An object-based method for mapping and change analysis in mangrove ecosystems

Object-based methods for image analysis have the advantage of incorporating spatial context and mutual relationships between objects. Few studies have explored the application of object-based approaches to mangrove mapping. This research applied an object-based method to SPOT XS data to map the land cover in the mangrove ecosystem of Low Casamance, Senegal. In parallel, the object-based method was tested to analyse the changes in the mangrove area between 1986 and 2006. The object-based method for mangrove mapping applied a multi-resolution segmentation and implemented class-specific rules that incorporate spectral properties and relationships between image objects at different hierarchical levels. The object-based approach for change analysis conducted the segmentation on the multi-date composite of the 1986 and 2006 images and applied a nearest neighbour classifier. The object-based method clearly discriminated the different land cover classes within the mangrove ecosystem. The overall accuracy of the land cover classification was 86%, the overall kappa value was 0.83 and the user's accuracy of the 'mangroves' class was higher than 97%. The estimated area of manoroves was 76,550 hectares in 2006. This result is an important update reference for mangrove studies in Senegal and the proposed method may represent a valid instrument for similar exercises in other regions. The image-to-image, object-based approach to change analysis clearly captured the fragmented and scattered pattern of change that prevails in the study area. The user's accuracy of the increase and decrease classes of transition produced results better than 85%. The overall accuracy. however, is lower due to the method's difficulties in detecting the small areas of change. To have conclusive evidence for the suitability of this method for change analysis of mangrove forest, this object-based approach should be tested in mangrove ecosystems where changes have different spatial patterns and modifications are more evident. Between 1986 and 2006, a small increase in the mangrove area was observed in Low Casamance. This was probably due to improved rainfall conditions after the droughts of the 1970s and 1980s. The pattern of change detected with the object-based approach corresponds to natural transitions and Suggests that anthropogenic influence was limited. (C) 2008 International Society for Photogrammetry and Remote Sensing, Inc. (ISPRS). Published by Elsevier B.V. All rights reserved.