Searching for Gold in the Andes with ASTER

Alvaro P. Crósta and Carlos Roberto de Souza Filho

Satellite remote sensing images have been widely and successfully used for mineral exploration since the launch of Landsat in 1972. This application relies mostly on the capability of the sensor to register spectral signatures and other geological features related to mineral deposits. Gold is one of the most important mineral commodities that have been searched with the use of satellite remote sensing images over the last 30 years.
Although gold cannot be “seen” directly by any remote sensor, the presence of minerals which form in association with this precious metal can be detected based on their spectral signatures. A group of minerals which occur in the alteration zones associated with gold deposits, generically referred as “clay minerals,” have diagnostic spectral signatures mostly in the shortwave infrared portion of the electromagnetic spectrum. These signatures can be used to locate sites most favorable to the occurrence of deposits, saving the mineral industry a great deal of time and costs in their exploration programs.
This capability of multispectral remote sensing is particularly useful in regions of the world where access is difficult and previous geological knowledge is limited. This is the case of the Andean Cordillera, in South America, a region where remote sensing has had a key role in the exploration efforts in the search for metals in the last decade. The Andes represent an ideal location for geological applications of remote sensing, due to the scarcity of vegetation and excellent bedrock exposure at the surface.
For nearly 20 years, the workhorse of mineral exploration has been Landsat Thematic Mapper/Enhanced Thematic Mapper+ (TM/ETM+), with its six spectral bands recording energy reflected by surface materials between the visible/near infrared (VNIR) and shortwave infrared (SWIR) portions of the electromagnetic spectrum, plus one band in the thermal infrared (TIR). Their use, in conjunction with appropriate data processing technology, allows the detection of ferric oxides/hydroxides, hydroxyl-bearing minerals and carbonates in the hydrothermal alteration zones associated with gold deposits. However, the broad band configuration of TM/ETM+, particularly in the SWIR, only allows the identification of sites with likely occurrence of hydrothermal alteration, without providing the necessary spectral resolution to identify specific minerals, a very important task in searching for potential mineral deposits.
With the launch of EOS/Terra platform in December 1999, a new sensor with enhanced capabilities for mineral exploration became available. This is the Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER), a state-of-the-art sensor built by Japan’s Ministry of Economy Trade and Industry (METI), with the collaboration of scientific and industry organizations, and launched by NASA. ASTER is a multispectral imaging radiometer that covers VNIR, SWIR and TIR wavelengths with 14 bands, ranging from 15 to 90m in spatial resolution. ASTER was conceived as a science instrument, aiming to improve the understanding of the processes occurring on or near the surface of the Earth and to address global change topics.

ASTER Characteristics and Implications for Geology and Mineral Exploration
ASTER has unique features which, in combination with its worldwide coverage, make it an excellent choice as a space-borne sensor for applications in geology and mineral exploration. Its 14 bands (compared to seven spectral bands of TM) are divided into three different subsystems, each one with a different spatial resolution: three bands in the VNIR (15m), six in the SWIR (30m), and five bands in the TIR (90m). In addition, ASTER has stereoscopic capability (Table 1).
The VNIR subsystem consists of two independent telescopes: a nadir- and a backward-looking telescope. The focal plane of the nadir telescope contains three detector arrays (Bands 1, 2, 3N), covering the green, red, and near-infrared regions respectively, while the backward telescope has only one (3B) covering the same spectral range of 3N. This nadir and backward-looking configuration is used for same-orbit stereo imaging acquisition (along-track stereo) with a base-to-height (B/H) ratio of about 0.6, and an intersection angle of about 27.7 degrees. These stereoscopic bands can be used for DEM generation with horizontal accuracy on the order of 25m, compatible with scales of 1:100,000 to 1:50,000 (see EOM, Jan. 2002). Stereo viewing capability brings a significant benefit to geological interpretation of remote sensing images, allowing a much better characterization of textures, which are used in identifying different rock types, and structures, such as faults that usually control the formation and are spatially related to mineral deposits.
The SWIR subsystem comprises six bands spanning from 1.60 to 2.43 µm These bands were carefully chosen to detect and distinguish among hydro­thermal alteration minerals. Band 4 covers the spectral region where all these minerals reflect strongly, whereas the other bands encompass wavelength ranges in which different alteration minerals show diagnostic absorption features:
-- 2.17-2.21 µm (bands 5 and 6): major absorption by Al(OH)–bearing minerals (kaolinite, montmorillonite, illite, pyrophyllite), plus alunite and buddingtonite;
-- 2.25-2.30 µm (band 8): major absorption by Fe(OH)-bearing minerals (jarosite);
-- 2.30-2.40 µm (bands 8 and 9): major absorptions by Mg(OH)–bearing (chlorite, talc) and carbonate (calcite, dolomite) minerals;
-- 2.40 µm (band 9): broad absorption feature by opaline silica.
The TIR subsystem comprises five spectral bands, making ASTER the first spaceborne sensor collecting multispectral data in the TIR, with medium spatial resolution. Despite the limitations in­herent to their 90-m resolution, these bands are capable of differentiating among several rock types, based on their chemical composition and crystal structure. Of particular importance is the characterization of rocks in the 8 to 12 µm range, based on the type of silicate crystalline structure. This range coincides with the window in the transmission of the atmosphere and is fully covered by ASTER’s thermal bands. In terms of mineral exploration, these bands are useful in recognizing areas of silicification (opaline silica) formed by hydrothermal alteration processes in association with mineral deposits.

Data Processing for
Alteration Mapping
Digital image processing techniques have been widely used in the search of mineral deposits using multispectral remote sensing images. The basic idea is that the spectral information related to minerals represents a very small fraction of the total information content of these images. Hence, the useful information is often “buried” within a vast amount of data, mostly unrelated to the minerals of interest, and is usually not identifiable unless the data is properly processed. The task is therefore to do a selective extraction of the information of interest.
There are two general approaches to this task. The first one is a semi-quantitative approach and can be generally referred to as “multispectral”. The second one is more quantitative, referred to as the “hyperspectral” approach.
The multispectral approach was developed in order to allow the processing of Landsat MSS and TM/ETM+ and it employs techniques such as band ratioing and principal component analysis (PCA). A technique called “feature-orientated principal component selection,” based on PCA, has proven more efficient than band ratioing and has played an important and successful role in hydrothermal alteration mapping using Landsat in the Andes and elsewhere in the world. It is a relatively simple and well-known technique and it does not require any ancillary spectral data to be available or any type of atmospheric correction previously applied to the data. In addition, it can be run on any commercial image processing software.
The hyperspectral approach was developed for very high spectral resolution sensors (up to hundreds of spectral bands covering the VNIR-SWIR range) and it is based mostly on spectroscopic methods, in which a quantitative comparison is made between pixel spectra and known materials (reference) spectra. It can also be applied to multispectral sensors like TM/ETM+ and ASTER, but it requires a sound knowledge of mineral spectroscopy, specific software, previous atmospheric correction, reference spectra, and it is considerably more complex than the multispectral approach.
The potential of using ASTER in combination with the PCA-based approach was assessed in the discrimination of alteration minerals related to a gold prospect in the southern Andean Cordillera. The ASTER scene covers the mining district of Los Menucos, in the Patagonia region of Argentina (Figure 1).

Mapping Alteration Minerals at Los Menucos District
At the region of Los Menucos, epithermal gold mineralization is known to occur at various locations, all related to hydrothermally altered Triassic-Jurassic volcanic rocks. This region is currently the focus of exploration efforts by mining companies.
The PCA-based method consists of calculating the PC images of different sub-sets of four ASTER bands, selected according to the position of characteristic spectral features of key alteration mineral endmembers in the VNIR and SWIR portions of the spectrum (Table 2).
The next step is to identify, from the four newly-generated PC bands, which one contains the spectral information related to the specific mineral endmember, based on the eigenvector statistics provided by the PCA. This is usually the 3rd or the 4th PC band and it will contain the desired information, as well as a fair amount of noise. This information usually represents, in quantitative terms, a very small fraction of the total information content of the four original bands. However, this is the information that matters when looking for mineral spectral signatures and it is revealed by the method, which selectively extracts the desired spectral information.
By proper manipulation of histograms and pseudo-coloring, the PC bands are converted into “mineral abundance” images, thus showing the areas most likely to contain the alteration minerals listed in Table 1. The PC mineral abundance images for the Los Menucos district are shown in Figure 2, with “hot” colors indicating areas of likely high abundances.
Another way to represent the spatial distribution of the alteration minerals is to use individual abundance images combined as a RGB colour composite. Figure 3 shows the RGB image of kaolinite, illite, and alunite, draped over ASTER band 3. All the altered areas known in the Los Menucos district are distinctively represented in different colors, which correspond to the major alteration minerals that occur at the various locations.
The stereo capability of ASTER can also be put into use for geological interpretation of relief features, together with the mineral information provided by PCA. This can be achieved by combining the image shown in Figure 2 with the digital elevation model (DEM) derived directly from ASTER bands 3N and 3B. The resulting image is shown in Figure 4.

Conclusions
The expensive and time-consuming task of finding new mineral deposits can benefit considerably from the introduction of new remote sensing systems. One of the best examples of these benefits is represented by ASTER, an instrument which has been truly built for geological applications. With its spectral bands strategically placed in wavelength ranges important for mineral identification, plus stereo capability, it represents an impressive advancement over Landsat TM/ETM+, the only previous space-borne sensor with worldwide coverage and spectral bands suitable for this task.
The results achieved at Los Menucos show how ASTER can be used to map mineralogical types in hydrothermally altered areas, establishing their spatial variability on a pixel basis over continuous ground surfaces. This is most remarkable when one considers the constraints imposed by ASTER’s spatial resolution of only 30m in the SWIR and limited (relatively to hyperspectral sensors) spectral resolution.
In terms of data processing, both, “multispectral” and “hyperspectral,” ap­proach­es can be used for extracting mineral information from ASTER data. However, for areas where ancillary spectral information is not available and with limited ground truth, the simple, yet robust, PCA-based method has shown to be very effective. Besides, this is an easy-to-use technique that has been widely employed for more than a decade, having become a standard in the mineral industry for alteration mapping using Landsat TM.

Acknowledgments
We thank IAMGOLD Corporation, Toronto, for releasing the results for publication and M. Abrams (NASA/JPL), for discussion on ASTER data characteristics and quality. DEM extraction from ASTER data was achieved through the AsterDTM software, developed by M. Steinmayer from Sulsoft, Brazil, as a plug-in to Envi image processing package. The authors were supported by grants from Fundaç˜ao de Amparo `a Pesquisa do Estado de S˜ao Paulo (FAPESP) and Conselho Nacional de Desenvolvimento Científico e Tecnológico (CNPq), Brazil.

About the Authors
Alvaro P. Crósta and Carlos Roberto de Souza Filho are senior lecturers at the Department of Geology and Natural Resources, Geosciences Institute, University of Campinas (UNICAMP), Brazil.

Back