.

Ore Geology Reviews 14 1999 157183

Remote sensing for mineral exploration
Floyd F. Sabins
)
Remote Sensing Enterprises, 1724 Celeste Lane, Fullerton, CA 92833, USA
Received 13 November 1998; accepted 20 April 1999
Abstract
Remote sensing is the science of acquiring, processing, and interpreting images and related data, acquired from aircraft
and satellites, that record the interaction between matter and electromagnetic energy. Remote sensing images are used for
. .
mineral exploration in two applications: 1 map geology and the faults and fractures that localize ore deposits; 2 recognize
.
hydrothermally altered rocks by their spectral signatures. Landsat thematic mapper TM satellite images are widely used to
interpret both structure and hydrothermal alteration. Digitally processed TM ratio images can identify two assemblages of
hydrothermal alteration minerals; iron minerals, and clays plus alunite. In northern Chile, TM ratio images defined the
prospects that are now major copper deposits at Collahuasi and Ujina. Hyperspectral imaging systems can identify individual
species of iron and clay minerals, which can provide details of hydrothermal zoning. Silicification, which is an important
indicator of hydrothermal alteration, is not recognizable on TM and hyperspectral images. Quartz has no diagnostic spectral
features in the visible and reflected IR wavelengths recorded by these systems. Variations in silica content are recognizable
in multispectral thermal IR images, which is a promising topic for research. q1999 Elsevier Science B.V. All rights
reserved.
. Keywords: remote sensing; mineral exploration; thematic mapper TM ; Goldfield mining district
1. Introduction
Remote sensing is the science of acquiring, pro-
cessing, and interpreting images and related data,
acquired from aircraft and satellites, that record the
interaction between matter and electromagnetic en-
.
ergy Sabins, 1997, p. 1 . This report reviews the use
of remote sensing for mineral exploration. Section 2
describes the remote sensing systems that are em-
ployed in mineral exploration and introduces the
)
Tel.: q1-714-879-4367; e-mail: ffsabins@home.com
computer techniques used to process digital data
acquired by the systems. Section 4 describes how
multispectral data are digitally processed to recog-

nize hydrothermal alteration minerals iron minerals,

.
clays, and alunite , using the Goldfield, NV, mining
district as a training site. The methods developed at
Goldfield were used in northern Chile to define
anomalies that are now world-class copper deposits.
Section 8 describes future remote sensing systems
and their potential applications to mineral explo-
ration. Most of this report is extracted from Sabins
.
1997 , to which the reader is referred for additional
information.
0169-1368r99r$ - see front matter q 1999 Elsevier Science B.V. All rights reserved.
. PII: S0169- 1368 99 00007- 4
( ) F.F. SabinsrOre Geology Reiews 14 1999 157183 158
2. Remote sensing technology
Table 1 lists characteristics of the principal re-
mote sensing systems that are currently available for
mineral exploration. Some systems are deployed only
.
on satellites Landsat, SPOT . Other systems are

currently deployed only on aircraft hyperspectral

.
systems . Radar systems are deployed on both satel-
lites and aircraft. Images acquired by satellite sys-
.
tems have the following advantages: 1 archives of
.
worldwide data are readily available; 2 images
.
cover large areas on the ground; 3 prices per square
kilometer are generally lower. Disadvantages of
.
satellite images are: 1 the latest hyperspectral tech-
.
nology is currently available only from aircraft; 2
aircraft missions can be configured to match the
requirements of a project. The following sections
summarize the major systems.
2.1. Landsat images
NASA has launched two generations of un-
manned Landsat satellites that have acquired valu-
able remote sensing data for mineral exploration and
other applications. Both generations were placed in
sun-synchronous orbits that provide repetitive im-
ages of the entire earth, except for the extreme polar
.
regions. The first generation Landsats 1, 2, and 3
operated from 1972 to 1985 and is essentially re-
placed by the second generation. Table 1 lists some

characteristics of the second generation Landsats 4,

.
5 and 7 , which began in 1982 and continues to the
present. Landsat 6 of the second generation was
launched in 1993, but failed to reach orbit. Images
.
are acquired by the thematic mapper TM which is

an optical-mechanical cross-track scanner Sabins,

.
1997, Fig. 1-12A . An oscillating scan mirror sweeps
the field of view of the optical system across the
terrain at a right angle to the satellite orbit path. A
spectrometer separates solar energy that is reflected
from the earths surface into narrow wavelength
intervals called spectral bands. Each band is recorded
as a separate image.
Fig. 1 shows reflectance spectra for vegetation
and three common sedimentary rocks. The vertical
axis shows the percentage of incident sunlight that is
reflected by the materials. The horizontal axis shows
wavelengths of energy for the visible spectral region
.
0.4 to 7.0 mm and the reflected portion 0.7 to 3.0
. .
mm of the infrared IR region. Reflected IR energy
consists largely of solar energy reflected from the
earth at wavelengths longer than the sensitivity range

of the eye. The thermal portion of the IR region 3.0

.
to 1000 mm consists of radiant, or heat, energy and
Table 1
Remote sensing systems for mineral exploration
Characteristic Landsat 4, 5 Landsat 7 SPOT SPOT AVIRIS
thematic mapper enhanced multispectral panchromatic hyperspectral
. . . TM TM scanner XS Pan scanner
Spectral region
Visible and reflected IR 0.45 to 2.35 mm 0.45 to 2.35 mm 0.50 to 0.89 mm 0.40 to 2.50 mm
Panchromatic 0.52 to 0.90 mm 0.51 to 0.73 mm
Thermal IR 10.5 to 12.5 mm
Spectral bands 7 8 3 1 224
Terrain coerage
East to west 185 km 185 km 60 km 60 km 10.5 km cross-track
North to south 170 km 170 km 60 km 60 km
Ground resolution cell
Visible and reflected IR 30 by 30 m 30 by 30 m 20 by 20 m 20 m
Panchromatic 15 by 15 m 10 by 10 m
Thermal IR 120 by 120 m 60 by 60 m
( ) F.F. SabinsrOre Geology Reiews 14 1999 157183 159
Fig. 1. Spectral bands recorded by remote sensing systems. Spec-
tral reflectance curves are for vegetation and sedimentary rocks.
. From Sabins 1997, Fig. 4-1 .
is not shown in Fig. 1. The TM system records three
.
wavelengths of visible energy blue, green, and red
and three bands of reflected IR energy, which are
indicated in Fig. 1. These visible and reflected IR
bands have a spatial resolution of 30 m. Band 6,
which is not shown on Fig. 1, records thermal IR
.
energy 10.5 to 12.5 mm with a spatial resolution of
120 m. Each TM scene records 170 by 185 km of
terrain. The image data are telemetered to earth
receiving stations.
Fig. 2 shows images for the six visible and re-
flected IR bands for a small subarea that covers the
Goldfield, NV, mining district. Any three of the
bands can be combined in blue, green, and red to
produce color composite images. Fig. 3A shows
bands 123 combined in blue, green, and red re-
spectively to produce a color image similar to that
observed by the eye or recorded by a color photo-
graph. Several alternate color combinations of TM

bands are commonly employed Sabins, 1997, Chap.

.
3 . The second generation of Landsat continued with
Landsat 7, launched in April, 1999, with an en-

hanced TM system. A panchromatic band 8 0.52 to

.
0.90 mm with spatial resolution of 15 m is added.
Band 8 can be combined with the visible and re-
.
flected IR bands 30 m resolution to produce a color
image with an apparent resolution of 15 m. Spatial
resolution of the thermal IR band 6 is improved from
120 m to 60 m.
TM data of the world are available for sale from
two sources. TM image data acquired in the past
decade are available from:
Space Imaging EOSAT
12076 Grant Street
Thornton, CO 80241
Phone: q1-303-254-2000
Fax: q1-303-254-2215
E-mail: -info@spaceimaging.com).
The Space Imaging-EOSAT archive of TM images
acquired during the past decade may be viewed
.
interactively and ordered on the Web at -
http:rrspaceimaging.com).
TM image data acquired prior to the past decade
and Landsat 7 data are available from:
U.S. Geological Survey EROS Data Center
Sioux Falls, SD 57198
Phone: q1-605-594-6511
Fax: q1-605-594-6589
E-mail: -custserv@edcmail.cr.usgs.gov).
The EROS Data Center archive of TM images may
.
be viewed interactively and ordered on the Web at
-http:rredcwww.cr.usgs.gov).
2.2. SPOT
Beginning in 1986 a French company, called
SPOT Image, has launched a series of unmanned sun
synchronous satellites that acquire image data in two
. .
modes Table 1 . The multispectral XS mode ac-
quires three bands of data at green, red, and reflected
.
IR wavelengths Fig. 1 with a spatial resolution of
.
20 m. The panchromatic pan mode acquires a
( ) F.F. SabinsrOre Geology Reiews 14 1999 157183 160
( ) F.F. SabinsrOre Geology Reiews 14 1999 157183 161
single band of data, primarily at green and red
wavelengths, with a spatial resolution of 10 m. Both
image modes cover 60 by 60 km of terrain and may
be acquired in a stereoscopic format.
2.3. Hyperspectral imaging systems
Conventional multispectral scanning systems, such
as Landsat TM and SPOT XS, record up to 10
spectral bands with bandwidths on the order of 0.10
mm. Hyperspectral scanners are a special type of
multispectral scanner that record many tens of bands

with bandwidths on the order of 0.01 mm Sabins,

.
1997, Chap. 1 . Many minerals have distinctive spec-
tral reflectance patterns at visible wavelengths and
.
especially at reflected IR wavelengths Hunt, 1980 .
Under favorable conditions, many minerals may be
identified on suitably processed hyperspectral data.
Fig. 1 shows the spectral region covered by the 224
spectral bands recorded by the airborne visiblerin-
.
frared imaging spectrometer AVIRIS which is a
hyperspectral system carried on high altitude aircraft
by NASA. AVIRIS image strips are 10.5 km wide
and several tens of kilometers long. The airborne
system is operated on an experimental basis, primar-

ily in the U.S. A website http:rrmakalu.jpl.nasa.

.
govraviris.html provides access to the archive of
.
AVIRIS images. Green et al. 1998 describe the
AVIRIS system and summarize a number of applica-
tion studies, including geology. Examples of AVIRIS
images are shown in the section on the Goldfield
.
mining district Section 4.3.1 . Sabins 1997, Tables
.
14 lists some airborne hyperspectral scanners that
are commercially available.
2.4. Radar systems
Radar is an active form of remote sensing that
provides its own source of electromagnetic energy to
illuminate the terrain. Radar energy is measured in
wavelengths of centimeters that penetrate rain and
clouds which is an advantage in tropical regions.
Another advantage is that radar images may be
acquired at a low depression angle that causes pro-
nounced highlights and shadows that enhance subtle
topographic features. These features are commonly
the expression of faults, fractures, and lithology.
Radar images of vegetated regions record the vegeta-
tion surface, rather than the underlying terrain. In
.
Indonesia, Sabins 1983 demonstrated that the forest
canopy conforms to the underlying terrain and that
geologic information can be interpreted from the
images. In Papua New Guinea, the Chevron Corpora-
tion relied on aircraft radar images to discover major
oil fields.
2.5. Digital image processing
Modern remote sensing systems record image data
in a digital raster format that is suitable for computer
processing using readily available software and per-
.
sonal computers. Sabins 1997, Chap. 8 groups
image-processing methods into three functional cate-
gories that are listed below, together with lists of
typical processing routines
1. Image restoration compensates for image errors,
noise, and geometric distortions introduced during
the scanning, recording, and playback operations.
The objective is to make the restored image re-
semble the scene on the terrain. Typical process-
ing routines include:
a. Restoring line dropouts
b. Restoring periodic line striping
c. Restoring line offsets
d. Filtering random noise
e. Correcting for atmospheric scattering
f. Correcting geometric distortions
2. Image enhancement alters the visual impact that
the image has on the interpreter. The objective is
to improve the information content of the image.
Typical processing routines include:
a. Contrast enhancement
b. Density slicing
c. Edge enhancement
Fig. 2. Landsat TM visible and reflected IR images of Goldfield mining district, NV. Fig. 4 is a map of the area which covers 7 by 7 km.
. . . . . . From Sabins 1997, Fig. 11-7 . A Band 1, blue 0.45 to 0.52 mm . B Band 2, green 0.52 to 0.60 mm . C Band 3, red 0.63 to 0.69
. . . . . . . mm . D Band 4 reflected IR 0.76 to 0.90 mm . E Band 5, reflected IR 1.55 to 1.75 mm . F Band 7, reflected IR 2.08 to 2.35 mm .
( ) F.F. SabinsrOre Geology Reiews 14 1999 157183 162
( ) F.F. SabinsrOre Geology Reiews 14 1999 157183 163
d. Making digital mosaics
e. Intensity, hue, and saturation transformations
f. Merging data sets
g. Synthetic stereo images
3. Information extraction utilizes the computer to
combine and interact between different aspects of
a data set. The objective is to display spectral and
other characteristics of the scene that are not
apparent on restored and enhanced images. Typi-
cal processing routines include:
a. Principal-component images
b. Ratio images
c. Multispectral classification
d. Change-detection images
The images in this report have been processed with
various combinations of these routines.
3. Mineral exploration overview
Table 2 lists representative recent mineral explo-
ration studies using remote sensing. These studies
describe two different approaches to mineral explo-
ration.
.
1 Mapping of geology and fracture patterns at
regional and local scales. Prospectors and mining
geologists have long recognized the importance of
regional and local fracture patterns as controls on ore
.
deposits. Rowan and Wetlaufer 1975 used a Land-
sat mosaic of Nevada to interpret regional linea-
ments. Comparing the lineament patterns with ore
occurrences showed that mining districts tend to
occur along lineaments and are concentrated at the
.
intersections of lineaments. Nicolais 1974 inter-
preted local fracture patterns from a Landsat image
in Colorado. The mines tend to occur in areas with a
high density of fractures and a concentration of
.
fracture intersections. Rowan and Bowers 1995
used TM and aircraft radar images to interpret linear
features in western Nevada. They concluded that the
linear features correlate with the geologic structures
that controlled mineralization.
.
2 Recognition of hydrothermally altered rocks
that may be associated with mineral deposits. The
spectral bands of Landsat TM are well-suited for

recognizing assemblages of alteration minerals iron

.
oxides, clay, and alunite that occur in hydrother-
mally altered rocks. In my experience the best explo-
ration results are obtained by combining geologic
and fracture mapping with the recognition of hy-
drothermally altered rocks.
4. Mapping hydrothermal alteration at epither-
mal vein deposits Goldfield, Nevada
Most epithermal vein deposits are accompanied
by hydrothermal alteration of the adjacent country
rocks. Not all alteration is associated with ore bod-
ies, and not all ore bodies are accompanied by
alteration, but the presence of altered rocks is a
valuable indicator of possible deposits. Prospectors
have long been aware of the association between
hydrothermally altered rocks and ore deposits. Many
mines were discovered by recognizing outcrops of
altered rocks, followed by assays of rock samples.
Prior to remote sensing, altered rocks were recog-
nized by their appearance in the visible spectral
bands. Today remote sensing and digital image pro-
cessing enable us to use additional spectral bands for
mineral exploration. In regions where bedrock is
exposed, multispectral remote sensing can be used to
recognize altered rocks because their reflectance
spectra differ from those of the unaltered country
rock. The Goldfield Mining District in south-central
Nevada is the test site where remote sensing methods
Fig. 3. Recognizing hydrothermally altered rocks at Goldfield mining district, NV. Image F courtesy F.A. Kruse, Analytical Imaging and
. . . . Geophysics, LLC, Boulder, CO. From Sabins 1997, Plate 21 . A TM 123 normal color image. B TM color ratio image. Ratio
. . 5r7sred, 3r1sgreen, 3r5sblue. C TM ratio 5r7 image with density slice. High ratio values shown in red. D TM ratio 3r1 image
. . with density slice. High ratio values shown in red. E TM unsupervised classification map. F Color composite image of AVIRIS
. endmember abundance images from Fig. 12 . Illitesred, alunitesgreen, kaolinitesblue.
( ) F.F. SabinsrOre Geology Reiews 14 1999 157183 164
Table 2
. Representative mineral exploration investigations using remote sensing. From Sabins 1997, Table 11-3
Locality Reference Comments
. Western North and Spatz and Wilson 1994 Summarizes published remote sensing studies of
South America 12 major mining districts from British Columbia to Chile.
. Altiplano, Bolivia Knepper and Simpson 1992 TM color ratio composite images used to recognize
hydrothermally altered rocks.
. Canada Singhroy 1991 10 papers on mineral exploration using Landsat and radar.
. Chile, Peru, Eiswerth and Rowan 1993 TM color ratio composite images used to recognize
and Bolivia hydrothermally altered rocks. Field studies evaluated results.
. Jordan Kaufmann 1988 Mapped hydrothermal alteration using digitally
processed TM images.
. Jordan Abdelhamid and Rabba 1994 A variety of digitally processed TM images identified
a historic CurMn deposit and located prospects.
. Sonora, Mexico Bennett et al. 1993 TM data were integrated with field and laboratory data
to discover several prospects.
. Nevada Watson et al. 1990 TIMS data were processed to recognize silicified
rocks associated with gold deposits.
. Spain Goosens and Kroonenberg 1994 TM ratio images were used to identify altered rocks
overlain by residual soil.
. Sudan Griffiths et al. 1987 Landsat MSS images and field work showed gold
occurrences are concentrated along regional shear zones
in mafic metavolcanics.
. Arizona Abrams et al. 1983 Mapped hydrothermal alteration using digitally processed
aircraft multispectral images.
. Montana Rowan et al. 1991 Compared the association of linear features with ore
deposits in Butte region.
. Idaho and Montana Segal and Rowan 1989 Mapped hydrothermal alteration in the Dillon region.
. Utah Murphy 1995 Used hyperspectral data to map jasperoid.
. Zaire, Zambia, Angola Unrug 1988 Major leadzinc vein deposits occur at intersections of
Landsat lineaments with folds and thrust faults.
Unexplored intersections are potential targets.
were first developed to recognize hydrothermally
.
altered rocks Rowan et al., 1974 .
4.1. Geology, ore deposits, and hydrothermal alter-
ation
.
The Goldfield district Fig. 4 was noted for the
richness of its ore. Over 4 million troy ounces
.
130,000 kg of gold with silver and copper were
produced, largely in the boom period between 1903
and 1910. The geology and hydrothermal alteration
of the district have been thoroughly mapped and

analyzed by the U.S. Geological Survey Ashley,

.
1974, 1979 , which makes Goldfield an excellent
locality to develop and test remote sensing methods
for mineral exploration.
Volcanism began in the Oligocene epoch with
eruption of rhyolite and quartz latite flows and the
formation of a small caldera and ring-fracture sys-
tem. Hydrothermal alteration and ore deposition oc-
curred during a second period of volcanism in the
early Miocene epoch when the dacite and andesite
flows that host the ore deposits were extruded. Heat-
ing associated with volcanic activity at depth caused
convective circulation of hot, acidic, hydrothermal
solutions through the rocks. Fluid movement was
concentrated in the fractures and faults of the ring-
fracture system. Following ore deposition, the area
was covered by younger volcanic flows. Later dom-
ing and erosion have exposed the older volcanic
center with altered rocks and ore deposits.
.
In the generalized map Fig. 4 , the hydrother-
mally altered rocks are cross-hatched and the unal-
tered country rocks are blank. Approximately 40 km
2
of the area is underlain by altered rocks, but less than
2 km
2
of the altered area contains economic mineral
( ) F.F. SabinsrOre Geology Reiews 14 1999 157183 165
. Fig. 4. Map showing geology and hydrothermal alteration of Goldfield mining district, NV. From Ashley 1979, Figs. 1 and 8 .
deposits, which are shown in black. The oval band of
altered rocks was controlled by the circular ring-frac-
ture system, with a linear extension toward the east.
The central patch of alteration shown in Fig. 4 was
controlled by closely spaced faults and fractures. The
most highly altered rocks are the veins of microcrys-
talline quartz with some alunite. The ore occurs in
the veins, but the majority of veins are barren.
Adjacent to the veins, the country rock is altered to
the clay minerals illite, kaolinite, and montmoril-
lonite plus alunite. This assemblage of alteration

minerals is called the argillic zone Harvey and

.
Vitaliano, 1964 . The hydrothermal solutions also
deposited jarosite and pyrite in the veins and argillic
rocks. The pyrite weathers to iron oxides which
impart pink and red hues to the altered rocks. The
hydrothermally altered rocks at Goldfield, and other
epithermal vein deposits, are characterized by two
mineral assemblages:
1. Alunite and clay minerals
2. Iron minerals
The following sections describe how Landsat images
are digitally processed to recognize these assem-
blages.
( ) F.F. SabinsrOre Geology Reiews 14 1999 157183 166
4.2. Recognizing hydrothermal alteration on Landsat
images
Fig. 2 shows the visible and reflected IR bands of
a TM subscene of the Goldfield district. Fig. 3A is
an enhanced normal color image of TM bands 123
shown in blue, green, and red, respectively. A yellow
patch directly northeast of the town of Goldfield is
caused by the mine dumps and disturbed ground of
the main mineralized area. A white patch 3 km north
of Goldfield is the dry tailings pond of the aban-
doned Columbia Mill, where gold was separated
from the altered host rock. The tailings pond is a
useful reference standard because it contains a con-
centration of altered rock material. The dark signa-
tures in the margins of the image are volcanic rocks
that are younger than the ore deposits and altered
rocks. Distinctive light blue signatures in the south-
east portion are outcrops of volcanic tuff. Neither the
normal color TM image nor alternate band color
combinations are diagnostic for recognizing the hy-
drothermally altered rocks. Therefore, additional dig-
ital processing is required in order to map hydrother-
mal alteration from TM data.
4.2.1. Alunite and clay minerals on 5r7 ratio im-
ages
Fig. 5A shows reflectance spectra of alunite and
the three common hydrothermal clay minerals illite,
kaolinite, and montmorillonite. These minerals have
.
distinctive absorption features reflectance minima
at wavelengths within the bandpass of TM band 7
which is shown with a stippled pattern in Fig. 5A.
The alteration minerals have higher reflectance val-
ues within TM band 5. Ratio images can emphasize
and quantify these spectral differences. A TM image
.
consists of picture elements pixels that represent a
ground resolution cell of 30 by 30 m. For each pixel
the reflectance values for all bands are recorded as
.
digital numbers DNs on an eight-bit scale from 0 to
Fig. 5. Recognition of hydrothermal clays and alunite from TM
. . data, Goldfield mining district. From Sabins 1997, Fig. 11-8 . A
. Laboratory reflectance spectra. TM bands 5 and 7 shaded are
. used to calculate 5r7 ratio image. B Ratio image of TM bands
. 5r7. C Histogram for 5r7 image.
( ) F.F. SabinsrOre Geology Reiews 14 1999 157183 167
255. Ratio images are prepared by dividing the value
for one band by that of another band, after atmo-

spheric corrections have been made Sabins, 1997,

.
Chap. 8 . Table 3 explains how TM ratio 5r7 distin-
guishes altered rocks containing clays and alunite
from unaltered rocks. Both rocks have similar values
in band 5. The reflectance of unaltered rocks in band
7 is similar to that in band 5. Therefore, the 5r7
.
ratio for unaltered rocks is unity 1.00 . Altered
rocks, however, have lower reflectance in band 7
because of the absorption caused by the minerals
shown in Fig. 5A. Therefore, the 5r7 ratio for
.
altered rocks is much greater than unity 1.45 . The
numbers in Table 2 are typical and will differ for
other examples. The decimal ratio values are con-
.
verted to 8-bit digital numbers DNs and displayed
as images.
Fig. 5B is a 5r7 ratio image of Goldfield with
higher ratio values shown in brighter tones. Compar-
.
ing the image with the map Fig. 4 shows that the
high ratio values correlate with hydrothermally al-
tered rocks. Fig. 5C is a histogram of the 5r7 ratio
.
image that shows the higher ratio values DNs)145
of the altered rocks. Low ratio values represent
unaltered rocks.
Fig. 3C is a color density slice version of the 5r7
image in which the gray scale is replaced by the
.
colors shown in the histogram Fig. 5C . Highest
.
ratio values DN)145 are shown in red, with the
.
next highest values DN 125 to 145 shown in
yellow. The red and yellow colors on the ratio image
.
Fig. 3C therefore correlate with the altered rocks.
4.2.2. Iron minerals on 3r1 ratio images
Iron oxides and sulfates are the second group of
minerals associated with hydrothermally altered
rocks. Fig. 6A shows spectra of the iron minerals
.
which have low blue reflectance TM band 1 and
.
high red reflectance TM band 3 . Iron-stained hy-
drothermally altered rocks therefore have high values
in a 3r1 ratio image. Fig. 6B is a 3r1 ratio image
with high DN values shown in bright tones. Fig. 3D
is a color density slice version of the 3r1 image,
with color assignments shown in the histogram of
.
Fig. 6C. Highest ratio values DN)150 are shown
.
in red, with the next highest values DN 135 to 150
shown in yellow. The red and yellow colors there-
fore correlate with the altered rocks.
4.2.3. Color composite ratio images
Color composite ratio images are produced by
combining three ratio images in blue, green, and red.
Fig. 3B shows ratios 3r5, 3r1, and 5r7 in red,
green, and blue, respectively. The orange and yellow
hues delineate the outer and inner areas of altered
rocks in a pattern similar to that of the density sliced
ratio images. An advantage of the color ratio image
is that it combines the distribution patterns of both
iron minerals and hydrothermal clays. A disadvan-
tage is that the color patterns are not as distinct as in
the individual density-sliced images.
4.2.4. Classification images
Multispectral classification is a computer routine
for information extraction that assigns pixels into
classes based on similar spectral properties. In a
supervised multispectral classification, the operator
specifies the classes that will be used. In an unsuper-
vised multispectral classification, the computer spec-

ifies the classes that will be used Sabins, 1997,

.
Chap. 8 . An unsupervised multispectral classifica-
tion was applied to the TM bands in Fig. 2 and
resulted in 12 classes. These classes were aggregated
Table 3
. Calculation of TM 5r7 ratio values. From Sabins 1997, Table 11-1
Band 5 reflectance Band 7 reflectance Ratio 5r7 DNs for
. . . typical typical typical ratio 5r7
Unaltered rocks 160 160 1.00 100
. without clays and alunite
Altered rocks 160 110 1.45 145
. with clays and alunite
( ) F.F. SabinsrOre Geology Reiews 14 1999 157183 168
into the six classes shown in Fig. 3E with colors that
are explained in Table 4. Two types of altered rocks
were classified. The class shown in red Altered
rocks, A is confined to altered rocks, but does not
indicate the full extent of alteration. The class shown
in orange Altered rocks, B includes all of the
remaining altered rocks, as well as some rocks out-
.
side the alteration zone. Basalt blue , volcanic tuff
. .
purple , and unaltered rocks green are reasonably
.
classified. Alluvium yellow is considerably more
.
extensive in the classification image Fig. 3E than in
.
the geologic map Fig. 4 . Field checking and com-
.
parison with the normal color image Fig. 3A shows
that much of the bedrock is thinly covered with
detritus and is correctly classed as alluvium by the
computer. The map, however, shows the lithology of
the underlying bedrock that was interpreted by the
field geologist.
4.3. Recognizing hydrothermal alteration on hyper-
spectral images
Because of their broad spectral band passes, TM
images cannot identify specific alteration minerals,
such as jarosite, alunite, and the individual clay
minerals. Such identifications could be valuable for
mapping details of hydrothermal zoning; these de-
tails can be mapped, however, from data acquired by
hyperspectral scanners. Fig. 7 shows laboratory spec-
tra of common alteration minerals in the atmospheric
window from 2.0 to 2.5 mm and the 50 spectral
bands recorded by the AVIRIS hyperspectral scanner
for this wavelength interval. The bandpass of TM
band 7 is also shown for comparison. Van der Meer
. .
1994 , Kruse 1996 and others have shown that
AVIRIS has the spectral resolution to identify indi-
vidual alteration minerals.
The following sections describe AVIRIS images
that show the abundance and distribution of individ-
Fig. 6. Recognition of hydrothermal iron minerals from TM data,
. . Goldfield mining district. From Sabins 1997, Fig. 11-9 . A
. Laboratory reflectance spectra. TM bands 1 and 3 shaded are
. used to calculate 3r1 ratio image. B Ratio image of TM bands
. 3r1. C Histogram for 3r1 image.
( ) F.F. SabinsrOre Geology Reiews 14 1999 157183 169
Table 4
Explanation of colors in classification image of Goldfield mining
. . district Fig. 3E . From Sabins 1997, Table 11-2
Color Classification Percent of image
Yellow Alluvium 39.2
Blue Basalt 14.0
Purple Tuff 6.6
Red Altered rocks, A 5.3
Orange Altered rocks, B 18.3
Green Unaltered rocks 16.6
ual alteration minerals. There are, however, two
major technical challenges to producing such images.
.
1 Some alteration minerals, especially the clays,
.
have similar spectra Fig. 7 . The major absorption
feature near 2.2 mm occurs at slightly different
wavelengths for the different clays and for alunite.
There are minor additional absorption features that
also help distinguish the spectra. Image processing
programs can identify the spectrum recorded for a
single AVIRIS pixel by comparing it with a library
of reference spectra for known minerals. This proce-
dure is a form of supervised classification. The
procedure is effective, however, only for the rare
ground resolution cells in which only a single min-
eral occurs.
.
2 Each ground resolution cell of AVIRIS typi-
cally measures 20 by 20 m. In areas of complex
geology, such as Goldfield, the 400 m
2
of a cell
includes a number of different minerals. The result-
ing pixel is called a mixed pixel because its spectrum
is a mixture of the spectra for the different minerals
that occupy the ground resolution cell. These indi-
vidual mineral species are called spectral endmem-
bers. Digital unmixing programs are used to derive
the spectra of the endmembers for each mixed pixel.
For each mineral, an endmember abundance image is
derived that shows the relative abundance of the
mineral.
4.3.1. AVIRIS images of Goldfield
AVIRIS hyperspectral images of the Goldfield
mining district were digitally processed at Analytical
Imaging and Geophysics LLC. Images showing spec-
tral endmember abundances of alteration minerals
were produced, using a spectral unmixing program
.
of Boardman 1993 . Fig. 3F is a color composite
image made by combining the endmember abun-
dance image of illite in blue, alunite in green, and
kaolinite in red. The black-and-white base is AVIRIS
.
band 30 visible red . The primary colors show areas
with high concentrations of the assigned mineral.
Other colors indicate co-occurrence of endmember
minerals. Yellow, for example, indicates a mixture
.
of kaolinite and alunite. Kaolinite red and illite
Fig. 7. Laboratory spectra of alteration minerals in the 2.0 to 2.5
mm band. Spectra are offset vertically. Note positions and band-
widths of the spectral bands recorded by AVIRIS and TM band 7.
. From Sabins 1997, Fig. 11-16 .
( ) F.F. SabinsrOre Geology Reiews 14 1999 157183 170
.
green are the most abundant alteration minerals;

their patterns coincide with the alteration map Fig.

.
4 that was prepared earlier by field-mapping.
The AVIRIS color image covers the western two-
thirds of the TM 5r7 ratio image shown in Fig. 3C.
It is instructive to compare these images. The red
and yellow colors of the TM 5r7 image show the
aggregate distribution of clays and alunite. The col-
ors of the AVIRIS image show the distribution of
individual alteration minerals. In summary, TM im-
ages show the broad pattern of hydrothermal alter-
ation; AVIRIS images show the distribution of the
individual alteration minerals.
4.3.2. Other AVIRIS examples
The Cuprite district, 25 km south of Goldfield,
consists of volcanic rocks that are intensely altered
to silica, opal, and clay. No significant mineral de-
posits occur, but the district has long been used as a
.
remote sensing test site. Goetz and Srivastava 1985
analyzed hyperspectral images from a precursor sys-
tem to AVIRIS. They recognized spectra of various
clay minerals, plus buddingtonite which is an ammo-
nium feldspar that had not previously been reported
at Cuprite. Fig. 7 shows the distinctive spectrum of
buddingtonite. Buddingtonite is associated with hy-
drothermally altered rocks in several localities in the
. .
western U.S. Krohn et al., 1993 . Hook et al. 1991
recognized the alteration minerals on images of
Cuprite acquired by AVIRIS and GEOSCAN, a
commercial hyperspectral system.
.
Crosta et al. 1998 analyzed AVIRIS images of
the Bodie mining district in eastern California, which
was an important goldsilver district in the second
half of the 19th century. Host rocks are intermediate
to mafic volcanic rocks. Gold occurs in quartz veins
and stockworks associated with hydrothermally al-
tered rocks. Silicification in the center is surrounded
by zones of potassic, argillic and sericitic alteration
and an outer zone of propylitic alteration. The
AVIRIS data were processed with algorithms that
classified the image spectra and compared them to
reference spectra. The resulting maps show the dis-

tribution of three separate iron minerals hematite,

.
goethite, and jarosite , four clay minerals
.
montmorillonite, kaolinite, halloysite, and illite ,
plus muscovite.
4.4. Summary

The spectra of alteration minerals Fig. 5A, 6A

.
and 7 were recorded in the laboratory using pure
minerals. Remote sensing images record data from
weathered outcrops of mixtures of rocks and miner-
als together with soil and vegetation. Despite these
complications, the digitally processed images give an
accurate picture of the alteration pattern at Goldfield.
In order to bridge the gap between laboratory and
.
outcrop, Rowan et al. 1979, Fig. 2A used a portable
spectrometer in the field to record spectra of several
hundred representative outcrops of altered and unal-
tered rocks at Goldfield. Fig. 8 summarizes their
results as average reflectance curves for altered and
unaltered outcrops. The average curves lack the fine
spectral detail of the laboratory curves, but the dif-
ferences between altered and unaltered rocks are
clearly shown. The altered rocks have distinctly lower
reflectance in band 7 than in band 5. Unaltered rocks
have similar values in those bands. In the visible
portion of the spectrum altered rocks have higher red
reflectance because of the hydrothermal iron miner-
als. These field spectra support the use of TM ratios
5r7 and 3r1 for recognizing alteration minerals.
Remote sensing studies of the Goldfield test site
developed techniques for recognizing hydrothermal
alteration from TM and hyperspectral data. Table 2
summarizes a number of projects that used these
. Fig. 8. Field spectra averaged of altered and unaltered rocks at
. Goldfield mining district. From Rowan et al. 1979, Fig. 2A .
( ) F.F. SabinsrOre Geology Reiews 14 1999 157183 171
techniques. The following section describes a suc-
cessful commercial application of digitally processed
TM images.
5. Mapping hydrothermal alteration at porphyry
copper deposits Collahuasi, Chile
Most of the worlds copper is mined from por-
phyry deposits, which occur in a different geologic
environment from vein deposits of the Goldfield
type. Hydrothermal alteration is also common at
porphyry deposits and may be recognized by the
same methods that were developed at Goldfield.
5.1. Alteration model
Fig. 9 is a model of hydrothermal alteration of
porphyry copper deposits that was developed by
.
Lowell and Guilbert 1970 . The most intense alter-
ation occurs in the core of the porphyry body and
diminishes radially outward in a series of concentric
zones described below.
. Fig. 9. Model of hydrothermal alteration zones associated with porphyry copper deposits. From Lowell and Guilbert 1970, Fig. 3
( ) F.F. SabinsrOre Geology Reiews 14 1999 157183 172
Potassic zone. Most intensely altered rocks in the
core of the stock. Characteristic minerals are quartz,
sericite, biotite, and potassium feldspar. The re-
.
flectance spectra not shown of biotite and sericite
have absorption minima in TM band 7, similar to the
spectra of clays. The TM ratio 5r7 is effective in
recognizing these micas, which have reflectance
spectra similar to those of clays.
Phyllic zone. Quartz, sericite, and pyrite are com-
mon.
Ore zone. Disseminated grains of chalcopyrite,
molybdenite, pyrite, and other metal sulfides. Much
of the ore occurs in a cylindrical shell near the
boundary between the potassic and phyllic zones.
Copper typically constitutes 1%, or less, of the rock,
but the large volume of ore is suitable for open pit
mining. Where the ore zone is exposed by erosion,
pyrite oxidizes to form a red to brown iron-stained
crust called a gossan, or leached capping. Gossans
can be useful indicators of underlying mineral de-
posits, although not all gossans are associated with
ore deposits.
Argillic zone. Quartz, kaolinite, and montmoril-
lonite are characteristic minerals of the argillic zone
in porphyry deposits, just as they are associated with
the argillic zone at Goldfield and elsewhere.
Propylitic zone. Epidote, calcite, and chlorite oc-
cur in these weakly altered rocks. Propylitic alter-
ation may be of broad extent and have little signifi-
cance for ore exploration.
Few porphyry deposits have the symmetry and
completeness of the model in Fig. 9. Structural de-
formation, erosion, and deposition commonly con-
ceal large portions of the system. Nevertheless,
Fig. 10. Geologic map of Collahuasi mining district, Chile. Hydrothermal alteration anomalies are edited from Landsat TM ratio images.
. Geology generalized from Vergara 1978A, B .
( ) F.F. SabinsrOre Geology Reiews 14 1999 157183 173
recognition of small patches of altered rock on re-
mote sensing images can be a valuable exploration
clue.
In the early 1980s, NASA and the Geosat Com-
mittee evaluated satellite and airborne multispectral
images of porphyry copper deposits in southern Ari-
zona. At the Silver Bell mining district, Abrams and
.
Brown 1985 used color ratio images to separate the
phyllic and potassic alteration zones from the argillic
and propylitic zones. A supervised classification map
defined the outcrops of altered rocks.
5.2. Geologic and exploration background
The Collahuasi Mining District is located in
northern Chile, 180 km southeast of the city of
Iquique. The district lies within a north-trending belt
of porphyry copper deposits that includes the major
mines at El Teniente, Disputada, El Salvador, Escon-
dida, and Chuquicamata. The Collahuasi District is
bounded on the west by a major regional fault
system that also passes through the open pit at the
Chuquicamata mine. Fig. 10 is a geologic map show-
ing distribution of the Macata, Capella, and Col-
lahuasi formations of Jurassic and Cretaceous age.
These country rocks are intruded by granitic stocks
of late Cretaceous to early Tertiary age that are hosts
for the porphyry copper deposits.
Mineral production in the Collahuasi District be-
gan in the late 1800s when copper was mined from
.
veins at Rosario Fig. 10 now known to be related
to the porphyry system. During the 1930s, these
veins were Chiles third largest producer of copper.
Modern exploration began in 1976 when a joint
venture of Superior Oil and Falconbridge acquired
the Collahuasi properties. The joint venture discov-
ered a porphyry deposit at Rosario. In 1985, owner-
ship of the district changed to a three-way joint
venture of Falconbridge, Shell Oil, and Chevron.
From 1985 to early 1991, exploration efforts were
concentrated on evaluating the Rosario deposit.
Rosario, however, occupies only a small portion of
the 28,000 ha of the Collahuasi District. There were
Fig. 11. Collahuasi mining district, Chile. Landsat TM bands 247 shown in red, green, and blue merged with SPOT pan image. From
. Sabins 1997, Plate 22 .
( ) F.F. SabinsrOre Geology Reiews 14 1999 157183 174
indications of other mineralized centers within the
district, but geologic information was incomplete and
additional exploration data were required.
5.3. Remote Sensing
The Remote Sensing Research Group of Chevron
processed satellite images of the Collahuasi District
and adjacent areas. Northern Chile is ideally suited
for such studies, because vegetation, soils, and clouds
are virtually absent in this arid environment of the
high Andes Mountains. Landsat TM bands 247
were combined in blue, green, and red to produce a
color image that is optimum for geologic interpreta-
tion in this arid terrain. A SPOT panchromatic image
.
10 m spatial resolution was merged with the TM
image to produce the version shown in Fig. 11.
TM 3r1 and 5r7 ratio images were produced
using the methods developed at the Goldfield Mining
District. The ratio images were interpreted to identify
areas with high concentrations of iron oxide miner-
als, clays, and alunite. These areas, called anomalies,
were plotted on a preliminary map. The TM anoma-
lies were evaluated to eliminate false anomalies.
Three major types of false anomalies are:
1. Sedimentary rocks, such as shale, that are rich in
clay
2. Rocks with an original red color, such as iron-rich
volcanic rocks and sedimentary red beds
3. Detritus eroded from outcrops of altered rocks;
these recent deposits in alluvial fans and channels
may indicate the proximity of altered rocks.
The edited anomalies are shown in black on the
.
geologic map Fig. 10 . A circular cluster of anoma-
lies, over 6 km in diameter, occurs south and west of
Collahuasi and Rosario and is now called the Col-
lahuasi Hydrothermal System. The Rosario deposit,
with a diameter of 1.5 km, occupies only a small
Fig. 12. Contour map of resistivity values, Collahuasi mining district. H high values. L low values. Hydrothermal alteration
. anomalies are edited from Landsat TM ratio images. From Sabins 1997, Fig. 11-13 .
( ) F.F. SabinsrOre Geology Reiews 14 1999 157183 175
portion of the north margin of the system. The
remainder of the Collahuasi hydrothermal system
was largely unexplored.
A second cluster of anomalies, 3 km wide, occurs
.
southwest of Ujina Fig. 10 and is called the Ujina
Hydrothermal System. Minor alteration had been
recognized earlier at Ujina, but the area has received
very limited exploration attention in the past. The
alteration shown on the ratio images is much more
extensive than previously recognized at Ujina.
5.4. Geophysical sureys
Geophysical surveys were made to evaluate the
.
Landsat TM anomalies. Dick et al. 1993 provide
details of the configuration and results of the geo-
physical surveys. The entire district was covered by
.
a helicopter-borne aeromagnetic survey not shown
that mapped subsurface geologic structures and the
distribution of magnetic minerals. The aeromagnetic
map shows that the Collahuasi and Ujina hydrother-
. Fig. 13. Landsat TM band 4 image of Salar de Uyuni and vicinity, southwest Bolivia. From Sabins and Miller 1994, Fig. 2 .
( ) F.F. SabinsrOre Geology Reiews 14 1999 157183 176
mal systems are localized at intersections of major
northeast- and northwest-trending faults. The Ujina
System has a circular rim of high magnetic values
that is interpreted as an ore shell within the porphyry
deposit, similar to that shown in the porphyry model
.
Fig. 9 .
A ground-based survey measured resistivity of the
rocks. Unmineralized rocks typically have high resis-
tivity values. Metallic minerals, such as copper sul-
fides, have low resistivity values; therefore, mineral-
ized rocks have low resistivity values. Fig. 12 is a
contour map of the resistivity survey at the same
. .
scale as the image Fig. 11 and map Fig. 10 . High
resistivity values are shown by H; the very important
low values are shown by L.
Results of the resistivity survey are outstanding.
Circular patterns of low resistivity contours occur at
both the Collahuasi and Ujina hydrothermal systems
.
Fig. 12 . These patterns are analogous to those of
classic porphyry copper deposits. At Collahuasi the
resistivity pattern is 5 km in diameter. The lowest
values form a marginal rim that may represent the
ore shell of the porphyry model. The very low
overall resistivity of the Collahuasi system is inter-
preted as an extensive development of veinlet miner-
alization.
The Ujina Hydrothermal System has a circular
pattern of low resistivity contours 3 km in diameter.
The eastern portion of the resistivity feature is cov-
ered by the Ujina tuff that post-dates the hydrother-
.
mal activity Fig. 10 . The Landsat anomalies coin-
cide with the exposed western portion of the system.
5.5. Ore discoeries
Core holes were drilled to evaluate the hydrother-
mal systems outlined by the remote sensing and
Fig. 14. Map of Salar de Uyuni. Triangles show high values for TM ratio 4r7 that correlate with high concentrations of ulexite. Contours

y1
. . show boron concentration mg l in near-surface brine. From Risacher 1989, Fig. 34 .
( ) F.F. SabinsrOre Geology Reiews 14 1999 157183 177
geophysical investigations. The first holes tested the
low resistivity values at Rosario, on the north rim of
the Collahuasi system, where the drills found zones
of structurally controlled copper mineralization.
These results led to the discovery of two major ore
bodies within the Collahuasi system that are shown
by stippled patterns in Fig. 12.
At Ujina, drilling of the resistivity feature discov-
ered a major new porphyry copper deposit shown by
the stippled pattern in Fig. 12. The primary ore
deposit is overlain by secondary enriched ore. By
early 1993, drilling had outlined over 150 million
tons of enriched ore with a grade of 1.8% copper.
In late 1992, Chevron decided to sell its mineral
properties in order to concentrate on its energy busi-
ness. Chevron sold its one-third interest in the unde-
veloped Collahuasi District to Minorco for US$190
million cash. Chevrons total investment in the prop-
erty is estimated at US$23 million. The remote
sensing work that contributed so much to the in-
creased value of the property cost less than US$50
thousand. In 1995, Minorco and Falconbridge pur-
chased Shells one-third interest for US$195 million.
Minorco and Falconbridge will spend US$1.3 billion
to develop Collahuasi into a world-class copper mine.
Production started in late 1998 and will last for 45
years. Total mineable reserves are 14 million tons of
copper with a value of US$36.4 billion at 1994
copper prices. Remote sensing played a key role in
defining this valuable property.
6. Borate minerals Salar de Uyuni, Bolivia
Boron and its compounds occur as borate miner-
als in the crust and brine of certain evaporite de-
posits and in modern dry salt lakes, called salars in
Spanish. Fig. 13 is a TM image of the Salar de
Uyuni in southwest Bolivia, which is the worlds
largest salar with an area of 10,000 km
2
. The Salar is
known to contain borate minerals, but the ore re-
serves and economic potential were incompletely
.
evaluated. Risacher 1989 analyzed brine samples
from 68 shallow drill holes and prepared a map of
boron concentration shown in Fig. 14. Had the holes
been uniformly distributed over the Salar, each hole
would represent an area of 147 km
2
, which is very
sparse sampling. Landsat TM, however, covers the
Salar with more than 1 million ground resolution
cells that represent 9=10
y4
km
2
each. The Boli-
vian government contracted with Intercontinental
Resources, to conduct a Landsat evaluation of the
.
Salar Sabins and Miller, 1994 .
A major question in the evaluation was whether
borate minerals in the crust of the Salar have spectral
features that can be recognized in TM data. Fig. 15

shows the reflectance spectrum of ulexite NaCaB O

5
.
P8H O which is the principal borate mineral in the
2
Salar. Fig. 15 also shows the spectrum of halite
.
NaCl , or rock salt, which constitutes more than
90% of the crust. TM ratio 4r7 should have high
values for ulexite and low values for halite. A 4r7
ratio image was generated and density sliced to
highlight the highest ratio values which are shown as
.
triangles in the map Fig. 14 . The highest ratio
values coincide with the contours of maximum boron
concentration in an embayment at the south margin
of the Salar. Additional triangles elsewhere around
the margin of the Salar indicate potential borate
. Fig. 15. Reflectance spectra for halite NaCl and ulexite
. NaCaB O 8H O . TM bands 4 and 7 are used to calculate 4r7
5 2
ratio image.
( ) F.F. SabinsrOre Geology Reiews 14 1999 157183 178
reserves that were not detected by the sample pro-
gram.
This ratio method should be useful for borate
exploration in other dry lakes.
7. Mineral exploration in covered terrain
The Collahuasi and Goldfield districts are in arid
terrain with extensive exposures of bedrock and little
soil or vegetation. Much of the world has temperate
to humid climates, however, and mineral deposits are
obscured or concealed by soil and vegetation. As a
rule of thumb, remote sensing cannot reliably recog-
nize hydrothermally altered rocks where vegetation
and soil cover exceeds 50%. Remote sensing, espe-
cially radar, can map lithology and structure in cov-
ered terrain.
Explorationists have long recognized the relation-
ship between vegetation, soils, and underlying min-
eral deposits that is shown diagrammatically in Fig.
16. Geochemical exploration techniques analyze the
metal content of samples of vegetation, soil, or wa-
ter. Areas with high metal concentrations are targets
for follow-up investigations. High concentrations of
metals in soils can cause changes in the vegetation
cover that include the following:
.
1 Lack of egetation. This may be caused by
concentrations of metals in the soil that are toxic to
plants. These areas are sometimes called copper
barrens where they are caused by high concentra-
tions of that metal. Areas that lack vegetation may be
seen on remote sensing images. These barren areas
may result from causes other than mineralization,
however.
.
2 Indicator plants. These are species that grow
preferentially on outcrops and soils enriched in cer-
.
tain elements. Cannon 1971 prepared an extensive
list of indicator plants. For example, in the Katanga
region of southern Zaire, a small blue-flowered mint,
Acrocephalus robertii, is restricted entirely to cop-
per-bearing rock outcrops.
.
3 Physiological changes. High metal concentra-
tions in the soil may cause abnormal size, shape, and
spectral reflectance characteristics of vegetation. A
relationship between spectral reflectance properties
of plants and the metal content of their soils could
form the basis for remote sensing of mineral deposits
in vegetated terrain.
It is reasonable to expect that vegetation growing
over mineral deposits should have different spectral
reflectance patterns from vegetation growing in non-
mineralized areas. The remote sensing of such spec-
. Fig. 16. Copper enrichment of vegetation and soil overlying a concealed copper deposit. From Sabins 1997, Fig. 11-19 .
( ) F.F. SabinsrOre Geology Reiews 14 1999 157183 179
tral differences could be an exploration method in
covered terrains. This concept was evaluated by
several research projects in the 1970s and 1980s.
Plants were grown hydroponically with metal salts
added to the nutrient solution. A control group was
grown with normal nutrients. Reflectance spectra of
the two groups were compared throughout the growth
cycle, but the results were inconclusive.
.
Yost and Wenderoth 1971 used the large, low-
grade, copper-molybdenum deposit at Catheart
Mountain, Maine, as a remote sensing test site. Field
spectrometers measured reflectance of trees growing
in normal soil and in mineralized soil overlying the
.
deposit Fig. 17 . Red spruce and balsam fir growing
in the mineralized soil both had higher metal concen-
trations than trees in unmineralized soil. In the re-
flected IR spectral region, the mineralized balsam
firs have a higher reflectance than the normal trees,
Fig. 17. Reflectance spectra of balsam fir and red spruce growing
in normal soil and in soil enriched in copper and molybdenum.
. From Yost and Wenderoth 1971, Figs. 5 and 6 .
whereas mineralized red spruce have a lower re-
.
flectance than the normal trees Fig. 17 . In the green
spectral region, the mineralized trees of both species

have a higher reflectance. Labovitz et al. 1983, Fig.

.
1 summarized other investigations of vegetation
spectra. With some exceptions, vegetation re-
flectance in the green and red bands generally in-
creased with increasing metal concentration in the
soil. In the reflected IR region, however, there is less
agreement; some studies show increased vegetation
reflectance and others show decreased reflectance.
.
Labovitz et al. 1983, p. 759 also noted that the
geobotanical model of Fig. 16 is not universally
correct. In Virginia, they found that the leaves of oak
trees growing in metal-rich soil may have a lower
metal content than leaves from trees in normal soil.
Geophysical Environmental Research used a non-
imaging airborne system that acquires detailed re-
flectance spectra. The spectra in Fig. 18 were ac-
quired for conifers growing in a mineralized area and
in an adjacent nonmineralized area. In the green
.
band 0.5 to 0.6 mm reflectance is higher for trees
in the mineralized area, which is consistent with
other studies. Beginning at a wavelength of about 0.7
mm, vegetation spectra have a steep upward slope to
the high reflectance values in the IR region. In Fig.
18, this steep slope is shifted slightly toward shorter
wavelengths for the conifers growing in the mineral-
ized area. This shift, called the blue shift, has been
noted in vegetation over several mineralized areas
.
Collins et al., 1983 and may have exploration
potential.
There is little research today on remote sensing of
vegetation spectra for mineral exploration, to my
knowledge. The original researchers are retired or
are working on environmental projects. The avail-
ability of hyperspectral data may encourage new
investigations.
8. Future technology
Secondary silica in the form of quartz is an
important component of hydrothermal alteration sys-
tems, but has no diagnostic spectral features in the
.
visible or reflected IR spectral regions Fig. 7 . This
( ) F.F. SabinsrOre Geology Reiews 14 1999 157183 180
Fig. 18. Airborne reflectance spectra of conifers in Cotter Basin, MT. Note the blue shift for conifers growing in a mineralized area.
. From Collins et al. 1983, Fig. 4B .
inability to detect quartz is a handicap for remote
sensing systems, regardless of their spectral resolu-
tion. A possible solution lies in the thermal IR region
.
8 to 14 mm where silica content is indicated by the
wavelength where the greatest energy absorption oc-
curs. Fig. 19 shows emissivity spectra of igneous
rocks in the thermal region from 8 to 14 mm. All the
spectra contain broad emissivity minima, called ab-
sorption bands, that are caused by the silica content
of the rocks. Arrows indicate the center of each
absorption band. Note that the arrows shift to longer
wavelengths as the silica content of the rocks de-
creases.
.
The thermal IR multispectral scanner TIMS is a
NASArJPL experimental aircraft system that ac-
quires six bands of imagery in the thermal IR region.
Fig. 19 shows the TIMS bands which are positioned
.
to record the absorption minima. Hook et al. 1992
processed TIMS data of the Cuprite, Nevada district
and recognized the high concentrations of silica that
occur in the hydrothermally altered rocks.
NASA plans to deploy the advanced spaceborne
.
thermal emission and radiation radiometer ASTER
.
on the first Earth Observation Satellite EOS-A that
may be launched in the future. Fig. 19 shows the five
thermal IR bands recorded by ASTER, which should
enable us to interpret variations in silica content.
TIMS and ASTER data can recognize high concen-
trations of silica, but cannot distinguish hydrothermal
silica from other forms such as igneous or sedimen-
tary silica. Hydrothermal silica can be recognized by
interpreting TIMS and ASTER images in conjunc-
tion with images showing geology and other alter-
.
ation mineral iron minerals, clays, and alunite .
Australia is organizing support for a satellite that
will include a hyperspectral scanner in the instru-
( ) F.F. SabinsrOre Geology Reiews 14 1999 157183 181
Fig. 19. Thermal emissivity spectra of igneous rocks with differ-
ent silica and quartz contents. Arrows show centers of absorption
bands. Note positions of spectral bands recorded by ASTER and
. TIMS. From Sabine et al. 1994, Fig. 3 .
ment package. The worldwide availability of hyper-
spectral images could be a major advance in mineral
exploration.
9. Summary
Remote sensing has proven a valuable aid in
exploring for mineral resources. Many ore deposits
are localized along regional and local fracture pat-
terns that provided conduits along which ore-forming
solutions penetrated host rocks. Landsat and radar
images are used to map these fracture patterns. Hy-
drothermally altered rocks associated with many ore
deposits have distinctive spectral features that are
recognizable on digitally processed TM images. In
the future, hyperspectral scanners may identify spe-
cific alteration minerals. Multispectral thermal IR
systems have the potential to map hydrothermal sili-
cification.
Detection of hydrothermally altered rocks is not
possible in vegetated areas, so this environment re-
quires other remote sensing methods. Reflectance
spectra of foliage growing over mineralized areas
may differ from spectra of foliage in adjacent non-
mineralized areas. The spectral differences, however,
are variable for different plant species. Additional
research and development is needed for remote de-
tection of mineral deposits in vegetated terrain.
Some explorationists object to the use of remote
sensing because Remote sensing is no substitute for
field mapping. We do not advocate remote sensing
as a substitute for field mapping. Our points are:
1. On a digitally processed TM image, a geologist
can interpret the rock types, structure, and hy-
drothermal alteration for a region of 31,000 km
2
.
2. Occurrences of important hydrothermal minerals
.
clays and alunite are expressed, using wave-
lengths that are undetectable by the eye.
3. The image interpretation will produce a map of
localities, or prospects, with favorable conditions
for mineral deposits. The image can also be used
to plan the best ground access to the prospects.
4. The field geologist can now efficiently locate,
evaluate, and sample the prospects. Some of the
image-derived prospects will not merit additional
investigation. Some potential deposits will not be
recognized on the image. Nevertheless, field work
can be concentrated in areas with higher mineral
potential.
In summary, remote sensing when properly em-
ployed is a valuable technical resource for mineral
exploration.
Acknowledgements
Much of my research on this topic was done
during my career with the Chevron. Many colleagues
in the remote sensing community allowed me to use
illustrations from their work and are acknowledged
in the figure captions.
( ) F.F. SabinsrOre Geology Reiews 14 1999 157183 182
References
Abdelhamid, G., Rabba, I., 1994. An investigation of mineralized
zones revealed during geological mapping, Jabal Hamra Fad-
dan-Wadi Araba, Jordan, using Landsat TM data. Int. J.
Remote Sensing 15, 14951506.
Abrams, M.J., Brown, D., 1985. Silver Bell, Arizona, porphyry
copper test site: The Joint NASArGeosat test case study,
Section 4, American Association of Petroleum Geologists,
Tulsa, OK, 73 pp.
Abrams, M.J., Brown, D., Lepley, L., Sadowski, R., 1983. Re-
mote sensing for porphyry copper deposits in southern Ari-
zona. Economic Geology 78, 591604.
Ashley, R.P., 1974. Goldfield mining district. Nevada Bureau of
Mines and Geology Report 19, 4966.
Ashley, R.P., 1979. Relation between volcanism and ore deposi-
tion at Goldfield, Nevada. Nevada Bureau of Mines and
Geology Report 33, 7786.
Bennett, S.A., Atkinson, W.W., Kruse, F.A., 1993. Use of the-
matic mapper imagery to identify mineralization in the Santa
Teresa District, Sonora, Mexico. Int. Geol. Rev. 35, 1009
1029.
Boardman, J.W., 1993. Automated spectral unmixing of AVIRIS
data using convex geometry concepts: Summaries of Fourth
Annual JPL Airborne Geoscience Workshop, Vol. 1, JPL
Publication 93-26, Pasadena, CA, pp. 1114.
Cannon, H.L., 1971. The use of plant indicators in ground water
surveys, geologic mapping, and mineral prospecting. Taxon
20, 227256.
Collins, W., Chang, S.H., Raines, G., Canney, F., Ashley, R.,
1983. Airborne biogeophysical mapping of hidden mineral
deposits. Econ. Geol. 78, 737749.
Crosta, A.P., Sabine, C., Taranik, J.V., 1998. Hydrothermal alter-
ation mapping at Bodie, California, using AVIRIS Hyperspec-
tral data. Remote Sensing of The Environment 65, 309319.
Dick, L.A., Ossandon, G., Fitch, R.G., Swift, C.M., Watts, A.,
. 1993. In: Romberger, S.B., Fletcher, D.I. Eds. , Discovery of
Blind Copper Mineralization at Collahuasi, Chile, Integrated
Methods in Exploration and Discovery, Society of Economic
Geologists, Abstracts.
Eiswerth, B.A., Rowan, L.C., 1993. Analyses of Landsat thematic
mapper images of study areas located in western Bolivia,
northern Chile, and southern Peru: Investigation de Metales
Preciosos en El Complejo Volcanico Neogeno-Cuaternario de
Los Andes Centrales, U.S. Geol. Surv. Projecto BIDrTC-88-
02-32-5, pp. 1944.
Goetz, A.F.H., Srivastava, V., 1985. Mineralogic mapping in the
Cuprite mining district: AIS Data Analysis Workshop Pro-
ceedings, Jet Propulsion Laboratory Publication 85-41,
Pasadena, CA, pp. 2231.
Goosens, M.A., Kroonenberg, S.B., 1994. Spectral discrimination
of contact metamorphic zones and its potential for mineral
exploration, province of Salamanca, Spain. Remote Sensing of
The Environment 47, 331344.
Green, R.O. et al., 1998. Imaging spectroscopy and the airborne
. visiblerinfrared imaging spectrometer AVIRIS . Remote
Sensing of The Environment 65, 227248.
Griffiths, P.S., Curtis, P.A.S., Fadul, S.E.A., Scholes, P.D., 1987.
Reconnaissance geological mapping and mineral exploration
in northern Sudan using satellite remote sensing. Geol. J. 22,
225249.
Harvey, R.D., Vitaliano, C.J., 1964. Wall-rock alteration in the
Goldfield District, Nevada. Journal of Geology 72, 564579.
Hook, S.J., Elvidge, C.D., Rast, M., Watanabe, H., 1991. An
. evaluation of short-wave-infrared SWIR data from the
AVIRIS and GEOSCAN instruments for mineralogical map-
ping at Cuprite, Nevada. Geophysics 56, 14321440.
Hook, S.J., Gabell, A.R., Green, A.A., Kealy, P.S., 1992. A
comparison of techniques for extracting emissivity information
from thermal infrared data for geologic studies. Remote Sens-
ing of the Environment 42, 123135.
Hunt, G.R., 1980. Electromagnetic radiation the communica-
tion link in remote sensing. In: Siegal, B.S., Gillespie, A.R.
. Eds. , Remote Sensing in Geology. Wiley, New York, NY.
Kaufmann, H., 1988. Mineral exploration along the AqabaLevant
structure by use of TM data; concepts, processing, and results.
Int. J. Remote Sensing 9, 16391658.
Knepper, D.H., Simpson, S.L., 1992. Remote sensing in Geology
and mineral resources of the Altiplano and Cordillera Occiden-
tal, Bolivia, U.S. Geol. Surv. Bull. 1975, pp. 4755.
Krohn, M.D., Kendall, C., Evans, J.R., Fries, T.L., 1993. Rela-
tions of ammonium mineral in several hydrothermal systems
in the western U.S. J. Volcanol. Geotherm. Res. 56, 401413.
Kruse, F.A., 1996. Cuprite, Nevada supplemental field trip
. information. In: Shaulis, L. Ed. , Remote Sensing Field Trip
Guidebook: Environmental Research Institute of Michigan
Eleventh Thematic Conference on Geologic Remote Sensing,
Guidebook, Ann Arbor, MI, pp. I92I103.
Labovitz, M.L., Masuoka, E.J., Bell, R., Segrist, A.W., Nelson,
R.F., 1983. The application of remote sensing to geobotanical
exploration for metal sulfides results from the 1980 field
season at Mineral, Virginia. Econ. Geol. 78, 750760.
Lowell, J.D., Guilbert, J.M., 1970. Lateral and vertical alteration-
mineralization zoning in porphyry ore deposits. Econ. Geol.
65, 373408.
Murphy, R.J., 1995. Mapping of jasperoid in the Cedar Moun-
tains, Utah, U.S.A., using imaging spectrometer data. Int. J.
Remote Sensing 16, 10211041.
Nicolais, S.M., 1974. Mineral exploration with ERTS imagery:
Third ERTS-1 Symposium, NASA SP-351, Vol. 1, pp. 785
796.
Risacher, F., 1989. Economic study of the Salar de Uyuni:
Institute Francis de Recherche Scientifique Pour le Developp-
ment en Cooperation, Informe no, 32, Translated by E. Jack-
son-Reardon, 67 p.
Rowan, L.C., Bowers, T.L., 1995. Analysis of linear features
mapped in Landsat thematic mapper and side-looking airborne
radar images of the Reno, Nevada 18 by 28 quadrangle,
Nevada and California implications for mineral resource
studies. Photogramm. Eng. Remote Sensing 61, 749759.
Rowan, L.C., Wetlaufer, P.H., 1975. Iron-absorption band analy-
sis for the discrimination of iron-rich zones. U.S. Geol. Surv.
Type III Final Report, Contract S-70243-AG.
Rowan, L.C., Wetlaufer, P.H., Goetz, A.F.H., Billingsley, F.C.,
( ) F.F. SabinsrOre Geology Reiews 14 1999 157183 183
Stewart, J.H., 1974. Discrimination of rock types and detec-
tion of hydrothermally altered areas in south central Nevada
by the use of computer-enhanced ERTS images. U.S. Geol.
Surv. Prof. Pap. 883, 35.
Rowan, L.C., Goetz, A.F.H., Ashley, R.P., 1979. Discrimination
of hydrothermally altered and unaltered rocks in the visible
and near infrared. Geophysics 42, 533535.
Rowan, L.C., Trautwein, C.A., Purdy, T.L., 1991. Maps showing
association of linear features and metallic mines and prospects
in the Butte 18 by 28 quadrangle, Montana. U.S. Geol. Surv.
Misc. Invest. Ser. Map I-2050-B.
Sabins, F.F., 1983. Geologic interpretation of Space Shuttle im-
ages of Indonesia. Am. Assoc. Pet. Geol. Bull. 67, 20762099.
Sabins, F.F., 1997. Remote Sensing Principles and Interpreta-
tion, 3rd edn., W.H. Freeman, New York, NY., 494 pp.
Sabins, F.F., Miller, R.M., 1994. Resource assessment Salar de
Uyuni and vicinity. Proceedings of Tenth Thematic Confer-
ence on Geologic Remote Sensing. Environmental Research
Institute of Michigan, Ann Arbor, MI, pp. I92I103.
Sabine, C., Realmuto, V.J., Taranik, J.V., 1994. Quantitative
estimation of granitoid composition from thermal infrared
. multispectral scanner TIMS data, Desolation Wilderness.
northern Sierra Nevada, California. J. Geophys. Res. 99,
42614271.
Segal, D.B., Rowan, L.C., 1989. Map showing exposures of
limonitic rocks and hydrothermally altered rocks in the Dillon
1 by 2 quadrangle, Idaho and Montana. U.S. Geol. Surv. Misc.
Invest. Ser. Map I-1803-A.
. Singhroy, V.H. Ed. , 1991. Geological remote sensing in Canada.
Can. J. Remote Sensing, 17, 71200.
Spatz, D.M., Wilson, R.T., 1994. Exploration remote sensing for
porphyry copper deposits, western America Cordillera. Pro-
ceedings Tenth Thematic Conference on Geologic Remote
Sensing. Environmental Research Institute of Michigan, Ann
Arbor, MI, pp. 12271240.
Unrug, R., 1988. Mineralization controls and source metals in the
. Lufilian fold belt, Shaba Zaire , Zambia and Angola. Econ.
Geol. 83, 12471258.
Van der Meer, 1994. Extraction of mineral absorption features
from high-spectral resolution data using non-parametric geo-
statistical techniques. Int. J. Remote Sensing 15, 21932214.
Vergara, H., 1978A. Cuadrangulo Ujina: Carta Geologica de
Chile, No. 33, Escala 1:50,000, Instituto de Investigaciones
Geologicas, Santiago, Chile.
Vergara, H., 1978B. Cuadrangulo Quehuita y sector occidental del
cuadrangulo Volcan Mino: Carta Geologica de Chile, No. 32,
Escala 1:50,000, Instituto de Investigaciones Geologicas, San-
tiago, Chile.
Watson, K., Kruse, F.A., Hummer-Miller, S., 1990. Thermal
infrared exploration in the Carlin trend, northern Nevada.
Geophysics 55, 7079.
Yost, E., Wenderoth, S., 1971. The reflectance spectra of mineral-
ized trees. Proceedings of Seventh International Symposium
on Remote Sensing of Environment, Vol. 1, University of
Michigan, Ann Arbor, MI, pp. 269284.