Automatic identification of saline blanks and pattern of related mangrove species on hyperspectral imagery

Abstract

This research attempts to apply hyper spectral imagery to identity saline blank patterns within mixed mangrove forest of the Sunderban Bio-geographic Province of West Bengal, India. Identification of patterns of changing mangrove species around saline blanks on hyperspectral image has also been attempted. This research uses derivative analysis to identify hyperspectral wavelengths that are sensitive to the presence of minerals comprising saline blanks. These wavelengths have been considered for development of a new saline blank identification model. Spectral signatures of mangrove species endmembers have been automatically extracted using NFINDR algorithm and considered as input for spectral unmixing of saline blank and mangrove endmembers. The fractional abundance estimates of spectral unmixing applied on time series hyperspectral data of the study area indicate that, over the years, as saline blanks become more abundant in the region, the presence of mangrove species (even the salt tolerant ones) gets reduced. However, in areas where the mangrove species still dominate and saline blanks are in its formative stage, there is still a possibility of re-growth of the saline prone mangrove species. It is found that mangrove species namely, Excoecaria agallocha and Ceriops Decandra are prevalent around the saline blank areas. Others like Avicennia marina and Avicennia alba also exist in certain locations of saline blanks.