Genus Level Classification of the Mangroves in Indian Sundarbans using Sentinel-2 Multispectral Imagery

Abstract

Listen

Translate article

Abstract Mangroves are the most productive ecosystems that provide stabilization to the coastlines, help in carbon sequestration, reduce storm surges, defend the coastal inhabitants and play a role in sustaining the local economy. Mangroves are halophytic in nature that typically thrive along tropical and subtropical coastlines in the saline intertidal zone. This paper explores the potentiality of Sentinel-2 MSI Imagery in separating different mangrove genus that has been evaluated using different classification algorithms like Maximum Likelihood Classifier (MLC), Mahalanobis Distance (MD), Minimum Distance (MD), Spectral Angle Mapper (SAM), knowledge-based classifier such as Random Forest (RF) and Support Vector Machine (SVM). The estimated accuracy is higher for Random Forest (88.50%) followed by Support Vector Machine (85.30%) and Maximum Likelihood Classifier (85.10%) as compared to Mahalanobis Distance (81.10%), Spectral Angle Mapper (71.20%), and Minimum Distance (75.30%). The result showed that Sentinel-2 Multispectral Imagery can efficiently discriminate 15 different mangrove genera distributed in the entire Indian Sundarbans and can simultaneously replace the barriers in terms of cost and availability of the spaceborne and airborne hyperspectral sensors.