Author : Puvadol Doydee

Summary : Various forms of coastal landuse covering the study area has been observed to have undergone changes as evidently detected between the satellite images sensed in 1994 and 2001 at Banten Bay. It is important to identify what these changes are. therefore, an appropriate change detection must be selected. In this study, three main objectives were set : (1) To determine the image preprocessing and image processing techniques that is needed for digital coastal landuse change detection, (2) To perform digital coastal landuse supervised classification, and (3) To study the coastal landuse change of Banten bay in two dates. The image preprocessing step involved removing errors from the raster data. This was done performing basic processes, such as, radiometric correction, geometric correction and image calibration. the image processing step comprised of supervised classification and change detection techniques. Supervised classification was employed in this study to transform multispectral image data into user defined thematic information classes and to serve as a reference for the quantitative results of the change detection techniques. On the other hand, change detection techniques tested on this study to show the best results included Red Green Method, Image Differencing Method, Image Rationing Method and Principal Component Analysis Method (PCA). Red Green Method gave the best result for detecting the coastal landuse change because the number of changed area closely resembled the total number of changed area reference. Through careful comparison it was observed that Red Green method is suitable for detecting areas changes in the paddyfields increase and settlement increase; Image Differencing Method is better to detect areas changes in agriculture increase, fishponds decrease, and natural area decrease; Image Rationing Method gave the best result for monitoring areas change in fishponds increase, paddyfields decrease and agriculture decrease.