EO-1 Satellite Imagery

The NASA Earth Observing-1 (EO-1) satellite is a multispectral/hyperspectral imaging instrument that is extremely capable but less well known than its older cousin: Landsat 7. EO-1 was launched on November 21, 2000 from Vandenberg Air Force Base, California. The satellite is a testbed platform for testing new multispectral and hyperspectral hardware concepts. The idea was to pioneer advanced technologies designed to give the satellite and its sensors increased capability at a low cost.

The satellite tracks the same orbit as Landsat-7, i.e. a polar sun-synchronous orbit at a height of 705 km, lagging 1 minute behind Landsat 7. EO-1 is fitted with the Advanced Land Imaging (ALI) multispectral and Hyperion hyperspectral imaging instruments with advanced capabilities such as the elimination atmospheric effects and sensor calibration. The Hyperion hyperspectral sensor acquires digital satellite imagery at a resolution of 30m in 220 visible and IR bands. The EO-1 ALI sensor acquires 9 multispectral bands at the same resolution, plus a panchromatic band at a resolution of 10m. Upon successful termination of its 1-year orbital mission, NASA decided to prolong its orbital lifetime indefinitely. It continues to operate as of this writing.

The EO-1 ALI instruments sample a wavelength spectrum that ranges from 0.43 to 2.40 microns. Its multispectral bands resemble those of Landsat in many respects. Hyperion is a hyperspectral sensor, meaning that its bands continuously span the same spectrum. The characteristics of the two sensor suites are summarized in the table at the upper right.

The Advanced Land Imager was developed at Lincoln Laboratory under the sponsorship of the National Aeronautics and Space Administration (NASA) and their New Millennium Program. The purpose of ALI was to validate new technologies that could be utilized in future Landsat satellites to realize economies of mass, size, power consumption and cost, and in improved instrument sensitivity and panchromatic image resolution.

ALI employs a fixed planar array of detectors operating in push-broom mode, replacing the mechanically scanned Landsat-type linear array. This solid-state array is expected to be more reliable than the current Landsat technology. (Recall that the scan line corrector that failed in 2003 was a part of the Landsat mechanical scanning system.) The planar detector array is coupled to an optical system with a 15° field of view that covers the full swath width of a typical Landsat image (185 km). Unfortunately the focal plane was only partially populated to reduce cost, providing a 3° cross-track coverage that corresponds to a 37 km swath width on the ground. The result is an image width that is only one fifth as wide as the 185km Landsat swath. Image length is either 42km or the full 185 km Landsat length. Typically four swaths are mosaiced and bundled with ALI data sets, resulting in an effective swath width of 148km.

EO-1 ALI data is offered free of charge at the USGS Earth Explorer and GLOVIS websites. The data may be accessed by selecting the EO-1 ALI check box and searching the UI map. (I found the GLOVIS site the better option as you can search by Landsat path and row numbers.) Although about 60,000 ALI scenes have been captured and archived, global coverage is far from complete. Some regions (for example populated areas of North America and Europe) are fairly well covered with multiple scenes of the same location. Other areas are covered very sparsely if at all. This is a major and serious drawback for this data set. As an example the image to the right shows the available coverage for the entire USA state of Utah. Global coverage as of 2005 is shown on the map to the right. According to the USGS EO-1 webpage: "On June 15, 2009, the USGS Earth Resources Observation and Science (EROS) Center will begin accepting EO-1 Data Acquisitions Requests (DAR) at no cost. The DAR forms can be found on the EO-1 Web page and on EarthExplorer. The EO-1 images not already in the archive will need to be requested by the customer at no cost using the DAR form. Once the data has been acquired the customer will be notified by email with image access information."

I found an EO-1 DAR form at the EROS Website. The form is straightforward and even has an embedded Google map as an aid in specifying the acquisition coordinates.

EO-1 ALI data can be used and processed similar to Landsat with a few differences. The first major difference is the dynamic range of the ALI band data: 16 bits versus 8 bits for Landsat. In order to be handled by standard desktop computing equipment, the 16 bit data must be scaled to accommodate an 8-bit grayscale or RGB color model. PANCROMA handles this the same way it handles 11 bit Digital Globe data, by scaling by a ratio of 255/MAX_VALUE where MAX_VALUE is the largest DN value in the image. Alternatively every DN can be divided by successive multiples of two (bit shift). The former is the PANCROMA default while the latter can be accomplished by checking the 'Bit Shift Enable' check box on the PANCROMA Image Processing Data Input screen. In either case, after extracting all of the band files from the .tar.gz file, each band must be loaded ('File' | 'Open' | 'Display One File' | 'Display One Grayscale Image'), scaled ('Activate Image Processing Routines' check box or alternatively accept default method), and then saved to disk after scaling.

The second difference is the number and arrangement of the band files. Landsat has 8-30m multispectral bands plus a 15m panchromatic band. ALI has 9-30m multispectral bands plus a 10m panchromatic band. As a result, bands 3, 4 and 5 correspond to Landsat bands 1, 2 and 3 while ALI band 1 (panchromatic) corresponds to Landsat band 8.

The EO-1 ALI multispectral band rows and columns are not even multiples of the panchromatic band's values. As a result, all of the band files (both multispectral and panchromatic) must be decreased by exactly one row and one column. You can accomplish this using the PANCROMA 'Pre Process' | 'Rescale' | 'Rescale Three Images' and 'Pre Process' | 'Rescale' | 'Rescale Single Image' selections.

The metadata is included in a file that has the suffix '...MTL_L1T.TIF'. Note that this file is NOT a GeoTiff file, despite the .TIF suffix, but instead is an ordinary text file. It will be necessary to rename to file with a more appropriate suffix (.txt) for many text editors to open it properly.

After resizing, you can pan sharpen the multispectal band files as usual. Since the panchromatic band is three times the resolution of the multispectral bands, the Laplacian interpolation will be the default and it will be necessary to use the appropriate number of iterations (at least 40) to process the hue and saturation images. Note that since the panchromatic band spectral range exactly matches the combined range of the RGB bands, NIR processing techniques such as XIONG and ENHG are not generally helpful. The image to the right shows an RGB color composite from bands 5, 4 and 3 (Equivalent to Landsat 3, 2, 1). Following it is the corresponding pan sharpened image. Pan sharpening Landsat multispectral band data with the ALI panchromatic band should not offer any big problems and of course would offer increased resolution as compared to the native 15m Landsat panchromatic band. The final image is a composite made from bands 6, 5 and 4 (equivalent to Landsat 4, 3, 2). The band files have been stretched for improved contrast.

EO-1 ALI data is a wonderful addition to the global multispectral database. The increased resolution of the panchromatic band and other improvements make it a very useful data set. The only major downside is the lack of anything approaching complete global coverage. Since the satellite is approaching the end of its useful life, attaining this goal is unlikely.