Analysis of Red Edge Position and Spectral Unmixing of Hyperspectral Data for Identification of Mangrove Species

Abstract

Application of hyperspectral data for species level mapping of mangroves is an area of active research in the present date. Linear Spectral Unmixing (LSU) of hyperspectral data has been successful in mangrove discrimination at sub pixel level. However, it only provides information about the fractional abundance of end members within each mixed pixel whereas their location within the pixel is unknown. This study has used spectral profile and data of pure end members extracted by automated target generation algorithm (NFINDR) and applied them to determine the fractional abundance of sub pixels in pure and mixed pixels. The locations of sub pixels have been determined using Simulated Annealing Algorithm. The study has been successful in finding out the spatial distribution of mixed mangrove species within a pixel and has greatly helped in obtaining classified maps at finer spatial resolution.The accuracy has been validated from field visits made in the study area. This paper also attempts to show that the red edge position of different mangrove species vary and thus can be an effective parameter to classify and distinguish different mangrove types in this study. Comparison of ground reference map and the classified output generated from REP suggests some degree of success in identifying species such as Ceriops, Excoecaria Agallocha, Avicennia Marina and Avicennia Alba. Analysis of classified output and accuracy assessment shows that LSU shows better classification accuracy than red edge values. LSU has been able to successfully discriminate between 8 dominant mangrove species of the study area namely, Excoecaria, Aegialitis, Avicennia Alba, Avicennia Marina, Ceriops, Phoenix, Avicennia Officinalis and Casurina.