Hal Shelton Revisted: Designing and Producing Natural-Color Maps with Satellite Land Cover Data

Tom Patterson, US National Park Service
Nathaniel Vaughn Kelso, National Geographic

Published in Cartographic Perspectives (No. 47, Winter 2004), the journal of the North American Cartographic Information Society (NACIS).



MODIS VCF (Vegetation Continuous Fields) is the second type of land cover data that we examine. It consists of three data layers representing forest, herbaceous, and bare land cover. Although three land cover categories may seem scant, VCF data possesses unique qualities that are amenable to making generalized natural-color bases. Afterwards, color modifications and additional data are applicable to the VCF bases as needed.

Fuzzy data

VCF is the product of two organizations. Like the “Blue Marble” discussed earlier, it originates from the MODIS sensor aboard NASA’s Terra satellite platform, which traverses the entire Earth every one to two days in a polar orbit. The University of Maryland, Global Land Cover Facility created VCF from raw MODIS data collected by NASA in 2000 and 2001. The final 500-meter-resolution land cover data derives from monthly composites (they use seven bands of spectral information with emphasis given to bands 1, 2, and 7) processed to remove clouds and cloud shadows (Hansen et al., 2003).

Coverage includes all terrestrial areas of the planet except Antarctica and the polar fringes of Canada, Greenland, and Siberia north of 80 degrees latitude. VCF land cover layers for each continent (up to several hundred megabytes apiece) are downloadable for free from the University of Maryland website in either the Geographic or the Interrupted Goode Homolosine projections. If you plan on reprojecting these data, choose the Geographic projection, which is better suited for use with most GIS and cartographic software. MODIS VCF data layers, provided in BIL (Band Interleaved by Line) format, readily open in Photoshop or GIS software. Note: Photoshop only opens single-channel (i.e. grayscale) BIL files in “Raw” file format, so make sure to note the row and column dimensions (in pixels) prior to opening the file. Downloads also include metadata and projection information (see Appendix B).

Unlike hard categorical land cover data such as NLCD, MODIS VCF consists of a matrix of continuous tone values. For any given 500 x 500-meter sample of Earth’s surface, grayscale pixels represent the three land cover categories as percentages. Together they add up to 100 percent. For example, Figure 14 shows forest, herbaceous, and bare land cover for Africa loaded into the Red, Green, and Blue channels of an RGB image—a quintessentially scientific choice of colors. A sample selected from the relatively lush savannah of East Africa shows the content as 38 percent forest and 62 percent herbaceous. By comparison, a sample from the Sahara registers as 100 percent bare, as one would expect, given the extreme aridity of that region. Elsewhere in Africa the three land cover categories blend softly with one another much as vegetation does in nature. They also combine to form intermediate categories. Bare desert gradually transitions to semi-desert, semi-desert to herbaceous grassland, herbaceous grassland to savannah, and savannah to forest. Compared to categorical land cover data, this model better represents nature and Shelton’s painted art where there are few stark boundaries between vegetation types. 

Figure 14. Blended lands cover categories in MODIS VCF. The combined values for any sampled pixel on the map are 100 percent.

Given the global extent, 500-meter resolution, and general nature of MODIS VCF land cover data, this product is most appropriate for making natural-color maps at small and medium-scales. Next, we will make a natural-color map of North America. With diverse natural environments ranging from tropical rainforests to ice caps, North America is a rigorous test of the capacity of VCF for natural-color map design.

Using MODIS VCF in Photoshop

Having downloaded, decompressed, and, perhaps, reprojected VCF, you will next need to open it in Photoshop as a raw raster file from the File/Open dialog. Note: to open Eurasian VCF layers that are more than 42,000 pixels wide requires Photoshop CS (v. 8.0) or later. North America and the other smaller VCF tiles are accessible to earlier versions of Photoshop, which are limited to a maximum file width of 30,000 pixels. When opened in Photoshop, VCF land cover appears as an ordinary 8-bit grayscale image. White areas on the image represent open water, so in effect VCF provides you with a bonus fourth category of information. The forest, herbaceous, and bare information appear as grayscale values with lighter values representing greater densities. They are analogous to photographic negatives. This trait makes VCF amenable for use as layer masks for modulating colors in Photoshop. After opening each VCF data layer as a separate Photoshop file, combine them into one multilayer image as follows:

  1. Create a new Photoshop document with exactly the same pixel dimensions as the VCF data you just opened.

  2. Create five new layers by selecting Layer/New Fill Layer/Solid Color in the drop menu or by clicking the “Create a new layer” button in the Layers palette.

  3. Name the layers white background, herbaceous, forest, bare and water respectively from bottom to top.

  4. Fill each layer with an exploratory color. Use white for the background, green for forest, yellow-green for herbaceous, beige for bare, and blue for water (these colors can be fine-tuned later in the design process).

  5. Create a layer mask for each of the layers by selecting Layer/Add Layer Mask/ Reveal All, or by clicking the “Add layer mask” button in the Layers palette.

  6. (a) To insert the VCF land cover data into their respective layer masks, copy and paste the data. Tip: you need to Option-click (Mac) or Alt-click (PC) on the Layer mask thumbnail to open the Layer mask itself for the pasting to occur.

    (b) Alternatively, you can use the Apply Image dialog (Image/Apply Image) to insert the VCF land cover data into Layer masks (all VCF files intended for insertion must be open). First click the Layer mask thumbnail to activate it. Then open the Apply Image dialog and choose one of the VCF files as the source image. The target is by default the Photoshop file you are currently working in. Set blending to normal and opacity to 100 percent. Repeat these steps to insert for the two remaining VCF data files.

  7. To color the land cover layers at the full intensity as chosen in step 4, activate the VCF layer mask for each layer mask as described in 6b above. Then use the Levels dialog (Images/Adjustments/Levels) to convert the grayscale data into a high contrast mask by adjusting the Input Levels settings to 0, 1.00, and 100 respectively from left to right.

  8. To prepare the water layer, insert any one of the three VCF data files into the Layer mask on that layer. First, activate the Layer mask. Then use the Brightness/Contrast dialog (Image/Adjustments/Brightness/Contrast) to convert the grayscale data into a high contrast land/water mask by setting the contrast slider to plus 100. Lastly, invert the mask so that water areas appear white (Image/Adjustments/Invert).

Color adjustments

When finished you should have a Photoshop file that looks similar to Figure 15 (left side). Although preparing MODIS VCF for use in Photoshop is complex, the resulting file permits the easy application of colors to the data. We will start by globally colorizing the vegetation colors. Double clicking the foreground color in the Tool palette brings up the Color Picker and using the Fill command (File/Fill) delivers the new color to the layer (remember to click on the Layer thumbnail before filling). Assigning new colors to the forest, herbaceous, bare, and water layers takes only minutes. Changing the master opacity (keep the blending mode as normal) or manipulating the VCF grayscale data in the layer masks permits even finer global color adjustments. For example, to bring more emphasis to low-density forests use Curves (Image/Adjustments/Curves) to increase the value of these areas. In the North American example, employing this technique made the arctic tree line more distinct.

Figure 15. (left) MODIS VCF in Photoshop presented as uniform colors. (right) With environmental color adjustments applied to the herbaceous layer.

Another even more powerful option is to locally adjust colors based on environmental factors. Doing this creates new land cover categories and adds geographically relevant color variations to the map. For example, in Figure 15 (left side) herbaceous land cover appears as the same yellow-green whether it shows cornfields in Iowa, rangeland in Montana, or tundra in Nunavut. Contrast this with Figure 15 (right side), where local color adjustments depict rangeland as yellow-gold and tundra as light gray-green. Applying local color adjustments is technically simple—just draw a selection boundary with the Lasso tool, apply feathering (Select/Feather) to taste, and fill with a new color. A more critical concern, however, is where the color adjustments are applied and the colors used. For accomplishing this task biogeography and climatic references are a must. Returning to the example of North American tundra, we considered a number of geographical definitions of the arctic for delineating this environmental zone. The 10-degree-centigrade isotherm for July average temperature, for example, generally defines the northern limit of trees worldwide. This definition, however, proved inadequate for subarctic regions, such as Labrador, where tundra-like muskeg and spruce-lichen woodland extend southward for hundreds of kilometers. To bring tundra coloration to these deserving areas we drew the diffuse southern boundary of the tundra zone to include the northern third of the boreal forest zone. Because the tundra coloration applies only to herbaceous land cover, the green forests remained undisturbed.

We applied similar environmental color adjustments to the bare VCF layer to accentuate the polar desert of the high arctic (muted purple), alpine areas (light gray), and the hot southwestern deserts (light red brown). Many other local adjustments are possible. For instance, according to the Köppen climate classification system, the 18-degree-centigrade isotherm for January average temperature defines tropical areas in the northern hemisphere. A slight increase in saturation for all land cover categories within this area (delineated by a diffuse boundary) would increase the vibrancy of tropical areas—bringing the colors on the map closer to our perceptions of geographic reality. Considering that tropical areas in North America account for a small percentage of the total area, as an added benefit (and depending on the design goals of the map) brighter natural colors could bring needed emphasis to the tropics. 

Accessorizing MODIS VCF

MODIS VCF is not a complete data solution for making natural-color maps. While the natural manner in which it blends colors into one another is highly effective, it lacks important land cover categories one would expect to find on a map. To bring a natural-color map based on VCF to final completion requires supplemental data. For example, in VCF the “bare” category does not differentiate between sand, rock, and permanent ice and snow, etc. Any surface without biomass receives the bare classification, be it a gravel bar in Alaska or the concrete skyscrapers of Manhattan. Looking again at the map of North America, glaciers in northern Canada and Greenland appear with the same white color as the surrounding bare ground, rendering them invisible (Figure 15, left and right). Fixing this problem involved three steps—and two days of work. First, using the color adjustment technique discussed previously, we assigned a muted purple tint to all bare areas (including the glaciers) in the high arctic to depict polar desert. Next, we reprojected, rasterized, and imported DCW (Digital Chart of the World) vector data at 1:2,000,000-scale as white polygons, which contrasted with purple background in VCF to show glacier shapes. As a final touch, in Photoshop we overprinted the white glaciers with gray-blue shaded relief to bring tonal modulation to their surfaces (Figure 16).

Although not applicable to our map of North America, large deserts completely devoid of vegetation, such as the Sahara and the Rub al Khali in Arabia, expose another problem with VCF data for the bare category. Completely lacking in tonality, these areas appear flat and homogenous when colorized. Furthermore, the addition of shaded relief to these areas is often not enough to break the monotony. To bring tonal variation to the deserts, selectively swap in the “Blue Marble” image discussed earlier in this paper. Also derived from MODIS, the “Blue Marble” is a close kin to VCF and merges easily with it. For best results lighten and recolorize the entire “Blue Marble” image as desert beige using the Hue/Saturation dialog (Image/Adjustments/Hue/Saturation). Then copy and paste the recolorized “Blue Marble” image to replace the flat bare color in the layered VCF file, retaining the bare VCF data as a mask. The Sahara will now appear as a mosaic of textured dune fields and rocky areas.

Figure 16. (left) The final map primarily based on MODIS VCF data. (right) The top five Photoshop layers contain supplemental data added to the MODIS VCF base.

Cities are another category requiring outside help on MODIS VCF. Although VCF depicts large urban areas mostly as bare, they appear indistinctly. For the North America map we used the USGS’s one-kilometer-resolution Global Land Cover based on AVHRR as an add-on (see Appendix B for URL). Placing the AVHRR city data in layer mask with an associated layer color permitted easy adjustments to color, opacity, and blending. Looking elsewhere, the water layer extracted from MODIS VCF may also require replacement. While it is reasonably good for delineating oceans and lakes, all but the widest rivers appear as discontinuous strings of pixels. Replacing open water areas and rivers with rasterized vector data from sources such as DCW is advisable in most cases. Doing this also requires the removal of the embedded water from the VCF land cover to prevent it from conflicting with the new water. We discuss a semi-automated Photoshop technique for removing embedded water in the next section. The portrayal of open water areas on natural-color maps also deserves consideration. The North America map in Figure 16 portrays open water with hypsometric tints and shaded relief derived from ETOPO2 (2-minute Worldwide Bathymetry/Topography) data. It is ironic that a mapping style that so stridently eschews hypsography for terrestrial areas happens to work so well with hypsography in water areas. On the other hand, since no human has ever seen the ocean basins without water, portraying them with blue hypsometric tints and shading is the closest approximation of natural colors that we have.

New MODIS VCF products are in the production pipeline based on more recent data. The University of Maryland, Global Land Cover Facility plans to expand the forest and herbaceous categories. Forest (woody vegetation, more precisely), will include needleleaf and broadleaf, and evergreen and deciduous subcategories. Herbaceous will include new subcategories for crops and shrubs. These new data will give cartographers even more tools to make refined natural-color maps.


Return to Shaded Relief home