Conifer Discrimination in the Sandhills of North Carolina Using High Spectral Resolution Data

Abstract

We investigated techniques to discriminate long leaf pine (Pinus palustris) and loblolly pine (Pinus taeda) in 126 band HyMap imagery with a 4 meter spatial resolution. Field assessment provided stand composition information, and crowns of known species were selected in the imagery to represent species types for model construction. A Quadratic Discriminant Analysis used with a likelihood ratio test was able to identify southern yellow pine with a producer's accuracy of 98% and a user's accuracy of 96%. The same test identified loblolly pine with a producer's accuracy of 80% and a user's accuracy of 49%. Longleaf pine identification had a producer's accuracy of 60% and a user's accuracy of 76%. Price of image acquisition, the relatively low accuracy of discrimination between longleaf and loblolly pine crowns, and inherent bias in the approach make this particular method unreliable as an option for targeting potential sites for RCW habitat restoration.