Context and objectives
The study aims to contribute to a better and systematic knowledge on the capabilities of ERS-SAR.PRI for forest monitoring in Central Africa.
Expected scientific results
The results showed the high capacity of ERS.SAR images for a forest/non-forest distinction, but the limited potential to distinguish different types of woody vegetation. No discrimination within the superclass 'forest' was possible. Plantations could not be distinguished. Distinctions of woody vegetation types in terms of soil cover could not be made. Addition of the texture information did not enhance the classification accuracies. These result correspond with other studies that mention the poor sensitivity of ERS.SAR for biomass. This is caused by its low incidence angle, 23° and its specific wavelength, the C-band. Higher wavelengths or radar interferometry could offer new possibilities.
Nevertheless a discrimination of shrub savannah and forest was possible for the study area in the humid tropics (Congo). Only a slight confusion existed for shrub savannah with a soil cover larger than 40 %. The multitemporal data sets of ERS. SAR images also allowed a good distinction of dry and humid grass savannah, of burned savannah and of certain agriculture types. Hence, additional information to optical satellite images could be provided. However, the pixel-based fusion of the optical and radar images gave no satisfactory results. Another method to combine multi-sensor data, like a decision-level fusion, should yield better results.