Spectral separability among invasive and native plant species for satellite image analysis

Abstract

Comprehending the spatial and temporal dynamics of the alien species infestation is critical to manage invasive plants in Hawai'i. Although recent advances in technologies have made a variety of remotely sensed images providing spatial and temporal information of invasive species available to resource managers, little spectral information are available for plant species on tropical island settings. In this study, we evaluated the capabilities of four major satellite sensors to distinguish nonnative species from native species of Hawai'i at leaf and canopy scales with the ultimate goal of establishing the baseline information for image analysis. The sensors were IKONOS, Quickbird, Landsat-7, Enhanced Thematic Mapper Plus and Advanced Spaceborne Thermal Emission and Reflection Radiometer. The spectral properties of nonnative species: Psidium cattleianum (strawberry guava), Schinus terebinthifolius (Christmas berry) and, Coffea arbacia (coffee) and native species: Acacia koa (koa), Diospyros sandwicensis (lama) and Metrosideros polymorpha ('ohia) were directly measured at the leaf scale, whereas the canopy spectral properties of the species were estimated using a radiative transfer model constrained with sampled biophysical and chemical components of the species. Statistical analyses determined the ability of these sensors to discriminate among different species. In particular, the sensors with the short wavelength bands resulted in the high discriminability. The relative capability of the spectral bands in determining species separability were different at the leaf and the canopy scales indicating the change in the primary biophysical characteristics affecting the separability from the leaf to the canopy scales. These results suggest the effectiveness of remote sensing techniques for invasive species management in tropical islands.