NASA Goddard Space Flight Center
|Figure 1. Nadir views of Mount Takahe,
Antarctica from (a) Landsat-7 ETM+, (b)
MODIS, using the ultra-low contrast
enhancement, and (c) Radarsat during IPY.
One goal of the International Polar Year is to learn more about the polar regions, but it may not be obvious to most people how we know as much as we do. These areas are remote and harsh, to say nothing of the obvious cold. Most people will never visit them, but for the early explorers and scientists, there was no alternative - to study them required that one traveled there, and survive. A very small percentage of scientists choose to take on these hardships for the sake of increasing our knowledge of these remote areas and scientific expeditions were costly and rare. These limitations led to the first two International Polar Years, in 1882-83 and 1927-28 and helped motivate planning for a third that grew into the International Geophysical Year in 1957-58.
Yet in just the last few decades our observations of these regions has gone from a data-poor environment to a data-rich environment. The key component of this reversal has been the increasing number of satellite sensors in near-polar orbit that silently and systematically gather data for scientists to pore over. No longer must all polar scientists travel to the frigid limits of the earth to contribute to our knowledge of its latitudinal limits.
The satellite era was ushered in as one of the grandest achievements of the International Geophysical Year (IGY) in 1957-58. The first satellites were more showpieces than scientific workhorses, but by 1960, the military was launching high resolution, for the day, cameras into space to collect classified intelligence [Richelson, 1990]. Fortunately, they also were engaged in mapping the polar regions. These early photographs are now declassified and when added to the now robust suite of ever-more-capable space-borne optical imagers make this visual record of the polar regions the longest satellite record in existence.
A much longer article is required to adequately cover the variety of operating sensors that image the polar regions. The variety includes differing combinations of spectral, spatial and temporal resolution. This report will focuses on the data from the Enhanced Thematic Mapper Plus (ETM+) carried on-board Landsat-7, the latest in this 35-year long satellite series, and contrast it with data from the MODerate resolution Imaging Spectrometer (MODIS) imager carried on board both the Terra and Aqua sensor platforms as well as synthetic aperture radar (SAR) imagery collected by the Radarsat sensor and satellite. These choices are made for two reasons: first, they represent many of the contrasting features of the existing suite of earth-imaging sensors; and second, because data from each of these sensors have been mosaiced (stitched together) into collections that cover the Antarctic ice sheet.
An unavoidable limitation to satellite remote sensing is the volume of data that can be transmitted back to receiving stations on the ground. ETM+ measures the surface reflectance within 8 spectral bands over a field of view 185 kilometers across at a relatively high resolution of 15, 30 or 60 meters (depending on spectral band). MODIS, on the other hand, measures 36 spectral bands, over a wider 2.3-km field of view, but with a spatial resolution of 250 or 500 meters (again, depending on spectral band). Radarsat provided its own radar illumination, at a single wavelength, collected data over a field of view of 100 km, but resolution is dependent upon processing parameters. A visual artifact of radar imaging, called “speckle”, due to coherent scattering, limits the effective resolution to about 100 meters.
The Antarctic mosaics created from data of these three sensors also vary. The Landsat Mosaic of Antarctica (LIMA) is the first benchmark data set created for the IPY [Bindschadler et al., in review]. It stitches together nearly 1100 individual Landsat scenes and recovers saturated pixels in one of the visible light spectral bands to produce the first true-color, high-resolution mosaic of Antarctica. The data can be explored and downloaded from http://lima.usgs.gov and educational materials about LIMA and how polar scientists use Landsat imagery is available at http://lima.nasa.gov. The MODIS Mosaic of Antarctica (MOA) was created from 260 scenes of a single spectral band by a process called data cumulation that increases spatial resolution by averaging multiple images together [Scambos et al., 1999]. MOA can be viewed and downloaded from http://nsidc.org/data/nsidc-0280.html. The Radarsat Antarctic Mapping Project (RAMP) created that mosaic from strips of Radarsat data [Jezek et al., 2002]. A description of the project can be found at http://nsidc.org/daac/ramp/ and the mosaic can be viewed through either the LIMA or MOA websites.
A good Antarctic feature to illustrate the differences in the characteristics of these three mosaics is Mount Takahe, an extinct volcano located at 76o17-S, 112o5-W, in West Antarctica. Figure 1 shows identical nadir views of the volcano. LIMA’s view is true color and shows how the mountain would actually appear. The sun’s illumination saturates the sensor’s view of much of the mountain’s sun facing slopes, but fine detail is seen throughout the remainder of the image. The MOA image of the 30-km wide mountain shows less detail and is illuminated from a different direction. The RAMP view is distinctly different, highlighting the tendency for radar reflections to amplify sharper edges. These can be exceedingly important to geologists: note, for example the outflow feature at the right edge of the crater.
A single feature does not highlight the strengths of each sensor, but makes a useful starting point. An important strength of the Landsat-7 ETM+ sensor used to construct LIMA is the high-spatial resolution. To illustrate this, Figure 2 zooms into the 3400-meter high crater of Mount Takahe. Many more details of the slopes and even within the non-active summit crater can be seen. The true-color rendering is lost in this sample because the data have also been enhanced to illustrate that the saturation on the sun-facing slopes is not as severe as it appeared in Figure 1. This type of enhancement is the digital equivalent of sunglasses: more detail of very bright areas can be seen at the expense of darkening less bright areas.
|Figure 3. Nadir view of the region around Mount
Takahe collected with MODIS sensor. Image taken from
MOA and enhanced with the high contrast stretch.
MOA’s strength lies in the increased radiometric resolution of the MODIS sensor. This is illustrated in Figure 3 where a broader view of the region around Mount Takahe is shown. MODIS measures the surface reflectance more precisely than ETM+, so that when the imagery is enhanced, very subtle details in the surface slope can be discerned. These patterns express the undulations of the surface as the ice flows over its rough bed (up to 2000 meters beneath the surface in this region) as well as elongated features that indicate the direction of flow. This information is valuable to glaciologists wanting to know how and where the ice is flowing.
Radarsat data’s emphasis on sharp edges is extremely useful in highlighting areas of crevasses, fractures in the ice that indicated large local stresses. This strength is amplified further by the fact that, unlike imagery using visible light wavelengths that show only the surface, radar energy penetrates the surface and even clouds. Thus, radar data show the surface despite cloud cover and can show crevasses buried beneath even thick layers of snow (if the snow is not wet). An illustration of this property is given by Figure 4 that capture the RAMP view of Whillans Ice Stream, a major outlet of ice flowing from West Antarctica into the Ross Ice Shelf. This view of this ice stream spans the 400-km length and shows details of its two branches coalescing into a single trunk interspersed with local patches of crevasses where the general flow field is disrupted.
|Figure 4. View of Whillans Ice Stream, West
Antarctica with Radarsat. Brightest areas are
crevasse zones at the margins of the ice stream and at
localized areas within the ice stream. Image size is
roughly 400 kilometers wide.
The storied exploits of Scott, Amundsen, Shackleton and others are nearly a century old, but exploration of Antarctica continues, driven, in large measure, by satellite imagery such as discussed here. In 1997, RAMP was the very first continental scale mosaic with spatial resolution better than one kilometer and its ability to show crevassed areas led to the discovery of new, large ice flows within the ice sheet. MOA arrived next, in 2005, and has been widely used by scientists to infer new information about ice flow. Most recently, the sensitivity of the imagery to extremely small changes in slope has led to the discovery of many new subglacial lakes, detected by the very smooth ice surface dictated by a subglacial lake. The presence of these lakes at the onset of fast ice flows suggests the subglacial water and/or the absence of basal friction over the lake play an important role in the initiation of fast flow. LIMA is very new, being released last fall, but already it has become very popular with scientists seeking the best maps of field areas and by the public who, for the first time can get a high-definition view of Antarctica. New discoveries are sure to follow.
Bindschadler, R.A., P.L. Vornberger, A.J. Fox, A.H. Fleming, B. Granneman, The Landsat Image Mosaic of Antarctica, International Journal of Remote Sensing, in review.
Jezek, K., and the RAMP Product Team, 2002. RAMP AMM-1 SAR Image Mosaic of Antarctica. Fairbanks, AK: Alaska SAR Facility, in association with the National Snow and Ice Data Center, Boulder, CO. Digital media.
Richelson, Jeffrey T.; America’s Secret Eyes in Space, Harper & Row Publishers Inc., 1990 – An in-depth look at the politics and history of America’s Keyhole satellite program
Scambos, T.A., Kvaran, G. and M.A. Fahnestock, 1999. Improving AVHRR resolution through data cumulation for mapping polar ice sheets, Remote Sensing Environment, Vol. 69, No. 1, pp. 56-66.