AVIRIS-NG HYPERSPECTRAL DATA FOR MAPPING MANGROVE FORESTS AND THEIR HEALTH SPATIALLY: A CASE STUDY OF INDIAN SUNDARBANS

Abstract

The present work deals with the classification of mangrove forests and determination of their health spatially using Airborne Visual Imaging Infrared Spectrometer–Next Generation data over a portion of Indian Sundarbans. The objectives were to test pixel-based classifiers and object-based approach for classification of mangroves into floristic composition classes and map the health status of the forests. Five known spectral indices were applied in Decision Tree algorithm for identifying mangrove forests, followed by creation of a mangrove mask of the study area. Spectral Angle Mapper, Support Vector Machine (SVM), and Object-Based Image Analysis classifiers were separately evaluated on the data within the mask. The overall classification accuracy was highest in SVM (99%). A total of 10 floristic composition classes were obtained. The classes were further classified into three health classes, viz. most healthy, moderately healthy and less healthy/stressed using different vegetation indices for greenness, light-use efficiency, leaf pigments and canopy moisture contents. The methodology presented here holds good opportunity to be applied to other mangrove forests for producing mangrove health maps at finer levels.