ES 771 Lab 3



This exercise is based on an outstanding, cloud-free Landsat TM dataset acquired on October 19, 1985. The subscene is centered on Lake Benton, located on the Prairie Coteau (upland) of southwestern Minnesota. Surficial sediments and the present landscape are products of glaciation during the Pleistocene (ice age) period. The Bemis end moraine runs across the scene from upper left to lower right. This end moraine was constructed along the western margin of the Des Moines ice lobe during the late Wisconsin glaciation, about 14,000 years ago. Lake Benton occupies a tunnel valley that was cut across the Bemis end moraine by a melt-water river flowing from the northeast out from under the ice lobe. See space-shuttle photograph for regional overview (on campus).

Ground views of Lake Benton.

Agriculture is the predominant landuse in this scene. Various row crops, pasture, and fallow fields form a patchwork across the area. Besides Lake Benton, several other smaller lakes are visible, and several small streams can be seen. They form a parallel drainage network, oriented NW-SE, following the Bemis moraine. The city of Lake Benton is located at the southwest end of the lake, and the town of Tyler is situated seven miles (10 km) to the east. Most highways and roads follow the township-and-range survey grid. Rural roads are spaced one mile apart across most of this scene.

The climate of this region is characterized by long, cold winters and short, warm summers. This scene was acquired in late autumn. Lakes are not yet frozen, and no snow is present on the ground. Most vegetation is turning dormant, and most agricultural crops have been harvested. Given the latitude of 44°N and time of year, the sun elevation is relatively low—only 32° above the horizon.

The spectral bands of the Landsat Thematic Mapper were selected for a variety of applications. The specific uses for each band are given in the table below. Bands 1-5 and 7 have a spatial resolution of 30 m (~100 feet). The thermal band (6) has a spatial resolution of 120 m. All bands are resampled to a standard cell size of 28.5 m in the dataset. Bands 1-5 and 7 are included in this exercise, named BENTON1.* etc. which you should FTP to your personal workspace.

Spectral bands of the Landsat Thematic Mapper and their intended applications.
Wavelength given in µm. Adapted from Mika (1997, table 4).
Color Band Wavelength Applications
Soil/vegetation and deciduous/coniferous forest differentiation, clear-water bathymetry.
Vegetation growth/vigor, sediment estimation, turbid-water bathymetry.
Crop classification, ferric iron detection, ice & snow mapping.
Near infrared
Biomass (vegetation) surveys, water-body delineation.
Mid infrared
Vegetation moisture, snow-cloud differentiation.
Mid infrared
Hydrothermal mapping, rock/soil type discrimination.
Thermal infrared
Thermal mapping, plant stress, urban/ non-urban landuse differentiation.


Begin the exercise by making a standard false-color composite of the scene: band 2 = blue, band 3 = green, and band 4 = red. Select "Linear with saturation points," and create a 24-bit composite with stretched values. This composite resembles color-infrared photography.

Sample image of Benton composite based on TM bands 2, 3, and 4 color coded as blue, green, and red. Your image should appear similar. Click on small image to see a full-sized version.

1. How many rows and columns are there? What is the resolution of each cell? What land area (km2) is represented by this image? How do the number of rows and columns and cell resolution relate to min/max x and y values for this scene?

2. Describe the appearance of the following features, paying attention to colors, patterns, shapes and sizes of objects.

Now make a composite image, as before, that simulates natural color: band 1 = blue, band 2 = green, band 3 = red.

Sample image of Benton composite based on TM bands 1, 2, and 3 color coded as blue, green, and red. Your image should appear similar. Click on small image to see a full-sized version.

3. Describe how the colors of features in the natural-color composite differ from those of the false-color composite. What features are more or less distinct in the natural-color composite?

Now make a composite image that includes only infrared energy: band 5 = blue, band 4 = green, band 7 = red, as before, but change the percent to be saturated to 2.5%.

Sample image of Benton composite based on TM bands 5, 4, and 7 color coded as blue, green, and red. Your image should appear similar. Click on small image to see a full-sized version.

4. Describe how the colors of features in the infrared composite differ from those of the false-color composite. What features are more or less distinct in the infrared composite?

Within the tropics and temperate regions, most land areas are covered by either vegetation (agriculture) or water bodies. These are the regions in which most of the world's human population resides. The identification and monitoring of vegetation and water bodies are, thus, particularly important for environmental studies related to global change and human land-use practices.

5. Of these three composite images, which do you think contains the most (and which the least) information about types of land cover and human land use? Explain your answer.

Many false-color composites render features in unnatural colors that distract from visual interpretation of the scene. The standard false-color composite is a typical example, in which active vegetation appears red and pink. On this basis, one goal of image processing is to produce composites that combine good feature definition with naturalistic coloration. To complete this portion of the exercise, you should experiment with different band combinations to create false-color composites that incorporate both visible and infrared bands. Your result should appear "natural" in terms of overall coloration.

Name your favorite result BEN-BEST. Prepare a composition that includes an appropriate title, subtitle (with your name and date), scale bar and true north arrow. Save an image file to submit via e-mail.

6. Describe the band combinations you utilized and the appearance of features in the false-color composite. Why do you prefer this composite image compared to others you made?

Filtering is an important method of image processing. Various filters may be employed to smooth or roughen the appearance of an image. Filters are neighborhood operations, in which the value for a cell is reassigned based on the values of surrounding or neighboring cells. For example, a mean filter, recalculates the value of each cell from the mean or average value of neighbors. The amount of averaging depends on how many neighboring cells are taken into account, as specified by the filter window. The window indicates the number cells to be included in the mean calculation for each new cell value.

Find the FILTER module (under Image Processing, Enhancement). Click the "Help" button to read more about the functions of filters. You will begin with the mean or low-pass filter. Use TM band 4 as the input image, and make three low-pass versions with 3x3, 5x5 and 7x7 windows. Click on "Mean" for the Filter type. Enter appropriate titles and value units for each. Display the original band 4 image using the "Grey scale" palette and autoscaling. Display the low-pass (mean) versions the same way (autoscaling will be invoked automatically).

7. Describe how the appearances of these images change with increasing window size. What happens to min and max cell values with increasing window size? Hint: use Idrisi Explorer to view file metadata.

Now use the band 4 image to create a high-pass filter with a 3x3 window. Display the result as you did for the low-pass images. View file metadata to examine the data type and cell values.

8. Describe the appearance of the high-pass filtered image. What features stand out clearly? In what way does high-pass filtering enhance the appearance of the image?

Next filter the band 4 image with the Sobel edge detector option. This calculation may take slightly longer to accomplish, depending on your computer. The result will display automatically in gray tones. Note the data type and cell values.

9. What do you think of this image; what features does it emphasize?

Filtered images can be used as inputs for false-color composites. To make composite images with Idrisi, the input images must be in byte-binary format. In this case the Sobel image has values 0 to >340. You need first to use SCALAR to multiple the Sobel image by a constant value, so that it falls in the byte-binary range (0-255). Next CONVERT it to byte-binary format (rounding of values).

Now make a composite as follows: Sobel byte-binary image = blue, band 4 = green, band 7 = red, linear stretch with saturation (1%). Name your image SOBEL_COMP.

10. Describe the appearance of the Sobel composite image, paying particular attention to vegetation and water bodies.

Make a composition of your Sobel composite image including a suitable title, subtitle (your name and date), and scale bar. Save an image file to turn in.

11. What are some advantages for including filtered images in a false-color composite? What are some possible disadvantages compared to a conventional composite image?

Turn in:


Return to course schedule.
ES 771 © J.S. Aber (2017).