c. 2 published into 111 leaves

A thesis in partial fulfillment of the requirement of the degree of Master of Science

Thesis (M.Sc.) - Asian Institute of Technology, 2005

Horizontal changes in shoreline positions in Pak Phanang in southern Thailand has been monitored using Remote Sensing and Geographic Information Systems. Topographic maps, Landsat MSS, TM and ETM+ imageries have been used to detect shoreline changes over a period between 1973 and 2003 on a 58 kilometer stretch of the coast. Map and imagery data were both temporally and spatially inconsistent. The coast was found to be geomorphologically very active showing geographically and temporally alternating trends of recession and pro gradation of the shoreline. The influence of the strong northerly long-shore current carrying sediments from the south to the north was prominent during the last decade of observation. A small net deficit of sediment in the coast is observed. Occurrence of alarmingly high rates of erosion (about 2.8 ha/km of coast or about 34m/year horizontal shift during 1994-2003) was observed, with corresponding accretions at other places (about 1.7 ha/km). Accuracy of the estimates based on remote sensing data was assessed using data from a recent survey. It is found that relatively rapid changes in shoreline (>20m/year) can be detected at fairly accurate levels, while detection of the small changes are not as accurate. Therefore, this kind of study is suggested to be useful as a low cost and rapid reconnaissance of a long coastal region, and alternative means of measurements can be deployed at places found to have significant rates of shoreline changes. Success and failure of manmade erosion control structures were observed. Based on the shifts of shoreline positions over years, the coast was classified into six different types, viz. severely eroding, recently eroding from a previously stable condition, recently eroding from a previously accreting condition, mildly eroding, and mildly accreting with noisy movements. Agglomerative Hierarchical Clustering was found to provide better results in coast classification than an Artificial Neural Network (ANN) based clustering, probably because of the limitations in parameterizations in that particular implementation of ANN. Impact of shoreline smoothing was examined. Except in case of very small changes, generalization was found to produce fairly accurate results. Three software modules were developed during the study. An Arc View extension to facilitate accurate viewing of shoreline changes, another Arc View extension to facilitate vector based land dynamics analysis, and a Visual Basic MapObjects application for easy viewing of GIS data by non-skilled end users were developed.