Landsat Image Processing

ES 351, ES 771, ES 775
James S. Aber

IntroductionSpring imagery
Autumn imageryWinter imagery
VegetationBand ratios
Burn ratioHigh contrast


Rapid developments of computer hardware and software are major factors that facilitate much greater use of Landsat imagery than was possible in the past. The integration of remotely sensed data and geographic information system (GIS) technology is one of the great ideas whose time has come (Faust et al. 1991). This integration occurs naturally in a raster-based GIS, because both data structures are essentially the same. Common GIS raster operations (overlay and neighborhood operations) are functionally the same as many image processing techniques.

Modern work stations based on high-speed processors have the computational power of former mainframe computers. High-capacity hard-disk drives and other compact data-storage devices greatly aid in the processing of digital images. Display devices have likewise improved dramatically. Large-screen, high-resolution monitors allow true-color display of processed images. Sophisticated software packages for combined image processing and GIS are readily available for various computer operating systems. It is safe to say that research based on Landsat imagery is now possible at moderate cost for almost any academic or governmental organization.

Careful selection of Landsat scenes and attention to special seasonal conditions are important for acquiring the best images of landscapes. Several image-processing techniques are also useful for enhancing images and creating dramatic displays of certain environmental conditions and landform assemblages. Seasonal selection and digital processing of Landsat imagery for geomorphology are described in more detail in the following sections.

Spring imagery

Spring imagery has been found most useful from the stand point of distinctive patterns in soil moisture and vegetation development, using geobotanical methods. This is a time of year when vegetation phenology and soil moisture conditions reflect subtle differences in soil texture and organic content, in land slope and aspect, and in land cover or agricultural practices. Sun position is at or near its highest elevation of the year, which provides for maximum illumination and minimum shadow effects.

Variations in image tone (or color) indicate the kinds of surficial sediments and landforms on which the soils have developed (Morrison 1976a). This is also a good time of year to see surface drainage and lake patterns clearly. The standard false-color composite image is quite effective for displaying geobotanical aspects of the landscape. Landsat images from later in the growing season (July and August) may be almost completely dominated by vegetation, which blankets the landscape. Geomorphic features are usually less distinct on such summer images.

Autumn imagery

Autumn is again a time of year when vegetation phenology reflects geomorphic variations in the landscape, as certain plants become dormant or die and crops are harvested during the season. However, autumn scenes generally differ from spring images in two aspects: surface water and sun elevation. Soils are usually drier, and consequently lakes and streams are often lower. In semiarid or arid regions, shallow lakes may partly or completely dry up, leaving bare mud flats or salt deposits. Drought conditions may enhance certain geomorphic features revealed on autumn images.

Sun elevation is significantly lower, compared to spring images, which imparts a moderate topographic shadow effect. These factors working in combination can result in excellent synoptic displays of geomorphology on Landsat images--see Devils Lake. As with spring imagery, timing is important, as determined by local climate, vegetation activity, and latitude.

Standard false-color composite of Devils Lake, North Dakota. Autumn scene showing strong variation in vegetation cover and water bodies. Low sun elevation creates some shadowing of terrain. Landsat MSS bands 1,2,4; acquired 9/88; image processing by J.S. Aber.

Winter imagery

Many investigators have noted that winter Landsat images can strongly emphasize the topographic character of the landscape (Morrison 1976b; Slaney 1981; Eyton 1989). This topographic enhancement is the result of three factors.

  1. Low sun elevation: Very low sun angle produces a strong shadowing and highlighting of terrain features. Valleys, escarpments, lineaments, dune fields, end moraines, drumlins, and other elongated geomorphic elements may be emphasized.

  2. Snow cover: Snow smooths the landscape and provides a homogeneous cover with uniform spectral properties, which mask distracting variations in soils and vegetation. This tends to accentuate the effect of shadows on low-relief terrain. The presence of snow usually means that lakes and streams are also frozen over.

  3. Lack of active vegetation: Vegetation is an important, often dominant, aspect of Landsat images during the growing season. Certain subtle topographic features may be more apparent when vegetation is not active.

Winter images with low sun elevation and heavy snow cover have a "pen-and-ink" quality, in which topography is clearly depicted (Eyton 1989). MSS band 3 (shortest infrared) has proven most effective for displaying winter snow scenes, and false-color composites are also quite useful. Analytical hill shading using DEMs has shown that illumination direction and angle are extremely important in determining which landscape features are portrayed (Lidmar-Bergström et al. 1991). However, sun position cannot be varied when viewing a Landsat image. Linear geomorphic features that trend NE-SW are enhanced (NW-SE in southern hemisphere); conversely, features trending NW-SE may not be readily apparent on Landsat images.

Single-band, winter image of southern Nain province, eastern Canada. Snow cover combined with a low sun elevation (11°) gives strong emphasis to landscape topography. Glacial erosion has etched out crustal fractures of several sizes and orientations. Landsat MSS band 7, acquired 1/73; image from NASA GSFC.

Continuity of snow cover is extremely important; variations in thin, patchy, or irregular snow cover may so dominate the appearance of an image that any topographic impression is lost. The character of vegetation cover is another important variable. Good topographic expression is most apparent in regions with cropland, grassland or sparse deciduous forest cover. Regions with heavy forest, especially conifer trees, retain a vegetated character even in winter. It should be noted, finally, that winter images may suffer from the relatively low level of solar illumination.

Vegetation indices

Active terrestrial vegetation strongly absorbs red light and strongly reflects near-infrared (NIR) energy. No other common materials at the Earth's surface have this spectral signature. On this basis, many red-NIR combinations have been proposed as vegetation indices (Tucker 1979; Lymburner et al. 2000). Further research has demonstrated that mid-infrared (MIR) bands depict the moisture content of active leaves (Pinder and McLeod 1999). Some of the more popular or useful vegetation indices are listed below--see Table 3.

Table 3. Vegetation indices for Landsat MSS and TM.
Formula Type of IndexReference
NIR ÷ Red = RVI Ratio vegetation index Tucker 1979
(NIR - Red) ÷ (NIR + Red) = NDVI Normalized difference vegetation index Tucker 1979
(NIR - Green) ÷ (NIR + Green) = GNDVI Green normalized difference vegetation index Lymburner et al. 2000
MIR ÷ NIR = RDI Ratio drought index Pinder & McLeod 1999
NIR ÷ (Red + MIR) = SLAVI Specific leaf area vegetation index Lymburner et al. 2000

Perry and Lautenschlager (1984) demonstrated that most of the popular vegetation indices are functionally equivalent to a simple NIR/red ratio (RVI). In general, the 4/2 MSS band ratio provides greater contrast than the 3/2 MSS ratio for vegetation display. The ratio for TM is bands 4/3. For the ratio drought index (RDI), Landsat TM bands 5/4 are utilized, and bands 3, 4 and 7 are employed for SLAVI. The latter measures an important ecological variable (Lymburner et al. 2000).

Creating these vegetation indices involves one or more raster overlay operations (+, -, *, ÷), which may result in output of fractional real numbers. Real-number data storage is a significant limitation when analyzing large data sets. To overcome this problem, the real-number output may be scaled and converted into byte-binary format (0-255). As an example, raw NDVI values range from +1 (maximum vegetation) to -1 (no vegetation). The standard conversion technique is: (NDVI + 1) x 100. The resulting range of values is zero to 200, which can be converted to byte-binary format. A vegetation index may be combined with other spectral bands to create striking false-color images of land vegetation.

Special composite image of Gdansk Bay region, northern Poland. Field of view is about 150 km across. The image incorporates green and NIR bands plus a ratio vegetation index (RVI). Fields with crops and pasture appear light yellow; fallow fields are light blue; deciduous forest is orange; conifer forest and peat bogs are dark brown/green. Landsat MSS bands 1, 3, 4/2; acquired 10/86; image processing by J.S. Aber.
Devils Lake, North Dakota depicted in a false-color composite that includes a ratio vegetation index (RVI). Active vegetation shown by bright green and yellow colors; zones within Devils Lake indicate floating mats of algae. Field of view about 50 km across. Landsat MSS bands 1, 4/2, 3; acquired 10/86; image processing by J.S. Aber.

Water strongly absorbs infrared energy and weakly reflects red light. Water is the only common surface material with this spectral signal. Thus, the infrared/red ratio also has the effect of depicting all surface water bodies, regardless of water depth or turbidity. The only exceptions in water are floating mats of algae or other emergent aquatic plants that produce spectral signals like terrestrial vegetation. The combination of surface drainage and vegetation patterns, depicted on infrared/red ratio images, may give valuable information about geomorphology.

The state of the world's vegetation has become a major subject of research during the past two decades. Vegetation is a key indicator for overall environmental conditions, and changes in vegetation are useful means for recognizing changes in other environmental factors. Regional and global analysis of both terrestrial and marine vegetation is now possible based on satellite remote sensing. The status of vegetation and vegetation changes over multiyear periods can be documented with Landsat imagery.

Band ratios

The simple vegetation index, infrared/red ratio, is a good example of the potential of band ratios. Band ratios have proven quite useful for discrimination of surficial rocks and minerals in arid or hyperarid regions with little or no vegetation cover (Rowan et al. 1976). In more humid regions, vegetation (or agriculture) usually covers most of the land surface. Here is a summary of some commonly used band ratios for Landsat MSS image processing and interpretation--see Table 4.

Table 4. Some commonly used Landsat MSS ratios and their applications. Adapted from Avery and Berlin (1992, p. 442).
MSS Ratios Applications
1/2, 1/4, 3/4 Characterizing rocks and soils
1/2 or 2/1 Suspended sediment in water
1/2 or 2/1 Iron-oxide content in rocks
3/1, 3/2 Vegetation and water bodies
4/1, 4/2 Vegetation and water bodies

Note: the inverse ratios create negative images, which
may be more pleasing visually for certain features.

Band ratios have the effect of removing shadows and illumination variations caused by differences in ground slope. In this manner, spectral signatures can be compared directly regardless of different illumination conditions. This effect also eliminates any topographic impression by removing shadows from the image. Band ratios tend to emphasize noise patterns; thus, haze correction is an important processing step prior to making ratios. Simple haze correction is based on the assumption that band 4 suffers minimal atmospheric scattering. Other bands are adjusted to the same minimum value as band 4.

Normalized burn ratio

A large number of destructive forest fires in recent years stimulated a need to quickly map burned areas. The U.S. Geological Survey and National Park Service developed a burn severity index based on Landsat TM/ETM bands 4 (near-infrared) and 7 (mid-infrared). These two bands provide the best contrast between photosynthetically healthy and burned vegetation (Howard et al. 2002). The two bands are combined in a ratio called the Normalized Burn Ratio (NBR).

NBR = (TM4 - TM 7) (TM4 + TM 7)

NBR is determined for pre-fire and post-fire scenes, and the difference is determined by subtraction (pre - post) to give the Normalized Difference Burn Ratio (NDBR). Historical and recent Landsat TM and ETM datasets allow for analysis of fire severity of forested regions during the past two decades.

High-contrast images

In many Landsat images, the land areas are relatively bright compared to water bodies. This high contrast limits the amount of enhancement that can be applied to the image as a whole. The land and water regions may be separated for special processing. Separation of land and water is based on the infrared/red ratio; water bodies generally have ratios <1 and land areas are >1. Land and water masks, prepared from the infrared/red ratio, are applied to original images to remove (zero) either water or land pixels while leaving the other pixels unchanged.

For example, a land mask would be applied to black out water areas. The remaining land portion of the image could be processed then to enhance only land features. The water portions of the image could be handled in a similar manner, and finally the land and water portions could be recombined for a complete image. Equivalent techniques may be applied to other kinds of high-contrast images to treat relatively light and dark portions separately. The resulting dual-processed images can be quite impressive (Lillesand and Kiefer 1987).

Lake Van, Turkey showing enhancement of lake features. Lake and land portions of the image were processed separately to improve display of suspended sediment in surface water. Note several circular eddies within the lake. Compare with normal composite image of Lake Van. Landsat TM special processing; acquired 9/84; image from NASA GSFC.

Return to Landsat remote sensing.

Notice: ES 351, 771 and 775 are presented for the use and benefit of students enrolled at Emporia State University. Any other use of text, imagery or curriculum materials is prohibited without permission of the instructor. All Landsat webpage material © J.S. Aber (2013).