Object-based Monitoring of Land Cover Changes in Mangrove Ecosystems of Senegal

This paper presents the results of object-based techniques applied to mapping and hi-temporal change detection of mangrove ecosystem in Low Casamance, Senegal. The methodology was applied on SPOT images for 1986 and 2006. Reference data was based on ground truth and visual verification of very high resolution images. High accuracy was obtained in mapping of mangrove cover. Other land cover classes showed however lower accuracy. The segmentation process for the change detection analysis was executed on a multi-date image and classification was then completed with a standard nearest neighbour classifier. Change detection analysis in mangrove ecosystems proved to be particularly difficult due to underlying presence of water and tide influences. Unlikely most mangrove ecosystems worldwide, important negative trends were not identified and low dynamicity was recorded during the two dates. This information questions the generalized preconception over mangrove retreat. The methodology will be further extended to Sine-Saloum the other main mangrove ecosystem of Senegal and should provide updated mangrove inventory for the country.