Development of a spectral‐pattern‐analysis‐based method for automated water body extraction using Landsat image data: A case study in central Vietnam and southern Laos

Abstract

AbstractWater body mapping is an important application of optical remote sensing. Methods for classifying water bodies using multispectral image data have been successfully developed for water resource monitoring and management. However, in most cases, these traditional methods provide only partial automation capabilities. In this study, we propose a new spectral‐pattern‐analysis‐based (SPAB) method for water body extraction using simplified spectral patterns (SSPs). Simplified spectral patterns are a generalized transformation of spectral patterns from analogue to digital format, realized by nonrepetitive pairwise comparison of reflectance values between two bands. Simplified spectral patterns allow for the direct incorporation of spectral patterns into the recognition process. The advantage of this method compared to traditional methods is that the SPAB method allows automation of water body classification using an SSP database. In this study, we selected a mountainous region in central Vietnam and southern Laos as a study site. Four Landsat 8 OLI scenes from 2015 and four Landsat 7 ETM+ scenes from 2001 and 2002 were used. The results of our proposed method were compared with visual interpretation, normalized differential water index (NDWI) thresholding, and the Global Inland Water (GIW) dataset provided by the Global Land Cover Facility of the University of Maryland. It is concluded that the classification results of the SPAB method agree by more than 98.0% with NDWI results for Landsat 8 OLI images, and by more than 95.0% with the GIW dataset for Landsat 7 ETM+ images.