Potential of hyperspectral AVIRIS-NG data for vegetation characterization, species spectral separability, and mapping

The present study deals with hyperspectral species mapping, utilization of optimal bands, and studying the spectral separability of vegetation species using AVIRIS-NG data. To reduce data redundancy, a wide range of optimal spectral bands (491 nm, 541 nm, 641 nm, 722 nm, 772 nm, 852 nm, 942 nm, 1047 nm, 1132 nm, 1443 nm, and 2475 nm) were selected and simulated for classification using artificial neural network (ANN) and support vector machine (SVM) techniques. Band-to-band correlative plots demonstrated huge data redundancy (correlation as high as R2 = 0.99, at 682 nm and 687 nm) owing to the need of spectral sub-setting. ANN classification exhibited dominance of low-elevation (0–800 m) evergreen forest (44.81%) followed by Tectona grandis plantation (25.97%). Mid-elevation (800–1450) evergreen forest contributes ~ 16.79% of area, and primarily consists of Coffee arabica (3.14%) and Camelia sinensis (0.7%). Spectrally, each vegetation species exhibited distinct spectral characteristics plotted against different wavelengths and unveiled unique spectral signatures and their classification provided comparable accuracy in species identification in ANN (86.45%) as well as in SVM (86.75%). With immense potential and applicability, spectroscopic AVIRIS-NG data is highly recommended for characterizing, quantifying, modeling, and mapping vegetation.