Introduction

Rapid industrialization, infrastructural growth and societal needs have concerned the mineral wealth to be of utmost importance. Mineral exploration has been considered as the backbone for raw materials and industrial growth of a region. The study of mineral alteration to assess the mineral deposit, via remote sensing, has been a widely used approach (Sabins 1999; Goetz 2009; Van der Meer et al. 2012). These mineral alterations can be studied using multispectral (Landsat TM/ETM + and ASTER) and hyperspectral imaging sensors ( Sabins 1999; Abrams 2000; Goetz 2009; Van der Meer et al. 2012).

The Landsat Thematic Mapper (TM)/Enhanced Thematic Mapper (ETM +) is often used for mapping mineral alteration, produced due to hydrothermal processes. These mineral alterations can be mapped using band ratio of Band 5 of short-wave infrared (SWIR 1) and Band 7 (SWIR 2) of TM and ETM + (Carranza and Hale 2002; Pour et al. 2017). However, because of the limited band (Band 5 and Band 7) in the SWIR band, Landsat image data cannot effectively resolve the various types of alteration (Cŕosta et al. 2003; Pour and Hashim 2012). ASTER with six spectral bands in the short-wave infrared (SWIR) has the capability to map different alteration minerals in comparison with Landsat Thematic Mapper (TM)/Enhanced Thematic Mapper (ETM +) (Abrams 2000; Cŕosta et al. 2003; Pour and Hashim 2012). For mapping mineral alteration, the most routinely used techniques are the principal component analysis (PCA) and the band ratio techniques. Principal component analysis (PCA) increases the contrast between the bands by decorrelating the band to band information and organizing the data into decreasing signal-to-noise ratio. The resultant increase in contrast will highlight the mineral of interest by selecting the appropriate principal component, based on the reflectance and absorption feature of the mineral (Sabins 1999; Cŕosta et al. 2003; Van der Meer et al. 2012; Pour and Hashim 2012). Band ratio is the technique where the ratio of band corresponding to the reflectance of the targeted mineral and band corresponding to the absorption feature of the same mineral is taken. For example, ASTER Band 7/6 is used to identify the muscovite, whereas Band 8/9 is used to highlight features pertaining to Al–OH vibrations (Pour and Hashim 2015; Pour et al. 2017). However, with fewer bands and the coarser spectral resolution, in SWIR region, the precise mapping of minerals is not possible (Hunt 1979; Mars and Rowan 2006).

Hyperspectral imaging spectroscopy, with its hundreds of narrow and continuous bands, is found to be more efficient in identifying Earth’s material, through their spectral signature compared to multispectral imaging sensor (Landsat and ASTER) (Shaw and Burke 2003; Taylor et al. 2007; Goetz 2009; Li et al. 2012; Van der Meer et al. 2012; Krishna Mohan and Porwal 2015; Ramakrishnan and Bharti 2016; Khanna et al. 2018). Out of hundreds of bands, the bands in short-wave infrared (SWIR) region of hyperspectral imaging sensors are very useful for mapping mineral alteration in hydrothermally altered minerals (Laakso et al. 2015; Graham et al. 2017). The hyperspectral data have been found suitable for identification and exploration of the mineral alterations in Alaska, USA (Kruse et al. 1993; Harraden et al. 2013; Swayze et al. 2014; Kokaly et al. 2016; Graham et al. 2017), Australia (Mauger et al. 2007; Laukamp et al. 2011), Canada (Laakso et al. 2015, 2016), India (Bharti et al. 2015; Bhattacharya et al. 2012; Bharti and Ramakrishnan 2014).

Ambaji–Deri, area located in the western part of India (Fig. 1), is a well-known area for its polymetallic massive sulphide deposits (Deb 1979, 1980; Tiwary and Deb 1997). The area is covered under the forest and has a rugged terrain, owing to which the conventional geological exploration is complicated. Hence, in the present study, we aim to assess the application of hyperspectral remote sensing in mapping of the hydrothermal mineral alteration. This is vital first step for more in-depth mineral exploration in the region.

Fig. 1
figure 1

Location map showing the study area. 1a shows the spatial distribution of the Aravalli belt within India. 1b shows the stratigraphic succession of Aravalli belt. The box near Ambaji area in Fig. 1b shows the location of study area modified after (Mishra 2015)

Full size image

Geological Setup of the Study Area

The Ambaji–Deri area falls under the Delhi supergroup of Aravalli metallogenic province (Gupta et al. 1980; Heron and Ghosh 1938; Porwal et al. 2006). The Delhi supergroup of rocks are of Meso- to Neo-proterozoic in age and contains an arenite-dominated assemblage in the east and a pelite-dominated assemblage in the west (Fig. 2) (Deb 1979, 1980; Gupta et al. 1980; Heron and Ghosh 1938).

Fig. 2
figure 2

Geological map of the area around Ambaji region along with the AVIRIS-NG flight path. It also shows the locations of interest where we found the distribution of alteration minerals after GSI (2012a; b)

Full size image

The study area comprises of calcareous, arenaceous and argillaceous meta-sediments belonging to Ajabgarh group of Delhi supergroup (Deb 1979, 1980; Tiwary and Deb 1997; Deb et al. 2001; Porwal et al. 2006). Calcareous meta-sediments like calc-gneiss, calc-silicate dolomite and marble form the basement of the Ajabgarh group. These calcareous sediments are overlain by arenaceous and argillaceous meta-sediments. Concordant bodies of meta-volcanic rocks lie within these meta-sediments. Ajabgarh meta-sediments were intruded by post-Delhi Erinpura granite. This Ajabgarh group has been subjected to the two phase of greenschist facies deformation and thermal metamorphism of hornblende-hornfels facies (Deb 1979, 1980; Tiwary and Deb 1997).

The ore in this region is deposited in the form of massive lenses and tabular bodies (Deb 1979, 1980). These deposits are hosted by pelitic and psammitic meta-sediments represented by chloritic hornfelses, diopside-forsterite marble, and quartz-mica schists and amphibolite. These deposits are classified under the category of stratiform massive sulphide deposits (Deb 1979, 1980), because the deposits are concordant to the strata of host rock and are in the form of tabular or lensoid bodies. The origin of these deposits is thought to be syngenetic origin based on the various geological evidence, including the poly-metamorphosed nature of the ores (Deb 1979, 1980; Tiwary and Deb 1997).

Methodology

We used the Level-2 (Reflectance) of Airborne Visible Infrared Imaging Spectrometer-Next Generation AVIRIS-NG data, with 425 spectral bands, ranging between 380 and 2510 nm, with a spectral resolution of 5 nm, spatial resolution of 8.1 mts. The data processing was carried out using Envi software Version 5.1. Since most of the absorption feature of alteration minerals can be diagnosed between 2000 and 2500 nm spectral range (Hunt 1979; Kruse 1988), the data were spectrally subsetted to 110 band out of 425 spectral bands. The spectral range of the subsetted data is 1954.1700 nm to 2500.1201 nm. We followed the standard protocol for data analysis (Fig. 3) (Incorporated Research Systems 2004; Kruse et al. 1993).

Fig. 3
figure 3

Flow chart showing the standard protocol for processing the image data modified after Kruse et al. (1993)

Full size image

Raw data are marred with lot of noise and dimensionality, which hinder the identification of specific minerals (Luo et al. 2016). In order to reduce the dimensionality of the data, Minimum Noise Fraction (MNF) algorithm was used. MNF uses two cascading principal component analysis (PCA), of which the first PCA decorrelates noise from the data and makes it as unit variance (i.e. noise whitening), whereas the second PCA is operated on noise whitened data, this will arrange the data in the decreasing signal-to-noise ratio with increasing the principal component (Amato et al. 2009; Green et al. 1988).

In order to reduce the spatial dimension, the MNF data are further processed for the extraction of the end-members using Pixel Purity Index (PPI) algorithm (Boardman et al. 1993; Wu et al. 2014). The PPI identifies the spectrally unique pixels within the image. These spectrally unique pixels are then used as end-members. It is generally done by projecting pixels on n-dimensional random vectors (Boardman 1994; González et al. 2010). The end-members extracted are then matched with the standard spectral library (USGS mineral library) using spectral analyst tool in Envi software version 5.1 (Nidamanuri and Ramiya 2013). The end-members identified by spectral library match are then used as a trainee for classifying the AVIRIS-NG image data. Finally, Spectral Angle Mapper (SAM) algorithm was used to classify the image data. SAM algorithm classifies the image data by calculating the cosine angle between the reference pixel and the image pixel vector. This method is widely used because of its simplicity and it also nullifies differential illumination effects (Kruse et al. 1993; Nidamanuri and Ramiya 2014).

Results

Data Analysis

The end-members employed for SAM classification contains the spectra of the rock present and were derived from the image data. The results of the end-member spectra matching with the USGS library spectra show the presence of carbonate (CaCO3), muscovite, chlorite/epidote minerals within the data. The carbonate minerals can be detected based on the absorption feature around 2300 nm, due to vibration of (CO3)−2 and moreover, the substitution of Ca++ and Mg++ cations will shift the absorption feature towards 2330 nm, and the presence of Mg will lead it to shift towards 2319 nm, respectively (Gaffey 1987; Laakso et al. 2018). The absorption feature of carbonate in the end-member (Fig. 4a) at 2339 nm suggests the presence of Ca++ in cation. Similarly, chlorite feature occurs at 2250 nm due to presence of Fe–O–H and 2340 nm due to presence of Mg–O–H; further it is to be noted that these features vary between 2255 and 2248 nm for absorption centred around 2250 nm and for absorption centre around 2340 nm, they will vary due to variation in Mg++ ions (Dalton et al. 2004; Laakso et al. 2016; Neal et al. 2018). The end-member retrieved from image data shows two absorption features, i.e. at 2249 nm and 2339 nm (Fig. 4b). Epidote can be detected by the absorption feature around 2250 nm and 2340 nm due to Fe–O–H and Mg–O–H. It should also be noted that epidote feature can be differentiated from chlorite due to presence of deeper absorption feature around 2250 (Neal et al. 2018). Here, the end-member of epidote shows absorption feature at 2254 nm and 2339 nm (Fig. 4c). Muscovite is detected based on 2200 nm due to presence of Al–OH; however, it should be noted that these features also vary between 2199 nm and 2204 due to substitution of Al in the octahedral site of the mica (Hunt 1979; Dalton et al. 2004; Laakso et al. 2015, 2016). In the end-member processed, the value of muscovite is 2199 nm (Fig. 4d), which indicates, according to Laakso et al. (2016), more Al might be present in octahedral site. Confirmation of such results can be done by geochemical studies of rock samples, however, which is not in scope of the present study.

Fig. 4
figure 4

Spectra of end-member derived from the AVIRIS-NG image data were made to match with the known standard library spectra of USGS. Here, a represents the calcite mineral which shows the absorption feature at 2339 nm, b represents the chlorite mineral which shows the absorption feature at 2249 nm and 2339 nm. c represents the epidote mineral which shows absorption feature at 2254 nm and 2339 nm. d represents the muscovite mineral which shows the absorption feature 2199 nm

Full size image

Spatial Distribution of Mineral Alterations and Calcite Marble

The mineral predominance map produced out of AVIRIS-NG (Fig. 5) depicts a laterally variable spatial distribution of these minerals. The pure calcite (in the form of marble) is concentrated in the NW–SE orientation for a length of 2 km near the Khokhar Bili village. The host rock in this segment is calcareous meta-sediments of Ajabgarh series of Delhi system (Heron and Ghosh 1938; Deb 1980). The chlorite, similar to calcite, is spatial present in same orientation and shows more concentration in the chloritic hornfels and diopside forsterite marble host rock around Jharivav (Deb 1980). The muscovite however on the contrary shows more scattered signature with its distribution in south, western and northwestern parts of the studied area, with relatively higher concentrations in Amblimal (Fig. 5). Muscovite is found in the host rock same as that of chlorite (i.e. chloritic hornfels and diopside forsterite marble) (Deb 1980). Nevertheless, the distribution of chlorite is restricted to a patch near Jharivav, whereas muscovite is scattered in a wider area (Fig. 5). Epidote, on the other hand, is present only in few patches north of Khokhar Bili village. This mineral is formed over the marble of Ajabgarh series. The host rock in this area is the calcareous meta-sediments of the Ajabgarh series (Deb 1980).

Fig. 5
figure 5

Spatial distribution of the carbonate minerals at northeast of Khokhar Bili and alteration minerals around Jharivav and Amblimal

Full size image

Discussion

The results show the distribution of the alteration minerals (muscovite and chlorite) and calcite/epidote. The spatial distribution of these minerals is correlatable with the proved mineral deposits, i.e. existing mine owned by Gujarat Mineral Resource Development Corporation (GMDC) area for polymetallic deposits. Moreover, the muscovite also shows its high concentration around Amblimal, which is the small ore body reported earlier by Deb et al. (2001). Similarly, the calcite is more concentrated near Khokhar Bili village which is well known for its pure marble deposits (GSI 1967, 2012a). Epidote shows its patchy appearance near Khokhar Bili. It is occasionally associated with the marble deposits (GSI 2012b).

Based on the studies on ancient sulphide deposits and its modern analogue, model proposed for explaining genesis of Volcanogenic Massive Suphide (VMS) deposits, such deposits are deposited in submarine environment due to the convection of the seawater (Deb 1979, 1980; Galley 1993; Galley et al. 2007; Large 1992; Ross and Mercier-Langevin 2014; Tiwary and Deb 1997). During the process of formation of VMS deposits, the cold seawater percolates down through the fractures in the volcanic sequences. With increase in depth, the temperature of the seawater increases and simultaneously pH deceases; this leads to leaching of the metal elements from the host rock. Subsequently, these metals are transported upward and deposited at or near sea floor by the fluid. During this convection journey, the fluid will interact with the host rock which will result in the alteration of the feldspar in felsic volcanic rocks to white mica. In cases with further increase in temperature, the host rock (felsic and mafic volcanic rocks) will alter to the chlorite (Galley 1993; Galley et al. 2007; Herzig and Hannington 1995; Large 1992; Ross and Mercier-Langevin 2014; Tiwary and Deb 1997). These convections are the major responsible factor for the alteration of the host rock to VMS (Galley 1993; Galley et al. 2007; Herzig and Hannington 1995; Large 1992; Ross and Mercier-Langevin 2014; Tiwary and Deb 1997).

The deposits in Ambaji–Deri region are believed to have been deposited under submarine condition (Deb 1979, 1980; Tiwary and Deb 1997). The ore metals are thought to be derived from the then basalt, which is now metamorphosed to the amphibolite, due to the result of the basalt-seawater interaction. (Deb, 1979, 1980). (Tiwary and Deb 1997) have proposed the hydrothermal alteration model for the Deri massive sulphide deposit in which the felsic and mafic volcanic rocks were altered by the basalt-seawater interaction. The model suggests that the cold seawater percolates through warmer and permeable felsic volcanic rock. As a result, felsic volcanic rock got altered into quartz-sericite and when the temperature reached to 250 °C, it got converted into chlorite-sericite with the addition of Mg. With deeper circulation, it leached Fe and Mg. Subsequently altering the previously formed chlorite assemblages into the Biotite-Chlorite-Sericite facies (Tiwary and Deb 1997).

Apart from polymetallic deposits, lenses of economic marble deposition are also common in this area (Deb 1979, 1980; GSI 2012a). This marble deposition is made up of recrystallized calcite and is believed to be deposited in near shore transitional shelf environment (Deb 1980; GSI 2012a). In the present airborne hyperspectral study, we have mapped the distribution of these marble deposits, based on the absorption feature of the calcite, i.e. at 2330 nm. The observed spatial distribution matches well with known mineral mines and deposits in the Ambaji–Deri region (Fig. 5). Further, it should be noted that since epidote and chlorite are spectrally similar minerals (Neal et al. 2018), it is difficult to differentiate between the two at many places.

Conclusion

In the present study, we successfully illustrate the application of hyperspectral remote sensing in mapping mineral alterations, which aid in hunting newer mineral deposits. The calcite, muscovite and chlorite have been shown to be matching with the ground geological records, whereas epidote shows patchy occurrences. However, due to spectral similarities between epidote and chlorite, it is difficult to differentiate between the two only on the basis of hyperspectral studies and without being aided by geochemical analysis. Nevertheless, hyperspectral technique provides powerful tool for mapping mineral alteration to mineral deposits. This enables scrutiny of a wider area of interest to being narrowed down for first-hand field investigations.