EVALUATING AIRBORNE HYPERSPECTRAL IMAGERY FOR RANGELAND ASSESSMENT IN SOUTH TEXAS

Authors: Everitt, James; YANG, CHENGHAI - TX AGRIC EXP STN-WESLACO; Davis, Michael

Submitted to: Geocarto International
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 12/5/2003
Publication Date: 7/20/2004
Citation: Everitt, J.H., Yang, C., Davis, M.R. 2004. Evaluating airborne hyperspectral imagery for rangeland assessment in south Texas. Geocarto International Journal 19(3):23-31.

Interpretive Summary: Interest in the use of airborne electronic imaging systems as remote sensing tools has greatly expanded over the past several years. A 128-band hyperspectral imaging system was evaluated for distinguishing among a diversity of rangeland ecological variables on two study sites in southern Texas. Field reflectance measurements made on the dominant vegetation types on the two sites identified six wavelengths where significant spectral differences occurred among the parameters. Spectral bands from the hyperspectral system that corresponded to these wavelengths were used to develop false color composite images of the study sites. Accuracy assessments performed on the classification maps of seven false color composite images of the two sites had accuracies ranging from 74% to 85%. These results should be useful to rangeland resource managers who are interested in mapping vegetation cover types.

Technical Abstract: Airborne hyperspectral imagery was evaluated for differentiating among rangeland ecological variables on two range sites (tight sandy loam and shallow ridge) on south Texas rangelands. Field spectroradiometric reflectance measurements made on dominant vegetation (plant species and vegetation mixtures) and soil land-use types on the two sites identified six wavelengths (five visible and one near-infrared [NIR]) where significant spectral differences occurred among the parameters. The spectral bands that corresponded to these wavelengths were extracted from the 128-band hyperspectral images obtained of each site and used to develop false color composites that were subjected to computer classifications and accuracy assessments. The bands selected for the composites included the green (498.6 nm), yellow-green (549.7 nm), orange (629.6 nm), red (651.3 nm), dark red (680.4 nm), and NIR (825.5 nm). Accuracy assessments performed on the classification maps of four false color composites of the tight sandy loam site had overall accuracies ranging from 76% to 85%; the highest accuracy was obtained from the composite made up of the NIR, red, and yellow-green bands. Accuracy assessments performed on the classification maps of three composites of the shallow ridge site had overall accuracies ranging from 74% to 77%; the highest accuracy was acquired from a composite comprised of the NIR, orange, and yellow-green bands. Overall accuracies were considered good and indicate that several band combinations can be used successfully to differentiate among rangeland parameters in southern Texas.