| dc.description.abstract | Currently, knowledge on conditions of forest and land cover as well as its dynamics changes, particularly in the form of deforestation and forest or land degradation in the region is highly important, especially to support sustainable forest management. Detection of forest and land condition could be performed either using terrestrial, remote sensing or a combination between remote sensing and terrestrial approach. In Indonesia, since the 1990s, the mapping and monitoring forest and land cover condition is generally performed using optical satellite remote sensing approach. However, in the tropics, the use of optical images often have serious obstacle due to the presence of clouds and haze in the rainy seasonand the presence of smog and smoke during the dry season. However, since the presence ofradar images, a newperspective is coming, which may give complementary and/or substitution information over optical image. Satellite ALOS (Advanced Land Observation Sattelite) with sensor PALSAR (Phased Array L-band Synthetic Aperture Radar) is a radar image that had been launched in 2006 by the Government of Japan. They have four polarizations, namely HH, HV, VH, and VV. However, the image released to the public is commonly consisted of only HH and HV polarization. The study objective are (1) to identify forest and land cover and (2) to detect patterns of change in forest and land cover in the Western Part of Jambi Province using 50-meter ALOS PALSAR. The results of this research are expected to be used for supporting the preparation of MRV or REDD++ as well as for basic information in regional spatial planning. The method used in this study includes (1) image classification using qualitative classification approach (visual) on the composite image ALOS PALSAR (HH-HV-HH/HV) of 2007 and 2009, (2) image classification using quantitative classification (digital analysis) with the supervised method and (3) post-classification comparison by comparing 2007 and 2009 classified imageries. For ground checking and accuracy analysis, the location of ground observations were determined using systematic sampling with random start with the distance range from 100 meters to 500 meters from the edge of the road. The study results show that there are 13 classes of forest and land cover that can be identified through visual classification of 50-meter ALOS PALSAR composite image, namely water body, dry land forest, swamp forest, dry land farming, mixed farms/crop, rubber plantations, oil palm plantations, bush, shrub swamps, marshes, settlements and bare area. The qualitative classification approach provides Kappa accuracy of 82,38% for 2007 and 83,48% for 2009. With quantitative classification approach, slightly lower accuracy and separability were identified, where the average of inter-class separability value of Transformed Divergence for the image of 2007 and 2009 are 1641.7 and 1611.78. The Kappa accuracy with quantitative approach for 2007 is 53,3% and for 2009 is 58.8%. Based on the pattern distribution analysis of forest and landcover changes occurred between 2007 and 2009, it was found that the pattern of forest and land cover changes within the study site is clustered particularly in Tebo and Sarolangun Regencies. Deforestation of 14,583.5 hectaresis mostly occurred in Tebo Regency, while forest and land degradation of 399.8 hectares is mainly found in Tebo Regency. This study conclude that 50-meter ALOS PALSAR images could be used for detecting forest/land cover as well as deforestation and forest degradation. With skillful interpreter, the visual interpretation provides better accuracy than semi-automated (digital classification). | en |
