After more than a decade of development and refinement, hyperspectral remote sensing technology is ready to take a larger role in the commercial remote sensing arena. Following the growing demand for data of high spatial resolution, hyperspectral analysis is poised to become a much more common tool in the remote sensing analyst's toolkit.
      Spatial or shape attributes of objects can be used to interpret what objects are. Spectral attributes aid in determining what the objects are made of. As such, the hyperspectral data represents a corroborating suite of diagnostic signatures to greatly increase the probability of accurate interpretation. As spatial resolution of satellite imagery increases, so does the need for higher spectral resolution - this is the domain of hyperspectral imagery.
      The largest impediment to the commercial acceptance of hyperspectral imaging is the present lack of satellite-borne sensors. Currently, commercially available hyperspectral data is acquired by sensors mounted on aircraft. Some sensors can only be flown on specialized aircraft at a cost upwards of $50,000 per flight. As they may conflict with other uses of the aircraft, flight schedules for some of these instruments have multi-year lead times. These stumbling blocks to wider acceptance of hyperspectral imagery will be gradually removed over the next few years as several satellites bearing hyperspectral sensors are launched.
      Some user's needs already met by the existing airborne sensors, such as the NASA/JPL AVIRIS; the ASTER Simulator from GER Corp. of Millbrook, NY; ITRES Instruments' (Calgary) casi; HYDICE of the Naval Research Laboratory, and imaging spectrometers from TRW (Redondo Beach, Calif.) and Science Applications International (San Diego).
      These airborne instruments are a solution for users that do not need new data urgently, or who have the resources to purchase an instrument and acquire their own data.

Applications of Hyperspectral Data
Hyperspectral instruments detect tens to hundreds of separate, narrow spectral bands. The NASA/Jet Propulsion Laboratory AVIRIS (Airborne Visible Infra-Red Imaging Spectrometer) instrument, flown on the ER-2 aircraft, acquires images with 20m spatial resolution in 224 bands, each 10 nm wide, over the spectrum from approximately 0.4 to 2.5 µm. In contrast, the multispectral Thematic Mapper (TM) sensor carried by the Landsat 5 satellite acquires imagery from 0.45 to 12.4 µm, in seven bands 60 to 270 nm wide. Six of the seven TM bands have 30m spatial resolution.
      Whenever a high level of detail of the composition of remote objects on the Earth's surface is required, such as environmental site characterization where a quantitative knowledge of contaminant distribution and/or surface properties is necessary, hyperspectral image analysis is appropriate. Hyperspectral image analysis is also well-suited to natural resource management, exploration and extraction, where the principal objective is to identify, map or produce specific minerals or rock types. In the case of petroleum exploration, the goal might be to map surface distributions of rocks with certain characteristics (e.g., organic content, porosity, etc.) and then tie that information with subsurface data. Forestry, agricultural and ecosystems analyses - where differentiations between species and levels of plant health within species are made - also can benefit from the application of hyperspectral image analysis.
      The utility of hyperspectral analysis extends far beyond the realm of the earth sciences. In military and intelligence applications, hyperspectral imagery can add confidence to interpretations based on very high spatial resolution data. A commander using hyperspectral data to plan a military operation has the potential to more accurately assess target identity, attributes and defenses than one using more conventional forms of remote sensing. This elevated confidence translates into reduced decision risk, or an increased chance of making the correct decision. In a military action, this may result in lives saved or goals achieved. The following scenario highlights how reduced decision risk can have monetary value.

The Case for Hyperspectral Imaging
The value of reduced decision risk
Consider a proposed mining project with a cost of $55 million and a net present value (NPV) of $100 million at certainty. We assume that field work, coring, chemical analysis and multispectral remote sensing analysis will result in a finding that the proposed activity will be economically viable, with 50 percent confidence in the reliability of the analysis. Further, we assume that adding hyperspectral image analysis and field spectrometer measurements to the suite of test information increases the study's reliability to 60 percent. Without hyperspectral analysis, the risk-weighted NPV of the project is $50 million. With hyperspectral analysis, the risked NPV is $60 million. Without the hyperspectral analysis, investing in this project is not a sound business decision; the cost of the project exceeds its risk-weighted NPV. With the hyperspectral analysis as one more tool available to the organization, this project could be authorized. To make this example more realistic, consider the scenario where two mining companies are concurrently evaluating the same area. The company that incorporates hyperspectral analysis into its prospect evaluation will be better positioned to act more confidently and aggressively than their competitor using conventional methods.

The Synergy of Hyperspectral Imaging in Conjunction with Other Analyses
Hyperspectral imaging is by no means a total remote sensing solution. In fact, the greatest power of hyperspectral analyses are realized when they are combined with other forms of imagery and analytical techniques. A very clear picture of an area of interest can be created by combining hyperspectral imagery with high resolution spatial images, spectral libraries and field analyses.
      Hyperspectral imagery leverages investments made in field measurements and sampling programs. Furthermore, the scale and cost of such field programs can often be reduced by using imagery to extend the area of an evaluation. The optimum way to capture these savings is to use imagery during the field program. By targeting fieldwork to those areas that are least clearly defined by image analysis alone, the understanding of the area is maximized while the time and money spent on the ground is minimized.
      Data fusion techniques allow the combination of data from different sensors to create imagery with very high spatial and spectral resolution. The spatial resolution of space-based sensors has been advancing at at rapid pace. The High-Resolution Visible imaging system carried by the SPOT satellites acquires panchromatic images with 10m pixel size. The Russian KVR-1000 camera flown aboard Kosmos spacecraft has 2-3m resolution, depending on flight conditions. By 1996, Worldview Imaging (Livermore, Calif.) plans to have two satellites in orbit carrying instruments with spatial resolution comparable to the Russian system. Space Imaging Inc. (Sunnyvale, Calif.) plans to launch a satellite in 1997 that will carry a sensor to acquire 1m resolution panchromatic and 4m resolution multispectral imagery.
      Multispectral imagery, such as that acquired by instruments on the JERS-1, IRS and Landsat satellites, is also very complementary to hyperspectral data. Though multispectral instruments have lower spectral resolution than hyperspectral sensors, many have a wider total spectral range. For example, while AVIRIS has a spectral range of approximately 2.1 µm, the range of the TM instrument on Landsat 5 spans nearly 12 µm, from 0.45 to 12.4 µm. Band 6, the longest wavelength TM channel, acquires data in the thermal infrared range of the spectrum. Thermal infrared data is useful for emissivity analysis, such as evaluations of the heat discharged by factories and power plants. Hyperspectral instruments do not currently support this range of the spectrum.
      Today, hyperspectral data is only acquired via airborne platforms where a wide variety of conditions, such as gusty winds, can interfere with airborne data acquisition that have no effect on satellite operation. The wider temporal and spatial coverage of these satellite-borne multispectral sensors can be used to fill in the gaps left by airborne hyperspectral acquisition programs. Furthermore, each airborne mission carries a significant incremental cost, while satellites, once launched, have lower operating costs on a per-hour basis.

The Bright Future of Hyperspectral Imaging
The upcoming launch of several satellite-borne hyperspectral instruments promises to greatly expand the practical uses of hyperspectral technology. Satellite systems promise wider data availability, with acquisition not limited by concerns about flight duration, the remoteness of the terrain, or political and military conditions. The Strategic Defense Initiative Mid-Course Space Experiment is slated for 1995 launch, carrying a hyperspectral tool. A tentative 1997 launch is planned for the University of Colorado/Department of Energy HIRIS instrument, for which AVIRIS is a prototype. NASA is scheduling the launch of two hyperspectral sensors, one in 1996, the other in 1998.
      Satellites are more stable platforms for data collection than airplanes. Even with gyroscopic stabilization, planes are subject to pitch, roll and altitude variation that affect data acquisition. For this reason, satellite data has a more constant spatial resolution and tends to be easier to rectify (for a given spatial resolution) than airborne data.
      The predictable, inevitable progression of remote sensing technology is that advances in instruments and image processing software will allow us to investigate areas of interest with ever-increasing detail. Heightened image resolution is a one-way street. Once remote sensing analysts become accustomed to the interpretive power of state-of-the-art technology, they won't settle for less. Hyperspectral remote sensing is one part of an inevitable technology progression. As the use of high spatial resolution data becomes more commonplace, the demand for increased spectral information content will accelerate. At present, hyperspectral imagery is acquired only by airborne instruments. While airborne acquisition fills the needs of some organizations, it excludes a host of other potential users. The impending launch of several satellite-borne hyperspectral instruments promises to finally bring this technology into the commercial mainstream.

About the Author:
David M. Uhlir, Ph.D. works for Research Systems Inc. (Boulder, Colo.) as product manager for the remote sensing image processing application, ENVI. He can be reached by phone at (303) 786-9900 or by email to [email protected].

For more information on the use of field measurements with hyperspectral imagery, see "Ground Truthing Spectrometers Unlock the Value of Multispectral Imaging" by Stephen Parsons, Earth Observation Magazine, Sept. 1994, pp. 36-39.

Back