Vegetation Analysis


Estonia is a small country located at the eastern end of Baltic Sea in north-central Europe. It is bordered by Finland to the north, Russia to the east, and Latvia to the south. The country experiences a sub-boreal climate that is modified by proximity to the Baltic Sea. Winters are long and cold; summers are short and cool. The land is heavily vegetated with forests, bogs, and agriculture.

Map of Estonia and surrounding countries.
Taken from Estonia in the Baltics.

Forest covers nearly half of the Estonian territory today, and peatland covers another one-fifth of the country. This amounts to approximately 31,500 km². The modern forest is a mixed assemblage dominated by pine (Pinus), birch (Betula) and spruce (Picea). Aspen, oak, and alder are also found in the forest. The combination of spruce, a boreal conifer, and oak, a temperate hardwood, demonstrates the transitional character of this forest. Mires, fens and bogs occupy water-logged soils and infilled lakes, in which organic (plant) matter accumulates without decay forming peat. Both forest and peatland have long histories of human usage and modification in Estonia.

The Landsat TM subscene includes the island of Vormsi in northwestern Estonia. Forest is indicated by green, pale yellow shows cleared agricultural land, and coastal marshes have a blue ^ pattern. Click on the small image to see a full-sized version. Map taken from Vormsi Info.

A cloud-free Landsat TM dataset from 28 July 1986 is the basis for this exercise. The subscene is centered on the island of Vormsi, which is Estonia's fourth largest island. It covers 93 km² in area and has 330 inhabitants. The island is a bedrock platform composed of fossiliferous Ordovician limestone overlain by a few meters of Quaternary glacial and coastal sediments. At the surface, limestone beach gravel is conspicuous over much of the island. The coastal marshes are important bird breeding areas. Many of the place names end in "by" (pronounced buu) which is the Scandinavian word for town. These place names reflect the Swedish history of the island.

Beach and lighthouse at Saxby, northwestern Vormsi.
Seen at low tide. Digital image © Jeremy Aber, 8/00.

Kite aerial photographs of Vormsi.

Set up a new project and "working folder" for this exercise on your computer. Copy all "VORMSI" files (via FTP) into your main working folder, where you will conduct the exercise.


Begin by creating a false-color composite based on band 3 (red), band 4 (near-infrared), and band 5 (mid-infrared) color coded as blue, green and red (default settings). This composite is widely employed and has proven quite useful for general vegetation analysis. Band 3 is strongly absorbed by active vegetation, whereas band 4 is strongly reflected. Band 5 is sensitive to moisture in leaves and soil. Examine your composite image and try to recognize different kinds of vegetation cover.

Sample image of Vormsi. Yours should look quite similar, but without captions or scale bar. Spruce forest is dark green, pine forest is olive-green to reddish brown, deciduous forest is yellow-green, and croplands are pale green, pink, and lavender. Click on the small image to see a full-sized version.

The entire island is less than 20 m above sea level. Higher, better-drained portions have been cleared for agriculture and villages. Lower, poorly drained sections are covered by forest and marsh.

1. Describe the following features paying particular attention to the appearance of vegetation.

Now prepare a composite image based on bands 1, 2, and 3 color coded as blue, green, and red. This composite simulates natural color of the visible spectrum. Examine this composite and compare it with the 345 composite.

2. What kinds of features are more obvious in the 123 composite? What features are less distinct in the 123 composite compared to the 345 composite?

Idrisi has many functions for analysis of vegetation. Next you will make an NDVI (normalized difference vegetation index) image for this scene. Photosynthetically active vegetation strongly absorbs red light (band 3) and reflects near-infrared radiation (band 4). No other common materials at the Earth's surface have this spectral signature. NDVI is among the most popular and widely used vegetation indices, based on the following formula.

(band 4 - band 3) ÷ (band 4 + band 3) = NDVI

The NDVI is a type of band ratio, so it will be necessary to perform haze correction on each band before continuing. Use the SCALAR (under GIS Analysis, Mathematical Operators) to perform haze correction by adding (or subtracting) a whole number value, so that the new minimum = one (1). Output documentation with appropriate title and value units (haze correction).

3. What haze-correction values did you select for each band (1-5, 7)?

Now create the NDVI image with VEGINDEX (under Image Processing, Transformation). Click the NDVI option and enter band 3 (red) and band 4 (infrared). Note: use haze-corrected images for this and all subsequent processing in this exercise.

Your NDVI values should fall within the range -1 to +1, for which positive values are active vegetation and negative values indicate a lack of vegetation. Your image should display automatically with the NDVI palette. Vegetated land areas appear in green; the sea is orange with distinct stripes. This line banding is an artifact of the Landsat TM scanner, which is often apparent over homogeneous features, such as large water bodies. The effect disappears over most land areas. Ratio images tend to enhance line banding and other "noise" in the imagery.

4. What range of values did you obtain for NDVI?

To be useful for further processing and analysis, the real-number NDVI values need to be rescaled and converted into byte-binary format. Three steps are required to do this: (1) add 1.0 to NDVI, (2) multiply NDVI by 125, and (3) convert to byte-binary (with rounding). After completing these steps, you may delete real-number NDVI files.

Now use your byte-binary NDVI image as part of a false-color composite as follows. Band 2 = blue, NDVI = green, and band 5 = red (1% saturation).

5. Compare the new NDVI composite with your earlier 345 composite. How is the NDVI composite different? Do any vegetation features stand out more clearly? Of these two images, which do you prefer for display of vegetation, and why?

Now read about the TASSCAP module (under Image Processing, Transformation). Tasseled-cap transformation derives brightness, greenness, and moisture images from the input Landsat TM bands (1-5 and 7). Run the TASSCAP module on your haze-corrected files. Display the resulting images with autoscale, title, legend, and default palette.

6. Compare the three tasseled-cap images with each other and with your original 345 composite image. Describe the appearance of vegetation for each.

By now, you should be familiar with this image and the types of vegetation. Your ultimate goal is to separate three forest types--spruce, pine, and deciduous--from other types of vegetation, agriculture and land cover. To do this, you will utilize unsupervised classification techniques CLUSTER and ISOCLUST (under Image Processing, Hard Classifiers). Read about these techniques.

Examine the following spectral plots for typical coniferous and deciduous trees. The plots indicate reflectivity for visible, near-infrared, and mid-infrared wavelengths (comparable to Landsat TM bands). Note positions and absolute values for reflectance peaks. Taken from USGS Spectroscopy Lab.

7. On the basis of these plots, which type of tree produces the greatest near-infrared reflection; which produces the least?

Now run the CLUSTER module based on six input files (haze-corrected bands 1-5, 7). Accept all default selections for the cluster technique. The result will display automatically with the qualitative palette.

8. How many clusters resulted from the classification? Which of the classes appear to correspond to vegetation cover?

Next run the ISOCLUST module with the same six input files (bands 1-5, 7). A histogram will appear showing the number of cells (pixels) for each seed cluster. It's apparent that most pixels are contained in only a few clusters (1-20). On this basis, you must decide how many clusters should be retained for the analysis. Usually 12 to 20 is an appropriate number. In this case, indicate 16 as the number of clusters desired.

Note: isocluster is numerically intensive, and may take a few minutes to run. If any error messages appear, click OK and continue.

9. Compare the results of isocluster to cluster. How are they similar; how are they different? Which do you prefer for classification of vegetation types?

Now create a display of your classified (cluster) image showing types of vegetation (agriculture, spruce, pine, deciduous) and other land cover. Use RECLASS (under GIS Analysis, Database Query) to reassign categories into identified land cover types. Then create a special palette to display your reclassed image, and add legend captions to the display. Finally create a map composition that includes title, scale bar, legend, and other appropriate components. Note: the legend should clearly identify each class. Name your composition VORMSI and make a digital image file (jpg) to turn in.

10. Based on the cluster classification, what areas (in km²) are covered by the each cluster of spruce forest, pine forest, deciduous forest, and agriculture?

Turn in

ES 771 schedule or ES 771 homepage.

© 2018. Notice: ES 771 is 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, J.S. Aber.