Revista de Ciencias de la Tierra y Cambio Climático

Abstracto

Tree Species Discrimination using Narrow Bands and Vegetation Indicesfrom Airborne Aisa Eagle Vnir Data in the Taita Hills, Kenya

Samuel Nthuni1*, Janne Heiskanen2, Faith Karanja1, Mika Siljander2 and Petri Pellikka2

Tree species inventory and mapping are important for the management and conservation of forests. Especially in tropical forests, field based inventories are very tedious and time consuming. Therefore, the crown-level spectral data collected by the high spatial resolution airborne imaging spectroscopy provides promising possibilities for improving the accuracy and efficiency of tree species inventory and mapping. In this study, the feasibility of AISA Eagle VNIR data for spectral discrimination of indigenous and exotic tree species in the Ngangao forest in the Taita Hills in south-eastern Kenya was examined. The airborne AISA Eagle VNIR data (400-876 nm, bandwidth approximately 4.6 nm) was acquired in January 2013. The data was georeferenced and atmospherically corrected with a final spatial resolution of 1 m. The field data consisted of 152 samples from 10 species (six indigenous and four exotic species), which were mapped both in the field and from the AISA images. Stepwise Discriminant Analysis was used for tree species classification using three sets of inputs: (1) all narrowbands, (2) a combination of narrowbands and selected vegetation indices (VIs), and (3) simulated blue, green, red and NIR broadbands. According to the results, both the narrowbands and VIs provided a cross-validated overall accuracy of 77.0%. The simulated broadbands provided considerably lower overall accuracy of 38.2%, which emphasizes the utility of hyperspectral data in tropical tree species discrimination. High overall accuracy (92.8%) was attained when separating only exotic and indigenous species.