nm, a spectral sampling of 6.3 nm, and 288 bands.
The samples are wider than the field of view of the imager. Thus, each sample is covered by three or four strips, depending on the width of the sample. Each image strip is calibrated to reflectance with dark current and white reference measurements. Noisy bands at the beginning and end of the spectral range are removed. De-striping and a spectral mean filter are then applied to remove artefacts originated by calibration errors and to smooth the spectra Hecker et al., 2019a). The strips are mosaicked, and a second spectral subset from 1100 to 2400 nm is made.

Airborne hyperspectral data
This data set consists of five flight lines of 1 m pixel size that were acquired over the Yerington district, Nevada (USA) during the Joint Airborne Collection using Hyperspectral Systems (JACHS) with the Aerospace Corporation's SEBASS instrument and SpecTIR LLC's Pro-SpecTIR sensor system (Hecker et al., 2019b). The ProSpecTIR sensor operates in a spectral range of 400-2500 nm, with a spectral resolution of 5 nm, a sampling interval of 10 nm, and 357 bands. The thermal infrared (TIR) data acquired with the SEBASS (Spatially-Enhanced Broadband Array Spectrograph System) instrument is not used in this study. The flight lines provided by SpecTIR LLC are pre-processed to surface reflectance by radiometric and spectral calibration (SpecTIR, 2008). A spectral subset from 1100 to 2400 nm is created, similarly to the laboratory data.
The flight lines are processed separately. The outputs are mosaicked and spatially subset to the limits of the study area.

Minimum wavelength mapper
This method is based on the characterisation of the spectral absorption features such as the position and depth. It generates a proxy for mineral composition, mineral abundance and spatial patterns in a single colour-coded map (Bakker et al., 2011;Hecker et al., 2019c). The Minimum Wavelength Mapper is included in the Hyperspectral Python (HypPy) software developed by Wim Bakker of the Faculty ITC -University of Twente, Enschede, The Netherlands (Bakker, 2020).
The Minimum Wavelength Mapper allows the calculation of a proxy for mineral composition and can detect compositional change via small shifts detection in the wavelength position. These aspects are essential aspects for the selection of the Minimum Wavelength Mapper method. Methods based on the use of mineral end members or library spectra such as spectral angle mapper (SAM) and linear spectral unmixing (LSU) cannot detect subtle changes in the wavelength position of absorptions features or represent it as different endmembers (van Ruitenbeek et al., 2006).
The Minimum Wavelength Mapper method pre-processing consists of a continuum removal (hull quotient) over the wavelength range of interest to characterise accurately the spectral absorption features. Continuum removal is a technique to remove the overall background (shape) of a spectrum and highlight individual absorption features for quantitative analyses (Clark and Roush, 1984;Hecker et al., 2019c). The processing consists of a 2-step procedure. First, the interpolated wavelength position and depth of the deepest absorption feature are calculated through parabolic interpolation. Second, these parameters are combined in a single colour-coded map.
Both steps allow spectral subset adjustments. In step 1, the spectral subset focuses on the spectral range where the deepest absorption feature is searched. In step 2, a narrow spectral subset and a depth stretch (brightness) definition are possible. This helps to suppress noise from indistinct features in the resulting images (Hecker et al., 2019c).
For surface mineralogy identification and spatial distribution determination, wavelength maps are generated in the 2100-2400 nm range for both data sets. Mineral groups like phyllosilicates, sorosilicates, sulphates and carbonates present their main absorption features at this range (Clark et al., 1990). Wavelength maps are produced between 1650 and 1850 nm and between 2150 and 2250 nm to cover most diagnostic absorption features. A 2150-2250 nm map is also produced to enhance the difference between the minerals present, especially pyrophyllite (2166 nm) and muscovite (2180-2228 nm) (Pontual et al., 1997).
The airborne wavelength maps are produced with an automatic depth stretch (30 to 90 percentile stretch) to create higher consistency between flight lines and reduce the display of indistinct features (Hecker et al., 2019c). The spectral subset of step 2 is at the same range as the spectral range in step 1 for all maps with exception to the 2150-2250 nm map. For this case, step 1 is 2100-2400 nm and the spectral subset on step 2 is 2150-2250 nm. For the laboratory wavelength maps, the only difference is the adjustment of the depth stretch from 0 to 50% depth.
The automatic depth stretch excludes the 30% pixels with the shallowest absorption feature by turning them "black" (Hecker et al., 2019c). The manual depth stretch prevents that and facilitates the display of the mineralogy in the sample, especially at the border of each crystal.

Spectral index
A spectral index indicates the relative abundance of absorption features by combining spectral parameters extracted from reflectance spectra (Oniemayin et al., 2016;van Ruitenbeek et al., 2005). We develop a spectral index for identifying the relative abundance of pyrophyllite and muscovite by enhancing the position of diagnostic absorption features and we named it the pyrophyllite-muscovite index (PMI).
The PMI is defined by the following equation: where r pyro (Al− OH) is the reflectance measured at the wavelength position of the Al-OH deepest absorption feature of pyrophyllite (2166 nm) and r musc(Al− OH) is the reflectance measured at the wavelength position of the Al-OH deepest absorption feature for muscovite (in the range 2180-2228 nm) (Fig. 2). For the muscovite Al-OH absorption feature, the deepest feature is derived from the results of the wavelength maps. The index is calculated based on a continuum removed spectra (hull quotient), that is similar to the Minimum Wavelength Mapper method previously mentioned.
The PMI results are sliced into ten classes. Threshold values for each class are determined from a synthetic linear mixture of the purest pyrophyllite and muscovite spectra. The wavelength maps are used to select the pixels which were most representative for a pure pyrophyllite and muscovite spectrum, based on the high intensity (deep) features. These pure spectra are characterised as the endmembers of the linear mixture, and then, mathematically mixed on a 10% interval basis. After the ratio is applied to the synthetic mixtures, the results are used to determine the threshold values for the ten classes ranging from pure pyrophyllite to pure muscovite.

Hydrothermal alteration mineralogy
The airborne wavelength map of deepest absorption features in the range of 2150-2250 nm detects a variation in the range of 2165-2225 nm (Fig. 3A). Within this interval, the absorption features related to the short wavelength ranges (2165-2175 nm) correspond to alunite ± pyrophyllite (light cyan). In contrast, features associated with the longer wavelength range (2180-2225 nm) correspond to muscovite (green to orange). Dark areas indicate pixels with a very shallow absorption feature.
From the central study area, two domains are defined: an inner domain characterised by alunite ± pyrophyllite (cyan hues) and an outer domain of short to long-wavelength muscovite (green to yellow to orange hues). Fig. 3B shows example spectra of these domains. The spectral curves are extracted from airborne data at different locations of the central study area, as indicated by the numbers in the wavelength map (Fig. 3A).
The laboratory measurements are used to produce wavelength maps of hand samples at 1650-1850 nm and 2150-2250 nm that detects additional hydrothermal alteration minerals compared to those already detected in the 1 m airborne pixels. In addition to pyrophyllite, muscovite and alunite, accessory dickite, kaolinite (weathering-related), chlorite, topaz and zunyite is also detected. Fig. 4 shows example spectra of the mineralogy identified in the laboratory domain. The wavelength maps also indicate that some samples are dominated either by  − OH) ). For muscovite, r musc (Al− OH) is the reflectance measured at the wavelength position (w musc (Al− OH) ) of the deepest Al-OH feature (musc (Al− OH) ), which lies in the range of 2180-2228 nm.
The hydrothermal alteration mineralogy is characterised in the laboratory to produce wavelength maps, following a similar procedure to generate the airborne 2150-2250 wavelength map and these spot sample are displayed atop the airborne map of the central study area (Fig. 6A). Muscovite-rich samples of intermediate (yellow circles, Fig. 6) to long-wavelength (orange circles, Fig. 6) form a halo around alunite ± pyrophyllite-rich samples (blue symbols, Fig. 6). Laboratory samples with short-wavelength muscovite (green circles, Fig. 6) verify regions  with short-wavelength muscovite in the airborne data (green, Fig. 6). Laboratory samples with short-wavelength range muscovite are also found at the contact between short-wavelength muscovite (green, Fig. 6) and alunite ± pyrophyllite identified in the airborne data (light cyan, Fig. 6). Similarly, the laboratory wavelength maps follow the pattern verified in the hydrothermal alteration map (Fig. 6B).
The airborne wavelength map suggests that long-wavelength muscovite does not occur in direct contact with the alunite ±  pyrophyllite in adjacent domains. The laboratory wavelength maps of hand samples also confirm that pyrophyllite and long-wavelength muscovite do not occur together, since pyrophyllite and muscovite cooccurrence is characterised by the presence of a short-wavelength muscovite (<2200 nm).

Laboratory scale
Textural relationships between pyrophyllite and muscovite are interpreted using the 2150-2250 nm laboratory wavelength maps and the classified images of the PMI (Fig. 7). PMI classified images highlight the occurrence of pyrophyllite and muscovite in each sample, as well as the transition from a pyrophyllite-to a muscovite-rich zone (Fig. 7).
To identify the intergrowth of pyrophyllite and muscovite at the laboratory scale, 2150-2250 nm wavelength maps are constructed, highlighting alteration textures for samples with both minerals. In the wavelength maps and PMI images, pyrophyllite and muscovite occur as discrete zones in which, one of the two minerals is dominant. Sample R875478 (Fig. 7.2A-C) is interpreted as a combination of veinletcontrolled and pervasive muscovite alteration that has replace a preexisting pyrophyllite-rich assemblage. The iron oxide occurrences (colourless/black pixels, Fig. 7.2B. C) also reinforce the interpretation of a veinlet-controlled texture. Samples R875367 (Fig. 7.1 A-C) and R875524 (Fig. 7.3 A-C) are interpreted as a nearly pervasive replacement by a muscovite-rich assemblage of a pre-existing pyrophyllite-rich assemblage. However, this last interpretation is a result of detections in BSE images (Fig. 9.1-2A, 1-2C) rather than the combination of wavelength maps and PMI results.
Therefore, since a clear muscovite replacement texture of pyrophyllite is only observed in one sample (R87478, Fig. 7.2A-C), an alternative method is developed to attempt to identify the temporal relationship between muscovite and pyrophyllite in samples where wavelength maps and PMI images are ambiguous (e.g. Fig. 7.1C and 7.3C). This method is based on the detection of a subtle wavelength shift in the position of the Al-OH absorption feature of muscovite from 2189 nm in pyrophyllite-rich zones to 2195 nm in muscovite-rich zones. This shift is detected in all samples where pyrophyllite and muscovite co-occur and have a common boundary.
In Fig. 8, the wavelength position of the absorption feature varies from 2189 nm in pyrophyllite-rich zones (as the second deepest absorption feature, since the first was at 2167 nm) to an absorption feature at 2189-2195 nm close to the boundaries between pyrophyllite and muscovite-rich zones (near the dashed red line, Fig. 8B). In muscoviterich zones where the Al-OH absorption feature of muscovite is the deepest absorption feature, the wavelength position shifts to 2195 nm. The shift is also indicated through a spectral curve of three different zones: muscovite-rich, boundaries between the two zones and pyrophyllite-rich (Fig. 8C, D). Besides, the spectra show that these zones are intimate mixtures of the two minerals and that even at 26 µm resolution pure pixels are rare.
The initial consideration of the shift of the wavelength position of the Al-OH absorption feature for muscovite is a mixed pixel of pyrophyllite and muscovite, especially at the boundary between both minerals. However, synthetic linear mixtures created for the PMI show that a mixture of the two minerals would result in a shift of the wavelength position of the Al-OH absorption feature for both pyrophyllite and muscovite. In Fig. 8.C-D, no shift is observed in the pyrophyllite feature at 2167 nm. Thus, the mixed pixel explanation is not very likely.
Therefore, the shift in the wavelength position of the Al-OH absorption feature of muscovite is interpreted as the result of the intergrowth of pyrophyllite and muscovite, independently of the temporal relationship of both minerals. Sepp et al. (2019) documented crystal lattice distortions in muscovite crystals due to the presence of interlayered pyrophyllite sheets. These crystal lattice distortions in the muscovite crystals likely affect the vibrational energy of Al-OH bonds, resulting in a change in the absorptive energy of the Al-OH absorption feature (Scott and Yang, 1997).
The timing of the intergrowth of muscovite and pyrophyllite in the samples analysed in Fig. 7 is verified using BSE images, with the composition of the phyllosilicates confirmed using EDS spectrometry. In Fig. 9, the crystals of pyrophyllite are covered by sheets of muscovite (red circle). By analysing the intergrowth of the crystals, the temporal relationship between both minerals is defined as muscovite replacing pyrophyllite in several cases. For example, pyrophyllite in the Buckskin Range commonly form characteristic coarsely crystalline radiating   8. Representation of the shift of the wavelength position of the Al-OH absorption feature of muscovite in boundaries between pyrophyllite-rich (red, panels B, C and, D) and muscovite-rich zones (green, panels B, C and, D). The transition zone is indicated by the blue spectral curve (panel c and d). The absorption features linked to each mineral are highlighted in the spectral profile (panel c and d). In panel d, the spectral curve was enlarged to indicate that the absorption feature is deeper at 2189 nm for pyrophyllite-rich zones and deeper at 2195 nm at muscovite-rich zones. sheaves. Replacement of this pyrophyllite by muscovite is characterised by coarse muscovite forming along pyrophyllite cleavages or as euhedral sheets with interstitial anhedral pyrophyllite ( Fig. 9.1A-2A; 1B-2B). Additionally, narrow muscovite overgrowths locally form atop pyrophyllite crystals (Fig. 9.1C-2C).
Whereas alunite and pyrophyllite have overlapping features at 2175 nm (with some variation depending on the spectral resolution of the sensor), only alunite has a feature at 1760 nm. Therefore, the combination of the 1650-1850 nm and 2150-2250 nm wavelength maps is used to indicate advanced argillic alteration areas where pyrophyllite has predominance over alunite (dashed magenta polygon; Fig. 10.A-B). The ability to effectively delineate alunite-versus pyrophyllitedominant zones of hydrothermal alteration from airborne hyperspectral imagery is a useful tool to aid in the understanding of the magmatic-hydrothermal fluid history of the study area. SO 2 sourced from upward ascending magmatic-hydrothermal vapours will be disproportionate to H 2 SO 4 as these vapours cool and condense into local ground water by the following equation:  (Kokaly et al., 2017). Grey bands in the background indicate the range of the wavelength maps (left band: 1650-1850 nm; right band: 2150-2250 nm). Light red bands in the background indicate the range of atmospheric interference due to hydroxyl (OH) and water absorption, respectively (left band: 1350-1550 nm; right band: 1880-2040 nm).
The formation of alunite-dominant hydrothermal alteration requires higher concentrations of H 2 SO 4 compared to pyrophyllite; accordingly, areas where magmatic-hydrothermal vapours have fluxed through for a protracted period will contain abundant alunite alteration (Einaudi et al., 2003;Hedenquist and Browne, 1989;Hedenquist and Lowenstern, 1994;Hemley et al., 1969). As these mixed magmatic-and groundwaterderived fluids move along paleo groundwater paths, they are progressively neutralised by wall rock interaction and further mixing with meteoric waters and the stable hydrothermal alteration changes to pyrophyllite (or dickite/kaolinite at temperatures < 260 • C) and then, muscovite/illite ± chlorite, forming the classic high-sulphidation alteration zonation pattern first described at Summitville, Colorado (Einaudi et al., 2003;Gray and Coolbaugh, 1994;Meyer and Hemley, 1967;Simmons et al., 2005;Steven and Ratte, 1960;Watanabe and Hedenquist, 2001). Accordingly, the ability to map the distribution of alunite versus pyrophyllite from airborne hyperspectral imagery allows for the identification of areas into which greater amounts of magmatic sulphur was introduced (alunite occurrence), which may be more prospective for Au-Cu mineralization (Sillitoe, 2010;Simmons et al., 2005).
The confinement of the advanced argillic alteration within a broader halo or zone of sericitic alteration characterises the outward zoning alteration pattern in high-sulphidation epithermal systems discussed by Simmons et al. (2005). By distinguishing alunite-versus pyrophyllitedominant zones of hydrothermal alteration, a proxy for pH variation can be established in this high-sulphidation alteration zonation pattern. Similarly, muscovite of different wavelength ranges would also work as a vectoring tool to characterise fluids with different pH gradients. Halley et al. (2015) discussed that muscovite of a shorter wavelength range (<2205 nm) indicates a more acidic environment while longer wavelength muscovite (phengite) indicates a more neutral environment.
Within the advanced argillic alteration, discrete zones of shortwavelength muscovite (2187-2200 nm) (green pixels, Fig. 11) are observed. Some of these discrete zones correspond to the presence of chlorite-muscovite-feldspar alteration of andesite dikes (red dashed circle, Fig. 11); however, other zones do not appear to be attributed to a change in rock type or mapped hydrothermal alteration (Fig. 1A-B). Laboratory samples where the late muscovite replacement of pyrophyllite has been documented correspond to these short-wavelength muscovite zones (2187-2200 nm) within the advanced argillic alteration (magenta ovals, Fig. 11). This observation suggests that at least some of the short-wavelength muscovite zones (2187-2200 nm) characterised by green pixels in Fig. 11 are due to muscovite replacement of pyrophyllite. However, care must be taken in interpreting these data due to the spectral similarity of chlorite-muscovite alteration and muscovite replacement of pyrophyllite. We propose that the muscovite replacement of pyrophyllite observed at the 26 µm laboratory scale is detectable in the 1 m airborne hyperspectral data set over Alunite Hill. Due to this similarity in spectral response, field observation and laboratory-based hyperspectral studies are required to be able to interpret the airborne hyperspectral data effectively. Nonetheless, with the aid of groundtruthing, it is possible to map this late replacement of pyrophyllite by muscovite with airborne hyperspectral imagery.

Replacement detection in hyperspectral imagery for mineral exploration
Intergrowths of muscovite and pyrophyllite have also been identified in the advanced argillic alteration at the El Salvador (Chile), Oyu Tolgoi (Mongolia), and Sunnyside (United States) deposits (Khashgerel et al., 2009;Sepp et al., 2019;Watanabe and Hedenquist, 2001). Though Watanabe and Hedenquist (2001) and Khashgerel et al. (2009) suggest that pyrophyllite replaces muscovite at El Salvador and Oyu Tolgoi, the textural relationships in these studies are difficult to interpret and do not show clear cross cutting relationships. Detailed mineralogic studies (SEM, QESCAM, XRD and EMPA) by Lipske (2002) and Sepp et al. (2019 and in prep.) have demonstrated and established that in the Yerington district, muscovite replaces pyrophyllite. The results of our detailed 26 µm laboratory hyperspectral imagery analysis agree with the paragenetic relationship of the muscovite replacement of pyrophyllite.
The recognition of a late muscovite replacement of pyrophyllite has significant implications for the understanding of the magmatichydrothermal fluid history of the Alunite Hill area. Numerous fieldbased and experimental studies have suggested that quartz and quartzalunite ± aluminosilicate clay alteration is a result of an intense acid leaching and cation removal by low-pH magmatic-hydrothermal derived fluids (Einaudi et al., 2003;Hedenquist and Lowenstern, 1994;Sillitoe, 1977Sillitoe, , 2010Simmons et al., 2005;Steven and Ratte, 1960). Experimental studies demonstrate that muscovite is not the stable aluminosilicate clay phase with quartz-alunite alteration under typical magmatic-hydrothermal conditions (Hemley et al., 1969;Meyer and Hemley, 1967). Therefore, the replacement of pyrophyllite by muscovite suggests that late fluids were near-neutral pH and chlorine-rich, adding K + and other alkalis, allowing for the formation of muscovite. Indeed, past fluid inclusion studies have suggested that late, moderate salinity fluids may overprint earlier advanced argillic alteration and transport metals, including Au, Ag and Cu into the epithermal environment (Bethke et al., 2005;Rottier et al., 2018;Rusk et al., 2008). Therefore, this study demonstrates that the ability to identify and map this replacement in airborne and laboratory-based hyperspectral imagery can be a useful tool in assessing the exploration potential of advanced argillic alteration for Au, Ag and Cu mineralisation.

Directions for future work
The quartz-dominated feeders structures (orange, Fig. 1C) can be an important aspect to the discussion on the reconstruction of the temporal changes in fluid chemistry. As discussed by Sepp et al. (2019), these feeder channels likely represent permeable zones, perhaps enhanced by intense acid leaching, that provided access for later fluids. Since quartz and K-feldspar are not spectrally active in the SWIR range, this study should be extended and complemented with the use of longwave infrared (LWIR) to improve the characterisation of these structures.
Combining airborne SWIR and LWIR hyperspectral imagery could potentially strongly increase the volume and detail of information on the study area.
The results and discussion presented above demonstrate a rapid and effective identification and interpretation of replacement textures through spectral data. This can be used to reduce target areas for sample collection in mineral exploration campaigns. Nevertheless, this methodology could also be extended to areas in which reverse replacement textures are observed. This way, the shift in the wavelength position of the Al-OH absorption feature would be characterised in both scenarios. It is recommended to use different testing sites for an improved understanding of the spatial and textural alteration patterns.
Since replacement texture identification requires a high spatial and spectral resolution data, a combination of hyperspectral images and drones could be an efficient tool for improving the detection of these textures.

Conclusions
Replacement of pyrophyllite by muscovite in the Yerington district has previously been documented. The intergrowths of muscovite and pyrophyllite has been detected as a shift in the wavelength position of the Al-OH absorption feature to approximately 2185-2197 nm. Our study confirms the results of these past studies and expands them by using a combination of high spatial resolution airborne and laboratorybased hyperspectral images to detect and visualise replacement textures of pyrophyllite-muscovite in SWIR imaging spectroscopy both in airborne and laboratory surveys.
The documentation of replacement textures at the laboratory scale is based on two aspects: (1) the definition of pervasive and veinletcontrolled textures and (2) a subtle shift detection in the wavelength position of the Al-OH absorption feature of muscovite from 2189 to 2195 nm by using a novel spectral index, the pyrophyllite-muscovite index (PMI). The 2189 nm muscovite is here interpreted as a replacement phase of pyrophyllite. The temporal relationship between overprinting muscovite and early-stage pyrophyllite is directly determined for several samples based on petrographic textures, 26 µm wavelength maps of the Al-OH position in hand samples and backscattered electron images of pyrophyllite-muscovite intergrowths and spatial zonation, respectively. Replacement textures are detected only in a few samples since the majority are dominated by either muscovite or pyrophyllite.
The airborne hyperspectral imagery survey (1 m) allows the identification and delineation of the broad zonation of advanced argillic, sericitic and chlorite-muscovite alteration zones and when combined with the laboratory hyperspectral data, additionally allows the identification of zones where muscovite has partially or completely replaced pyrophyllite. The 1650-1850 nm wavelength range is used to delineate alunite-versus pyrophyllite-dominant zones of advanced argillic hydrothermal alteration and provide a proxy for pH variation in these zones. Discrete zones of short-wavelength muscovite (2187-2200 nm) are observed within the advanced argillic alteration. Some of these discrete zones are associated with the presence of chlorite-muscovitefeldspar alteration of andesite dikes. By contrast, at least some zones of short-wavelength muscovite (2187-2200 nm) indicate areas representing the muscovite replacement of pyrophyllite, as verified in the laboratory samples. The recognition of a late-stage muscovite replacement of pyrophyllite suggests that advanced argillic alteration and intense acid leaching are followed by near-neutral pH and chlorine-rich late-stage fluids, adding K + and other alkali elements and metals into the epithermal environment.
Our study demonstrates that hyperspectral SWIR imagery is a useful tool to characterise and detect the muscovite replacement of pyrophyllite at different scales. Importantly, it also reveals that the application of these methods both contributes to an improved understanding of epithermal systems and potentially to a better assessment of the exploration potential of advanced argillic alteration and lithocaps for Au, Ag and Cu mineralisation.

Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.