Application

Journal of of remote
Mineralogical sensing and GIS
and Petrological in mineral
Sciences, resource
Volume mapping
99, page 83 ─ 103, 2004 83

REVIEW

Application of remote sensing and GIS in mineral

resource mapping − An overview
H.M. RAJESH

Department of Geographical Sciences and Planning, Chamberlain Building,

University of Queensland, St Lucia 4072, QLD, Australia

Remote sensing, as a direct adjunct to field, lithologic and structural mapping, and more recently, GIS have
played an important role in the study of mineralized areas. A review on the application of remote sensing in
mineral resource mapping is attempted here. It involves understanding the application of remote sensing in
lithologic, structural and alteration mapping. Remote sensing becomes an important tool for locating mineral
deposits, in its own right, when the primary and secondary processes of mineralization result in the formation of
spectral anomalies. Reconnaissance lithologic mapping is usually the first step of mineral resource mapping.
This is complimented with structural mapping, as mineral deposits usually occur along or adjacent to geologic
structures, and alteration mapping, as mineral deposits are commonly associated with hydrothermal alteration of
the surrounding rocks. In addition to these, understanding the use of hyperspectral remote sensing is crucial as
hyperspectral data can help identify and thematically map regions of exploration interest by using the distinct
absorption features of most minerals. Finally coming to the exploration stage, GIS forms the perfect tool in
integrating and analyzing various georeferenced geoscience data in selecting the best sites of mineral deposits
or rather good candidates for further exploration.

Introduction regions of the earth have been found, current emphasis is

on the location of deposits far below the earth’s surface or
“Geologists seem to have rosy prospects in remote in inaccessible regions. Geophysical methods that
sensing for the next decade. This period is likely to be provide deep penetration into the earth are generally
one of consolidation rather than innovation, giving the
needed to locate potential deposits and drill holes are
majority of geologists the time to get to grips with what
has been happening over the last three decades in geolog- required to confirm their existence. However, much
ical remote sensing research, to apply the new data to information about potential areas for mineral exploration
exciting new geological problems instead of repeatedly can be provided by interpretation of surface features on
pawing over tiny test areas, and to catch up with their aerial photographs and satellite images. For example, in
colleagues in other fields.” (Drury, 2001, p. 67) Australia (where 70% of the continent is covered by sedi-
ments), a comparison of all significant gold deposits
Mineral resource mapping is an important type of geolo- currently known with the areas of weathered basement
gic mapping activity and usually covers a great part of rocks (where gold and other mineral deposits are likely to
varied studies, focused on spectral analysis (e.g. Longhi be found) indicates the potential for remote sensing in
et al., 2001), geological mapping (e.g. Harris, 1991), discovering new deposits under the sedimentary cover
structural mapping (e.g. Liu et al., 2000), identification of (Fig. 1). While use of remotely sensed images cannot
hydrothermal alteration zones (e.g. Podwysocki et al., replace direct ground observation or data derived from
1983), mineral alteration mapping (e.g. Tangestani and field and laboratory studies, they can form valuable
Moore, 2002), ferric oxide and oxyhydroxide mineral supplements to more traditional methods and provide
mapping (e.g. Farrand, 1997), gold exploration (e.g. information and a perspective not otherwise available.
Spatz, 1997), hyperspectral imagery (e.g. Neville et al., It should be appreciated that there is one important
2003), integration with geographic information systems limitation of remote sensing data in mineral exploration −
(GIS) (e.g. Akhavi et al., 2001) etc. Because most of the the depth aspect. Remote sensing data have a depth pene-
surface and near − surface mineral deposits in accessible tration of approximately a few micrometers in the very
near infrared region, to a few centimeters in the thermal
H.M. Rajesh, [email protected] Corresponding author infrared and some meters (in hyper arid regions) in the
 84 H.M. Rajesh

Figure 1. a: Digital Elevation

Model (DEM) of Australia.
b: Map of Australia showing
areas of weathered basement
rocks (dark), areas where
basement is covered by
younger sediments (light),
and all significant gold
deposits currently known in
Australia (circles). (Courtesy:
CSIRO, Australia)

knowledge of the general geological make up of an area.

Therefore, a little basic geology about the assumptions of
mineral deposits is relevant:
• A particular mineral deposit occurs in a particular
rock type (e.g. diamond usually occurs in kimber-
lite).
• Mineral deposits usually occur along or adjacent to
geologic structures (e.g. Fig. 2).
• Mineral deposits usually show strong alteration on
the surface.
• Mineral deposits are usually (spatially) associated
Figure 2. Proximity analysis (buffering) of anticlinal folds from an
with a high temperature rock (e.g. granite).
area, part of the Meguma terrane, Nova Scotia, Canada. The
Meguma terrane consists of predominantly Cambrian and Ordo- • Mineral deposits usually occur near the contact
vician sedimentary rocks intruded by Devonian granites. The between favorable rock types (e.g. porphyry copper
fold axes were digitized from the geological map and buffered. deposits have a direct spatial association with the
This map has 24 distance buffers spaced at 250 m intervals. The contact of granitic to intermediate intrusive rocks;
reported gold occurrences occur close to the axial trace of the
Guilbert and Park, 1996).
anticlines (Bonham − Carter, 1994).
A multiple combination of any of the above
mentioned hypotheses is used to locate mineral resources
microwave region. Therefore, in most cases, a remote in this paper. For example, the Chalice gold deposit,
sensing data interpreter has to rely on indirect clues, such Yilgarn Craton, Western Australia, occurs in a sequence
as general geologic setting, alteration zones, associated of intercalated mafic and ultramafic amphibolites, is
rocks, structure, lineaments, oxidation products, spatially and temporally related to granitic rocks, is
morphology, drainage, and vegetation anomaly, since only controlled by localized asymmetric folds, and is charac-
rarely is it possible to directly pinpoint the occurrence and terized by high − temperature silicate and sulfide alteration
mineralogy of a deposit based solely on remote sensing assemblages (Bucci et al., 2002). Because mineral
data. In this perspective, both multispectral and hyper- resources are associated most frequently with very small
spectral sensors, which can define mineralogy, are but highly anomalous areas where a great many processes
expected to play a greater role in mineral exploration, by have all acted together to concentrate the metals involved
helping to delineate ore minerals or their pathfinders. The above their normal abundances, it is very easy to miss
results are being more and more integrated into opera- even very high value deposits in the field. However, the
tional exploration models based on geographic informa- anomalous processes involved produce unusual rocks and
tion systems (GIS) technology, which plays a relevant minerals associated with mineral deposits. It is toward
role in mineral exploration (e.g. Bonham − Carter, 1994; these that remote sensing is directed. Remote sensing
Memmi and Pride, 1997). becomes a powerful exploration tool in its own right
Locating mineral resources relies primarily on when the primary and secondary processes of mineraliza-
 Application of remote sensing and GIS in mineral resource mapping 85

tion result in the formation of spectral anomalies. of mineral deposits. All the discussions in this paper on
To evaluate the different aspects of the application of satellite imagery refer to either electro − optical sensors
remote sensing in mineral resource mapping, this paper [measuring reflectance in the visible and near − infrared
attempts a review approach, focusing on the application (VNIR; 0.3 − 1.0 μ m), short − wave infrared (SWIR; 1.0 −
of remote sensing in lithologic, structural, and alteration 2.5 μ m) and mid − infrared or thermal infrared (TIR; 3 − 5
mapping. In addition to these an appreciation of the use μ m; 8 − 14 μ m) portions of the electromagnetic spectrum],
of hyperspectral remote sensing in mineral resource the most common type carried aboard remote sensing
mapping is important as hyperspectral data can help iden- satellites, or synthetic aperture radar (SAR) measuring
tify and thematically map regions of exploration interest reflectance in the microwave or radar portion [2 − 100
by using the distinct absorption features of most minerals. cm; typically at 2.5 − 3.8 cm (X band), 4.0 − 7.5 cm (C
Finally coming to the exploration stage, it is clear that the band), and 15 − 30 cm (L band)] of the spectrum. The
remote sensing data has to be integrated with other common satellites in the former category, detecting
geoscience data like geochemical, geophysical data, etc. reflected sun − source energy, include Landsat Multispectral
This demands a multithematic approach, and GIS forms Scanner (MSS), Landsat Thematic Mapper/Enhanced
the perfect tool as they allow more effective integration Thematic Mapper (TM/ETM), Système Probatoire
and analysis of large numbers of spatial data with d’Observation de la Terre (SPOT), Indian Remote Sensing
different attributes and formats in selecting the best sites Satellite (IRS), Advanced Spaceborne Thermal Emission

Figure 3. Spectra of common iron − bearing

minerals (above dashed line) and minerals
containing chemically bound water (below
dashed line) (a), iron oxides and hydroxides (b),
clay minerals and micas (above dashed line) and
carbonate minerals (below dashed line) (c) (Hunt
and Salisbury, 1970b; Hunt and Salisbury 1971;
Hunt, 1979). Absorption features are shown by
T − shaped symbols. Spectra of selected silicate
(above dashed line) and non − silicates (below
dashed line) in the mid − or thermal − infrared
part of the spectrum are given in (d) (Hunt and
Salisbury, 1970a; Kahle et al., 1996). The
progressive shift of the short wavelength peak
(arrows) and the main trough towards longer
wavelengths in (d) corresponds to a transition
from felsic to increasingly mafic minerals.
Thick horizontal ranges indicate the widths of
spectral bands sensed by the Landsat TM. The
spectra are offset for clarity.
 86 H.M. Rajesh

and Reflection Radiometer (ASTER) and IKONOS, while the spectrum the most important vibrational transitions in
those in the later category, which transmits and detects minerals are those associated with the presence of OH −
radiation, include RADARSAT, ERS, and the Japanese ions or water molecules. Absorption features at 1.9, 1.4,
Earth Resource Satellite (JERS, now FUYO). 1.14 and 0.94 μ m indicate the presence of molecular
water in minerals (Fig. 3a). The bending of Al − OH and
Spectral signature of minerals and rocks

Since Hunt (1979) and co − workers, a number of papers

provided both exhaustive libraries of laboratory spectra of
minerals and rocks, and accurate analyses of the absorption
features of specific minerals in the visible − short wave
infrared, mid − infrared and thermal infrared intervals,
establishing the scientific background for the interpreta-
tion of remotely sensed spectroscopic data (e.g. Gaffey,
1985; Clark et al., 1990; Salisbury et al., 1991;
Christensen et al., 2000). Spectroscopic criteria are
widely applied in hyperspectral image analysis for
mineral and alteration identification and mapping (see
later section). Comparing spectra of freshly cut rocks
with those of exposed surfaces gives an insight into the
relationship between original rock and superficial altera-
tion products, allowing the development of reconnais-
sance criteria that may also be applied in other areas with
similar environmental conditions.
Mineral structures are such that numerous absorption
bands exist due to electronic transitions and ion vibrations
(Hunt, 1977). Although minerals are of widely varying
types, electronic transitions are most often created by
iron, while vibrational ones are often created by water,
hydroxyl ions or carbonates. Figure 3a shows reflectance
spectra of several iron − bearing minerals that display
features resulting from electronic transitions in ferrous
(Fe2+) ions. The typical spectra of iron oxides are in the
range 0.35 to 1.5 μ m (Hunt and Ashley, 1979). It has
been observed that the occurrence of absorption anoma-
lies at wavelengths less than 0.9 μ m is a good indication
that hematite is the predominant mineral, when the anom-
alies are found at wavelengths close or larger than 0.9 μ m,
then jarosite or goethite are more abundant (Hunt and
Ashley, 1979).
The most common charge − transfer is involved in the
migration of electrons from iron to oxygen, and results in
a broad absorption band at wavelengths shorter than about
0.55 μ m. The most noticeable effect is with iron oxides
and hydroxides (Fig. 3b), and is the reason why these
minerals and the rocks containing them are colored
yellow, orange, red and brown. Vibrational transitions
produce reflectance anomalies in the near − infrared region
of the spectrum, between 1.1 and 2.5 μ m, and they Figure 4. a: Bidirectional reflectance spectra of some sedimentary
rocks (Salisbury and Hunt, 1974). b: Thermal − infrared transmis-
provide more information about the mineralogical rock
sion spectra of some igneous rocks (Vickers and Lyon, 1967). c:
composition than the spectra features observed in the Bidirectional reflectance spectra of metamorphic rocks (Hunt and
visible and near − infrared regions. In the SWIR part of Salisbury, 1976).
 Application of remote sensing and GIS in mineral resource mapping 87

Mg − OH, producing distinctive absorption features in the absorption due to ferrous ion is prominent in rocks such
reflectance spectra, are prominent in aluminous micas and as termolite schists (Fig. 4c). The features displayed by
clay minerals (Fig. 3c) and dominate signatures of marbles are strong carbonate absorptions (1.9 μ m and 2.35
hydroxylated minerals that contain magnesium, such as μ m) and high reflectivity in the near infrared. Water and
talc, chlorites, serpentines and magnesium − rich clays hydroxyl bands are found in schists, marbles and quartz-
(saponites). Carbonates give rise to a number of absorp- ites (Hunt and Salisbury, 1976). Figure 4 based on labo-
tion features in the SWIR of which that around 2.3 μ m is ratory experiments, confirms that rocks possess the poten-
most prominent (Fig. 3c). The mid − infrared region tial to be classified from airborne or satellite sensor data if
contains high reflectance anomalies for most rocks (basalt, sufficient spectral detail is generated.
gabbro, etc.) and minerals (clays, micas, sulphates, Field spectroscopy is a tool to perform feasibility
carbonates) at around 1.65 μ m and high absorption at studies to help in understanding the nature of the spectral
approximately 2.2 μ m (Hunt, 1979). characteristics of surface materials and their spectral sepa-
Quartz shows trough in the emittance curve between rability. Remote sensing has been used in combination
8 and 9 μ m as a result of Si − O bond − stretching vibra- with field spectroscopy as an aid in alteration mapping
tions. This and related spectral structures are best shown leading to mineral exploration (e.g. Van der Meer et al.,
in transmission spectra, and they occur in both silicates 1997; Mazzarini et al., 2001; Ferreir et al., 2002). A
(Fig. 3d) and non − silicates (Fig. 3d). Within the range 8 weathered rock surface (with modified mineralogical
to 14 μ m the emission spectra of silicate minerals contain composition) will mask some of the spectral properties of
a prominent, broad absorption trough and associated the original surface (fresh surface). In such cases, it is
features caused by Si − O bond stretching (Fig. 3d). In this necessary to study the spectral differences between the
region of the spectrum, various vibrational transitions in exposed surface in the field and the fresh one. Younis
non − silicates produce spectral features that are different et al. (1997) showed that the spectral regions where the
from those of silicates (Fig. 3d). The most important are fresh and weathered surfaces show minimum spectral
those associated with carbonate and iron oxides, which differences can be used to better characterize and discrim-
are so distinct that even small amounts of these non − sili- inate the lithological units. Van der Meer et al. (1997)
cates in dominantly silicate rocks drastically alter their used field spectra to study the spectral characteristics of
spectra. unweathered rocks samples and the alteration minerals
Rock spectra are mixtures of those for each of their that formed due to low − grade metamorphism in the
constituents, proportional to their abundance. It has long Troodos ophiolite complex, Cyprus. Further they investi-
been known that rocks can be distinguished from each gated the spectra of soils developing on the different
other under ideal conditions by their spectral signatures in lithologies and showed that with reflectance spectroscopy
the thermal emission region of the spectrum (e.g. Lahren it is theoretically possible to discriminate the different
et al., 1988; Sabine et al., 1994). Representative spectra lithological units based on the soils that develop on them.
of sedimentary, igneous and metamorphic rocks are given They used the information to extract characteristic TM
in Figure 4. In general, the dominating features in sedi- spectra from the Landsat TM image of the area. Thus
mentary rocks are due to the additional presence of the field spectroscopy can eventually lead to the selection
carbonate radical, which produces absorption bands areas that are spectrally representative for the different
between 1.9 and 2.3 μ m (Fig. 4a). All sedimentary rocks lithologies that can be detected in the image data.
generally have water absorption bands at 1.4 μ m and 1.9
μ m. Clay − shales have an additional absorption feature at Lithologic mapping
2.1 − 2.3 μ m. Limestones and calcareous rocks are char-
acterized by absorption bands of carbonates (at 1.9 μ m Reconnaissance lithologic mapping is usually the first
and 2.35 μ m, the latter being more intense); the ferrous step of mineral resource mapping studies. Broad litho-
ion bands at 1.0 μ m are more common in dolomites, due logical information is deduced from a number of parame-
to the substitution of Mg2+ by Fe2+. Falling SiO2 content ters observed in remote sensing images, viz. general
of igneous rocks (and metamorphic rocks of the same geologic setting, weathering and landform, drainage,
range of compositions) results in a progressive shift of the structural features, soil, vegetation, and spectral character-
Si − O bond stretching absorption feature to longer wave- istics. In the case of sedimentary rocks, especially those
lengths in thermal emission spectra (Fig. 4b). The simi- that are exposed at hillsides or by folding or faulting,
larity of the spectra for a class of rocks, such as the gran- bedding is one of the strongest clues to lithologic compo-
ites, allows a composite signature to be generated, which sition in images. These linear features are long, even −
may be used as representative of all granites. The broad spaced, and few in number (in comparison to those
 88 H.M. Rajesh

Figure 5. a: Landsat image covering part of Pilbara region, Western Australia. The different features include lighter colored areas of Archaean
granite batholiths (G), dark green colored metamorphosed basalt (B), Archaean sediments (AS), Archaean sediments and lavas (SL), Protero-
zoic sediments (PS), limestone (L), alluvium (A), Tertiary sediments (TS), Quaternary/Tertiary sediments (QTS) and dykes (d). (Courtesy:
ACRES, Geoscience Australia.) b: Landsat image covering part of the shield in the northeast corner of Sudan, close to the border of Egypt.
The region, part of the Sahara desert, has little soil and vegetation cover; so the rock bodies are exceptionally well exposed. Igneous rock
bodies form circular patterns ranging in size from almost 8 km across the body that is exposed in the large, white area, to small structures that
appear as circles or dots. Tonal variations in the circular patterns reflect both composition and sand cover. Dark areas are probably rocks
containing enough iron and magnesium to color the rock. Gold is commonly associated with such igneous rocks and has been mined in the
area since the time of the ancient Egyptians. (Courtesy: NASA and Earth Satellite Corporation.) c: Satellite image of the Snake River plain,
southern Idaho, USA vividly show how weathering modifies a rock body. Floods of older basalt form the smooth flat surface that extends
diagonally across the area. The younger extrusions are fresh and black, and retain the original features of the flows. Older flows (irregular
patches with tan hue, not black) have been subjected to longer periods of weathering and have developed a thin soil that supports sparse
vegetation. The oldest flows appear as light reddish brown areas. (Courtesy: NASA and Earth Satellite Corporation.) d: Satellite image of
Canadian Shield with dark tones indicating metamorphic rocks and light tones indicating areas of granitic rock. The complex folds and
contortions in the rock units show the degree to which metamorphic rocks have been deformed. The long linear lakes, ridges, and depression
are major fracture systems. Small lakes, shown on the image as black patches, occur in innumerable depressions. (Courtesy: NASA and
Earth Satellite Corporation.)
 Application of remote sensing and GIS in mineral resource mapping 89

produced by foliation in metamorphic rocks), and consti-

tute rather continuous ridges and valleys. Examination of
drainage density, drainage pattern and vegetation patterns
may also provide clues to lithology even when beds are
not directly exposed. The bareness of sandstone outcrops
in arid areas, combined with their well − drained nature,
usually permits them to show their true photographic tone
or color. Silica − and carbonate − cemented sandstones are
usually pale, but those containing iron compounds can be
any shade of yellow, orange, brown or red. In panchro-
matic black and white photographs red sandstones appear
dark. In humid climates, the main clue to the presence of
sandstone is the wide spacing of drainage and the round- Figure 6. SIR − A image showing part of the Hammersley Range,
ness of topography. Gently dipping mudstones and silt- western Australia. The area is underlain by lower Proterozoic
volcanics, which are interbedded with some of the world’s largest
stones typically develop dendritic and closely spaced
iron − ore deposits, in the form of banded ironstone formations.
drainage patterns because of their poor internal drainage. The two large structures, one in the top (Rocklea Dome) and one
Intrusive igneous rocks are generally massive, in the right, are with a core of Archaean granite. Bright linear
isotropic and homogenous, which can be easily observed features (including fine lines) cutting these granites are dykes.
on the remote sensing images. Their shapes (batholiths, The dark unit with bright ridges surrounding the dome is weath-
ered sandstone. Pillow basalts, the rough surface of which
laccoliths, dykes, sills, etc.), dimensions, distinct image
results in a bright signature, occurs between the large structures.
tones and topographic expression may also help in identi- The synform near the dome contains a layered sequence of iron-
fication (e.g. Fig. 5a, b). Extrusive igneous rocks can be stones, shales and carbonates. The rugged area in the top running
delineated by distinctive landforms, such as cones, craters, to the right of the image is underlain by massive ironstone,
and flows, especially if they are young, and differences in weathering of which has produced huge iron ore deposits. Bright
linear features, NW of the dome, are thick quartz veins, many of
image tone or by drainage patterns and vegetation distri-
which follow minor faults. (Courtesy: NASA.)
butions (e.g. Fig. 5c). When they are closely integrated
with surrounding sedimentary strata, intrusive igneous
rocks are more subtly expressed than are extrusive rocks qualified, and that, in some cases, the thickness of surfi-
and can be interpreted with less confidence. Information cial formations could be quantified. In addition, radar’s
concerning metamorphic rocks can also be interpreted advantage was shown by the fact that the substratum
from remotely sensed images, but typically only with under weathering material could be determined under
great difficulty because of complex local structures certain conditions. Marble is characterized by relatively
induced by metamorphism (e.g. Fig. 5d). Metamorphism flat topography often occupied by farmland or low − lying
may also reduce the differences in resistance to erosion vegetation. Landsat TM band 4 imagery (0.76 − 0.90 μ m)
that are so important in inferring lithology. As in the case typically shows dark homogenous gray tones for low −
of igneous rocks, the response is controlled mainly by the lying vegetation growing in areas underlain by marble.
relative proportion of quartz and other stable silicates, The distribution of marble can also be mapped on radar
compared with the amounts of unstable, usually ferro- imagery based on its relatively uniform flat topographic
magnesian silicates. signature but this task is more difficult. Fairly massive
Topographic patterns including drainage patterns gneissic complexes have a uniform vegetation cover and
often reflect geologic structure and lithology. Remote their radar signature usually reflects a mottled regional
sensing systems operating in the microwave spectrum, topography with substantial relief, and lack of regional
such as SAR, is known to provide good topographic ductile fabric. In some instances the radar imagery allows
enhancements useful for geomorphic and structural inves- the precise delineation of subcircular intrusive rocks. For
tigations while Landsat provides good mapping capabili- example, the plutons in Central Metasedimentary Belt,
ties of cover types including vegetation. Although radar Quebec, Canada, offer a unique signature in radar imagery
is used chiefly to map structure, it can be applied to litho- because they lack a regional fabric, show positive relief
logic mapping due to variations in outcrop patterns, surfi- within the surrounding domain, and deflect the regional
cial character of rocks, and the size and frequency of foliation (Rivard et al., 1999).
coarse rock detritus (Blom and Daily, 1982; Rebillard and Combining Landsat data with a higher spatial resolu-
Evans, 1983; e.g. Fig. 6). The advantage of radar was tion data, such as SPOT or IRS panchromatic data, will
demonstrated by the fact that bedrock nature could be provide a better discrimination of lithological units than
 90 H.M. Rajesh

either dataset alone (e.g. Fraser et al., 1997; Mickus and

Johnson, 2001; Chatterjee et al., 2003). Wiart et al. (2000)
mapped the extent of pumice deposits from a previous
(1861) eruption of the Dubbi volcano, located in the
northeastern part of the Afar triangle, Eritrea, using
JERS − 1 SAR and Landsat TM imagery. SAR imagery
revealed old lava flows buried below tephra deposits,
emphasizing the ground penetrating property of the L −
band. The merged SAR and TM imagery provided
insights into the differences between weathered lava flows
and the extent of the lava flow boundaries, and in the
process helped in revising the geological map of Dubbi.
The use of multispectral optical sensors (like Landsat) is
particularly troublesome in tropical environments, due to Figure 7. Flowchart depicting the general processing steps to
continuous adverse atmospheric conditions and terrain obtain a modified geologic map using remote sensing data.
characteristics. SAR imagery is applied to tropical
regions both because it is able to sense through clouds
and because the morphologies of the upper surface of of lineaments, and their networks. A flowchart depicting
forest canopies reflect underlying topography, which in the general steps adopted to obtain a modified geologic
turn relates to structure and lithology. Geophysical data map from remote sensing data is given in Figure 7. The
can be of key importance for regional studies in places classified image is usually sufficiently geometrically
where insufficient geologic knowledge is available, due to accurate to allow accurate location of sample sites for
their ability to provide lithologic and structural informa- further studies on mineral resource mapping.
tion. Integration and fusion of geophysical data with
spaceborne SAR imagery is an effective method for Structural mapping
enabling correlation of surface topographic expression
with subsurface geological characteristics (e.g. Pedroso et As a corollary, most mineral deposits are related to some
al., 2001). Stern et al. (2002) showed that the folded and type of deformation of the lithosphere, and most theories
faulted Neoproterozoic sedimentary rocks in northern of ore formation and concentration embody tectonic or
Ethiopia can be better characterized by images from the deformational concepts. As linear features (lineaments;
MOMS − 2P multispectral scanner (aboard the International O’Leary et al., 1976) shown on remote sensing imagery
Space Station, orbiting at a ultra low orbit from the earth) of increasingly smaller scale (greater extent) reflect
than the Landsat TM. The use of remote sensing data increasingly more fundamental structures, their study will
clearly adds the textural component that is so important in provide insights not only to the location of the mineral
detecting the subtle features related to lithologic changes deposits, but also to metallogenic theories as well. Many
and is therefore important to confirm and strengthen the studies have emphasized the importance of lineament
field data interpretation. interpretations and digital lineament analysis in localizing
Maps produced from remotely sensed images can the major mineral deposits and notes that there is a strong
serve as adequate reconnaissance geologic maps for correlation between mineral deposits and lineaments (e.g.
mineral resource mapping. Further many studies have Kutina, 1969; Katz, 1982; Liu et al., 2000; Rein and
shown that remotely sensed data could be employed to Kaufmann, 2003).
improve the existing maps of an area (e.g. Rothery, 1987; Linear features, 10 km or less in length, are clearly
Abrams et al., 1988). These studies have shown that discernible on aerial photographs, but are poorly seen in
many of the discrepancies found between the published satellite imagery. They indicate the form and position of
maps and those derived from the imagery were mostly individual folds, faults, joints, veins, lithologic contacts,
due to omissions or errors in the field geological mapping and other geologic features that may lead to the location
as proved through subsequent fieldwork. Krishnamurthy of individual mineral deposits. In general, they reflect
(1997) used digitally enhanced IRS Linear Imaging Self only immediate surface and near − surface conditions, and
Scanner (LISS) − I and II sensor data to revise/modify the are poor guides to concealed deposits. Linear features, 10
existing geological map of two areas in Karnataka, India, to 200 km or more in length, indicate the general geom-
to considerable extent in terms of refined lithological etry of folds, faults and other structures of an area,
boundaries, delineation of unmapped rock units, mapping providing a regional structural pattern. They are less
 Application of remote sensing and GIS in mineral resource mapping 91

Figure 8. Lineament and fracture map of Paraiba State, northeast Brazil, extracted and classified by the analysis of the Landsat TM image (Liu
et al., 2000).

abundant than the shorter linear features by probably an through the occurrence of ridges in the stereo model and
order of magnitude and are useful in defining target areas differences in tonal response where beds differ in their
− local settings in which mineral deposits may be concen- mineral constituents. The dip slopes of rocks often can be
trated, and which merit more detailed study in the field. recognized more reliably on a stereo model than on the
Linear features, more than 200 km long (some being ground because of the synoptic view of an area of dipping
poorly identifiable for short stretches along their length), sediments obtained from the air. Accurate measurements
are most effectively studied on mosaics of Landsat of dip can be made by photogrammetric measurements on
imagery. These linear features seem to be globally ubiq- stereo pairs. Several studies have determined the attitude
uitous, and usually display a nearly orthogonal pattern. of faults and lithological units of fold structures from
Some lineament patterns have been defined to be the stereoscopic SPOT images (Berger et al., 1992; Bilotti
most favorable structural conditions in control of various et al., 2000). A fold can be delineated by tracing the
mineral deposits, such as: the traces of major regional bedding/marker horizon along the swinging strike, and
lineaments, the intersection of major lineaments or both the recognition of the dips of the beds. Broad, open,
major (regional) and local lineaments, lineaments of longitudinal folds are easy to locate on satellite images
tensional nature, local highest concentration (or density) (e.g. Fig. 9). On the other hand, tight, overturned,
of lineament, between en echelon lineaments, and linea- isoclinal folds are difficult to identify on satellite images,
ments associated with circular features. For example, Liu owing to small areal extent of hinge areas (which provide
et al. (2000) utilized lineament analysis from satellite the only clues to their presence); therefore, such folds
imagery to delineate the following structural features of need to be studied on appropriately larger scales, such as
considerable interest in search for mineral deposits in aerial photographs. Fu et al. (2004) used Landsat
northeast Brazil: the warping (or dragging) part of the TM/ETM stereoscopic images in combination with high −
minor shear zones, which splay out (or branch off) from resolution IRS − IC PAN (5.8 m) satellite images to delin-
the major wrench belt, is of extensional nature favorable eate Quaternary deformational structures, including
for hydrothermal emplacement; the swollen parts along spatial distribution and arrangement of fold structures and
the extending lineament (shear) zones are also favorable fault scarps, along the Tian Shan orogenic belt, northwest
for magmatic fluid intrusion; the intersections of short and China.
regional lineaments; the periphery, and the margin of One of the greatest advantages of remote sensing
circular or ring structures, the internal and external data from aerial and space platforms lies in delineating
peripheral parts of small rings in a large circular structure, vertical to high − angle faults or suspected faults. Where
and the en echelon diagonal fractures crosscutting the mineralization has taken place along a fault line, however,
circular features (see Fig. 8). there may be positive rather than negative surface feature.
Aerial photographs provide evidence of bedding In most cases the most reliable evidence of faulting is
 92 H.M. Rajesh

displacement of bedding along negative linear surface

features. Low − angle faults are difficult to interpret, since
the images provide planar views from above. Such faults
have strongly curving or irregular outcrop and can be
inferred on the basis of discordance between rock groups.
Joints form patterns in rocks, which are very similar in
photographic appearance to faults, i.e. they often provide
fairly straight negative features. If relative movement can
be seen then the feature can be classified as a fault, and
conversely if no movement can be detected it is better to
record the feature as a joint. Shear zones involve changes
in angular relationships, so that linear features outside the
shear system are seen to swing into it. If marker units are
present then their displacement is immediately apparent.
Radar imagery (e.g. RADARSAT) and Landsat data
can be used to generate valuable structural information
during the study of structurally complex terrains, particu-
larly when the interpretation of imagery is conducted in
Figure 9. Landsat image covering part of the Anti − Atlas Moun- conjunction with the regional field work (e.g. Fig. 10).
tains of southern Morocco. By virtue of the climate, degree of
Topographic features observable in RADARSAT images
erosion, and contrast in rock types (both in terms of erodibility
and spectral response), it is one of the most spectacularly reflect either changes in slope, such as erosional escarp-
exposed fold belts in the world. Large − scale basement cored ments due to variation of resistant and recessive litholog-
domes with irregular shapes form a continuous area of positive ical units (Evans et al., 1986). Since SAR sensors provide
structural relief. The rocks are complexly folded and faulted. their own illumination source, the look direction can
Rock sequences are repeated, anticlines encroach upon anti-
influence the information content of the imagery. In low −
clines, and synclines encroach upon synclines. Numerous zones
of structural discontinuity or disharmonic folding are obvious in relief environments, the look direction can be used to
the image. In general, the older rocks to the northwest appear to provide a greater enhancement of lineaments. With infor-
be more highly deformed than the younger rocks to the south and mation from previously collected field measurements,
southeast. The lighter colored yellowish brown Precambrian complex structural patterns like details of the regional
rocks in the northwest are tightly folded, highly fractured,
trends of the foliation, shear zones, etc., can be mapped
intruded, and metamorphosed. The white sinuous band against a
fold ridge is a dry stream. Recently, Helg et al. (2004) showed and traced over large areas using radar imagery (e.g. Fig.
that the Anti − Atlas of Morocco is a special type of foreland fold 10). Simple photogeologic principles like image texture
belt lacking any evidence for thrust faults other than the occa- can be used to outline foliation patterns, folds, brittle and
sional steep reverse fault found near basement inliers. (Courtesy: ductile shear zones and lithology, while image tone can be
NASA and Earth Satellite Corporation.)
traced to cover type and could be traced and related to

Figure 10. Structural interpretation of an area from

Mont − Laurier in the Central Metasedimentary Belt,
Quebec, Canada: A) RADARSAT imagery; B) folia-
tion trends interpreted from the imagery and selected
field measurements of gneissosity; C) foliation trends
(light line, as shown in B) and distribution of shear
zones with shear indicators (dark line) interpreted from
the radar imagery with lineaments and gneissosity
(intermediate colored line) measured in the field.
Interpretation of the satellite imagery, combined with
field traverses, shows that NNE − striking dextral shear
zones dominate in the northern part of the mapped area
whilst SSE − striking sinistral ductile shear zones occur
mainly in the south. Here the sense of shear was
inferred from the rotation of the foliation trends seen in
imagery and then confirmed in the field from rotation
of gneissosity into the shear zones and S/C relation-
ships (Rivard et al., 1999).
 Application of remote sensing and GIS in mineral resource mapping 93

lithology. In the presence of a near complete forest field efforts. There are difficulties in integrating linea-
canopy, radar imagery largely conveys topographic infor- ment maps with mineral exploration models, as some of
mation without the distraction offered by tonal variations the features mapped as lineaments may not be of struc-
associated with varying ground cover in Landsat imagery. tural − geologic nature, and it may not be possible to distin-
In such areas, field observations are usually limited by guish between post − mineralization and pre − mineraliza-
outcrop availability and often consist of densely popu- tion structures. Many studies integrated lineament
lated observations separated by substantial distance structures derived from satellite (Landsat TM, SAR, etc.)
(hundreds of meters to kilometers). The imagery allows data with a database of known occurrences in GIS for a
separation of isolated structures from more common ones, more fruitful interpretation of lineaments for mineral
and provides a regional framework for the regional inter- exploration (e.g. Akhavi et al., 2001). Other studies
pretation of structures documented in the field. The correlated lineament intersection density to alteration and
important feature of radar imagery is manifested in the observed that lineament intersection density was nearly
capacity of images to provide stereoscopic view, which twice as dense in altered zones as compared to unaltered
further facilitates the work of lineament identification (e.g. zones (e.g. Zakir et al., 1999). Use of lineament intersec-
Sharma et al., 1999). tion relationships for mineral exploration gains validity
Recently different algorithms have been published when defined by multi − technique approaches, such as
describing the extraction of lineaments directly from combinations of remote sensing, geophysical, and geolog-
digital images and aerial photos (e.g. Budkewitsch et al., ical methods (e.g. Chernicoff et al., 2002). In addition,
1994; Raghavan et al., 1995; Koike et al., 2001; Costa applicability can be tested by inclusion of known metallo-
and Starkey, 2001). These techniques can reduce to a genic information, which may help to identify favorable
minimum the bias in the manual interpretation. The auto- structural settings in a given regional context.
mated methods in extracting lineaments provide a flow of
data, which requires in turn elaborate and exact methods Alteration mapping
for the analysis and presentation. Principal component
analysis (PCA) is a classical statistical method that Mineral deposits are commonly associated with hydro-
produces images (components) that are a linear combina- thermal alteration of the surrounding rocks (e.g. Fig. 11),
tion of multiband images. When applying PCA, the rela- the style and extent of the alteration reflecting the type of
tive image variance is a measure of the amount of infor- mineral deposit. The host rocks of hydrothermal mineral
mation observable in each image. PCA of SAR from deposits invariably show the results of their chemical
Seasat − SAR and the shuttle imagery radar (SIR − B) has interactions with the hydrothermal fluids that caused
been used successfully to enhance topographic informa- mineral deposition (Pirajno, 1992). Such alteration
tion for structural and lineament mapping (e.g. Masouka commonly forms a halo around the mineralization,
et al., 1988). Paganelli et al. (2003) recognized four providing an exploration target considerably larger than
lineament trends (N − NE, NW, NE, and E − NE) in each of the deposit itself. The delineation and characterization of
the RADARSAT − 1 principal component images from the hydrothermal alteration can therefore be of great value in
Buffalo Head Hills area, Alberta, Canada. The intersec- mineral exploration and assessment of new targets. The
tion and offset relationships between the various linea- spatial distribution of hydrothermally altered rocks is a
ment groups in the images enabled definition of a relative key to locating the main outflow zones of hydrothermal
succession of events in which the N − NE lineaments was systems, which may lead to the recognition of mineral
recognized as the oldest, followed by NW and NE linea- deposits.
ments, which define a conjugate set, and the E − NE − Several airborne and orbital imagery studies have
trending lineaments interpreted as the latest as it show shown the feasibility of remote sensing techniques to
crosscutting relationships with all the previous linea- detect hydrothermal altered areas. These studies are
ments. The lineaments interpreted by Paganelli et al. based on the fact that certain diagnostic minerals associ-
(2003) and their tectonic − geologic implications provided ated with hydrothermal processes, such as iron − bearing
a basis for kimberlite exploration in the Buffalo Head minerals (e.g. hematite, goethite, jarosite), hydroxyl −
Hills area. bearing minerals (e.g. clays, micas), and hydrated
At the initial stage of image analysis, focus must be sulphates (e.g. gypsum, alunite), show diagnostic spectral
given to well − mapped areas where detailed maps had features that permit their remote identification (e.g. Hunt
been completed, and predictive capabilities should be and Ashley, 1979; Prost, 1980; Podwysocki et al., 1983;
developed for image − based reconnaissance mapping of Gladwell et al., 1983; Townsend, 1987; Clark et al., 1990;
various structural features, thereby optimizing planning of Fraser, 1991). Weathering processes produce the same
 94 H.M. Rajesh

minerals as hydrothermal alteration processes and mask

the spectral response of underlying rocks with coatings
and internal mineralogical transformations (Buckingham
and Sober, 1983). Hence it is important that careful field
verification should be performed in the areas marked as
hydrothermally altered by satellite image processing. The
color of the rock is a good key to the identification of
these minerals: when iron oxides are present, the rock
color is red, brown, orange or yellow; and the presence of
clay minerals usually gives pale colors (yellow, violet,
green, beige).
Landsat TM images are useful for hydrothermal
mineral identification because of the availability of mid −
infrared bands in which the characteristic spectral features
of most hydrothermal minerals are present. The tradi-
tional processing of an image with an aim of identifying
alteration minerals includes application of band combina-
tions, band ratios, and/or PCA. Using Landsat TM data,
Figure 11. Landsat image covering part of the Transvaal craton, an image incorporating ratios of bands 5 and 7 (TM5/7;
South Africa. Important features in the image include the Bush-
clay mineral index), bands 3 and 1 (TM3/1; iron oxide
veld layered basic − ultrabasic intrusion in the north − east, the
circular Pilansberg intrusion (syenite and foyaite) at the top index), and bands 5 and 4 (TM5/4; ferrous index) will
center, strongly folded (dipping northward) Proterozoic Trans- highlight areas where concentrations of these minerals
vaal Supergroup rocks in the central part, and the northern edge occur, thereby discriminating altered from unaltered
[east − west bands of green − grey (darker portions in the ground. In arid and semi − arid regions, outcropping
southwestern part of the image)] of the Witwatersrand basin,
hydrothermal alteration zones are mineralogically
South Africa’s most productive area of gold and uranium, near
the bottom. A variety of important mineral deposits − such as conspicuous enough to be detected successfully from
chromium, vanadium, nickel, copper and tin − occur in and around Landsat TM data (e.g. Amos and Greenbaum, 1989;
the Bushveld complex. (Courtesy: NASA and Earth Satellite Fraser, 1991; Spatz, 1997). In tropical regions, however,
Corporation.) high vegetation density can critically limit the successful
application of Landsat TM data to the detection and

Figure 12. Scatter plot in the TM3/TM1 vs.

TM4/TM1 space of hematite, goethite, and
vegetation training sets for the Newman
area, Australia. The inset is a similar plot
with the eigen vectors superimposed (Fraser,
1991).
 Application of remote sensing and GIS in mineral resource mapping 95

mapping of hydrothermally altered rocks (e.g. Siegal and altered minerals in the visible, NIR and MIR regions of
Goetz, 1977). In this case, techniques for digital enhance- the spectra. Tangestani and Moore (2002) showed that
ment of Landsat TM data to map hydrothermally altered the application of Crósta technique (Crósta and Rabelo,
rocks commonly aim at the identification of clay and iron 1993), another variant of PCA, on TM bands 1, 4, 5 and 7
oxide alteration zones (Fraser and Green, 1987). The enhances the hydroxyl − rich altered haloes around the
remote detection of iron oxide and clay zones in the pres- porphyry copper deposits of the Meiduk area, Iran. The
ence of vegetation, however, is difficult due to similarities Crósta technique (using four TM bands) involves the
in the reflectance spectra of the materials. If TM data are analysis of the eigenvector values allowing identification
to provide information regarding the distribution of ferric of principal components that contain spectral information
oxide minerals (hematite, goethite), the effects of vegeta- about specific minerals, as well as the contribution of each
tion need to be minimized. Different techniques for of the original bands to the components in relation with
image processing of Landsat TM to detect and map the spectral response of the mineral of interest. Remote
hydrothermally altered rocks are hence aimed at sepa- sensing of limonitic and clay alteration by the different
rating or reducing substantially the spectral effects of PCA techniques proves inadequate where the hydro-
vegetation from the spectral effects of the underlying thermal alteration mineral assemblages are not iron oxides
substrate (Fraser and Green, 1987). Spectral unmixing and clays, as pointed by Carranza and Hale (2002) for the
(Smith et al., 1985) is one such technique and endeavors Baguio district, Phillippines. They developed a mineral
at searching the abundances or fractions of pure spectral imaging methodology using Landsat TM data. It includes
components, so − called end − members, which best explain four steps: a) first to use the selective PCA technique to
the observed mixed pixel spectra. PCA of images has enhance the spectral response of each alteration mineral
been shown to be a successful tool to minimize the vege- into a separate mineral image based on published reflec-
tation effect in the resulting images (Abrams et al., 1983; tance spectra of minerals, b) second to extract training
Kaufman, 1988; Loughlin, 1991; Fraser, 1991; Bennett et areas for known hydrothermal alteration zones, c) third to
al., 1993; Ruiz − Armenta and Prol − Ledesma, 1998; carry out a supervised classification of the mineral images
Tangestani and Moore, 2001). The principal components to map hydrothermal alteration zones, and d) fourth to
can be analyzed using the standard or selective method. incorporate a DEM for improving the results of the classi-
In the standard analysis all available bands of an image fication. In the Baguio district, the accuracy of the classi-
are used as input for the principal components calculation, fied hydrothermal alteration map based on the mineral
while in the selective analysis only certain bands are images reached 69%, while inclusion of a DEM in the
chosen. Fraser (1991) used the selective method classification enhances the accuracy to 82%.
involving directing a principal component analysis at two In order to improve the definition (both spatial and
selected input bands TM3/TM1 and TM4/TM1 to spectral) of the target areas, Landsat images can be
discriminate between ferric oxide (hematite and goethite) merged with a digitized aerial photograph through inten-
and vegetation from the Newman area, western Australia. sity, hue and saturation (IHS; here intensity represents
Here band ratios are chosen because they are more useful brightness, hue represents color and saturation represents
than the TM bands as they compensate for the variations the purity of the color) transform. Although multi − dimen-
caused by topographic features and illumination condi- sional images are typically portrayed in RGB (red, green,
tions in the scene. A pixel containing hematite and blue), IHS transformed images incorporate more informa-
goethite would plot on the hematite − goethite line (Fig. tion because they include more easily defined and identifi-
12). If vegetation were added to that pixel, its position in able color attributes, greater control over the chromatic
3/1 versus 4/1 space would tend to move away from the and achromatic components of the image, and the ability
hematite − goethite line in the direction of vegetation (Fig. to create images utilizing information from more than
12). Similarly, for an area in central Mexico, Ruiz − three input data channels when the image is returned to
Armenta and Prol − Ledesma (1998) used two selected RGB color space (Harris et al., 1990). IHS incorporating
input ratio pairs (TM3/TM1 and TM4/TM3 ratios, and the concept of high pass filter (HPF; Chavez et al., 1991)
TM4/TM5 and TM5/TM7 ratios), each chosen because of technique is used for merging high spectral resolution
its effectiveness at highlighting and/or separating hema- multi − spectral satellite remote sensor data (e.g. Landsat)
tite, goethite, hydroxyl minerals, and vegetation, in with higher spatial resolution panchromatic (e.g. IRS)
n − dimensional ratio − versus − ratio space. data, to increase the spectral and spatial frequency distri-
The enhancement of the iron oxide and hydroxyl − bution. With its high spatial resolution, and bands
bearing areas around intrusive bodies in an area relies covering a wide part of the electromagnetic spectrum,
mainly on the spectral characteristics of the dominant ASTER data is known to provide accurate alteration maps
 96 H.M. Rajesh

thermal − IR portions of the spectrum. These systems typi-

cally collect 200 or more bands of data, which enables the
construction of an effectively continuous reflectance
(emittance in the case of thermal − IR energy) spectrum for
every pixel in the scene. Thus hyperspectral sensors can
produce data of sufficient spectral resolution for direct
identification of minerals, whereas the broader band TM
cannot resolve these diagnostic spectral differences.
Most hyperspectral remote sensing is still performed
using airborne systems such as the Compact Airborne
Spectrographic Imager (CASI), the Airborne Visible and
Infrared Imaging Spectrometer (AVIRIS), the SWIR Full
Spectrum Imager (SFSI), HYperspectral Digital Imagery
Collection Experiment (HYDICE), Probe − 1 and HyMap.
Examples of satellite sensors include Hyperion, FTHSI
and Australian Resource Information and Environment
Satellite (ARIES − 1). Reflectance spectra of minerals
measured by different spectroradiometers with different
spectral resolution are stored in spectral libraries that are
Figure 13. ASTER image of the Escondida copper, gold, and available in digital format (e.g. Grove et al., 1992; Clark
silver open − pit mine, Chile. The hydrothermal alteration mineral et al., 1993) against which the data can be correlated and
zones include porpylitic, phyllic and potassic zones. A high − classified.
grade supergene cap overlies primary sulfide ore. The top image Several approaches to extract information from
is a conventional 3 − 2 − 1 (near infrared, red, green) RGB
high − spectral resolution image data have been reported,
composite. The bottom image displays short − wave infrared
bands 4 − 6 − 8 (1.65 μ m, 2.205 μ m, 2.33 μ m) in RGB, and high- processing spectral information on a pixel − by − pixel basis,
lights the different rock types present on the surface, as well as including binary encoding (Goetz et al., 1982), spectral
the changes caused by mining. (Courtesy: NASA GSFC, MITI, unmixing (Adams et al., 1986), relative absorption band −
ERSDAC, JAROS, and U.S./Japan ASTER Science Team.) depth mapping (Crowley et al., 1989), waveform charac-
terization (Okada and Iwashita, 1992), classification (Cetin
et al., 1993), spectral angle mapping (SAM; Kruse et al.,
(e.g. Fig. 13). Aster 4 operates in the spectral range 1.6 − 1993), decorrelation stretching (Abrams and Hook, 1995),
1.7 μ m, which is a general high − reflectance band, and is probability density function (Nedeljkovic and Pendock,
coded in red. Aster 6 (2.225 − 2.245 μ m) is absorbed by 1996), constrained energy minimization (CEM; Farrand
the Al − OH minerals, whereas Aster 8 (2.295 − 2.365 μ m) and Harsanyi, 1997), cross correlogram spectral matching
corresponds to the absorption by Mg − OH minerals (and (Van der Meer and Bakker, 1997), back − propagation
carbonates, if present). Therefore in this color coding neural network (BNN; Yang et al., 1999), and geophysical
scheme, Al − OH − bearing minerals appear in shades of inversion (Van der Meer, 2000). These methods all use
blue − purple, Mg − OH − bearing minerals appear in shades imaging spectrometer data corrected to reflectance and
of green − yellow, and Al − Mg − OH − bearing minerals quantitatively test the similarity of unknown imaged
appear in shades of red (e.g. Fig. 13). spectra with known spectra measured in the field or labo-
ratory or extracted from the image data at locations of
Imaging Spectrometry known ground truth. This often results in mineral maps,
which portray the probability that a pixel is composed of
Typically, the number of spectral bands of remote sensors a certain mineral.
determines the amount of spectral information that remote Most of the image analysis algorithms developed
sensors can acquire. Early remote sensors only have specifically to exploit the extensive information contained
several spectral bands (e.g. Landsat 7 has 7 spectral in hyperspectral imagery also provide accurate, although
bands), and thus limited spectral information can be more limited, analysis of multispectral data. The different
obtained from such remote sensors. Hyperspectral algorithms can be grouped as either whole − pixel or sub −
sensors (also referred to as Imaging Spectrometers) can pixel analysis methods. Whole − pixel analysis methods
acquire images in many, very narrow, contiguous spectral calculates the spectral similarity between a test (or pixel)
bands throughout the visible, near − IR, mid − IR, and reflectance spectrum and a reference (or laboratory)
 Application of remote sensing and GIS in mineral resource mapping 97

Figure 14. a and b: Mineral maps derived

from AVIRIS data obtained over Cuprite
in 1995. The image in a is derived from
analyzing the vibrational absorption
features in minerals (typically in the 2 − 2.5
μ m spectral region) common to OH − ,
CO3 − , and SO4 − bearing minerals. The
image in b is derived from analyzing the
electronic absorption features in minerals
(typically in the 0.4 − 1.2 μ m spectral
region) common of Fe2+ and Fe3+ bearing
minerals (Clark et al., 2003).
c. SAM classification map over Cuprite
derived from SFSI − 2 image. Collected
end − member spectra with corresponding
JPL library spectra are shown on the right
(Borstad et al., 2000).

reflectance spectrum and include standard supervised tral feature fitting. Back − Propagation neural Network
classifiers such as minimum distance or maximum likeli- (BPN) is a method, which examine all the pixels in the
hood, as well as tools developed specifically for hyper- image in parallel. For BPN method, first each classifica-
spectral imagery such as SAM and spectral feature fitting. tion resulting from a hidden − layered neural network
The SAM computes a spectral angle between each pixel containing hidden units is trained using a back − propaga-
spectrum and each reference spectrum. The outcome of tion algorithm. The result is an image classified by the
SAM − algorithm gives a qualitative estimate of the pres- acquired network, which responds to each new unseen
ence of absorption features, which can be related to vector with the knowledge gained from the training stage
mineralogy. Van der Meer et al. (1997) demonstrated the (e.g. Yang et al., 1999). Sub − pixel analysis methods
potential of SAM technique for a first assessment of calculate the quantity of reference materials in each pixel
mineral potential in ultramafic terrains. Another approach of an image and include tools such as linear spectral
to matching reference and pixel spectrum is to examine unmixing and matched filtering. Linear spectral unmixing
specific absorption features in the spectra, as used in spec- exploits the theory that the reflectance spectrum of any
 98 H.M. Rajesh

pixel is the result of linear combinations of the spectra of

all end members inside that pixel.
Beraton et al. (1997) showed that AVIRIS could be
successfully used for mapping the occurrence of potas-
sium metasomatism (usually the alteration marked by
pervasive replacement of diverse rock types by adularia +
hematite ± quartz ± illite − montmorillonite; Roddy et al.,
1988) in well − exposed areas. The AVIRIS derived
mineral distribution maps for the Cuprite mining district,
Nevada, well known in the remote sensing community as
a test site for many years because of its exposed altered
bedrock, alluvial deposits and minimal vegetation cover
(e.g. Abrams et al., 1977; Kruse et al., 1990; Rast et al.,
1991), is given in Figure 14 (a, b). Figure 14 (a, b)
utilizes the tricorder algorithm (Clark et al., 2003), that
uses a digital spectral library of known reference minerals
and a fast, modified least − squares method of determining
if a diagnostic spectral feature for a given mineral is
present in the image. SAM classification map (2.0 − 2.4
μ m) over Cuprite derived from the modified and updated
SFSI (SFSI − 2) is shown in Figure 14c. Resmini et al.
(1997) illustrated the application of CEM, which requires
only the spectrum of the mineral to be mapped and no
prior knowledge of background constituents, as a rapid
technique for mineral mapping from the Cuprite area.
Yang et al. (1999) illustrated the high classification accu- Figure 15. Flow chart illustrating the initial database building,
racy of the BPN over SAM for the Curpite area, and is decision support system, and final modeling used for mineral
potential mapping in a GIS. The possible layers included within
attributed to its ability to deal with complex relationships the GIS database are given in the dashed − line box. Depending
and the nature of the data set. Neville et al. (2003) on the availability of data, different steps illustrated in the flow-
showed the ability of constrained linear unmixing proce- chart will be modified.
dure to create mineral fraction maps, using SFSI and
AVIRIS data sets collected over Cuprite. The resulting
mineral abundance maps are usually merged with other effective integration and analysis of large numbers of geo-
exploration data such as the digital elevation models to referenced spatial data with different attributes and formats.
get a better perspective. As far as the mineral potential mapping using GIS is
concerned, it generally involves four main steps: building
Application of GIS a spatial digital database, extracting predictive evidence
for a particular mineral deposit type, calculating weights
Mineral exploration has traditionally been based on the for each predictive evidence map, and combining the
application of a variety of prospecting methods, namely maps to predict mineral potential. The probability that a
geochemistry, geophysics, geological mapping, aerial mineral deposit exists increases where the predictor
photo interpretation and ground surveys, operatively inte- themes areally overlap. Here the initial database building
grated within exploration phases. The main criterion of step is the most time consuming step and can include
the prospector is to identify anomalies associated with among others, remote sensing data, geophysical data
target mineral areas, gradually reducing the original (magnetic, gravity), geochemical data (for each element
extent area to a small set of anomalies. This process is of interest), geological data (structural, lithological), topo-
complex and it needs both analysis and integration of the graphical data (DEM), and mineral occurrence data.
above multithematic exploration information, from which A general scheme of the methodology followed for
decisions must be made over time and at different stages. GIS implementation is given in Figure 15 and consists of
In the applied context remote sensing, and more recently, two main parts: the relational database and spatial anal-
GIS have shown their usefulness as support tools of great ysis leading to a decision support system (geognostic
interest. GIS is considered because they allow for more variable selection and weighting of variables). The rela-
 Application of remote sensing and GIS in mineral resource mapping 99

tional database contains and integrates the georeferenced

information coming from remote sensing imagery,
existing maps and exploration/prospecting data, expressed
as tables and maps. A major function of a GIS is the
ability to analyze the spatial relationships between data
sources. Here spatial analysis is based on the criterion on
combining (overlaying) multi − class maps (Bonham −
Figure 16. Example of the structure of a network model used for
Carter, 1994) leading to the mineral potential map or gold potential mapping.
mineral favorability map. It can also be carried out with
reference to a practical concept named Mineral Potential
Index (MPI) or Mineral Potential Score (MPS). This tion function, a lineament density can be calculated for
index is numerically a linear expression in which such a the concerned area. Lineament density, in similar fashion
calculation requires us to assign score to each class − map to fracture density, but on a larger scale, provides an
and weights to each input − map according to some criteria, approximate location of mineral concentration and
leading to the final MPI or MPS map (see Fig. 15). measure of how broken is the rock mass (McGregor et al.,
Data integration methods in GIS may be divided into 1999). This is a very good example for the application of
two main groups: knowledge − driven models, such as GIS in extracting accurate geologic information including
weighted index overlay, fuzzy logic and multi − criteria linear features (lineaments) for mineral resource mapping.
evaluation, estimate the parameters of a function for
combining datasets on the basis of expert opinion; and Concluding remarks
data − driven models, such as regression, weights of
evidence modeling, Dempster − Shafer belief functions, The important question that a geologist can ask is “Why
certainty factors, Bayesian statistics and artificial neural should one use remotely sensed imagery for mineral
networks (ANN), estimate model parameters from resource mapping?” The answer is, because it can provide
measured data (Bornham − Carter, 1994). The data − driven information that is not available any other way. Geologic
approach, unlike the knowledge − driven approach, maps (assuming that they are even available) may be
requires a number of known deposits to exist in the area inaccurate and, at best, are usually generalizations. The
of interest, and the parameters identified for a particular geologist doing the mapping may have painstakingly
deposit type in an area can only be applied to other mapped the boundaries of a particular formation and
regions with similar geology. It is now known that both missed an obvious alteration zone within it. Figures
knowledge − driven data integration and data − driven drawn on the ground by ancient peoples are overlooked
models involve assumptions that are difficult to satisfy by those walking on them; however, they stand out clearly
when dealing with geological variables such as linear when viewed from the air. Similarly, geologic structures
relationships. ANNs, however, are an exception and seem and mineral alteration patterns can be quite vivid on satel-
to offer advantages over other methods because they lite imagery. In a hilly terrain, where poor accessibility
make no assumptions about the data. Although ANNs are and dense vegetation cover hinder fieldwork, it is not
used frequently as remote sensing data classifiers, their always possible to map or collect data from field
potential as spatial modeling tools in mineral potential throughout any structural feature. The data are collected
mapping is illustrated in Figure 16. from accessible places and geologists then interpolate
A growing number of studies evaluated the potential those data to obtain continuity. In this context, image
for extracting geologic information (e.g. lithologic interpretation of structural features is more reliable than
mapping, structural mapping, etc.) by integrating tradi- their interpolation from field data. Subtle color variations
tional remote sensing, geophysical data and ancillary data that would go unnoticed on the ground can be made quite
(e.g. DEM) in a GIS environment. Structural analysis of bold in false − color renditions, begging to be assayed, as it
satellite imagery and aerial photograph can generate large were. Known mineral resource areas can sometimes be
population of interpreted linear features. This population extended by tracing lineaments outside the areas. Remote
can be reduced through lineament interpretations, based sensing data, by virtue of its synoptic overview, multi-
on the identification of features which are geologically spectral and multi − temporal coverage, can help to rapidly
significant, within a GIS; thus allowing the lineament delineate metallogenic provinces/belts/sites over a larger
vectors to be properly registered immediately. The linea- terrain, based on known models of commercial ore occur-
ments interpreted from the satellite imagery can be rences. This can help to isolate potential areas from non −
converted to a raster grid and, using a local block summa- interesting areas for further exploration.
 100 H.M. Rajesh

The main considerations for remote sensing applica- References

tions in mineral resource mapping are:
• Reconnaissance lithologic mapping is usually the Abrams, M.J. and Hook, S.J. (1995) Simulated Aster data for
first step of mineral resource mapping projects. geologic studies. IEEE Transactions of Geosciences and
Remote Sensing, 33, 692 − 699.
Better results can be achieved when remote sensing
Abrams, M.J., Ashley, R.P., Rowan, L.C., Goetz, A.F.H. and
and potential field data are used for the extraction of Kahle, A.B. (1977) Mapping of hydrothermal alteration in
lithologic information. the Cuprite mining district, Nevada, using aircraft scanner
• Structural mapping to supplement the lithologic images for the spectral region 0.46 − 2.36 μ m. Geology, 5,
information. Deduction of information regarding 713 − 718.
Abrams, M.J., Brown, D., Lepley, L. and Sadowski, R. (1983)
localization of mineral deposits by certain types of
Remote sensing for porphyry copper deposits in southern
geologic structures (including lineaments) is vital in Arizona. Economic Geology, 78, 591 − 604.
planning detailed mineral resource exploration. Abrams, M.J., Rothery, D.A. and Pontual, A. (1988) Mapping in
• Alteration mapping to deduce the information about the Oman ophiolite using enhanced Landsat Thematic
the possible mineral resource. Mineral deposits are Mapper images. Tectonophysics, 151, 387 − 401.
commonly associated with hydrothermal alteration Adams, J.B., Smith, M.O. and Johnson, P.E. (1986) Spectral
mixture modeling: a new analysis of rock and soil types at
of the surrounding rocks, the style and extent of the
the Viking lander site. Journal of Geophysical Research, 91,
alteration reflecting the type of mineral deposit. 8098 − 8112.
• Hyperspectral data can help identify and thematically Akhavi, M.S., Webster, T.L. and Raymond, D.A. (2001)
map regions of mineral resource interest by using RADARSAT − 1 imagery and GIS modeling for mineral
distinct absorption features of most minerals. With exploration in Nova Scotia, Canada. Geocarto International,
16, 55 − 61.
little doubt, the advent of hyperspectral imagery is
Amos, B.J. and Geenbaum, D. (1989) Alteration detection using
the most exciting thing to happen to geology in a TM imagery: the effects of supergene weathering in arid
decade. The detection of subtle differences in reflec- climate. International Journal of Remote Sensing, 10, 515 −
tance (or, more correctly, radiance), measured across 527.
a large number of narrow bandwidths, allows classi- Bennett, S.A., Atkinson, W.W. and Kruse, F.A. (1993) Use of
fication of rock types and analysis of mineral Thematic Mapper imagery to identify mineralization in the
Santa Teresa district, Sonora, Mexico. International Geology
contents with far greater accuracy than ever before.
Review, 35, 1009 − 1029.
• Finally coming to the exploration stage, it is clear Beraton, K.K., Peer, B., Dunbar, N.W. and Blom, R. (1997) A
that the remote sensing data gathered has to be inte- remote sensing approach to alteration mapping: AVIRIS data
grated with other geoscience data like geochemical, and extension − related potassium metasomatism, Socorro,
geophysical, etc. This demands a multithematic New Mexico. International Journal of Remote Sensing, 18,
3595 − 3609.
approach, and GIS form the perfect tool to integrate
Berger, Z., Williams, T.H.L. and Anderson, D.W. (1992) Geologic
the data sets and do spatial analysis leading to stereo mapping geologic structures with SPOT satellite data.
mineral potential maps. AAPG Bulletin, 76, 101 − 120.
Bilotti, F., Shaw, J.H. and Brennan, P.A. (2000) Quantitative
Acknowledgments structural analysis with stereoscopic remote sensing imagery.
AAPG Bulletin, 84, 727 − 740.
Blom, R.G. and Daily, M. (1982) Radar image processing for
Stuart Phinn, David Pullar and Pramod Sharma are
rock − type discrimination. IEEE Transactions of Geoscience
thanked for comments and discussions. University of and Remote Sensing, 20, 343 − 351.
Queensland is thanked for facilities. Natsumi Takao is Bonham − Carter, G.F. (1994) Geographic information systems for
thanked for support and discussions. Steve Drury is geoscientists: Modeling with GIS. pp.398, Pergamon, UK.
thanked for the clarification regarding his quote on the Borstad, G.A., Lim, J., Neville, R.A. and Hauff, P. (2000) SFSI− 2 −
a new source of airborne hyperspectral SWIR imagery. 22nd
unending search for the hitherto undiscovered potato field
Annual Canadian Remote Sensing Symposium, Victoria,
and the interest in the work. Bonham − Carter GF (and British Columbia, 21 − 25 August 2000.
Anna Ypma from Elsevier), Benoit Rivard, Gary Borstad, Bucci, L.A., Hagemann, S.G., Groves, D.I. and Standing, J.G.
Gregg Swayze, Manoel Araújo Sousa (Jr.), and Roger (2002) The Archean Chalice gold deposit: a record of
Clark are thanked for the permission to reproduce figures complex, multistage, high − temperature hydrothermal activity
from their book/article. The two journal reviewers and gold mineralisation associated with granitic rocks in the
Yilgarn Craton, Western Australia. Ore Geology Reviews,
provided constructive and encouraging comments.
19, 23 − 67.
Koichiro Fujimoto is thanked for editorial efforts. This Buckingham, W.F. and Sober, S.E. (1983) Mineralogical charac-
is a contribution to the Gondwana Institute for Geology terization of rock surfaces by hydrothermal alteration and
and Environment (GIGE) and IGCP 453. weathering: Application to remote sensing. Economic
 Application of remote sensing and GIS in mineral resource mapping 101

Geology, 78, 664 − 674. mapping and vegetation discrimination. IEEE Transactions
Budkewitsch, P., Newton, G. and Hynes, A.J. (1994) Character- of Geosciences and Remote Sensing, GE − 24, 246 − 257.
ization and Extraction of Linear Features from Digital Farrand, W.H. (1997) Identification and mapping of ferric oxide
Images. Canadian Journal of Remote Sensing, 20, 268 − 279. and oxyhydroxide minerals in imaging spectrometer data of
Carranza, E.J.M. and Hale, M. (2002) Mineral imaging with Summitville, Colorado, USA, and the surrounding San Juan
Landsat Thematic Mapper data for hydrothermal alteration Mountains. International Journal of Remote Sensing, 10,
mapping in heavily vegetated terrane. International Journal 1543 − 1552.
of Remote Sensing, 23, 4827 − 4852. Farrand, W.H. and Harsanyi, J.C. (1997) Mapping the distribution
Cetin, H., Warner, T.A. and Levandowski, D.W. (1993) Data clas- of mine tailings in the Coeur d’Alene River valley, Idaho,
sification, visualization, and enhancement using n − dimen- through the use of a Constrained Energy Minimization
sional probability density functions (nPDF): AVIRIS, TIMS, Technique. Remote Sensing of Environment, 59, 64 − 76.
TM and geophysical applications. Photogrammetric Ferreir, G., White, K., Griffiths, G., Bryant, R. and Stefouli, M.
Engineering and Remote Sensing, 59, 1755 − 1764. (2002) The mapping of hydrothermal alteration zones on the
Chatterjee, R.S., Prabakaran, B. and Jha, V.K. (2003) Fusion of island of Lesvos, Greece using an integrated remote sensing
surface relief data with high spectral and spatial resolution dataset. International Journal of Remote Sensing, 23, 341 −
satellite remote sensor data for deciphering geological infor- 356.
mation in a mature topographic terrain. International Journal Fraser, A., Huggins, P., Reens, J. and Cleverly, P. (1997) A satel-
of Remote Sensing, 24, 4761 − 4775. lite remote sensing technique for geological horizon
Chavez, P.S., Sides, S.C. and Anderson, J.A. (1991) Comparison mapping. International Journal of Remote Sensing, 18,
of three different methods to merge multiresolution and 1607 − 1615.
multispectral data: Landsat TM and SPOT panchromatic. Fraser, S.J. (1991) Discrimination and identification of ferric
Photogrammetric Engineering and Remote Sensing, 57, 295 − oxides using satellite Thematic Mapper data: a Newman case
303. study. International Journal of Remote Sensing, 12, 635 −
Chernicoff, C.J., Richards, J.P. and Zappettini, E.O. (2002) 641.
Crustal lineament control on magmatism and mineralization Fraser, S.J. and Green, A.A. (1987) A software defoliant for
in northwestern Argentina: geological, geophysical, and geological analysis of band ratios. International Journal of
remote sensing evidence. Ore Geology Reviews, 21, 127 − Remote Sensing, 8, 525 − 532.
155. Fu, B., Lin, A., Kano, K. − I., Maruyama, T. and Guo, J. (2004)
Christensen, P.R., Bandfield, J.L., Hamilton, V.E., Howard, D.A., Application of stereoscopic satellite images for studying
Lane, M.D., Piatek, J.L., Ruff, S.W. and Stefanov, W.L. (2000) Quaternary tectonics in arid regions. International Journal of
A thermal emission spectral l ibrary of rock − forming Remote Sensing, 25, 537 − 547.
minerals. Journal of Geophysical Research, 105, 9735 − 9739. Gaffey, S.J. (1985) Reflectance spectroscopy in the visible and
Clark, R.N., King, T.V.V., Klejwa, M., Swayze, G. and Vergo, N. near − infrared (0.35 − 2.55 microns): applications in carbonate
(1990) High Spectral Resolution Reflectance Spectroscopy petrology. Geology, 13, 270 − 273.
of Minerals. Journal of Geophysical Research, 95, 12653 − Gladwell, D.R., Lett, R.E. and Lawrence, P. (1983) Application of
12680. reflectance spectroscopy to mineral exploration. Economic
Clark, R.N., Swayze, G.A., Gallagher, A., King, T.V.V. and Geology, 78, 699 − 710.
Calvin, W.M. (1993) The US Geological Survey Digital Goetz, A.F.H., Rowan, L.C. and Kingston, M.J. (1982) Mineral
Spectral Library, Version 1: 0.2 to 3.0 mm. US Geological identification from orbit: initial results from the Shuttle Multi-
Survey Open File Report 93 − 592. spectral Infrared Radiometer. Science, 218, 1020 − 1031.
Clark, R.N., Swayze, G.A., Livo, K.E., Kokaly, R.F., Sutley, S.J., Grove, C.I., Hook, S.J. and Payloe, E.D., II. (1992) Laboratory
Dalton, J.B., McDougal, R.R. and Gent, C.A. (2003) reflectance spectra of 160 minerals, 0.4 to 2.5 micrometers.
Imaging spectroscopy: Earth and planetary remote sensing pp.300, NASA − JPL Publication 92 − 2, Psadena.
with the USGS Tetracorder and expert systems. Journal of Guilbert, J.M. and Park, C.F. (1996) The geology of ore deposits.
Geophysical Research, 108(E12), 5131, DOI 10.1029/ W.H. Freeman, New York.
2002JE001847. Harris, J., Murray, R. and Hirose, T. (1990) IHS transformation
Costa, E.D.d. and Starkey, J. (2001) Photolin: a program to iden- for integration of radar imagery with other remotely sensed
tify and analyze linear structures in aerial photographs, satel- data. Photogrammetric Engineering and Remote Sensing,
lite images and maps. Computers and Geosciences, 27, 527 − 56, 1631 − 1641.
534. Harris, J.R. (1991) Mapping of regional structures of eastern
Crósta, A.P. and Rabelo, A. (1993) Assessing Landsat TM for Nova Scotia using remotely sensed images: implications for
hydrothermal mapping in central − western Brazil. Proceed- regional tectonics and gold exploration. Canadian Journal of
ings of the Ninth Thematic Conference on Geologic Remote Remote Sensing, 17, 122 − 135.
Sensing, Pasadena, California, USA, 1993, 1053 − 1061. Helg, U., Burkhard, M., Caritg, S. and Robert − Charrue, C. (2004)
Crowley, J.K., Brickey, D.W. and Rowan, L.C. (1989) Airborne Folding and inversion tectonics in the Anti − Atlas of
imaging spectrometer data of the Ruby Mountains, Montana: Morocco. Tectonics, 23(TC4006), 1 − 17.
mineral discrimination using relative absorption band − depth Hunt, G.R. (1977) Spectral signatures of particulate minerals, in
images. Remote Sensing of Environment, 29, 121 − 134. the visible and near − infrared, Geophysics, 42, 501 − 513.
Drury, S. (2001) Image interpretation in geology. pp. 290, Hunt, G.R. (1979) Near − infrared (1.3 − 2.4 μ m) spectra of altera-
Blackwell Science, USA. tion minerals − potential for use in remote sensing.
Evans, D.L., Farr, T.G., Ford, J.P., Thompson, T.W. and Werner, Geophysics, 44, 1974 − 1986.
C.L. (1986) Multipolarization radar images for geologic Hunt, G.R. and Ashley, R.P. (1979) Spectra of altered rocks in the
 102 H.M. Rajesh

visible and near infrared. Economic Geology, 74, 1613 − (1988) Digital processing of orbital radar data to enhance
1629. geologic structure: examples from the Canadian shield.
Hunt, G.R. and Salisbury, J.W. (1970a) Visible and near − infrared Photogrammetric Engineering and Remote Sensing, 54, 621 −
spectra of minerals and rocks: I silicate minerals. Modern 632.
Geology, 1, 283 − 300. Mazzarini, F., Pareschi, M.T., Sbrana, A., Favalli, M. and
Hunt, G.R. and Salisbury, J.W. (1970b) Visible and near − infrared Fulignati, P. (2001) Surficial hydrothermal alteration
spectra of minerals and rocks: III oxides and hydroxides. mapping at Vulcano island using MIVIS data. International
Modern Geology, 2, 195 − 205. Journal of Remote Sensing, 22, 2045 − 2070.
Hunt, G.R. and Salisbury, J.W. (1971) Visible and near − infrared McGregor, R.G., Bruton, P., Singhroy, V.H. and Brown, A. (1999)
spectra of minerals and rocks: II carbonates. Modern An applications of remote sensing and GIS towards geolog-
Geology, 2, 23 − 30. ical site characterization of a Canadian shield terrain.
Hunt, G.R. and Salisbury, J.W. (1976) Visible and near − infrared Canadian Journal of Remote Sensing, 25, 245 − 257.
spectra of minerals and rocks: XII. Metamorphic rocks. Memmi, J.M. and Pride, D.E. (1997) The application of Diamond
Modern Geology, 5, 219 − 228. Exploration Geoscientific Information System (DEGIS) tech-
Kahle, A.B., Morrison, A.D., Tsu, H. and Yamaguchi, Y. (1996) nology for integrated diamond exploration in the north −
Geologic remote sensing in the thermal infrared. Interna- central United States of America. International Journal of
tional Archives of Photogrammetry and Remote Sensing, 31, Remote Sensing, 18, 1439 − 1464.
327 − 330. Mickus, K. and Johnson, E. (2001) Mapping sedimentary and
Katz, M. (1982) Lineament analysis of Landsat imagery applied volcanic units within and surrounding Petrified Forest
to mineral exploration. In Mineral exploration techniques in National Park, Arizona, using Landsat − 5 and SPOT panchro-
tropical forest areas (Laming, D.I.C. and Gibbs, A.K. Eds.). matic data. International Journal of Remote Sensing, 22,
Hiden Wealth, 157 − 166. 1919 − 1935.
Kaufman, H. (1988) Mineral exploration along the Aqaba − Nedeljkovic, V. and Prndock N. (1996) Finding spectral anoma-
Levanat structure by use of TM data, concepts, processing lies in hyperspectral imagery using nonparametric probability
and results. International Journal of Remote Sensing, 9, density estimation techniques. Proceedings of the Eleventh
1630 − 1658. Thematic conference and Workshop on Applied Geologic
Koike, K., Nagano, S. and Ohmi, M. (2001) Lineament analysis Remote Sensing, Las Vegas, Nevada, 1, 197 − 204.
of satellite images using a Segment Tracing Algorithm (STA). Neville, R.A., Lévesque, J., Staenz, K., Nadeau, C., Hauff, P. and
Computers and Geosciences, 21, 1091 − 1104. Borstad, G.A. (2003) Spectral unmixing of hyperspectral
Krishnamurthy, J. (1997) The evaluation of digitally enhanced imagery for mineral exploration: comparison of results from
Indian Remote Sensing Satellite (IRS) data for lithological SFSI and AVIRIS. Canadian Journal of Remote Sensing, 29,
and structural mapping. International Journal of Remote 99 − 110.
Sensing, 18, 3409 − 3437. O’Leary, D.W., Friedman, J.D. and Pohn, H.A. (1976) Lineament,
Kruse, F.A., Kierein − young, K.S. and Boardman, J.W. (1990) linear and lineation: some proposed new standards for old
Mineral mapping at Cuprite, Nevada with a 63 − channel terms. Geological Society of America Bulletin, 87, 1463 −
imaging spectrometer. Photogrammetric Engineering and 1469.
Remote Sensing, 56, 83 − 92. Okada, K. and Iwashita, A. (1992) Hyper − multispectral image
Kruse, F.A., Lefkof, A.B., Boardman, J.W., Heidebrecht, K.B., analysis based on waveform characteristics of spectral curve.
Shapiro, A.T., Barloon, J.P. and Goetz, A.F.H. (1993) The Advanced Space Research, 12, 433 − 442.
spectral image processing system (SIPS) − interactive visual- Paganelli, F., Grunsky, E.C., Richards, J.R. and Pryde, R. (2003)
ization and analysis of imaging spectrometer data. Remote Use of RADARSAT − 1 principal component imagery for
Sensing of Environment, 44, 145 − 163. structural mapping: a case study in the Buffalo Head Hills
Kutina, J. (1969) Hydrothermal ore deposits in the western United area, northern central Alberta, Canada. Canadian Journal of
States: a new concept of structural control of distribution. Remote Sensing, 29, 111 − 140.
Science, 165, 1113 − 1119. Pedroso, E.C., Rivard, B., Crósta, A.P., Souza Filho, C.R. and
Lahren, M.M., Schweickert, R.A. and Taranik, J.V. (1988) Miranda, F.P. (2001) Reconnaissance geologic mapping in
Analysis of the northern Sierra accreted terrain, California, the Tapajós mineral province, Brazilian Amazon, using
with airborne thermal infrared multispectral scanner data. spaceborne SAR imagery and airborne geophysics. Canadian
Geology, 16, 525 − 528. Journal of Remote Sensing, 27, 669 − 678.
Liu, C.C., Sousa Jr., M.A. and Gopinath, T.R. (2000) Regional Pirajno, F. (1992) Hydrothermal mineral deposits, Principles and
structural analysis by remote sensing for mineral exploration, fundamental concepts for the exploration geologists.
Paraiba state, Northeast Brazil. Geocarto International, 15, Springer, Berlin.
69 − 75. Podwysocki, M.H., Segal, D.B. and Abrams, M.J. (1983) Use of
Longhi, I., Sgavetti, M., Chiari, R. and Mazzoli, C. (2001) Multispectral scanner images for assessment of hydrothermal
Spectral analysis and classification of metamorphic rocks alteration in the Marysvale, Utah, Mining area. Economic
from laboratory reflectance spectra in the 0.4 − 2.5 μ m Geology, 78, 675 − 687.
interval: a tool for hyperspectral data integration. Interna- Prost, G. (1980) Alteration mapping with airborne multispectral
tional Journal of Remote Sensing, 22, 3763 − 3782. scanners. Economic Geology, 75, 894 − 906.
Loughlin, W.P. (1991) Principal component analysis for alteration Raghavan, V., Masumoto, S., Koike, K. and Nagano, M. (1995)
mapping. Photogrammetric Engineering and Remote Automatic lineament extraction from digital images using a
Sensing, 57, 1163 − 1169. segment tracing and rotation transformation approach.
Masouka, P.M., Harris, J., Lowman, P.D., Jr. and Blodget, H.W. Computers and Geosciences, 21, 555 − 559.
 Application of remote sensing and GIS in mineral resource mapping 103

Rast, M., Hook, S.J., Elvidge, C.D. and Alley, R.E. (1991) An tance spectra using principal component analysis. Journal of
evaluation of techniques for the extraction of mineral absorp- Geophysical Research, 90, 797 − 804.
tion features from high spectral resolution remote sensing Spatz, D.M. (1997) Remote sensing characteristics of the sedi-
data. Photogrammetric Engineering and Remote Sensing, ment − and volcanic − hosted precious metal system: imagery
57, 1303 − 1309. selection for exploration and developments. International
Rebillard, Ph. and Evans, D. (1983) Analysis of coregistered Journal of Remote Sensing, 18, 1413 − 1438.
Landsat, Seasat and SIR − A images of varied terrain types. Stern, R.J., Beyth, M., Bodechtel, J. and Wetzel, H − U (2002)
Geophysical Research Letters, 10, 277 − 280. Potential of the International Space Station for imaging
Rein, B. and Kaufmann, H. (2003) Exploration for gold using Earth: Lessons from MOMS − 2P aboard Mir. Geology, 30,
panchromatic stereoscopic intelligence satellite photographs 851 − 854.
and Landsat TM data in the Hebei area, China. International Tangestani, M.H. and Moore, F. (2001) Comparison of three prin-
Journal of Remote Sensing, 24, 2427 − 2438. cipal component analysis techniques to porphyry copper
Resmini, R.G., Kappus, M.E., Aldrich, W.S., Harsanyi, J.C. and alteration mapping: A case study, Meiduk area, Kerman, Iran.
Anderson, M. (1997) Mineral mapping with HYperspectral Canadian Journal of Remote Sensing, 27, 176 − 182.
Digital Imagery Collection Experiment (HYDICE) sensor Tangestani, M.H. and Moore, F. (2002) Porphyry copper altera-
data at Cuprite, Nevada, U.S.A. International Journal of tion mapping at the Meiduk area, Iran. International Journal
Remote Sensing, 18, 1553 − 1570. of Remote Sensing, 23, 4815 − 4825.
Rivard, B., Corriveau, L. and Harris, L.B. (1999) Structural Townsend, T.E. (1987) Discrimination of iron alteration minerals
reconnaissance of a deep crustal orogen using RADARSAT in visible and near − infrared reflectance data. Journal of
and Landsat satellite imagery and airborne geophysics. Geophysical Research, 92, 1441 − 1454.
Canadian Journal of Remote Sensing, 25, 258 − 267. Van der Meer, F. (2000) Geophysical inversion of imaging spec-
Roddy, M.S., Reynolds, S.J., Smith, B.M. and Rui, J. (1988) K − trometer data for geologic modeling. International Journal of
metasomatism and detachment − related mineralization, Remote Sensing, 21, 387 − 393.
Harcuvar mountains, Arizona. Geological Society of Van der Meer, F. and Bakker, W. (1997) Cross Correlogram
America Bulletin, 100, 1627 − 1639. Spectral Matching (CCSM): application to surface mineral-
Rothery, D. (1987) Improved discrimination of rock units using ogical mapping using AVIRIS data from Cuprite, Nevada.
Landsat Thematic Mapper imagery of the Oman ophiolite. Remote Sensing of Environment, 61, 371 − 382.
Journal of the Geological Society of London, 144, 587 − 597. Van der Meer, F., Vazquez − Torres, M. and Van Dijk, P.M. (1997)
Ruiz − Armenta, J.R. and Prol − Ledesma, R.M. (1998) Techniques Spectral characterization of ophiolite lithologies in the
for enhancing the spectral response of hydrothermal altera- Troodos Ophiolite complex of Cyprus and its potential in
tion minerals in Thematic Mapper image of central Mexico. prospecting for massive sulphide deposits. International
International Journal of Remote Sensing, 19, 1981 − 2000. Journal of Remote Sensing, 18, 1245 − 1257.
Sabine, C., Realmuto, V.J. and Taranik, J.V. (1994) Quantitative Vickers, R.S. and Lyon, R.J.P. (1967) Infrared sensing from
estimation of granitoid composition from Thermal Infrared spacecraft − a geological interpretation. Proceedings,
Multispectral Scanner (TIMS) data, Desolation Wilderness, Thermophysics Special Conference, American Institute of
northern Sierra Nevada, California. Journal of Geophysical Aeronautics and Astronautics, Paper 67 − 284.
Research, 99, 4261 − 4271. Wiart, P.A.M., Oppenheimer, C. and Francis, P. (2000) Eruptive
Salisbury, J., Walter, L., Vergo, N. and D’Aria, D. (1991) Infrared history of Dubbi volcano, northeast Afar (Eritrea), revealed
(2.1 − 2.5 μ m) spectra of minerals. pp.267. Johns Hpokins by optical and SAR image interpretation. International
University Press, Baltimore. Journal of Remote Sensing, 21, 911 − 936.
Salisbury, J.W. and Hunt, G.R. (1974) Remote sensing of rock Yang, H., Van der Meer, F., Bakker, W. and Tan, Z.J. (1999) A
types in the visible and near − infrared. Proceedings of the back − propagation neural network for mineralogical mapping
Ninth International Symposium on the Remote Sensing of from AVIRIS data. International Journal of Remote Sensing,
the Environment, III, 1953 − 1958. 20, 91 − 110.
Sharma, K.N.M., Singhroy, V.H., Madore, L., Lévesque, J., Younis, M.T., Gilbert, M.A., Meliá, J. and Bastida, J. (1997)
Hébert, C. and Hinse, M. (1999) Use of radar images in the Weathering process effects on spectral reflectance of rocks in
identification of major regional structures in the Grenville a semi − arid environment. International Journal of Remote
Province, Western Quebec. Canadian Journal of Remote Sensing, 18, 3361 − 3377.
Sensing, 25, 278 − 290. Zakir, F.A., Qari, M.H.T. and Mostafa, M.E. (1999) A new opti-
Siegal, B.S. and Goetz, A.F.H. (1977) Effect of vegetation on mizing technique for preparing lineament density maps.
rock and soil discrimination. Photogrammetric Engineering International Journal of Remote Sensing, 20, 1073 − 1085.
and Remote Sensing, 43, 191 − 196.
Smith, M.O., Johnston, P.E. and Adams, J.B. (1985) Quantitative (Manuscript received; 16 February, 2004)
determination of mineral types and abundances from reflec- (Manuscript accepted; 28 June, 2004)