Change Detection of Tropical Mangrove Ecosystem with Subpixel Classification of Time Series Hyperspectral Imagery

Abstract

This chapter aims to use hyperspectral imagery to categorize saline blank classes amidst mangrove mixtures and analyze its changing patterns in the Sunderban Mangrove Forests of West Bengal, India. This research derives fractional abundance of mangrove endmembers at subpixel level with Fully Constrained Linear Spectral Unmixing (FCLSU) based on Least Square Error optimization criteria. NFINDR algorithm has been applied on time series hyperspectral image data of 2011 and 2014 to recognize pure saline blank and mangrove endmembers in the thickly forested study area followed by FCLSU to estimate mangrove species distribution maps of 2 years. The estimates in location 21° 34′ 24.81′′N and 88° 17′ 36.89′′E indicate a pure saline blank patch showing 74.47% occurrence with Phoenix paludosa, Avicennia alba, and Ceriops decandra showing 9.87%, 12.67%, and 2.99% presence in 2011. In 2014, the coordinate shows an increase in occurrence of saline blanks and Ceriops decandra but reduction in Phoenix paludosa and Avicennia alba. Ceriops decandra are salt-tolerant mangrove species that show an increase in abundance with increase in saline blanks. Phoenix paludosa which is salt intolerant shows a decrease in abundance with increase in saline blank areas. It is observed that mangroves, namely, Excoecaria agallocha and Ceriops decandra, are common and dominant around the saline blank areas. Salt-tolerant mangroves such as Avicennia marina and Avicennia alba are also observed to survive in certain locations of saline blanks.