MANGROVE DETECTION FROM HIGH RESOLUTION OPTICAL DATA Emmanuel Christophe, Choong Min Wong, Soo Chin Liew Centre for Remote Imaging, Sensing and Processing , National University of Singapore 1. INTRODUCTION Mangroves are a speciļ¬c type of forest present in tropical regions. They develop in the intertidal region along coastal areas and are present extensively in Southeast Asia. Besides the unique and interesting biodiversity of these forests, mangroves provide valuable environmental services such as coastal protection [1]. In addition, mangroves are also forests with high levels of productivity and have been exploited extensively by man for its numerous natural products. Due to increasing losses of mangroves from exploitation, it is necessary to monitor the status of mangroves regularly to determine the sustainability of exploitation, as well as the preservation of existing mangroves. Due to the extensive coastline present in Southeast Asia (Fig. 1), the work required to map and monitor mangroves for the whole region is very tedious. Moreover, most mangroves are located in developing countries, where changes to the land cover can be quite rapid. Therefore easing the task of mapping would speed up the production of data over a huge region such as Southeast Asia. (a) (b) Fig. 1. Area of interest: mangroves location in Southeast Asia (a) and one example of the desired output over a small example area on the west coast of Malaysia (b) (result of manual extraction from low resolution images) 2. OBJECTIVES OF THE ALGORITHM Currently, in a yet to publish study, most of the extraction is done manually as no satisfactory algorithm are available. As the region of interest is huge, it is not possible to process the whole region at the highest resolutions available. A reasonable compromise between resolution, accuracy and the speed of manual map production, for the whole region of Southeast Asia, is to use a resolution of 80 to 100 meter per pixel. This relatively low resolution has an impact on the detection of mangroves on satellite imagery. To provide accurate results, as well as locality information, it is important to produce higher resolution maps (e.g. 20 meters).