You are on page 1of 9

International Journal of Earth Sciences and Engineering

ISSN 0974-5904, Vol. 04, No 03 - Spl issue, May 2011, pp. 30-38

Landsat ETM+ Image Analysis for Alteration

Mapping in Sukinda Chromite Deposit, Orissa: Case
of a Vegetated Mineralized Region
Frontier Basins, ONGC, Dehradun, India
RQA Division, CIMFR Digwadih Campus, P.O. FRI Dhanbad - 828 108 INDIA
Dept. of Earth Sciences, Indian Institute of Technology Roorkee, Roorkee - 247 667
Uttarakhand, INDIA

Abstract: In this paper we examine the potential of a freely available Landsat ETM+
dataset of the Sukinda Chromite deposit in Orissa, India. The deposit is hosted in the
layered ultramafics (dunite and peridotite rocks) and is co-folded with iron ore group (IOG)
quartzites, capped with volcanic tuffaceous rocks at places. The Sukinda valley is covered
by thick veneer of latertized and /or limonitized ultramafics. Given these geologic
characteristics, the ETM+ data has been processed to enhance spectral features related to
anomolous concentrations of ferric oxides and ferrous oxides and hydroxides, alteration
minerals with rich hydroxyl components, such as clays and laterites. Due to extensive and
thick vegetation cover, the simple methods of band ratios are unable to provide a very clear
picture of the alteration anomalies; however, a modified form of the Principal Components
Analysis, popularly called the Crosta technique is able to provide a more unambiguous
discrimination related to the mineralogic spectral differences among outcrop lithologies and
soil cover.
Keywords: vegetation, alteration, mineralized, Crósta technique

Introduction: minerals typical of alteration in the visible

and the middle-infrared sections of the
The primary role of remote sensing in
electromagnetic spectrum. The
mineral exploration activity is in enabling a
hydrothermal minerals used in mineral
focused ground exploration through
exploration can be classified into three
synoptically examining a large area for the
groups: hydroxyl (clays and micas),
presence of favorable combinations of
hydrated sulphates (gypsum and alunite),
lithology, alteration minerals and anomalies
and minerals containing iron (haematite,
expressed through vegetal stress and
goethite, and jarosite). Some of these
structural patterns, etc. As a matter of fact,
minerals are also related to weathering, and
remote sensing images in carefully selected
therefore a careful field check is warranted.
spectral regions significantly reduce the cost
Laterites are an alteration product of
and time involved in exploration. The
bedrock subjected to intense and prolonged
occurrence of hydrothermally altered rock is
chemical weathering in tropical and
one of the features that can be utilized to
subtropical climates. Primary minerals are
determine the localization of an ore deposit.
transformed into a characteristic
Structural features are very important in the
assemblage of the secondary minerals,
localization of a mineralized zone, but many
kaolinite, hematite, goethite, and residual
structures are not necessarily mineralized so
quartz. The outcome of lateritisation is a
that their relation to hydrothermal alteration
build-up of iron and aluminium oxides at the
is highly important in exploration (Armenta
top of the palaeosol profile, resulting in the
and Ledesma, 1998). Mapping of
formation of ferricretes (iron crust or
hydrothermally altered rock has been made
duricrust). Below this, a clay-rich leached
possible with the use of sensors that can
zone, first mottled and then saprolitic,
detect the spectral response of groups of
indicates progressive degeneration of the

#02040405 Copyright © 2011 CAFET-INNOVA TECHNICAL SOCIETY. All rights reserved.


original bedrock (Schellmann 1986, Butt and spreading over an area of 40sq km in NE-
Zeegers 1992, Nahon and Tardy 1992, SW direction in Jajpur and Dhenkanal
Deller, 2006). This constitution of a lateritic districts of Orissa. It is flanked by the
terrain is amenable to detection and Tomka (782.42m amsl)-Daitari (847.77m
characterization using remote sensing amsl) range to the north and Mahagiri
techniques. Initial success in operational use (707.69m amsl) range to the south. The
of remote sensing for lithologic mapping was Valley is largely covered with alluvium and
achieved using passive optical remote thick mantle of laterite (Mohanty 2007). The
sensing data from the Landsat series of main drainage in the Valley is contributed by
sensors, viz. Multispectral Scanner (MSS), perennial Damsal stream that flows from
Thematic Mapper (TM) and Enhanced east to west along the centre of the Valley
Thematic Mapper Plus (ETM+) (e.g., Abrams following the regional synclinal fold axis
et al., 1983; Podwysocki et al., 1983; plunging WSW.
Hutsinpillar and Taranik, 1988). Landsat
Remote Sensing Image Data and
data continues to be the most widely used
remote sensing image data for geologic
applications. Goetz et al. (1983) and Sabins The remote sensing image data used in this
(1999) have provided a general overview of study include the six solar reflective bands
the use and potential of multispectral of the Landsat ETM+ multispectral imagery
passive remote sensing in mineral (WRS Path 140, Row, 45) acquired on
exploration activity, with emphasis on the 2/11/2002, and the panchromatic image
application of Landsat TM data. data of the same sensor (15m ground
resolution). The data has been obtained
Study Area and Geologic Setup:
from the Global Land Cover Facility, at the
The majority of India’s Chromite production University of Maryland, USA which provide
(~95%) comes from two ultramafic the Landsat data as orthorectified image
complexes of Orissa namely Sukinda data with real-world geodetic positional
ultramafic complex and Boula-Nausahi accuracy of >50m. The salient aspects of
ultramafic complex (~70km in NE of Landsat ETM+ data are provided in Table-1
Sukinda Valley and relatively much smaller and the spectral bands of ETM+ data
than SUC), which are geologically related relevant to the major surface cover types in
and geographically in close proximity. The the present study area are provided in
Valley extends from Tomka (85°55’E: Table-2. The data has been processed using
21°7’N) in the east to Kathpal (85°41’E: ERDAS Imagine 8.7 software suite.
21°1’N) on the west (Banerjee 1972)
Table 1: Salient Characteristics of Landsat ETM+ Data.

Band Spectral Ground Image size

Number Range Resolution (km)
ETM1 0.45-0.52 Blue 30
ETM2 0.52-0.60 Green 30
ETM3 0.63-0.69 Red 30
ETM4 0.76-0.90 Near Infrared 30
Shortwave 185 x 170
ETM5 1.55-1.75 30
ETM6 10.4-12.5 Thermal Infrared 60
ETM7 2.08-2.35 30
ETM8 0.45-0.90 Panchromatic 15

International Journal of Earth Sciences and Engineering

ISSN 0974-5904, Vol. 04, No 03 - Spl issue, May 2011, pp. 30-38
Landsat ETM+ Image Analysis for Alteration Mapping in Sukinda
Chromite Deposit, Orissa: Case of a Vegetated Mineralized Region

Table 2: Landsat Bands with Characteristic Spectral Properties for Major Surface
Components Expected in the Study Area

Material High reflectance Absorption

Iron oxide ETM-3, ETM-5, ETM-7 ETM-1, ETM-2
Hydroxyl ETM-5 ETM-7
Vegetation ETM-2, ETM-4 ETM-1, ETM-3, ETM-7

Image Processing: arithmetic rather than full spectral

Atmospheric Correction:
Resolution-Merged False Color
Some of the important reasons for
Composite and Band Ratios:
atmospheric correction of passive optical
remote sensing image data, such as that of Figure 1 shows the resolution enhanced
Landsat ETM+, in geological studies are (Lu modified FCC (bands 7-4-2 coded in R-G- B)
et al., 2002): displaying the best lithological
discrimination. The FCC was generated by
1. within-scene comparison of spectrally
fusing the high resolution panchromatic
similar materials and hence in spectral
image (15m) with the multispectral data. As
much of the study area is dominated by
2. quantitative analysis of remotely sensed
thick vegetation cover which has the highest
data with comparison and use with
reflectance in band 4, green hues dominate
laboratory derived reflectances for known
the image display. The vegetation cover
greatly limits the discrimination capability of
3. accurate material mapping using band-
surface lithology using remote sensing
arithmetic, etc.
approach as the reflected radiation forming
Many methods for atmospheric correction the signal is dominated by the interference
exist, ranging from simple band-minimum due to vegetation. This effect is best
subtraction for removal of path radiance to demonstrated by the use of band ratio
highly rigorous physical radiative transfer ETM5/ETM7. Ideally, this ratio generates a
theory-based methods (for details see, Lu et clear discrimination of hydroxyl (OH-)-
al., 2002). For most common applications, bearing rocks. However, as vegetation also
the primary atmospheric influence on has higher values for the same band ratio,
passive optical data in solar reflective region the discrimination is badly affected. A
is due to short wavelength scattering normalized difference vegetation index
leading to atmospheric path radiance. In the (NDVI) image [(ETM4-ETM3)/(ETM4+ETM3)]
present study area, due to the presence of is used to identify healthy vegetation from
clear water bodies, the simple method of the background (Tucker 1979). A difference
band minimum subtraction has been of the ETM5/ETM7 and NDVI image for
adopted to remove the effects of sparsely vegetated areas yields good
atmospheric path radiance from the dataset, discrimination for hydroxyl-bearing rocks,
since other atmospheric effects are but for thickly vegetated areas the
relatively less important, and also because interference is not fully eliminated. Figure 2
the processing is mainly based on band shows a difference image of ETM5/ETM7 and
NDVI of the ETM+ data for the study area.

International Journal of Earth Sciences and Engineering

ISSN 0974-5904, Vol. 04, No 03 - Spl issue, May 2011, pp. 30-38

Figure 1: PAN-Sharpened ETM+ Image of Sukinda Chromite Deposit. (FCC, 7-4-2 in RGB)

Figure 2: Difference Image of Ratio 5/7 and NDVI, Color Coded to Depict Areas of Enriched
Hydroxyl-Bearing Rocks/Regolith and Related Alteration Mineralogy. Blue-Cyan-Green Pixels
represent Hydrothermal Alteration Minerals Rich in Hydroxyl Ion.

Figure 3: Color Ratio Composite (CRC: 7/5-3/1-5/4).

Interpretation is given in the Color Legend.

International Journal of Earth Sciences and Engineering

ISSN 0974-5904, Vol. 04, No 03 - Spl issue, May 2011, pp. 30-38
Landsat ETM+ Image Analysis for Alteration Mapping in Sukinda
Chromite Deposit, Orissa: Case of a Vegetated Mineralized Region

Table 3: PCA Stats for the 6-Band ETM+ Dataset

Basic Stats Min Max Mean Stdev Num Eigenvalue

Band 1 0 239 70.48083 7.198288 1 1009.061
Band 2 0 216 56.31558 8.687165 2 215.2878
Band 3 0 248 49.37797 13.84802 3 79.99165
Band 4 0 143 71.40717 13.64555 4 9.82631
Band 5 0 227 78.58263 22.4566 5 7.960294
Band 7 0 255 43.3394 17.7736 7 3.323382
Covariance Band 1 Band 2 Band 3 Band 4 Band 5 Band 7
Band 1 51.81535 56.85084 81.27299 15.19902 100.9549 80.85227
Band 2 56.85084 75.46683 109.3826 20.37694 141.6219 111.6633
Band 3 81.27299 109.3826 191.7676 -13.8008 237.0823 207.9934
Band 4 15.19902 20.37694 -13.8008 186.201 101.8279 5.095413
Band 5 100.9549 141.6219 237.0823 101.8279 504.2987 369.9134
Band 7 80.85227 111.6633 207.9934 5.095413 369.9134 315.9009
Correlation Band 1 Band 2 Band 3 Band 4 Band 5 Band 7
Band 1 1 0.909138 0.815322 0.154737 0.624531 0.631957
Band 2 0.909138 1 0.909248 0.171898 0.725953 0.723198
Band 3 0.815322 0.909248 1 -0.07303 0.762373 0.845058
Band 4 0.154737 0.171898 -0.07303 1 0.332301 0.021009
Band 5 0.624531 0.725953 0.762373 0.332301 1 0.926789
Band 7 0.631957 0.723198 0.845058 0.021009 0.926789 1
Eigenvector Band 1 Band 2 Band 3 Band 4 Band 5 Band 7
PC 1 0.165663 0.226056 0.382571 0.090951 0.689864 0.539359
PC 2 -0.04061 -0.04865 -0.30683 0.894972 0.240253 -0.20771
PC 3 -0.46861 -0.5128 -0.49678 -0.23281 0.331737 0.326182
PC 4 -0.73952 -0.0505 0.584827 0.065964 0.100838 -0.30661
PC 5 0.178743 0.122245 -0.17575 -0.35856 0.586225 -0.67082
PC 7 0.415289 -0.81613 0.373888 0.060334 0.049886 -0.12468
Table 4: Eigenvector Matrices for FPCS Analysis using Bands 1-3-4-5 and 1-4-5-7,
respectively for the Iron Oxides and Hydroxyl Minerals related to Hydrothermal Alteration.

Band 1 Band 3 Band 4 Band 5

PC 1 0.201507 0.449097 0.171905 0.85332
Eigenvector Matrix
PC 3 0.097975 0.433803 -0.8928 -0.07159
for bands 1-3-4-5
PC 4 -0.5663 -0.54663 -0.36747 0.495445
PC 5 0.793161 -0.55796 -0.19576 0.145787
Band 1 Band 4 Band 5 Band 7
PC 1 0.165998 0.132543 0.777223 0.592285
Eigenvector Matrix
PC 4 -0.01947 0.935786 0.101458 -0.33709
for Bands 1-4-5-7
PC 5 -0.98095 -0.02899 0.189214 0.033119
PC 7 0.099021 -0.32542 0.591463 -0.73107

International Journal of Earth Sciences and Engineering

ISSN 0974-5904, Vol. 04, No 03 - Spl issue, May 2011, pp. 30-38

Figure 4a: PC4-PC2-PC5 (RGB), Showing Iron-Oxide-Rich Areas in Red Tones, Vegetation
in Green Tones, and Hydroxyl-Rich Areas in Blue Tones.

Figure 4b: PC5 (FPCS using Bands 1-3-4-5), Showing Iron Oxide and Hydroxide Alteration
in Bright White Tones.

International Journal of Earth Sciences and Engineering

ISSN 0974-5904, Vol. 04, No 03 - Spl issue, May 2011, pp. 30-38
Landsat ETM+ Image Analysis for Alteration Mapping in Sukinda
Chromite Deposit, Orissa: Case of a Vegetated Mineralized Region

Figure 4c: PC7 (FPCS using Bands 1-4-5-7) Showing Hydroxyl-Bearing Areas in Black-Dark
It can be readily observed from the image PCs according to the magnitude and sign of
that the vegetation cover severely affects the eigenvectors. The Crósta Technique was
the discriminability of the ETM5/ETM7 band applied to the six ETM+ bands, and to four
ratio and poses problems in clear bands selected on the basis of the spectral
identification of OH- anomalies. In order to characteristics of the hydrothermal minerals
distinguish between the iron oxide and and the vegetation. The band statistics
ferrous hydroxides, the image through PCA for a full 6-band dataset are
transformations in the form of band ratios provided in Table 3. Whereas, the
ETM3/ETM1 for FeO and ETM5/ETM4 for Fe- eigenvector matrices generated through PCA
OH were calculated. Figure 3 shows the using bands 1-3-4-5 and bands 1-4-5-7 are
band ratio composite (7/5-3/1-5/4 in RGB). given in Table 4. The RGB composites
corresponding to the PCs providing the best
PCA & FPCS Analysis:
hydrothermal alteration mineral distribution
In order to differentiate the areas containing in the area are shown in Figures 4a, b & c.
hydrothermal minerals covered with
Interpretation and Discussion:
vegetation from those that contained only
vegetation and no alteration minerals, Landsat ETM+ bands which best express the
statistical techniques based on principal distinctive features of laterite minerals are
component analysis (PCA) were applied to bands 2, 3, 4, 5 and 7, in various RGB band
the images. The Feature Oriented Principal and band-ratio colour combinations. Band 2
Components Selection (FPCS) Technique is distinguishes hematite from goethite
also known as Crósta Technique (Crósta and because it is less reflective in the green
Moore 1989). Through the analysis of the region. Band 3 highlights high reflectance
eigenvectors values it allows identification of for quartz and kaolinite. Band 4 is important
the principal components (PCs) that contain in that the iron oxides and hydroxides with
spectral information about specific their distinctive crystal-field absorption
materials, as well as the contribution of each feature are separated from kaolinite and
of the original bands to the components in quartz. Band 5, with high reflectance for all
relation with the spectral response of the minerals but low reflectance for vegetation,
materials of interest. This technique is useful for band-ratios. Band 7 indicates
indicates whether the materials are the presence of hydroxyl minerals,
represented by bright or dark pixels in the particularly the high Al-OH absorption

International Journal of Earth Sciences and Engineering

ISSN 0974-5904, Vol. 04, No 03 - Spl issue, May 2011, pp. 30-38

feature for kaolinite around 2.2 µm, but is Acknowledgements:

so broad that the C-O absorption of
Authors thankfully acknowledge Prof. R. P.
carbonates and the Mg-OH absorption of
Gupta, Dept. of Earth Sciences, IIT Roorkee
hydrated ferromagnesian minerals such as
for his constructive suggestions during the
chlorite, also affect it (Drury 1993). Landsat
preparation of the earlier manuscript. DM is
TM 742 images (figure 4) provide an
grateful to CSIR for funding in the form of
excellent basis for overall geological
research fellowship to carry out the work.
interpretation and also highlight laterite
occurrences in the study area. A rich green References:
color is indicative of conspicuous dense
[1] Abrams, M. J., Brown, D., Lepley, L. and
foliage in the study area. Bright pink pixels
Sadowski, R., (1983). Remote sensing
represent areas rich in quartz+clayey soil,
for porphyry copper deposits in southern
whereas darker pink and light maroon pixels
Arizona. Economic Geology, 78, pp. 591–
are likely to represent ferricrete and Fe-rich
kaolinitic soil. Cyanish-blue pixels are most
[2] Armenta, J. R. and Prol Ledesma, R. M.,
likely to represent pure kaolinite-rich
(1998). Techniques for enhancing the
regolith. On the other hand the color ratio
spectral response of hydrothermal
composite (CRC: 7/5-3/1-5/4; Fig. 3) bring
alteration minerals in thematic mapper
out a clearer contrast among the main
images of central Mexico. International
spectral classes in the area, and the effects
Journal of Remote Sensing, 19(10), pp.
of topography are also subdued. An
interpreted color index gives the main
[3] Banerjee, P. K., 1972, Geology and
spectral classes inferred. Bright orange-red
geochemistry of the Sukinda Ultramafic
pixels and reddish-pink shades are indicative
Field, Cuttack District, Orissa, Mem Geol
of water bodies and shadows respectively.
Surv Ind 103: 1-171.
Examination of the eigenvector matrix for
[4] Butt C.R.M. and Zeegers, H. (Eds), 1992,
the PCA of the full 6-band dataset, reveals
Regolith exploration geochemistry in
that PC1 carries positive loadings from all
tropical and subtropical terrains
bands and is essentially represents the
(Amsterdam and London: Elsevier).
effects of surface albedo and topography.
[5] Crósta, A. P. and Moore, J. M., (1989).
PC2 carries the highest loading from band 4,
Enhancement of Landsat thematic
which means that the brighter pixels in PC2
mapper imagery for residual soil
correspond to vegetation-rich surfaces.
mapping in SW Minas Gerais State,
Similarly, PC4 has the highest negative and
Brazil: a prospecting case history in
highest positive loadings from bands 1 and
greenstone belt terrain. In: Proceedings
3, respectively, which implies that the dark
of the 7th Thematic Conference on
pixels in this PC correspond to areas
Remote Sensing and Exploration
enriched in iron-oxides and hydroxides. In a
Geology, Vol. II, Environmental Research
similar fashion, PC5 relates to the hydroxyl-
Institute of Michigan, Ann Arbor,
rich pixels, owing to highest positive and
Michigan, pp. 1171–1187.
negative loadings from bands 5 and 7,
[6] Deller, 2006, Facies discrimination in
respectively. PC color composite,
laterites using Landsat Thematic Mapper,
corresponding to PC4-PC2-PC5 has been
ASTER and ALI data—examples from
shown in figure 4a. It is readily observed
Eritrea and Arabia, IJRS, Vol. 27, No. 12,
that this image affords the best
20 June 2006, 2389–2409.
discrimination between the dominant
[7] Drury, S.A., 1993, Image Interpretation
surface spectral classes. The FPCS analysis
in geology, second edition (London:
also brings out similar enhancements, and a
Chapman &Hall).
relatively less ambiguous mapping for the
[8] Goetz, A. F. H., Rock, B. N. and Rowan,
iron-oxides (Fig. 4b) and hydroxyl-bearing
L. C., (1983). Remote sensing for
rock/regolith (Fig. 4c) could be made.
exploration: an overview. Economic
Geology, 78 (4), pp. 573–590.

International Journal of Earth Sciences and Engineering

ISSN 0974-5904, Vol. 04, No 03 - Spl issue, May 2011, pp. 30-38
Landsat ETM+ Image Analysis for Alteration Mapping in Sukinda
Chromite Deposit, Orissa: Case of a Vegetated Mineralized Region

[9] Hutsinpillar, A. and Taranik, J. V., Exploration Geochemistry in Tropical

(1988). Spectral signatures of and Subtropical Terrains, C.R.M. Butt
hydrothermal alteration at Virginia City, and H. Zeegers (Eds), pp. 41–56
Nevada. In: Bulk minerable precious (Amsterdam and London: Elsevier).
metal deposits of the western United [13] Podwysocki, M. H., Segal, D. B. and
States, Robert W. Schafer (ed.), Abrams, M. J., (1983). Use of
Geological Society of Nevada Symposium multispectral scanner images for
Proceedings, pp. 505–530. assessment of hydrothermal alteration
[10] Lu, D., Mausel, P., Brondizio, E. and in the Marysville, Utah mining area.
Moran, E., (2002). Assessment of Economic Geology, 78, pp. 675–687.
atmospheric correction methods for [14] Sabins, F. F., (1999). Remote Sensing
Landsat TM data applicable to Amazon for Mineral Exploration. Ore Geology
basin LBA research. International Reviews, 14, pp. 157–183.
Journal of Remote Sensing, 23(13), pp. [15] Schellmann, W., 1986, A new definition
2651–2671. of laterite. Geological Survey of India,
[11] Mohanty, D., 2007, Ore genesis of PGE Memoir, 120, pp. 1–7.
mineralization in the Sukinda chromite [16] Tucker, C. J., 1979, Red and
deposite, Orissa, India, unpub. Ph. D. Photographic Infrared Linear
thesis, Indian Institute of Technology Combinations for Monitoring Vegetation.
Roorkee, Roorkee, India, 213pp. Rem Sen Envir 8, 2: 127-150.
[12] Nahon, D. and Tardy, Y., 1992, The
ferruginous laterites. In Regolith

International Journal of Earth Sciences and Engineering

ISSN 0974-5904, Vol. 04, No 03 - Spl issue, May 2011, pp. 30-38