Tropical forest mapping and change detection using ALOS PALSAR data

Abstract

Tropical forests represent a significant carbon store in the global carbon balance. Over recent years, substantial forests have been cut or burned. To investigate the changes of deforestation in tropical forests, ALOS PALSAR 50 m Orthorectified mosaic product (FBD) in Borneo from 2007 to 2009 was used, as SAR data can overcome the shortcoming of optical images. Pixel based unsupervised classification and object-oriented classification methods were applied to choose better method for classification. The comparison showed the object-oriented classification got the better result and its accuracy was higher. In the following step, object-oriented classification method was employed to monitor the changes of deforestation in Borneo during two years. The results indicated the deforestation increased sharply, especially the areas in the coastal and next to non-forest land cover/use. The change law of land cover/use in Borneo is forest to shrub, grassland and bare soil. There are also inverse changes from non-forest to forest by reforestation around settlement and water areas. Overall, forest decreases gradually and non-forest areas increases simultaneously in Borneo. This research indicates clearly that ALOS PALSAR Ortho-rectified FBD product can be used to classify land cover/use and two different time images could be employed in change detection without in situ data.