Mapping savanna tree species using carnegie airborne observatory hyperspectral data resampled to worldview-2 multispectral configuration

The advent of hyperspectral remote sensing has provided new opportunities for species mapping. However, the high dimensionality of hyperspectral data limits the application of parametric classifiers for species mapping because of the demand for a large number of training samples. This situation could change with the arrival of new spaceborne multispectral sensors such as WorldView-2 and RapidEye designed with new bands in the yellow and red-edge spectrum. We assessed the spectral configuration of WorldView-2 for discriminating eight savanna tree species in the Kruger National Park, South Africa. Carnegie Airborne Observatory Alpha imagery (72 bands in the visible-Near infrared sampled at 9.23 nm and spatial resolution of 1.12 m) acquired was spectrally resampled to WorldView-2 consisting of 8 bands in the visible-near infrared. The results showed a higher classification accuracy (77%) for maximum likelihood classification involving all WorldView-2 bands compared to the traditional blue, green, red and NIR bands (61.8%).