ES 351/771/775 Landsat image interpretation

Landsat Image Interpretation
ES 351, ES 771, ES 775
James S. Aber

Three general methods have been used to interpret and extract information from Landsat and other remote-sensing images (Singhroy 1992).

Photointerpretation: Visual interpretation of images based on feature tone (color), pattern, shape, texture, etc.
Spectral analysis: Identification of surface materials on the basis of spectral signatures.
Data integration: Merging of remote-sensing data with other types of data, such as digital elevation models (DEM).

Visual interpretation

The outstanding aspect of Landsat imagery is complete, small-scale coverage of the Earth at moderate resolution. Computers are used routinely to handle the large volume and quantitative nature of Landsat data. However, the images can still be usefully analyzed with classical techniques of photointerpretation (Short 1982). Black-and-white images represent light intensity variations for a single band. Color-composite images include three bands, each of which is color coded--see Table 2.

**Table 2.** Some common band combinations
for Landsat color composite images.
Instrument	Bands	Colors	Type
MSS (1-3)	4, 5, 7	B, G, R	Std. false color
MSS (4-5)	1, 2, 4	B, G, R	Std. false color
TM (4-5, 7)	1, 2, 3	B, G, R	Natural color
TM (4-5, 7)	2, 3, 4	B, G, R	Std. false color
TM (4-5, 7)	1, 2, 7	B, G, R	Special composite
TM (4-5, 7)	2, 3, 5	B, G, R	Special composite
TM (4-5, 7)	3, 7, 5	B, G, R	Special composite
TM (4-5, 7)	4, 5, 7	B, G, R	Infrared composite

	Single-band, black-and-white image of southern Saskatchewan, Canada. Field of view about 150 km across. Short-infrared band depicts active vegetation in light gray tones; bare/fallow ground is dark gray; water bodies are black. Landsat MSS band 7, acquired 5/78; image from EROS Data Center.
	Metropolitan Kansas City region. This image is a "natural color" composite that simulates normal color appearance of the scene. The urban area is depicted in white (highways, buildings, railroads), muddy water is brown, and vegetated areas are shown in dark green. Compare with the standard false-color version (below) of this same scene. Landsat TM bands 1,2,3; acquired 3/83; image from NASA GSFC.
	Kansas City metropolitan region. Standard false-color composite image formed from green, red, and short-infrared bands, color coded as blue, green, and red. Active vegetation appears red and pink, and water bodies are black. Urban structures appear in white to light blue shades. Compare with the natural-color version (above) of this same scene. Landsat TM bands 2,3,4; acquired 3/83; image from NASA GSFC.
	Special composite of visible and mid-infrared bands. The visible bands penetrate water and show suspended sediment in surface water. Volcanic mountains are visible around the lake. Landsat TM bands 1,2,7; acquired 9/84; image from NASA GSFC.
	The edge of the Andes Mountains is seen flanked by extensive alluvial fans that form relatively flat surfaces. Straight lines and geometric designs (left center) are the archeologic "Nazca lines" of prehistoric origin. This special composite employs visible and mid-infrared bands. Landsat TM bands 2,3,5; image from NASA GSFC.
	Special false-color composite of Paris, France. Field of view is about 40 km across. Active vegetation appears brown/orange; suburban area is green/yellow; urban core is blue. Landsat TM bands 3,7,5; acquired 6/84; image from NASA GSFC.
	Infrared false-color composite of White Sands, New Mexico. The White Sands are dunes composed of gypsum sand blown from a nearby dry lake bed. In this image the gypsum dunes appear light blue, and the sand source area is dark blue. Landsat TM bands 4,5,7; image from NASA GSFC.

An appropriate band or combination of bands should be selected for each interpretive use. Short (1982) provided these general guidelines for interpreting Landsat imagery.

Look for differences in gray tones and/or color. Such differences are often paramount in recognizing the presence of specific surface features and defining boundaries between different classes.
Rely on shapes of ground features and their spatial patterns as further clues to identify different features.
Compare two or more scenes from different seasons or years to resolve uncertainties about some unknown features. If a feature remains similar in tone or color from season to season, it is probably not related to vegetation or to transient events. Features in short-IR that appear light during the growing season and noticeably darker in winter scenes are almost certainly vegetation.
Seek out supporting information from maps, aerial photographs, reports, and other publications, as well as ground observations, as aids to feature identification.
Draw upon your own experiences, observations, and knowledge of the region in question to help in recognizing features and interpreting the geographic and geologic aspects.

Many features change with a seasonal regularity--deciduous vegetation, soil moisture, and sun angle; hence the season and time of year must always be taken into account. Other features, such as forest fires, volcanic dust, drought, defoliation due to insects, and so on, are transient events. Interpretation of Landsat imagery requires an appreciation of all aspects of the scene in order to recognize and analyze the features of interest.

Spectral signatures

The combination of energy reflected and emitted from an object is its spectral signature. Under clear, sunlit conditions, many objects have characteristic spectral signatures in the visibile and short infrared portions of the spectrum.

Spectral signatures are how we "see" the natural or false color of objects. Various objects may be discriminated or classified--rock types, vegetation, water bodies, snow and ice, etc. This is a powerful method for interpreting Landsat data, but classification techniques must be used with some caution. The spectral signatures of objects are not constant; signatures vary with time of year, amount of solar illumination, geographic locations, and many other factors. Thus, a classification technique that works well in one situation may not be appropriate in other circumstances.

Merging Landsat data with other data

Many innovative techniques can be used to create dramatic images of Landsat data combined with other kinds of remotely sensed or ground-based data. Digital elevation models (DEM) are especially useful in this regard. DEMs may be displayed as either contour maps, block diagrams, or shaded-relief images. Possible combinations include:

Landsat image draped over DEM block diagram; perspective view of the landscape with color coding from Landsat image.
Landsat image added to shaded-relief image; enhancement of topography visible in Landsat image.
Separation of water and land areas; color-coded DEM to show bathymetry of water bodies; land areas depicted by remotely sensed data (Shanks 1992).

Mt. McKinley, Alaska portrayed with Landsat image draped over a block model of landscape topography. The oblique view is created from a digital elevation model (DEM), and color coding comes from the Landsat image. Large valley glaciers with medial moraines are visible in foreground. Landsat TM bands 2,3,4 + DEM; image from NASA GSFC.

Any of these approaches requires considerable computation, in which the remotely sensed data must be rectified for accurate fit onto the ground coordinate system of the DEM. Resampling of pixel size, shape, and position may be necessary. The effort necessary to produce such merged images is well worth the investment of time. Landsat/DEM images may be particularly valuable for geomorphic interpretation.

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 (2008).