Hyperspectral image data for mapping wetland vegetation

A. Hirano,M. Madden,R. Welch
DOI: https://doi.org/10.1672/18-20
2003-05-15
Abstract:Data acquired by the Airborne Visible/Infrared Imaging Spectrometer (AVIRIS) with 224 bands, each with 0.01-μm spectral resolution and 20-meter spatial resolution, were used to produce a vegetation map for a portion of Everglades National Park, Florida, USA. The vegetation map was tested for classification accuracy with a pre-existing detailed GIS wetland vegetation database compiled by manual interpretation of 1∶40,000-scale color infrared (CIR) aerial photographs. Although the accuracy varied greatly for different classes, ranging from 40 percent for scrub red mangroves (Rhizophora mangle) to 100 percent for spike rush (Eleocharis cellulosa) prairies, the Everglades communities generally were successfully identified, averaging 66 percent correct for all classes. In addition, the hyperspectral image data proved suitable for detecting the invasive exotic species lather leaf (Colubrina asiatica) that is sometimes difficult to differentiate on aerial photographs. The findings from this study have implications for operational uses of spaceborne hyperspectral image data that are now becoming available. Practical limitations of using such image data for wetland vegetation mapping include inadequate spatial resolution, complexity of image processing procedures, and lack of stereo viewing.
What problem does this paper attempt to address?