One of the tasks of UNESCO is to endorse World Heritage Sites and Man and Biosphere Reserves. Of these areas, close to 60 of the World Heritage Sites and around 20 of the Biosphere Reserves are located in Tropical Forests. There is a need to monitor these protected areas, in order to make sure that illegal logging not is taking place and that the land is not used for other purposes than intended. These changes of land use are especially frequent in the edges of the protected areas. Furthermore, UNESCO has a need to verify that giving UNESCO-status for these tropical forests has stopped or slowed down the rate of deforestation compared to the situation before. The main objective for this study was to suggest a strategy for monitoring of forest changes in the context of World Heritage Sites and Man and Biosphere Reserves in tropical rain forest areas. Preferably, methods and recommendations should be worked out that also allows the change detection to be done at a local level, in the respective countries, and preferably also to be as automatic and easy as possible. In specific this would include recommendations about image selection, image processing steps and validation of results.
The most suitable type of satellite data is usually Landsat TM or similar. These images have large swath width, a spatial resolution that enables the detection of most man-induced land cover changes and are now also distributed free of charge from several sources, including the USGS in USA. Our screening of UNESCO forested sites in the tropics gave at hand that usually, even though these large swath width satellites were used, several scenes were needed in order to completely cover one World Heritage Site. Most often, images from more than one year were needed in order to obtain a cloud free data set for a whole World Heritage Area. When using satellites with a smaller swath width the problem to cover the sites with data from one time point grew even more. Cloud free satellite data for parts of a protected areas are however much easier to accomplish. Hence, it is with current data availability much more realistic to concentrate the analysis to small areas of importance, rather than expecting simultaneous full coverage of the protected areas.
To achieve the necessary image geometry to perform a digital change analysis, an image-to-image matching is usually needed. The geometry of the freely downloaded images are seldom good enough for direct comparison thus an image-to-image matching is recommended before change detection. We successfully used ERDAS Imagine Autosync 9.3 for this step.
Based on Rio Platano in depth analysis, Landsat TM band 2 (green) or band 3 (red) could be recommended for change detection of tropical forests, and especially for burnt areas also band 4 (NIR). To acquire radiometric relative normalization of the images, histogram matching on cloud masked images has been used. For the actual image differencing and change analysis, pixel wise, band wise subtraction of the late minus the early image, followed by manual thresholding has been used. It is a simple and easy to understand, and still one of the more powerful and also one of the most described methods around.
Most human induced abrupt changes in the landscape, like small clearings, are visible when imagery from different years are compared. The problem is instead to determine the type of changes.
Based on our findings it is our opinion that this type of change detection analysis should be done on a relatively local level, preferably at regional or at least, national level. This will make the access to local knowledge easier and the results or the analysis more legitimate. A hands-on recommendation for using remote sensing in the monitoring and evaluation of world Heritage and MAB areas is to construct a package of hardware and software needed to perform a change analysis. Add satellite imagery if it is not possible to download locally and the instructions to perform a change analysis. What should be added locally is then somebody to perform the analysis, add ground control points and the local knowledge to evaluate interesting phenomena’s.