|Exercise 2: Quantifying the deforested area|
Grazing cattle at one of the many cattle farms in Rondonia
The data that will be used in this exercise is the following:
- Rondonia_1989_band1.tif; Rondonia_1989_band2.tif; Rondonia_1989_band3.tif; Rondonia_1989_band4.tif; Rondonia_1989_band5.tif; Rondonia_1989_band7.tif
- Rondonia_1999_band1.tif; Rondonia_1999_band2.tif; Rondonia_1999_band3.tif; Rondonia_1999_band4.tif; Rondonia_1999_band5.tif; Rondonia_1999_band7.tif
- Rondonia_2009_band1.tif; Rondonia_2009_band2.tif; Rondonia_2009_band3.tif; Rondonia_2009_band4.tif; Rondonia_2009_band5.tif; Rondonia_2009_band7.tif
Now that we’ve inspected the images, we can start analysing them by performing a classification. The main features we want to distinguish are the forested and deforested areas. We will look for differences in their spectral signature, measured by the satellite, to create a map of the deforestation in Rondonia.
To be able to do the classification, you need to open all bands of the Landsat image, except for channel 6, which is not commonly used for classification purposes. For now, only do this for the year 1989, in order to avoid too many open images.
Do not enhance the images you have just opened. It is important that we work with the original pixel values. Enhancing images can only be used when you want to create a better visualisation.
We will conduct an unsupervised classification. This means that the software will gather information about the spectral characteristics of each land cover class (i.e. the areas in the image with similar pixel values) from the different bands, to separate them into different classes. Every pixel is assigned to the class that it resembles most closely.
Classified image, 1989
Open Unsupervised classification and select all the individual Rondonia_1989 bands. We want to distinguish between forested and deforested areas. We therfore want two classes. Let the algorithm run for ten iterations.
If the colours are not clear, you can change them by adding a Legend. A Legend window appears. In the left column, you can select every class and change the name and the colour.