A Canonical Variate Analysis Methode to Classify Forest and Non Forest using Landsat Image for Central Kalimantan Province in 2000-2008

The information of forest and non forest forCentralKalimantan Province in year 2000-2008 derived using Landsat 5 and 7 images. This research is conducted by Indonesia's National Carbon Acounting System (INCAS) project in order to support a mechanism scheme plan of decreasing emission from Reducing Emission from Deforestation and Forest Degradation/REDD. The INCAS activities in LAPAN are inventoring LAndsat data, scene selection, orthorectification, radiometric correction, cloud masking, mosaicing wall to wall per year, and classification. This paper is only discussed of classification of forest and non forest. The criterion of forest in this terms is a trees, having canopy cover more than 20%, and 2 meter tall. The data are Landsat 5 and 7 in year 2000 and 2008. The secondary data areIkonos, Quickbird, geology and ground truth data. The stages of process are decided training sample, forest base probability, forest matching, and stratification zones. Yhe mrthodology to make forest base probability is using canonical variate analysis (CVA) to derive the indices and threshold. The result are forest extend and changes per year