DETECTING FRAGMENTED MANGROVES IN THE SUNDARBANS, BANGLADESH USING OPTICAL AND RADAR SATELLITE IMAGES

摘要

S: This study shows that Landsat TM, JERS-VNIR (optical) and JERS-SAR (radar) data can be used for detecting edges of fragmented mangrove in the Sundarbans, Bangladesh. The classification JERS-SAR image gives highest overall classification accuracy of 85.53% and per class accuracy of 86.49% in detecting high contrast edges. Whereas overall classification accuracy of Landsat TM and JERS-VNIR image are 80.52% and 82.27% respectively. Mapping of high contrast edges can be done through producing segment map from on screen digitising of class high contrast edges on classification map of all three sensors mentioned above. It is believed that the high contrast edges, with their specific spatial characteristics and pattern (e.g., linear feature shape and network like pattern) can be detected spatially on remotely sensed data.