Rapid Assessment of Forest Cover in Indonesia using MODIS Satellite Data: Forest Fragmentation and HCVF Identification

Land surface conditions are changing rapidly in Southeast Asia as forests are cleared for plantations and as frequent wildfires burn vegetation in different parts of the region. Of particular concern is the recent loss of lowland tropical forest cover in Indonesia. This loss of forest not only exacerbates flooding in urban areas and increases soil erosion and sediment yield in coastal regions, but it also poses a severe threat to the region's terrestrial biodiversity. As part of a new effort to develop an operational, annual land-surface monitoring system (see http://www.ifmn.org for further information), composite images from the Moderate Resolution Imaging Spectrometer (MODIS) bands 1-7 were used from 2002, 2003, and 2004. These images were then classified using a semi-automated approach to identify different forest cover types. The accuracy of the classified images was assessed using 15 SPOT-5 false colour composite images and Landsat imagery. The error matrices generally indicated greater than 85 percent accuracy overall; although the accuracy increased to approximately 91 percent when large validation polygons (gt 5,000 ha) were considered. The remaining mid-2004 forest area for Indonesia was calculated using GIS data to reveal approximately 95.6 million ha. Using similar MODIS data, researchers at Center for International Forestry Research (CIFOR) estimated Indonesia's 2003 forest cover to be approximately 98 million ha. These two estimates suggest that Indonesia experienced a net forest loss of approximately 2.4 million ha in one year (i.e., an annual loss rate of about 2.5%). Thus, not only do MODIS data provide an excellent source of information for identifying large, intact forest remnants, which are important reservoirs of biodiversity, but they allow regular monitoring of forest cover to assess the long-term efficacy of conservation policies and practices.