LANDSAT 9 IMAGERY BASED MANGROVE FOREST MAPPING USING MAXIMUM LIKELIHOOD CLASSIFIER AND SUPPORT VECTOR MACHINE ALGORITHM

Abstract

The successful launch of Landsat 9 in 2021 emphasizes the commitment to the continuity of the Landsat mission to provide global medium resolution satellite records. The benefit of the availability of the Landsat dataset is the extraction of information related to mangrove forest cover through mapping, especially in Gili Lawang. The main challenge in mangrove mapping is selecting a classification method that provides the most accurate results. This article aims to explore the use of the Maximum Likelihood Classifier (MLC) and Support Vector Machine (SVM) algorithms in the context of mangrove mapping using Landsat 9 imagery on Gili Lawang. The two algorithms show differences in detecting the area of mangrove, non-mangrove and water bodies, respectively for SVM 371.89 ha, 58.11 ha, 3.73 ha; while the MLC algorithm is 328.88 ha, 98.41 ha, and 6.45 ha. The MLC algorithm identifies non-mangrove objects at the outer boundary of Gili Lawang, while SVM identifies the same area as mangrove objects due to the influence of the training area created and low separability. The SVM algorithm has better accuracy with a Kappa statistic of 0.85 compared to MLC with a value of 0.73.

Keywords: Gili Lawang, Landsat 9, Mangrove, Support Vector Machine, Maximum Likelihood Classifier