Advanced Image Processing of the Konza Prairie
using Landsat 5 data


Jake Stevenson

A Student Presentation from
Emporia State University
Earth Science Department

This webpage project was created in partial completion for the Advanced Image Processing course in spring of 2006 at Emporia State University. The assignment was to demonstrate Advanced Image Processing techniques learned throughout the semester.


The Konza Prairie Preserve south of Manhattan, KS provides the unique oportunity to study long term effects of management on a tall grass prairie ecosystem. Once a cattle ranch, this area has become a haven for native animals and plants. Combined with Kansas State University, The Nature Conservancy has conducted tests and research programs throughout the years. Due to the preserve's unspoiled land this area has been the focus of study by many on the tall grasses that are present here and that once thirved across the midwest. This webpage is about one study of these grassses. Remote sensing is a tool that allows researchers to conduct experiments about ground cover while not actually being in the field and its powers of data processing will be demonstrated on this site.

Table of Contents


The area of study for this project was the Konza Prairie Preserve. Jointly owned by the Nature Conservancy and Kansas State University the 8,600 acre preserve is operated by the Kansas State Department of Biology as a field research station. The Konza Prairie Preserve is a diverse landscape with the dominant ecosystem being tall grass. In addition, forest, claypan, shrub and riparian communities also exist throughout the landscape (The Nature Conservancy).

The original purpose of this project was to, by use of image processing; study the effects of spring burning on the grasslands of the Konza Prairie. However this proved to be unobtainable, due to the regrowth of the grasses and date of when the images were taken. Nevertheless, while processing the imagery it was noticed that entirely different and useful information could be created.

What was created was an image showing grasslands of the Konza Prairie without clutter from surrounding areas or forest vegetation. Below are two images showing the boundries of the study area. The first is a true color composite using bands 1, 2, and 7 from Landsat 5 imagery. The second image is from the Konza Prairie LTER program and depicts the watershed bourdaries of the prairie. These watersheds are what the Kansas State Department of Biology uses as areas to burn on a rotational basis for research use.

Natural-color composite (Left), Watershed Map from Konza LTER website(Right).
Click on Images to Inlarge.

The imagery used in this project is from the Landsat 5 Satellite from Sept. 11th, 1997. This satellite is capable of acquiring 7 bands of visible, reflective-infrared, middle-infrared, and thermal infrared regions of the electromagnetic spectrum. The usefullness of such data will be evident later in the process.

Back to the table of contents


The first step towards processing images of the Konza Prairie was obtaining the images. Dr. James Aber at Emporia State University was gracious enough to supply me with this data. The imagery was raw and had to be converted into a format that Idrisi could use. To do this the suffixes were changed from *.I1-*.I7 to *.rst. After this was done documentation files (*.rdc) were created using data from the header file that was supplied on the cd which was provided. The header file was saved with a suffix of *.H1 and had to be opened using notepad. contained within the file was information such as the number of columns and rows in the image, resolution, and units of the pixels.

Once the images were in raster format and a documentation file was created using the metadata within Idrisi, they needed to be clipped down. In order to accomplish this task the WINDOW function was used. The Window function clips the images based on a user defined box, which are the top left and bottom right rows and column values. The values used for this exercise are listed below in the table.

Upper left column 3960
Upper left row 4020
Lower right column 4420
Lower right row 4500

Once all the images were clipped to highlight the study area, composites were created. Each of these composites showcases a different feature on the landscape. The first image on this page is a true color composite using bands 1, 2 and 7. This type of image is helpful in showing what the ground looks like at the area. Below are three other types of composite images. The first image is a false color composite composed of bands 2, 3, and 4. This type of image is used in showing active vegetation in shades of red. The darker red represents dense vegetation such as a stand of trees while the lighter shades represent less dense areas such as agricultural fields. The second image is comprised of bands 2, 5, and 7. What is good about this type of composite is that the vegetation has different levels of green to represent the health of the plant. The darker the green the better, on some of the agricultural fields barren ground in depicted by shades of tan. Another good feature about this image is that the river, roads and towns show up nicely against the green prairie background. The last image was created using bands 3, 4, and 5. This false color composite is useful in identifying water bodies such as the river and farm ponds. Looking closely in all the images, but mostly the 257 composite, you are able to discern the outline of the watersheds used by the Konza Prairie as boundaries for burning cycles.

234 Composite (Left), 257 Composite (Center), 345 Composite (Right)

As mentioned before, originally planned was an image showing burn area from the previous spring burn. However after determining that this was not feasible with the data available, other conclusions could be drawn after processing of the data was preformed.

The next step in processing the images was to create an NDVI, which stands for Normalized Difference Vegetation Index. This index value is created using the OVERLAY function in Idrisi. This function calculates the NDVI image based on the following equation.

NDVI = (Band 4 - Band 3) / (Band 4 + Band 3)

The above NDVI image of the Konza Prairie has been manipulated and enhanced using the SCALAR and CONVERT operations. This was done to convert the image from "real" numbers to "byte-binary". This is done mainly because of the storage space difference between the two formats. The steps involved in this process are listed below.

After determining that the information needed could not be obtained from the NDVI more methods of image processing were used. An alternative for vegetation analysis other than the NDVI is called a Tasselled Cap which is run using the TASSCAP function. This function creates three images; greenness (green vegetative cover), brightness (soil brightness), and moisture (soil moisture). Of the three images created the Moisture image shows the most definition between the Konza Prairie and the surrounding landscape.

Greenness (Left), Brightness (Center), Moistness (Right)

The next process was to isolate the grasslands area, separate from wooded vegetation in the Konza Prairie. For this the ISOCLUST function was used. This function separates features based on their spectral reflectance. Using 6 of the 7 bands the ISOCLUST function combined like features into 16 groups.

Konza Isocluster

These groups can be further isolated so that the grasslands area can be isolated. To do this further isolation a set of steps were preformed. Because a Boolean image has cell values of either 1 or 0, a reclassification of the existing values was done. Using the Isocluster image and the EDIT function an attribute file was created by combining the cell values of the desired grassland area. Listed below are the general steps taken to create the final image.

  • RECLASS the grassland clusters to equal 1 and all other clusters to equal zero.
  • GROUP the clusters into areas with unique ID numbers.
  • RECLASS the resulting image to extract the Konza Prairie Grassland.

    First Boolean Image (Left), Grouped Image (Right)

    Back to the table of contents


    After all those steps were performed the final image was created showing the grassland areas of the Konza Prairie Preserve. In this image are parts of the surrounding landscape, which due to there shared values when reclassifying were unable to drop.

    Lastly the area was calculated for the image of the Konza Prairie by creating a histogram of the scene. The pixels were multiplied by the resolution and then converted to acres. These values are listed below and show the extended area around the preserve as well, leading to the higher numbers.

    Category Acres
    0 37540
    1 6785

    Back to the table of contents


    Aber, J.S. Advanced Image Processing. ES775 Advanced Image Processing. Emporia State University. 2006

    Clark Labs 2006. Geographic Analysis and Image Processing Software.

    Konza Prairie LTER Program 2006. Program Website Kansas State University, Division of Biology.

    The Nature Conservancy 2006. Konza Prairie Partnership

    Related Links

    Emporia State University Earth Science at ESU
    Advanced Image Processing Konza Prairie LTER

    For more information email