Because all materials are formed by chemical bonds, solids, liquids, and gases that pervade the Earth can be detected by the use of imaging spectroscopy. In fact, data collected and analyzed from implementing this technology are being applied to produce surface maps for mineral exploration, vegetation discovery, and land-management studies.

Th Vsd51579 23

AIRBORNE imaging spectrometer MAPS EARTH`S SURFACE

By David Wilson

Because all materials are formed by chemical bonds, solids, liquids, and gases that pervade the Earth can be detected by the use of imaging spectroscopy. In fact, data collected and analyzed from implementing this technology are being applied to produce surface maps for mineral exploration, vegetation discovery, and land-management studies.

Detection of a particular material depends mainly on spectral coverage, spectral resolution, and the signal-

to-noise ratio of the spectrometer used. Other parameters include the abundance of the surface material and the strength of absorption features in the wavelength region measured. For aircraft-borne applications at high altitudes, a spectrometer can quickly image large surface areas (around 2 sq km/s) and generate large amounts of data that can be analyzed for specific absorption bands and, therefore, specific materials.

The US Geological Survey (USGS; Denver CO) is using such an imaging spectrometer--the NASA/ Jet Propulsion Laboratory (JPL) Airborne Visual and Infrared Imaging Spectrometer (AVIRIS) system--to collect surface mineral data on several western states (see figure on p. 41). The spectral-range capabilities of the AVIRIS system allow it to detect minerals, vegetation, and vibration absorptions due to their inherent lightweight elements. Vibration absorptions from heavier elements such as quartz, however, occur in the mid-infrared spectral range and are not detected.

The data obtained by this imaging spectrometer contain the solar flux spectrum, the Earth`s atmospheric absorption bands, and the Earth`s surface reflectance. The data are calibrated by using a combination radiate transfer-atmospheric transmission model. Data verification is obtained by measuring the reflectance of relatively large homogeneous areas on the ground with a field spectrometer.

Actual samples from the observed site are also measured on a laboratory spectrometer to confirm the spectral details of, and to provide information for, removing artifacts from the field data. Spectra from the imaging spectrometer data are then compared with the field and laboratory data. Then, a set of correction multipliers and offsets are derived and applied to the data (see Fig. 1).

The USGA is currently using two 0.4- to 2.5-µm field spectrometers to measure samples and provide for in situ calibration of surface reflectance for terrestrial imaging-spectroscopy studies. These instruments also serve as laboratory spectrometers when rapid spectra are needed. For standards calibration in the field, the USGS usually obtains hundreds of spectra at a site and averages them to provide a spectrum similar to what the remote-sensing instruments would measure.

Imaging-spectrometer analysis of the data produces a several-hundred-layer geographical information survey database. The various layers are assembled into materials maps that are verified in the field to ensure that the data have been analyzed correctly. This process can take several weeks to verify the data results and for researchers to study the mapped complexities.

Imaging analysis

Although many groups around the country are developing methods for analyzing imaging-spectrometer data, the USGS uses a Tricorder algorithm that can analyze and determine hundreds (or thousands) of materials simultaneously (see Fig. 2). This software tool can help resolve which materials are present on the Earth`s surface, whether they be minerals, environmental contaminants, dryland, or water.

Before applying the Tricorder software, the data obtained by the AVIRIS system, while flown in aircraft at altitudes to 65,000 ft, are grouped in 10.5-km-wide by 17.5-km-long segments (614 ¥ 1024 pixels). These segments cover 184 sq km, which is called a double segment, and occupy as much as 598 Mbytes of disk space. Prior to analysis, the data are processed into an acceptable storage-tape format. Next, the data tapes from the aircraft are sent to the JPL, where the tapes are translated and data-written onto other tape drives. These tapes are then returned and loaded into the USGS system.

The USGS computer system comprises a server that works in an Ethernet network and is connected to many desktop X-terminals. The peripherals in the system, such as the tape and CD-ROM drives, are connected into the system via SCSI interfaces. Currently, the USGS is using a Hewlett-Packard HP9000 K250 server and 672 Mbytes of RAM, five read/write optical disks, consisting of three 1.3-Gbyte drives and two 2.6-Gbyte drives, three CD-ROM drives, several tape drives, and 83 Gbytes of hard-disk space. The laboratory spectrometers feed data in real time into the computer system via RS-232 lines. The spectra data from the spectrometers are analyzed and then placed into a spectral library for further analysis by the Tricorder algorithm.

"A high-speed interface is not required to hook the spectrometers into the system because they aren`t generating data that fast," says Roger Clark at the USGS. The X-terminal configuration was selected because the main system data flow from the disks to the CPU (for analysis) and then back to the disks. "If the data were on a remote system so that all the data had to go over the network, then the system would roll over and die," says Clark. In the USGS setup, on the other hand, the data flow is kept on the server`s fast internal bus, and traffic is kept to a manageable level. "From the user`s perspective, the main data flow is coming to the screen, and that flow is several orders of magnitude less than the amount of data being analyzed while moving from the CPU to disk and back," he adds.

Using their desktop X-terminals, USGS scientists can monitor spectrometer functions and data from their offices or at remote locations. The USGS maintains tens of gigabytes of data, often analyzing gigabytes at a time.

Running on the HP computer, the Tricorder software identifies conditions in a spectrum, and then, based on the results found, applies other algorithms to the data. Because many different spectral signatures exist in a single scene, the USGS uses a spectral-identification algorithm as a first analysis step. "Spectral identification is the main algorithm that we use; there are others that perform functions such as edge detection, but most of them support the spectral feature algorithm. That`s the core of what we are doing," says Clark.

Fitting algorithms

A new fitting algorithm was developed by the USGS in 1991 to aid the spectrum-classification process. The algorithm performs a fit of each spectral feature in the spectral library to the spectrum undergoing analysis. Multiple features from multiple materials are compared to determine which material has the closest match to the data under analysis. The algorithm does not force a positive match, which makes it different from other algorithms in use. In operation, for example, it would attempt to map only minerals included in the reference database.

The HP UNIX-based computer system at the USGS delivers more than 130 MFLOPS and employs a 965-Mbyte/s backplane. According to the USGS, that data delivery is about four times faster than a Sun Microsystems` Ultra II 200-MHz system while running the analysis software. This high data speed allows the USGS to analyze about 300 materials in a 614 ¥ 1024-pixel imaging-spectroscopy (NASA AVIRIS) data set in about one hour.

The accuracy of the data produced is determined by comparing the results of the spectroscopy analysis with the results obtained by studying hand samples from the field using laboratory x-ray diffraction analysis. Basic verification, however, can take a few weeks, depending on how much of the mapping area needs to be checked. In many cases, verification identifies additional minerals that were not included in the original imaging spectroscopy analysis. Therefore, a second round of analysis and verification may be required.

DAVID WILSON is a science writer in London, England.

Th Vsd51579 23
Click here to enlarge image

The objective of the Airborne Visible Infrared Imaging Spectrometer (AVIRIS) is to identify, measure, and monitor constituents of the Earth`s surface and atmosphere based on molecular absorption and particle scattering signatures. The instrument uses a scanning mirror to sweep back and forth in whisk-broom fashion, producing 614 pixels for the 224 detectors in the spectrometer. Delivering wavelengths from 400 to 2500 nm, the instrument flies aboard a NASA ER-2 (a modified U2 airplane) 20 km above sea level at about 730 km/h.

Th Vsd51579 24
Click here to enlarge image

FIGURE 1. Mapping data obtained by the Airborne Visual and Infrared Imaging Spectrometer is shown for the Summitville mining district and the adjacent San Luis Valley in Colorado. A combined method of radiate transfer modeling and empirical ground calibration site reflectance were used to correct the flight data to surface reflectance. This method corrects for variable water vapor in the atmosphere and produces smooth spectra with spectral channel-to-channel noise approaching the signal to noise of the raw data. Therefore, the data can be compared to standard laboratory measurements.

Th Vsd51579 25
Click here to enlarge image

FIGURE 2. Mapping data obtained by the AVIRIS show the mineral distribution of the waste rock and tailing piles at the California Gulch Superfund Site near Leadville, CO. Each color identifies iron-bearing minerals in each 17 ¥ 17-sq m area (pixel) on the ground. Blue areas show minerals high in dissolved metals such as cadmium, zinc, and lead. Areas in green have minerals that are more neutral but are still of concern. Other colors are minerals not contributing to water contamination. No iron-bearing minerals were found in areas shown in black.

More in Cameras & Accessories