CHAPTER 6 HYDROTHERMAL ALTERATION ZONES MAPPING USING PRINCIPAL COMPONENT ANALYSIS AND BAND RATIO TECHNIQUES 150 CHAPTER – 6 HYDROTHERMAL ALTERATION ZONES MAPPING USING PRINCIPAL COMPONENT ANALYSIS AND BAND RATIO TECHNIQUES 6.1 INTRODUCTION Hydrothermal alteration zones mapping is one of the most common applications of remote sensing for mineral exploration in which the presence of altered rocks is the main indicator of the possible ore deposit (Sabins, 1999; Rajesh, 2004). Many techniques have been developed to get significant performance and image quality enhancement of specific features (Achalakul and Taylor 2000; Novak and Soulakellis, 2000; Ferrier et al., 2002; Moghtaderi et al., 2007; Tommaso and Rubinstein, 2007). Principal Component Analysis (PCA) and band ratios are two examples of these techniques. The use of PCA and band ratios during the early stages of mineral exploration has been very successful in pointing to hydrothermally altered rocks. Many researchers have proposed that PCA is an effective approach to delineate anomalous concentrations (e.g. Chica-Olmo and Abarca, 2000; Tangestani and Moore, 2000; Ranjbar et al., 2004). Initially PCA and band ratios were applied to different Landsat sensors by many researchers in order to study the alteration zones. Band ratio images are generated from bands in which specific geological materials have either relatively high or relatively low total reflectance (Abdelsalam et al., 2000b). The band ratio technique is well proven to identify geological materials through the detection of diagnostic absorption bands of the component materials (Vincent, 1997). It has been used successfully since the advent of
27
Embed
CHAPTER – 6 HYDROTHERMAL ALTERATION ZONES MAPPING …shodhganga.inflibnet.ac.in/bitstream/10603/74721/11/chapter 6.pdf · CHAPTER 6 HYDROTHERMAL ALTERATION ZONES MAPPING USING PRINCIPAL
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
CHAPTER 6 HYDROTHERMAL ALTERATION ZONES MAPPING USING PRINCIPAL COMPONENT ANALYSIS AND BAND RATIO TECHNIQUES
150
CHAPTER – 6
HYDROTHERMAL ALTERATION ZONES MAPPING
USING PRINCIPAL COMPONENT ANALYSIS AND BAND
RATIO TECHNIQUES
6.1 INTRODUCTION
Hydrothermal alteration zones mapping is one of the most common applications
of remote sensing for mineral exploration in which the presence of altered rocks is the
main indicator of the possible ore deposit (Sabins, 1999; Rajesh, 2004). Many techniques
have been developed to get significant performance and image quality enhancement of
specific features (Achalakul and Taylor 2000; Novak and Soulakellis, 2000; Ferrier et al.,
2002; Moghtaderi et al., 2007; Tommaso and Rubinstein, 2007). Principal Component
Analysis (PCA) and band ratios are two examples of these techniques.
The use of PCA and band ratios during the early stages of mineral exploration has
been very successful in pointing to hydrothermally altered rocks. Many researchers have
proposed that PCA is an effective approach to delineate anomalous concentrations (e.g.
Chica-Olmo and Abarca, 2000; Tangestani and Moore, 2000; Ranjbar et al., 2004).
Initially PCA and band ratios were applied to different Landsat sensors by many
researchers in order to study the alteration zones. Band ratio images are generated from
bands in which specific geological materials have either relatively high or relatively low
total reflectance (Abdelsalam et al., 2000b). The band ratio technique is well proven to
identify geological materials through the detection of diagnostic absorption bands of the
component materials (Vincent, 1997). It has been used successfully since the advent of
CHAPTER 6 HYDROTHERMAL ALTERATION ZONES MAPPING USING PRINCIPAL COMPONENT ANALYSIS AND BAND RATIO TECHNIQUES
151
multispectral scanners in 1970. Science 2000 ASTER data enable alteration zones to be
identified before field work is undertaken.
Both ASTER and ETM+ data are suitable for mapping the altered rocks, but with
ETM+ data it is less effective comparing with the ASTER data because of its limited
spectral resolution. ASTER image is more suitable to map silica alteration because of its
5 thermal bands as compared to the single band of ETM+. The main objective of this
chapter to map the hydrothermal alteration zones in north east of Hajjah using PCA and
band ratios techniques.
6.2 Principal Component Analysis (PCA)
PCA, which is also called as Principal Component Transformation (PCT), is an
image processing technique which transforms the original remotely sensed data set into a
substantially smaller and easier to interpret set of uncorrelated variables that represent
most of the information present in the original data set (Fig. 6.1) (Jensen, 2005). The
main objective of PCA is to remove redundancy in multispectral data and extract new
information. It builds up a new set of axes orthogonal to each other (i.e. non-correlated
data) (Gupta, 2003), and can be performed on as many spectral bands as possible. PCA is
also widely used for mapping of alteration in metallogenic provinces (Abrams et al.,
1983; Kaufman, 1988; Loughlin, 1991; Bennett et al., 1993; Tangestani and Moore,
2001; 2002; Crosta et al., 2003; Ranjbar et al., 2004; Zhang et al., 2007). By applying
PCA a new data set with fewer variables is created (Lillesand and Kiefer, 2000). Crosta
technique is known as a feature of oriented principal component selection; it indicates
where the materials are represented as bright or dark pixels in the PC (Ranjbar et al.,
2004).
CHAPTER 6 HYDROTHERMAL ALTERATION ZONES MAPPING USING PRINCIPAL COMPONENT ANALYSIS AND BAND RATIO TECHNIQUES
152
The idea of applying PCA to derive mineral abundance maps using high spectral
resolution data was proposed by Crosta et al. (1996) and Prado and Crosta (1997). The
PC1 is generally the weighted average of all data and represents albedo and topographic
effects found in the scene (Drury, 1993; Ranjbar et al., 2004). The PC1 contains the
largest percentage of data variance and the PC2 contains the second largest data variance,
and it continuous like that. The last PC appears noisy because it contains very little
variance, much of which is due to noise in the original spectral data or uncertainty in the
data. It is possible to combine any three of different PCs in R-G-B to create a colour
image (Sabins, 1997).
A limitation of PCA is the gray-tone statistics of a PC image which are highly
scene dependent and can not be extrapolated to other scenes. Further, geologic
interpretation of PC image also requires great care as the surface information dominates
the variation (Gupta, 2003). In some cases different materials are enhanced with same
brightness as for example, in the study area vegetation cover and altered clays are
enhanced with the same brightness.
PCA technique was applied to VNIR and SWIR bands of ASTER (1, 2, 3, 4, 5, 6,
7, 8 and 9) and ETM+ (1, 2, 3, 4, 5 and 7) imagery using ERDAS Imagine 9.1 model
(Fig. 6.2).
CHAPTER 6 HYDROTHERMAL ALTERATION ZONES MAPPING USING PRINCIPAL COMPONENT ANALYSIS AND BAND RATIO TECHNIQUES
153
Fig. 6.1 Diagram showing principal component transformation technique. The creation of principal component by shifting and rotating the coordinate system a) scattering of two original highly correlated variables x1 and x2 with means µ 1 and µ 2. b) The new coordinate system x′′′′ found by shifting the original x axis system. c) the principal components (PC1-PC2) are the new coordinate system by rotating the shifted axis system x′′′′ about the point (µ 1, µ 2) (Jensen, 2005).
Fig. 6.2 ERDAS model which was used in PCA of the study area
CHAPTER 6 HYDROTHERMAL ALTERATION ZONES MAPPING USING PRINCIPAL COMPONENT ANALYSIS AND BAND RATIO TECHNIQUES
154
6.2.1 Principal Component Analysis of Landsat-7 ETM+
PCA of ETM+ data is shown in Table 6.1. The PC1 contains the largest amount
(93.817%) of total variance of six bands, which decreases until it reaches 0.108 % for PC
7. PC1 has positive loadings from all bands (1, 2, 3, 4, 5, and 7); with the highest loading
is in band 5 (0.621) and band 7 (0.575). The other components showed decreasing
variance caused by differences between spectral regions and individual bands.
PC1 does not contain spectral feature relevant to this analysis as it is a
combination of all bands (Fig. 6.3-PC1). It provides information mainly on albedo and
topographic effects. In the PC2, the spectral bands are separated into visible and infrared
bands with the negative sign for bands 1, 2, and 3 and positive sign for bands 4, 5, and 7.
This PC shows the contrast between the visible red and the near infrared (de Jong and van
der Meer, 2004).
PC2 contains 3.355% of the total variance of six bands. Vegetation cover, Amran
limestone, Kohlan sandstone and granitic rocks are enhanced in this component with
bright pixels. Akbra shale and quartz-graphite-biotite-sericitic schist are represented by
dark pixels (Fig. 6.3-PC2). The PC3, which contains 1.978% of the total variance of six
bands, has positive loading in band 7 (0.453) and low negative loading in band 4
(-0.875). In this component the vegetation cover is enhanced with dark pixels (Fig. 6.3-
PC3), because it has a higher reflectance in band 4 of ETM+ and lower reflection in band
3 (Fig. 4.1).
CHAPTER 6 HYDROTHERMAL ALTERATION ZONES MAPPING USING PRINCIPAL COMPONENT ANALYSIS AND BAND RATIO TECHNIQUES
155
Table 6.1 Principle Component Analysis of ETM+ data
In PC3, band 5 represents the negative loading and band 7 the positive loading so
that the altered clay minerals appear in this component as dark pixels. The PC4 contains
0.433 % of the total variance, and has positive loading in band 1 (0.690) and low positive
loading in band 3 (-0.467). This component shows areas with iron oxide as dark pixels
because they possess higher reflection in the red region of the spectrum and absorption in
the blue region (Fig. 6.3-PC4).
The PC5 explains 0.309% of the total variance, and has the highest positive
loading value in band 5 (0.550) and the lowest loading value in band 7 (-0.595). In this
component the hydroxyl bearing altered zones are seen as bright pixels (Fig. 6.4-PC5).
PC5 has higher loading of band 5 with positive sign and band 7 with negative sign so that
the hydroxyl bearing altered zones are enhanced with bright pixels. The spectral feature
of the clay minerals exhibits absorption feature in band 7 of ETM+, and higher
PC Band No. PC1 PC2 PC3 PC4 PC5 PC7
Band 1 0.159 -0.363 -0.036 0.690 -0.304 -0.523
Band 2 0.250 -0.424 -0.145 0.268 -0.123 0.806
Band 3 0.410 -0.646 0.012 -0.467 0.362 -0.254
Band 4 0.196 0.145 -0.875 -0.241 -0.324 -0.111
Band 5 0.612 0.450 -0.085 0.336 0.550 0.004
Band 7 0.575 0.217 0.453 -0.250 -0.595 0.010
variance % 93.817 3.355 1.978 0.433 0.309 0.108
Cumulative variance %
93.817 97.173 99.151 99.583
99.892
100
CHAPTER 6 HYDROTHERMAL ALTERATION ZONES MAPPING USING PRINCIPAL COMPONENT ANALYSIS AND BAND RATIO TECHNIQUES
156
reflectance in band 5 (Fig. 4.3). This explains the higher pixel values of the altered areas
in this PC. PC7 contains 0.108% of the total variance and has high positive and negative
loading in band 2 (0.806) and band 1 (-0.523), respectively. In this component the
features are not clear probably due to the noise (Fig. 6.4-PC7). The colour combination of
PC5, PC3 and PC2 as R-G-B, respectively highlights altered area represented by the red
color (Fig. 6.5).
As discussed in PC3 the vegetation cover is displayed with dark pixels, but after
inversing (inverse PC3) using the following equation it displays the vegetation cover with
bright pixels (Fig. 6.6).
)1.6()7(453.0)5(085.0
)4(875.0)3(012.0)2(145.0)1(036.03
bandband
bandbandbandbandPC
−++−+=
CHAPTER 6 HYDROTHERMAL ALTERATION ZONES MAPPING USING PRINCIPAL COMPONENT ANALYSIS AND BAND RATIO TECHNIQUES
157
Fig. 6.3 PC1, PC2, PC3, PC4 of ETM+ of the study area
CHAPTER 6 HYDROTHERMAL ALTERATION ZONES MAPPING USING PRINCIPAL COMPONENT ANALYSIS AND BAND RATIO TECHNIQUES
158
Fig. 6.5 Band color combination of ETM+ PC (5-3-2) as R-G-B of the study area
Fig. 6.6 Result of inversing PC3 of ETM+ of the study area
Fig. 6.4 PC5 and PC7 of ETM+ of the study area
CHAPTER 6 HYDROTHERMAL ALTERATION ZONES MAPPING USING PRINCIPAL COMPONENT ANALYSIS AND BAND RATIO TECHNIQUES
159
6.2.2 Principal Component Analysis of ASTER data
PCA of subsystems VNIR and SWIR bands (1, 2, 3, 4, 5, 6, 7, 8 and 9) of ASTER
data is shown in Table 6.2. The PC1 explains the largest amount (90.456%) of total
variance among nine bands. It has positive loadings of all bands with the highest loading
in band 2 (0.460). This component generally represents the albedo and the topographic
effects (Fig. 6.7-PC1). PC2 accounts for 6.081% of the total variance, and has the highest
positive loading in band 4 (0.314), and the negative loading in band 1 (-0.526). In this
component the sedimentary and granite rocks appear with bright pixels and the
Precambrian basement rocks with the dark pixels (Fig. 6.7-PC2). PC3 accounts for
1.655% of the total variance, and has the highest positive loading in band 3 (0.907) and
high negative loading in band 2 (-0.277).
Vegetation cover is enhanced in this component and appears as bright pixels, as
this PC3 has higher loading of band 3 with the positive value and lower loading with
negative value in band 2 (Fig. 6.7-PC3). The spectral feature of vegetation cover (Fig.
4.1), exhibits strong absorption feature at 0.45µ m and 0.65µ m and maximum
reflectance at 0.55µ m, and as such it appears in this component as bright pixel. The PC4
explains 0.585% of the total variance, and has the highest positive loading in band 4
(0.722) and negative loading in band 7 (-0.401). Hydroxyl bearing area is enhanced in
this component by bright pixels, as this PC has positive loading of band 4 and negative
loading of bands 7, 8 and 9 (Fig. 6.7-PC4). The spectral feature of different types of clay
minerals (Fig. 4.3) shows, that clay minerals exhibit higher reflectance in band 4 and
higher absorption in bands 7, 8 and 9.
CHAPTER 6 HYDROTHERMAL ALTERATION ZONES MAPPING USING PRINCIPAL COMPONENT ANALYSIS AND BAND RATIO TECHNIQUES
160
PC5 explains 0.384% of the total variance, and has the highest positive loading in
band 6 (0.563) and negative loading in band 2 (-0.454). Iron oxides are enhanced in this
component with dark pixels, because this PC has higher loading in band 2 with negative
sign and positive loading in band 1 (Fig. 6.8-PC5). Iron oxide with bright pixels can be
seen when the PC5 has been inversed as mentioned in equation 6.1. PC6 explains 0.339%
of the total variance, and has the highest positive loading in band 6 (0.513) and negative
loading in band 1 (-0.475). Iron oxide is enhanced in this component with bright pixels
(Fig.6.8-PC6). PC7, PC8, and PC9 explain 0.248%, 0.126% and 0.125%, respectively of
the total variance. PC7, PC8 and PC9 are predominantly noisy images and do not
discriminate any features (Fig. 6.8-PC7, PC8; Fig. 6.9-PC9). Rocks rich in massive
sulfide materials are represented in PC9 by the bright pixels.
Table 6.2 Principle Component Analysis of ASTER data