Feature extraction of coastal surface inundation via water index algorithms using multispectral satellite on North Jakarta

Abstract

The study aims to characterize water index algorithms using Sentinel 2A-MSI and Landsat 8-OLI sensors to generate the performance of CSI and its proper uses in science. Feature extraction of the 35 combinations of CSI specifically uses information deriving from a number of water index spectral bands. Supervised classification is used to categorize water and non-water features using Spectral Angle Mapper (SAM). Furthermore, Dynamic Surface Water Extent (DSWE) algorithm was taken to mapping inundation and non-inundation. Result of this research shows that the best water index algorithms is exhibited by the MNDWI algorithm where it can enhance the water reflectance to be positive value. Re-classification on all water index algorithms indicates a similar pattern to the feature extraction of water and non-water. It is determined by overall accuracy (99.56-99.99%), Kappa coefficient (0.97-1.00), producer accuracy (99.05-99.99%) and user accuracy (93.43-99.94%). Substituting all water index algorithms into DSWE algorithm points out that the MNDWI is still the good performance and other substituted algorithms generate low performances. Overall feature extractions of CSI in North Jakarta at this research manifests that the MNDWI algorithm using multi-spectral and satellites is the best algorithm.