The Use of Polarized L-band Alos Palsar for Identifying Forest Cover in Peninsular Malaysia

Currently there are about 5.89 million ha of forest occur in Peninsular Malaysia, which covers about 44.7% of its land area. Out of this, 4.92 million ha falls under Permanent Reserved Forest (PRF), 0.58 million ha is Totally Protected Area (TPA) and the remaining 0.39 million ha belongs to state/alienated land. These figures are officially used by the Forestry Department Peninsular Malaysia (FDPM) in its reports. These official figures are derived from several mapping techniques from several responsible departments in Peninsular Malaysia for management and reporting purposes, which are sometimes not representing the forest cover on the ground at real state. Information on the distribution of these forests is critical for decision-making and significant in climate stabilization, biodiversity conservation and social-related issues. The majority of tropical nations need high-resolution, satellitebased maps of their forests managing and maintaining their forests intact and therefore addressing their forests accordingly. This study was conducted to identify the extents of forests in Peninsular Malaysia, which comprise mainly inland dipterocarp, peat swamp and mangroves. Advanced Land Observing Satellite (ALOS) featuring the Phased Array L-Band Synthetic Aperture Radar (PALSAR) was used for this purpose. Fine Beam Dual mosaic of PALSAR images that has spatial resolution of 25 m was acquired in year 2010 and was provided by the Japan Aerospace Exploration Agency (JAXA) through Kyoto and Carbon (K&C) Initiatives. A combination of texture measure and polarizations manipulation of dual polarizations, HH and HV from PALSAR was applied to the images. Object-based classification or segmentation method was adapted to classify forest cover in the study area. The ability of PALSAR to support large area of tropical forest cover was assessed. The strong performance achieved indicates that the L-band dual polarization from PALSAR has greatly improved forest identification and classification. The result also indicated that the use of texture measure on dual-polarization were able to average out random speckle noise, and the use of ratio instead of absolute quantities, due to its well known ability to reduce forest structural and terrain effects.