Mapping blackberry thickets in the Kosciuszko National Park using airborne video data
Paul Frazier, School of Science and Technology, Charles Sturt University, PO Box 588, New South Wales 2678, Australia.
Summary
High resolution, multi-spectral airborne video data were used successfully to map blackberry thickets (Rubus fruticosis spp. agg.) in the Kosciuszko National Park. The digital data with a spatial resolution of one metre and spectral resolution which includes channels of blue, green, red and NIR light were capable of detecting patches of blackberry as small 2 x 2 m. Five different mapping techniques were compared including manual interpretation, thresholding the NIR band, thresholding a ratio of NIR/red bands, unsupervised classification, and supervised classification.
Manual interpretation was able to successfully identify 97% of the known blackberry sites with no errors of commission. The two thresholding techniques were able to highlight the areas of known blackberry but were not able to clearly differentiate between blackberry and woodland areas. Similarly, the unsupervised technique showed good agreement with the regions of known blackberry thickets but was unable to adequately separate blackberry spectral response from woodland spectral response. Supervised classification was the best of the digital techniques for discriminating blackberry from surrounding land cover types, and achieved a 79% success rate for identifying known blackberry sites.
Plant Protection Quarterly (1998) 13 (3) 145-148.