Efficacy of AVIRIS-NG data for species-specific recognition towards a comparative analysis by hyperspectral classifiers

Due to increasing anthropogenic pressure on the forest ecosystem, many plant species face a threat, and many of them are rapidly getting extinct. Proper identification of tree species and understanding their geographical distribution are prerequisites towards its conservation and developing management plans. Species mapping and distribution have typically had restrict due to limitations in spectral information. Recently, hyperspectral imageries have solved this problem, but more satisfactory resolution and exact identification by different methods are still in discussion. Here, we used airborne AVIRIS-NG hyperspectral data to identify-five tree species, namely Butea monosperma, Diospyros melanoxylon, Tectona grandis, Terminalia bellirica, and Terminalia tomentosa. In this study, Spectral Angle Mapper (SAM) and Spectral Information Divergence (SID) classifiers identify these tree species using spectral signatures with 92.44% and 74.55% overall accuracy; and 0.89 and 0.67 kappa coefficient, respectively. A comparative analysis is executed for both applied species identification algorithms, which shows importance of used identification techniques for a specific dataset and a relationship was also stabilized between overall accuracy of tree species and number of ground samples. (c) 2023 COSPAR. Published by Elsevier B.V. All rights reserved.