Title: Satellite Imagery for Geological Mapping, Mongolia.
Presenter: Mr Bob Walker
Geoimage Pty Ltd
P.O. Box 789
Bob graduated in Geology in 1968 and after a decade in mineral exploration became involved in remote
sensing leading to the founding of Geoimage in 1988. Geoimage is a privately owned Australian
company and a leading supplier of satellite imagery and image processing services to the user
community in Australia. Bob is a leading implementer of commercial remote sensing technology in
Australia and has extensive experience in developing image product solutions to meet client needs.
Page 1 of 13
SATELLITE IMAGERY FOR GEOLOGICAL MAPPING, MONGOLIA
R.N. Walker1 and C.R. Nash2
Since the first LANDSAT sensor was launched in 1972, mineral and petroleum
exploration geologists have been the most enthusiastic in recognising the value of the
imagery for surface mapping. The availability of the SWIR bands in the more recent
sensors on the LANDSAT Thematic Mapper (1982) and the Enhanced Thematic
Mapper (1999) have greatly improved the ability to discriminate rock units. Two recent
changes have seen major advances in the ability to remotely map terrains such as in
The first was the global availability of a reliable Digital Elevation Model (DEM). Initially
this was the release of the SRTM DEM which covers the majority of the world’s land
mass at a resolution of 90m. Subsequently the routine capture of ALOS PRISM triplets
from which can be derived DEMs with a 5m resolution and a 5m vertical accuracy has
provided considerably improved spatial resolution. Orthorectification using these surface
models reduces geometric distortion in satellite imagery and pseudo-stereo imagery
created using them greatly enhances the ability to interpret structural detail.
The second has been the general availability of satellite imagery with ground resolutions
between 0.5m and 10m. These satellites generally have multispectral sensors however
the spectral bands are in the visible and NIR and are not ideal for geological
discrimination. The panchromatic bands, when geometrically corrected, can be used to
pan-sharpen the more geologically useful LANDSAT band combinations 741 or 543
which are then interpreted in pseudo-stereo imagery at scales up to 1:20 000.
In Mongolia, these developments in satellite imagery have proven successful in
exploration programs for oil and gas, coal and porphyry copper deposits. A case study
in which merged ALSO/Landsat data have been used to re-examine the structure of
petroleum traps in the Zuunbayan Oilfield region is presented.
Geoimage Pty Ltd, PO Box 789, Indooroopilly 4068, Brisbane, Australia
(firstname.lastname@example.org) Colin Nash & Associates Pty Ltd, PO Box 519, Mt Gravatt
Plaza 4122, Brisbane, Australia (email@example.com)
Page 2 of 13
The spectral bands of the LANDSAT Thematic Mapper are ideal for the preparation of
colour composite imagery for geological discrimination of rock units. The pixel size of
the imagery however limits its usefulness to scales of 1:70k and smaller. If good Digital
Elevation Models (DEMs) and higher resolution imagery are available, it is possible to
effectively use the LANDSAT spectral information in the preparation of pseudo-stereo
imagery to 1:20 000 scale and possibly larger.
LANDSAT THEMATIC MAPPER
The LANDSAT satellites, managed by NASA and the U.S. Geological Survey, have
routinely imaged the Earth’s surface since the Launch of LANDSAT 1 in 1972. The first
three satellites contained the Multispectral Scanner Sensor (MSS) which had 4 spectral
bands in the visible and near IR at a pixel resolution of 80m (Fig.1). The imagery from
this sensor was comparable to false colour infrared aerial photography in its spectral
characteristics and the bands were selected for agricultural purposes rather than
geological. The characteristics of the imagery restricted its use to ‘lineament” studies
rather than geological mapping at scale of about 250 000.
Figure 1. Spectral reflectance curves of common cover types with spectral bands of the
main remote sensing satellites.
Page 3 of 13
LANDSATs 4 and 5 (1982 to current) contain the Thematic Mapper sensor (TM) which
provide improved spectral and spatial resolution. The TM has seven bands and
includes a visible blue and two mid infrared bands of which band 7 (2.08 – 2.35um) is
of particular importance to geologists. These extra bands enables the preparation of
three band colour composites which give geologists unprecedented powers to
discriminate rock-types (Fig.2). The 30m resolution of the sensor is useful for
geological mapping to scales to 1:120 000.
Figure 2. Landsat Thematic Mapper – B432 – simulated MSS bands (left) and
B741(right) showing the increased spectral information on the Thematic Mapper
imagery available for geological mapping. Each image 10km by 10km.
LANDSAT 7 launched in 1999 included an Enhanced Thematic Mapper ( ETM+) with
the same spectral bands as the TM but with a panchromatic sensor at 15m resolution.
This band is routinely used to pan-sharpen the multispectral bands. Although sampled
at 15m resolution, it probably has a real resolution somewhere between 15 and 20m
and the pan-sharpened imagery can be enlarged to between 1:60 000 and 1:80 000
Apart from the better spectral bands, the other advantages of the TM and ETM+
1. The large archive of worldwide data that has been collected particularly between
1999 and 2003 when LANDSAT 7 was working without problems, and
2. This worldwide archive where available from the USGS is now free for internet
download. This data has been orthorectified to the Global Land Survey (GLS)
2000 dataset and is generally accurate to better than 50m in E and N.
The only disadvantages, if they can be called that, are-
Page 4 of 13
1. The pixel size restricts the scale that the imagery can be used independently to
about 1:70 000 scale, and
2. The imagery is not available in stereo apart from the overlap of the north-south
image strips and this varies from about 7% at the equator to complete overlap at
latitudes greater than 60 degrees.
The Japanese ASTER instrument is an imaging system onboard Terra – the first Earth
Observation System (EOS) satellite. ASTER consists of separate visible-near infrared
(VNIR), short wave infrared (SWIR) and thermal (TIR) sensors at 15m, 30m and 90m
resolution respectively, and is capable of providing colour composite imagery
comparable in spectra and resolution to the ETM+.
The Advanced Land Observation Satellite (ALOS) developed by the Japan Aerospace
Exploration Agency (JAXA) was successfully launched on January 24, 2006. The
satellite has two optical sensors, a Panchromatic Remote-sensing Instrument for Stereo
Mapping ( PRISM) and an Advanced Visible and Near Infrared Radiometer type 2
(AVNIR-2). The PRISM instrument has three independent optical systems for nadir,
forward and backward looking to achieve along-track stereoscopy all at a pixel
resolution of 2.5m. This stereo imagery can be used to prepare DEMs at a cell size of
5m and with errors of the order of 5m in height. The AVNIR-2 sensor has 4 bands in the
visible and near Infrared and has a 10m pixel size at Nadir. These bands limit the
application of such imagery for geological mapping application. The normal collection
angle is near vertical although the instrument can collect up to +/- 44 degrees off nadir.
On-demand programming is not available for any of the ALOS sensors, however they
have now been collecting imagery for three years and there is a large archive of
imagery available. The PRISM imagery produces the most accurate DEM from any
satellite imagery available from archive.
The SPOT5 satellite has sensors similar to PRISM and AVNIR-2 however the ALOS
was used in preference because of the requirement to create a DEM.
VERY HIGH RESOLUTION (VHR) SATELLITES
Since the launch of the first commercial VHR satellite IKONOS in 1999, there has been
a steady progression of launches and planned launches of satellites with sub-metre
resolution. Commensurate with this has been an increasing amount of archive imagery
available through the world usually at prices considerably lower than new capture
prices. The main satellites are those operated by DigitalGlobe – QuickBird and
WorldView1, and by GeoEye – GeoEye1 and IKONOS (See Fig. 1).
These satellites commonly have a panchromatic sensor at a sub-metre pixel size and
the multispectral bands are collected at four times the spatial resolution of the pan.
Depending on the application, a user may purchase the pan, multispectral or a pan-
sharpened colour product at the spatial resolution of the pan band. The spectral bands
Page 5 of 13
are the same for all the multispectral sensors with a visible blue, green and red and a
near Infrared. This limits the application of such imagery for geological mapping
The narrow swath width of these satellites ( 10km - 17km at nadir) often requires that
imagery is collected at large off-nadir angles to get timely capture for clients. Therefore
to get plannimetrically correct imagery it is necessary to use a reliable Digital Elevation
Model (DEM) to take out the height distortion in the imagery due to this oblique viewing.
The fall back world-wide DEM is the SRTM however DEMs produced from ALOS
PRISM produce a much more accurate product.
The Shuttle Radar Topography Mission (SRTM) was a joint project of NASA and the
Department of Defence’s National Imagery and Mapping Agency (NIMA). The project
involved the collect of single-pass interferometer using the Spaceborne Imaging Radar
(SIR-C) on the Space Shuttle Endeavour in February 2000. Data covered all the
Earths land masses between 60°N and 50°S and the DEM was released at a resolution
of 30m over the US and at a resolution of 90m over the rest of the world. The absolute
horizontal and vertical accuracy was expected to be 20m and 16m respectively however
the vertical accuracy has been shown by ground truthing to be between 5m and 10m
RMSE (Root Mean Squared Error).
The Ministry of Economy, Trade, and Industry (METI) of Japan and the United States
National Aeronautics and Space Administration (NASA) released Version 1 of the
Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) Global
Digital Elevation Model (GDEM) on June 29, 2009. The methodology used to produce
the ASTER GDEM involved automated processing of the entire 1.5-million-scene
ASTER archive, including stereo-correlation to produce 1,264,118 individual scene-
based ASTER DEMs, cloud masking to remove cloudy pixels, stacking all cloud-
screened DEMs, removing residual bad values and outliers, averaging selected data to
create final pixel values, and then correcting residual anomalies before partitioning the
data into 1°-by-1° tiles.
The ASTER GDEM covers land surfaces between 83°N and 83°S and is composed of
22,600 1°-by-1° tiles. The ASTER GDEM is in GeoTIFF format with geographic lat/long
coordinates and a 1 arc-second (30m) grid of elevation postings. Pre-production
estimated accuracies for this global product were 20m at 95 % confidence for vertical
data and 30m at 95 % confidence for horizontal data. Initial validation studies
concluded that the ASTER GDEM generally meets the pre-production accuracy
predications, but results do vary and included are areas where GDEM accuracy does
not meet the pre-production estimates. The ASTER GDEM is available at no charge to
users worldwide via electronic download from the Earth Remote Sensing Data Analysis
Center (ERSDAC) of Japan and from NASA’s Land Processes Distributed Active
Archive Center (LP DAAC).
Page 6 of 13
Pan-sharpening is an image processing technique where the spatial resolution of
multispectral images is increased using a co registered panchromatic image and at the
same time the spectral information in the multispectral data is preserved. There are
several techniques for pan-sharpening available but we have found the best to be that
of Zhang(2002) which is implemented in the PCI Geomatics Software. For best results,
the PCI pan-sharpening algorithm requires the identification of the multispectral bands
which cover the same spectral bandwidth as the panchromatic band.
For pan-sharpening to be effective, the pan and multispectral datasets must be co
registered. This is not a problem with LANDSAT ETM+ as the orthorectified
panchromatic and multispectral files supplied by the USGS are perfectly coregistered.
The ALOS AVNIR and PRISM are collected by different sensors and are not
coregistered. However if the AVNIR image has been collected at nadir , and as the
nadir PRISM is collected at a maximum few degrees off the vertical, the SRTM DEM is
sufficient to take all the height distortion out of the raw data. If accurate ground control
are not available for the orthorectification, correction to the EROS orthorectified
LANDSAT 7 panchromatic data will give a final location accuracy of the order of 50m
anywhere in the world. The correction of the geometrical distortion in the VHR satellite
imagery is more of a problem and depends on the off-nadir collection angle and the
accuracy of the DEM available for the orthorectification. For large off-nadir angles, the
SRTM with its 90m cell size will not be sufficient and an ALOS DEM or similar is
PAN IMAGE PIXEL SIZE DEM BEST SCALE
LANDSAT PAN 15m SRTM 60 000 – 80 000
10m ALOS/SRTM 40 000
2.5 ALOS 15 000 – 20 000
0.6-0.8m ALOS or better 10 000
0.5m ALOS or better 10 000
Table 1. Combinations of pan imagery to sharpen LANDSAT multispectral imagery and
the best usable scale. The DEM is required for the orthorectification and to prepare the
Page 7 of 13
Table 1 provides a guide to the types of panchromatic imagery that can be used to
sharpen LANDSAT 30m multispectral imagery and the best scale that these can be
used for interpretation. The inclusion of the multispectral AVNIR-2 and SPOT5 datasets
should not be surprising since they are just multiple “panchromatic” bands. Tests have
shown that a single band consisting of an average of all four AVNIR-2 bands produces
an aesthetically more pleasing image than using any of the individual bands (Fig.3).
Figure 3. Landsat Thematic Mapper B741 in RGB. Left image pan-sharpened with
LANDSAT pan and right image pan-sharpened with averaged AVNIR.
LANDSAT data that has been pan-sharpened with PRISM can be used in hardcopy
prints at 1:15 000 scale, a 5 times improvement in the scale that the LANDSAT can be
used by itself. The improvement in spectral discrimination that such a combination gives
over an AVNIR/PRISM combination is truly remarkable(Fig. 4).
The inclusion of the VHR satellite imagery with sub-metre pixel sizes to pan-sharpen
30m LANDSAT may surprise give that a ratio of about 1:5 (pan to multispectral pixel
size ratio) is commonly considered the maximum for such pan-sharpening. Such a
maximum ratio may be appropriate in urban areas where the variation in colour “texture”
may be of the order of metres, however in areas of geological interest, variations in rock
and soil colouration are often on the scale of tens of metres, and merging of the
datasets is appropriate. In such instances, the use of a term such as “colourisation”
rather than pan-sharpening may be more appropriate and we would also not advocate
over enlargement of such composite images.
Page 8 of 13
Figure 4. Landsat Thematic Mapper B741 in RGB pan-sharpened with PRISM (left)
and AVNIR natural colour pan-sharpened with PRISM (right)
Most geologists and geographers are familiar with air photos and the increased
information that can be obtained from viewing them in three dimensions using a
stereoscope. This information mainly relates to the ability to estimate dip angles and to
interpret structural information however it is also important in understanding the
development of the physical landscape. With the advent of two dimensional satellite
imagery, geologists learnt to rely more heavily on spectral information in their
interpretations. It is now possible to integrate two dimensional satellite imagery with
readily available digital elevation data from the SRTM and PRISM DEMS to produce
Although modelled on the stereo viewing of air-photos, the stereo imagery produced by
GEOIMAGE is slightly different. Air photos have radially oriented stereo (or height
induced) distortion by virtue of the instantaneous capture of a single air photo using a
centrally-focused lens system. In the pseudo stereo, the input is an orthorectified three
band colour image (or a black/white single band image) and the height offset is
introduced into the data in the east-west direction. It is normal to prepare a left and right
stereo pair, where the height offsets are made in equal amounts but in opposite
directions. The amount of introduced parallax is linearly related to the range of height
values in the image and will be dependent on the scale of the images, the type of
stereoscope used and the viewing needs of the interpreter (the amount of the stereo
offset can be fine tuned for an individual). One of the problems of a stereo left-right
combination is that each image is height distorted so that neither image will produce an
undistorted interpretation. This can be overcome by preparing a left-vertical stereo pair
Page 9 of 13
with double the height offset on the left stereo image and a vertical or undistorted image
containing the coordinate information. A second problem relates to the size of the image
prints and trying to position them under a stereoscope. This problem is best handled by
cutting the left stereo print into vertical strips and interpreting a strip at a time. The
vertical print can be left intact. The digital stereo images can also be examined on a
computer screen. This is most easily performed in ERMapper with the images displayed
in adjacent geolinked windows. If the windows are set up with a spacing similar to the
viewers eye separation distance, it is possible to view the stereo without a stereoscope.
Figure 5. Pseudo-stereo prints at 1:50k scale. LANDSAT B741 pan-sharpened with
PRISM. Vertical image has the gridlines and is superimposed on the left image.
CASE STUDY: STRUCTURE OF THE NORTHERN PART OF THE ZUUNBAYAN
OILFIELD, SOUTH-EASTERN MONGOLIA
The Zuunbayan oilfield, discovered in the 1940’s, is located within the East Gobi Basin
of southeastern Mongolia (Graham et al., 2001), and has recently been investigated by
major oil companies (Prost, 2004). Basin location, local geology and seismic structure
are shown in Fig. 6. A test study using merged stereoscopic Landsat and ALOS
PRISM imagery has recently been carried out (Nash and Walker, in prep.), covering an
area of some 1,500 km2 over the northern part of the oilfield. Interpretation was carried
out under a mirror stereoscope, using 1:50 000 scale hardcopy ALOS
Page 10 of 13
PRISM/LANDSAT 741 stereopair images (Vertical and Left Stereo), prepared as
described in the preceding section. Fig. 7 shows the locations of anticlines, synclines
and faults mapped from pan-sharpened PRISM-LANDSAT imagery, and compares
these to structures mapped by Prost (2004).
In the area shown in the figure, Cretaceous sedimentary rocks of the East Gobi basin
rest upon a basement of Paleozoic and Late Jurassic sediments. The new ALOS-
LANDSAT interpretation suggests a major structural offset along a major NNE-SSW
lineament (locality 4 in Fig. 7). The position of the ’NZF lineament’ indicates that it could
be the northern extension of the North Zuunbayan Fault (NZF) as determined from
seismic data (Prost, 2004). The lineament may be a late-stage left-lateral fault, as fold
axes northeast of the fault and of the Zuunbayan Anticline (locality 5) adjacent to the
NZF lineament can be seen in the figure. The trace of the Zuunbayan Anticline, as
indicated by bedding trends and dip directions seen on images, is seen to lie between 1
and 2km north of the location shown on Prost’s map. Finally, our study has revealed
the presence of WNW-ESE trending cross-folds (localities 1 and 3 in Fig. 7). In the
northernmost fold set of locality 3, the younger cross-folds clearly deform earlier ENE-
These results demonstrate the value of merging of the spectral content of LANDSAT
Thematic Mapper data with the higher spatial resolution of ALOS PRISM data.
Preparation of 3D pseudo-stereo ALOS PRISM images by incorporating SRTM/DEM
topographic data provides excellent hardcopy products for stereoscopic geological
interpretation, resulting in significant re-assessment of the locations of oil-prone
structures in the Zuunbayan oilfield,
Graham, S.A., Hendrix, M.S., Johnson, C.L., Badamgarav, D., Badarch, G., Amory, J.,
Porter, M., Barsbold, R., Webb, L.E. and Hacker, B.R., 2001. Sedimentary record and
tectonic implications of Mesozoic rifting in southeast Mongolia. Geol. Soc. America
Bull., 113, 1560-1579.
Nash, C.R., and Walker, R.N., (in prep). Integration and interpretation of ALOS and
LANDSAT Imagery: Sainshand-Zuunbayan Test Study, Southeastern Mongolia.
Prost, G.L., 2004. Tectonics and hydrocarbon systems of the East Gobi Basin,
Mongolia. AAPG Bull., 88, 483-513.
Zang, Yun. 2002. Problems in the Fusion of Commercial High-resolution Satellite
Images as well as LANDSAT 7 Images and Initial Solutions. International Archives of
Photogrammetry and Remote Sensing (IAPRS), Vol 34, Part4. “Geospatial Theory,
Processing and Applications”.
Page 11 of 13
Figure 6. Location, geological map and seismic section of study area, after Prost,
(2004). Pz - Paleozoic; JKl – Late Jurassic-Early Cretaceous; Ku – Late Cretaceous;
TQ – Cenozoic; ZA – Zuunbayan Anticline; SB – Sainshand belt of early Mesozoic
Page 12 of 13
Figure 7. Anticlinal structures mapped in the Sainshand-Zuunbayan area during
present study, compared to those mapped by Prost (2004). Numbered structures are
discussed in text. NZF – North Zuunbayan Fault.
Page 13 of 13