Comparing the new generation world view-2 to hyperspectral image data for species discrimination

×

Error message

User warning: The following theme is missing from the file system: journalijdr. For information about how to fix this, see the documentation page. in _drupal_trigger_error_with_delayed_logging() (line 1138 of /home2/journalijdr/public_html/includes/bootstrap.inc).

International Journal of Development Research

Comparing the new generation world view-2 to hyperspectral image data for species discrimination

Abstract: 

Discriminating indicator species in mountainous rangelands is critical for better understanding the condition of the rangeland and their levels of degradation. The objective of this study was to compare whether canopy reflectance spectra, resampled to WorldView-2 and HyMap resolution could discriminate four increaser species representing different levels of rangeland degradation. Canopy spectral measurements were taken from the four indicator species: Hyparrhenia hirta, Eragrostis curvula, Sporobolus africanus, and Aristida diffusa. The random forest algorithm and a forward variable selection method were applied in order to identify optimal variables (HyMap and WorldView-2 wavelengths) for discriminating the species. Results revealed that 8 optimal wavelengths from HyMap and 6 from Worldview-2 yielded the lowest OOB error (15.82%) and (17.36%) for HyMap and WorldView-2, respectively, in discriminating among the four increaser species. The random forest algorithm could discriminate species with an overall accuracy of 84.1% (KHAT =0.79) using HyMap wavelengths and an overall accuracy of 82.9% (KHAT = 0.77) using the WorldView-2 wavelengths. Overall, the study demonstrated the potential of WorldView-2 in terms of cost effectiveness compared to HyMap data for mapping indicator species of rangeland degradation.

Download PDF: