Hyperspectral Imaging of Kazakstan
The search for minerals with the ESSI Probe 1
By William Ferrand

The collapse of the Soviet Union left in its wake numerous independent republics. While most public attention in the West has focused on the largest post-Soviet republic, Russia, the second largest, Kazakstan, has generated great enthusiasm in the business community. This central Asian country is a land of great natural resources and is the site of many exploration programs for oil and minerals. A unique exploration approach is being taken by Earth Search Sciences Inc. (ESSI), a McCall, Id. based company. Through its involvement in SEMTECH, a joint U.S.-Kazakstani stock company, ESSI is participating in a program of exploration over a 4 million acre former Soviet nuclear test site known as "the Polygon." Given the size of the Polygon, remote sensing was deemed crucial to a successful mineral exploration program. Even before its involvement with SEMTECH and Kazakstan, ESSI had realized the added value of hyperspectral remote sensing. Consequently, ESSI took the unprecedented step of having its own airborne hyperspectral sensor developed, the ESSI Probe 1.

Heritage of the Probe 1
First, some background. What is hyperspectral remote sensing, and where did it originate? People have become familiar with multispectral remote sensing, the use of some small number of bands covering relatively broad sections of the visible and infrared wavelength range to image the Earth. Multispectral imagery has been used quite successfully to create maps that consist of land cover units with discernable spectral differences in the sensor's band set. The shortcoming of multispectral systems is that they undersample the spectrum so severely that no unique identification of those land cover units can be made from the data. Also, many units with subtle spectral variations might be grouped with inappropriate classes because a multispectral system lacks the spectral resolution necessary to discern such subtle differences. In contrast, hyperspectral imagery fully samples the visible and near infrared wavelength region collecting anywhere from 10s to 100s of spectrally narrow, and spectrally contiguous, bands of data. Thus from any spatial element of a hyperspectral image a complete visible and near infrared reflectance spectrum can be extracted, allowing a trained analyst to determine the identity of land cover units based on diagnostic absorption bands and other reflectance characteristics.
     Workers at NASA's Jet Propulsion Laboratory under the direction of Dr. Alexander Goetz (now head of the Center for Study of Earth from Space (CSES) at the University of Colorado) built the first Airborne Imaging Spectrometer or AIS in 1982. The AIS was a pushbroom instrument of limited swath width that flew on the NASA Ames C-130. It collected 64 channels, and its signal to noise ratio (SNR) was relatively poor by today's standards. Nonetheless, it was valuable as a proof of concept and made it possible for the same team to oversee the development of the successor to AIS, the Airborne Visible/InfraRed Imaging Spectrometer (AVIRIS). AVIRIS was a much more ambitious instrument consisting of a whiskbroom scanner and four spectrometers with four linear focal plane arrays to cover the 0.4 to 2.44 mm region. AVIRIS flies on the NASA ER-2 aircraft at 60,000 feet and the resulting imagery mimics what could be seen using an orbital imaging spectrometer. After a shaky first flight season in 1987, AVIRIS has proven very successful. It has proven the utility of imaging spectrometer data for everything from mineral mapping, to determinations of canopy chemistry.
     Amongst this field of government financed and operated imaging spectrometers, entered ESSI. ESSI founder Larry Vance realized the added value of hyperspectral remote sensing for mineral exploration. In order to learn about and utilize this technology, ESSI became a participant in NASA's Visiting Investigator Program (VIP). The VIP was available to companies that submitted qualified scientific projects of merit to the agency through NASA's Space Act Agreement program. Through this program ESSI was able to participate in several AVIRIS data collections. The wealth of information in the AVIRIS data convinced Vance and his associates that ESSI needed its own hyperspectral capability. Consequently, ESSI contracted with Dr. Terry Cocks of Integrated Spectronics Pty. Ltd. (ISPL) of Sydney, Australia to build a series of airborne imaging spectrometers that make use of the successful AVIRIS whiskbroom design with linear focal plane arrays for image acquisition.
     Much lighter than AVIRIS, the ESSI Probe 1 weighs less than 200 lbs (see Figure 1). and can be flown on platforms ranging from a Cessna 404 to a DeHaviland DHC-7. The Probe 1 is mounted on a stabilized platform and collects GPS data to reference the image data to GPS coordinates. Data from the Probe 1 has a somewhat lower spectral resolution than AVIRIS; 130 channels for the Probe 1 versus 224 for AVIRIS. However, its spectral resolution is more than sufficient to enable the data analyst to recognize a wide range of spectral features in minerals and vegetation. Like AVIRIS, the Probe 1 utilizes four spectrometers and four linear focal plane arrays to cover the 0.4 to 2.45 mm wavelength region in 128 channels. Unlike AVIRIS, the Probe 1 also has a single middle infrared (3 - 5 µm) and a single thermal infrared (8 - 10 µm) channel. The Probe 1 can be flown over a range of altitudes to provide pixel sizes ranging from 3 to 10 m and swath widths from 1.5 to 6 km. Given the Probe 1's transportability between platforms, finer pixel sizes can be achieved by transferring the instrument to a platform such as a helicopter.

New Data Processing Tools
In order to deal with the expected flood of data from the Probe 1, ESSI has teamed with Applied Signal and Image Technology Inc. (ASIT) of Millersville, Md. ASIT founders Dr. Joseph Harsanyi and Joe Hejl have developed a suite of data processing algorithms for the extraction of information from hyperspectral imagery. This suite of algorithms has been dubbed the "OSP Family" of algorithms after the central or "parent" orthogonal subspace projection algorithm. One advantage that the OSP family of operators has over most existing multispectral processing techniques, is that several of the key routines in the OSP family can be run without any a priori knowledge of materials that exist in the imaged area. ASIT has also recently demonstrated a prototype of the Prospectreª automated hyperspectral data processing system that will be used in conjunction with Probe 1 data collections to provide near-real time processing.

The Mission
The first major mission for the ESSI Probe 1 was as one of a suite of sensors that was flown on a U.S. Department of Energy (DOE) mission over Kazakstan. The DOE mission to Kazakstan came about as a direct consequence of ESSI's involvement with SEMTECH and representatives of the Kazakstan National Nuclear Center (NNC). The NNC wanted to develop the resources of the Polygon, but did not have the knowledge of, or access to the appropriate remote sensing technology to begin an exploration program. Thus, it reached an agreement with ESSI which resulted in the formation of SEMTECH.
     Initial work done with a Landsat TM scene that covered most of the polygon highlighted some interesting areas with probable mineralization; however, the multispectral undersampling referred to previously prevented identification of actual mineral types. Clearly, either more detailed, hyperspectral, data or an extensive, and costly, ground data collection effort would be required.
     Discussions about the use of hyperspectral technology spurred the interest of the Kazakstani scientists in the Ministry of Science who were interested in learning more about this new advance in remote sensing technology. The interest of the Kazakstan government in doing a remote sensing mission over the Polygon for mineral exploration and in learning more about remote sensing technology led to the involvement of the DOE and ultimately to an invitation to the DOE to fly their AMPS (Airborne Multispectral Pod System) aircraft with the hyperspectral ESSI Probe 1 over the Polygon and other select sites in Kazakstan.
     The AMPS pods carried by the P-3 include a number of remote sensing instruments. However, the Probe 1 was the only instrument on-board the AMPS P-3 aircraft to provide coverage in the 1 to 2.5 µm (ShortWave Infrared or SWIR) region. Thus, it provided unique information both to ESSI in their search for minerals and to other DOE researchers who are studying the Kazakstan mission data.
     The Kazakstan mission took place in late June 1997. Flights were conducted out of the capitol city of Almaty and out of Semipalatinsk in the northeast. Unfortunately, engine trouble with the P-3 cut the mission short. However, a great deal of data was collected before that mishap. While analysis of that data is still on-going, initial results are promising and show the advantage of the hyperspectral Probe 1 data over the multispectral TM data. While the biggest advantage is the Probe 1's high spectral resolution, the Probe 1 data has a greater spatial resolution as well. Figure 2 shows part of a Probe 1 flightline and the corresponding area from the TM data. The relatively small segment of the TM data has been expanded three times to approximately match the scene dimensions of the Probe 1 data. The ground sample distance for this portion of the Probe 1 data is approximately 10 m, three times finer than that of the TM.

Finding Minerals in Hyperspectral Imagery
The areas overflown by the Probe 1 were identified in the TM data as having surface materials with absorptions in TM band 7 (2.08 - 2.35 µm). Again, from the TM data it is impossible to say with any certainty what minerals could be causing those absorptions; however, using the Probe 1 data these minerals can be identified. Minerals and other surface materials with unique spectral signatures were automatically detected in the Probe 1 data using proprietary software developed by ASIT. These "endmember" spectra were then extracted directly from uncorrected Probe 1 data (i.e., no atmospheric correction was applied). Using field reflectance spectra collected during the June 1997 mission, the endmember spectra were corrected to apparent surface reflectance and identified on the basis of diagnostic spectral features. Using the image pixel spectra as target signatures, these endmember materials were then mapped using ASIT abundance mapping software and then, using a series of abundance images, were classified on the basis of some threshold abundance.
     An example of the stages in this process is shown in Figure 3. In Figure 3a, a color composite of a segment of Probe 1 data is shown. In Figure 3b, a gray level abundance image of just one of the endmembers, in this case the mineral calcite, is shown. Finally in Figure 3c, a classification map derived from multiple abundance images is shown.
     The abundance images and classification maps derived from this process will be used by field geologists to go directly to the most mineralogically promising areas to collect samples for assays of precious metal content. The mineral maps resulting from analysis of the Probe 1 data can also be used to aid in the construction of more detailed geologic maps which will be used in conjunction with existing maps to better understand the area's structural geology and thus provide some hint of subsurface trends.

The Future
The June 1997 mission to Kazakstan was just the first in an on-going campaign of exploration. More are planned. There are also more missions for the Probe 1 on the horizon. A second instrument of the Probe 1 design is being completed by ISPL and will soon enter onto operation. Major mining companies are taking a hard look at hyperspectral remote sensing in general and the Probe 1 specifically. ESSI recently signed a letter of intent with Noranda Inc., a $10 billion per year Canadian company, to use the Probe 1 on mineral prospects around the world. Through this agreement and other projects, the added value of hyperspectral remote sensing should be demonstrated time and again not only over mineral prospects but for other applications as well such as mine waste remediation, forestry, fish and range management, and agriculture.
     While many saw the fall of the Soviet Union as a coming of chaos, out of the chaos arose opportunity. A very serendipitous opportunity was the June 1997 mission to Kazakstan and the opening it provided for ESSI to put its new Probe 1 hyperspectral sensor to use and to demonstrate, on an international scale, the fact that hyperspectral technology can be successfully used for commercial applications.

About the Author:
William Farrand has a Ph.D. in the Geosciences with a minor in Remote Sensing from the University of Arizona. He has been working with data acquired by airborne hyperspectral imaging systems for over eight years. He may be reached at [email protected] or through ASIT at 410-729-3108.

Back