SATELLITE IMAGES—SEEING THINGS DIFFERENTLY

If the growth of commercial aviation compelled Hal Shelton to paint natural-color maps, going higher still, the advent of space flight largely defines the genre today as we settle into the digital era. The digital solutions offered in this article for making natural-color maps employ raster land cover data derived from satellite images. But more than just a technical means to an end, space imagery has forever changed how people visualize Earth in a realistic context from afar. The aptly named “Blue Marble” photograph taken by the Apollo 17 crew heading to the moon in 1972 gave us the first full view of oasis Earth set against the emptiness of outer space (Figure 6). Embraced as a symbol of environmental awareness, the “Blue Marble” has become one of the world’s most recognized images. And as a natural-color geographical image with pop culture cachet, it is a rarity.

Figure 6. NASA’s “Blue Marble” photograph shows Earth from a distance of nearly 48,000 kilometers (30,000 miles). Dominated by the Sahara and Kalahari deserts, Africa is usually the most cloud-free continent. Antarctica is also visible in this image for the first time (NASA, 2002)


Starting with the launch of Landsat (originally called ERTS—Earth Resources Technology Satellite) in 1972, remotely sensed images of Earth in both natural and false colors have become increasingly abundant with each passing year. The sheer volume of remotely sensed data collected by a multitude of platforms and sensors is difficult to comprehend. A Google search in early 2004 using the key words “satellite images” yields 4,810,000 hits, a coarse, but nevertheless telling barometer of current popularity. Satellite images have also replaced maps as the dominant method of depicting natural color on geospatial products. However, as we shall discuss shortly, quantity and quality are not always directly related. Off-the-shelf satellite images are less than ideal for the presentation of natural color on maps.

Intended primarily for scientific, military, and economic use, satellite images find secondary (and sometimes gratuitous) application in television and print news, posters, coffee table books, and even silk screened on coffee mugs and t-shirts. That atlases sometimes use satellite images and space shots of Earth as cover art points to their marketing potential. These colorful images attract a buyer’s attention and deliver a subliminal message that the contents within are scientifically based, accurate, environmentally concerned, and global in scope. Perhaps for similar reasons, a satellite image appears on the cover of the Spring 2002 issue of Cartographic Perspectives. Satellite images also serve as fine art. The traveling exhibit “Our Earth as Art” prepared by the Library of Congress, NASA, and USGS is currently showing in museums and other public venues. The 41 images in the exhibit (selected from 400,000 taken since 1999) treat audiences to eye-catching views of Earth from above that appear both surreal and abstract (Stenger, 2002). Trying to identify the images without reading the captions is a pleasurable challenge.

Several popular products containing global satellite data in natural color are available today. “The Living Earth” is one of the older and better-known commercial images (Figure 7, left). The original image derives from 1.1-kilometer-resolution AVHRR (Advanced Very High Resolution Radiometer) data, which NASA began collecting in 1978. While technically sophisticated when first introduced, AVHRR products are beginning to show their age. Contrary to what the name suggests, AVHRR images, judged by contemporary standards, appear coarse, with too much contrast, and contain tonal variations seemingly unrelated to land cover and vegetation. The Living Earth, Inc. now sells updated and improved natural-color images made from Landsat 5 and 7 data. They also have other promising products under development.

Figure 7. (left) “The Living Earth.” (right) NASA’s new “Blue Marble.”

Another notable product is NASA’s new “Blue Marble,” named after the famous photograph taken by the Apollo 17 crew. The “Blue Marble” derives from MODIS (Moderate Resolution Imaging Spectro-radiometer) data collected at 500-meter resolution and distributed to the public at one-kilometer resolution (Figure 7, right). The “Blue Marble” portrays arid regions with convincing natural colors but is less effective in humid regions, which appear as monotonous green tones. Meteorological interference in the form of snow cover and sea ice in the polar regions is visible, and areas of shallow water appear to have been arbitrarily added and appear discordant with rest of the image.

The “Blue Marble” comes pre-composited with shaded relief generated from GTOPO30 (Global 30-Arc-Second Elevation Data Set) that employs standard cartographic illumination from the northwest. “The Living Earth” image discussed previously also contains shaded relief made from GTOPO30. While the inclusion of shaded relief is a step in the right direction, its legibility diminishes at scales smaller than 100 percent of original size, and, with more significant reduction, disappears entirely. Shaded relief is not nearly as tolerant of reduction as the underlying color tones found in satellite images. While colors tend to blend pleasingly into one another when reduced, shaded relief becomes an unattractive dark smudge. These shortcomings aside, the “Blue Marble” is a remarkable dataset and is invaluable for making 3D visualizations and animations. It is also free.

Pictures vs. maps

Based on the popularity of satellite images, many of which appear in natural color, it would appear that Shelton’s idea about depicting Earth more realistically has finally taken hold. Well, not quite. The graphical merging of satellite images and maps, two closely related geospatial products, has not happened to any great extent. It is a rare map that uses a satellite image as a raster base in conjunction with other map information, such as labels, lines, area tones, and point symbols. In the cartographic context, satellite images are perhaps most common in atlases as stand-alone thematic insets, decorative introductory pages, and the aforementioned cover art.

The lack of satellite images used as cartographic backdrops, in the same manner as shaded relief or hypsometric tints, has a simple explanation: they are not maps. Remotely sensed images contain traits that are incompatible with basic cartographic design conventions (Figure 8). Dense, detailed, colorful, and contrasting, satellite images tend to dominate all other classes of map information to the point of illegibility. Just as photographic snapshots often contain undesirable visual elements—a jet contrail in the sky above the Grand Canyon—satellite images are similarly afflicted. For example, clear-cut forests in the US Pacific Northwest appear on satellite images as a mottled pattern that distracts from everything else.

Figure 8. A natural-color Landsat image of the Grand Canyon made from bands 2, 4, and 7. Even the handsomest satellite images contain graphical elements inconsistent with cartographic design goals. Courtesy of the USGS.

Other common problems with satellite images include:

• Meteorological interference — Because clouds on average cover 64 percent of Earth’s surface (54 percent of land areas) at any given time, the odds of finding satellite images completely free of clouds and their shadows are slim at best (Warren, 1995). Even one small cloud on an image requires a cartographer to make a difficult choice—an exercise in cartographic situational ethics. Is it best just to leave the unsightly blemish on the image, or is it proper to quickly remove it with the Clone Stamp (Rubber Stamp) tool in Photoshop? After all, who would ever notice or object? Other meteorological interference encountered on satellite images includes snow-covered ground, frozen water bodies, smog, smoke plumes from wildfires, and lowland flooding. Such undesirable traits plague many of the satellite images available online for free, which are provided by organizations that monitor the environment and natural disasters. When using satellite images as backdrops on maps, boring is better.
  
  
• Inappropriate color variations — It is typical for natural color satellite images to contain at least some colors that are decidedly unnatural. Despite appearances, natural-color satellite images are not truly photorealistic. They typically consist of “bands” of electromagnetic radiation data from outside the visible spectrum. By inserting these data, in the form of 8-bit grayscale images, into the red, green, and blue channels of, say, a TIF image, the results are a concocted scene with ersatz natural colors—if the mixing and matching of data and colors are done correctly. The inappropriate color variations flagged in Figure 8 are in an area of uniform desert shrub. The cause of these pronounced variations, which are invisible to humans on the ground, might be differing mineral or moisture content in the exposed soil—information only of interest to a few people in specialized fields. Regardless of what the colors represent, they are too prominent compared to other colors on the image and do not belong on a general-purpose map.
  
  
• Relief inversion — The southeast lighting (lower right) found on most high-resolution satellite images when north-oriented causes an optical illusion known as relief inversion, whereby mountains appear as valleys, and vice versa. Because being able to tell up from down is so fundamental, the presence of relief inversion is completely unacceptable on maps. Removing the embedded shadows that cause relief inversion on satellite images is technically difficult and yields poor results, especially in areas of high, sharp relief where shadows are entirely black. Substituting neutral colors and textures to replace the shadows, whether by automated means or the Clone Stamp tool in Photoshop, is an inelegant solution at best. On another tack, overprinting the satellite image with shaded relief generated from a DEM (digital elevation model) using standard northwest illumination only serves to flatten and darken the topography. The shadows, based on opposing light sources, cancel each other out.
  

Land cover—cartographically friendly data

In contrast to satellite images, raster land cover data derived from satellite images is well suited for making natural-color maps. Raster land cover data with its classified structure and generalization brings order to the tonal anarchy of satellite images. The product of sophisticated image processing techniques and often multiple data sources, raster land cover data is generally free of the problems that afflict raw satellite images, such as relief inversion. The result is reconstituted data that facilitates the production of natural-color maps. A satellite image distilled into a land cover classification of, for instance, a dozen categories is merely the raster equivalent of the polygons etched into zinc plates that guided Shelton’s painting. Shelton colored his maps with brushes and acrylic paints. We will use software and pixels. Although the means of production has changed, the process remains essentially the same today as it was a half century ago—a cartographic version of painting by the numbers.

Next, we will examine how to make natural color maps from two different types of raster land cover data. Both are in the public domain and available online for free. Our primary software for manipulating the data is Adobe Photoshop. We will begin with National Land Cover Dataset.