Change Detection

Satellite imagesRemote sensing can be used to monitor (via a time series of images ) and manage World Heritage sites. For an overview see Remote Sensing and other associated links. Some specific applications of remote sensing in land cover and land use analysis within the Virunga National Park are illustrated by two Landsat images taken in 1987 and in 2001. The knowledge of the changes in time of the vegetation status and its dynamics allows to take the appropriate management options regarding the habitat of certain species. The changes can include vegetation disappearance, vegetation degradation, replacement of forests by agriculture, regrowth of certain vegetation types.

Image analysis and interpretation involves initially checking the grey-scale (black and white) image of each individual channel (band) to obtain a sense of the spectral characteristics of each type of land cover features in every single band. In the second step, false colour composites are obtained by assigning Red, Green or Blue colour to one of the three component bands (given in the order of RGB). Image enhancement procedures are applied to the images to improve their appearance.

Image Analysis

The images below show a border region of the Virunga National Park, situated near Rutshuru. The left satellite image was taken in 1987 in the dry season (Landsat 5 TM), and the right satellite image was taken in 2001 during the rainy season (Landsat 7 ETM+). Visual interpretation of the 1987 figure shows the following land cover features. The dark twisting line in the image is the Rutshuru River. The dark sickle-shaped black spot is a volcanic lake. Notice the tonal differences between dense forest (dark red-brown) in circle on right and the sparse or secondary vegetation (light brown) above the river in the centre of the image. The predominant savannas are represented by the darker shade of blue in the bottom centre of the image. Bare soil is characterised by light blue to the left of the lake.

Landsat Image (1987)	Landsat Image (2001)

Comparisons of the two images in the upper left corner show that, in the 2001 image (rainy season) almost all of the dense vegetation (reddish-brown colour) had disappeared. This means that the vegetation on the right side of the river had almost disappeared over the 14-year period. The upper section of the river forms the natural border of the park, so this remarkable vegetation transition occurred outside the park, up to the border edge. The right circle situated outside the park, also experienced loss of vegetation between 1987 and 2001.

Generally, we notice more dense vegetation (dark red-brown) in the 2001 picture compared to 1987. The reason is not because of a vegetation increase from 1987 to 2001, but is due to seasonal changes, with the 2001 image taken in the rainy season. For the same reason, we see that the fields which become bright orange in the 2001 image where bare land in the1987 image (circle under the lake).

Image Classifications

Supervised classification is a method that “trains” the computer to recognize the spectral signature of each image pixel as belonging to one of a number of learned signatures. Training sites are developed by locating (in video) pixels, which are examples of each land cover that is to be classified based on the interpreters’ knowledge of the scene. Here are some classification results of the 1987 (on the left) and 2001 (on the right) images using the following classes:

• dense forest (green)
• sparse/secondary vegetation (yellow)
• savannas (light brown)
• croplands/regrowth (dark pink - 2001)
• bare soil (beige)
• water (blue)

Classification 1987	Classification 2001

When we compare the two images, we notice the replacement of dense vegetation (green) by savannas or secondary vegetation in 2001. This process known as bush encroachment is common were the primary forests have been cut down. We can also clearly see areas overtaken by deforestation or sparse vegetation after the 14-year period.

This vegetation transition is more clearly seen in the change analysis image.

Change of dense vegetation

In 2001, all dense vegetation loss is shown in red. The seasonal differences are noticeable through the replacement of bare soil (beige) by cropland and by savannas, and in few cases by dense vegetation.