Comparison of Normalized Difference Water Index (NDWI) and Sobel Filter Methods in Landsat 8 Imagery for Coastline Extraction

Abstract

For a country, the existence of a coastline has an important value because it acts as a protection for marine resources and determines maritime boundaries between countries. There are various methods used for coastline analysis, both manual by digitizing and automatic by edge detection. This study compares the Normalized Difference Water Index (NDWI) method with the Sobel filter for coastline extraction in the South Coast of Sampang, Madura using Landsat 8 imagery. The best approach is then applied to the image to determine changes in coastlines from 2015 to 2020. This research shows that visually, the NDWI method produces better edges than the Sobel filter because the resulting lines are close to the original conditions in Landsat 8 or Basemap World Imagery. Sobel filter, the resulting accuracy is not very good. It does not approach field conditions, but this filter has the advantage of a relatively fast processing time because it can use a single band. Then the NDWI value generated in this study has a range of -0.497121 to 0.377046. The first class, which is a non-water body object, has a value of -0.497121 to 0. Then the second class, which is a body of water object, has a value of 0 to 0.377046. The coastline change for five years shows a shift in the coastline with a range of 0.62 to 2.75 meters. The Landsat 8 pixel size is 30 meters, while the shift is only <3 meters. So that this experiment does not show any significant coastline changes. Suggestions for further research: It is necessary to conduct a study using high-resolution imagery to confirm changes in the coastline accurately.