Document: RIC-3-89-2

Accuracy of a computer-based tree canopy delineation.

LANDENBERGER, R.E.*, J.B.MCGRAW and T.A.WARNER

West Virginia University, Morgantown, WV 26506 USA ¹

Abstract:
High resolution digital imagery, with a pixel size of less than 1 meter, offers great potential for delineation of individual deciduous trees in heavily forested, closed canopy stands. Automation of the delineation process will allow rapid censusing of canopy trees for population and community studies. Textural canopy patterns, characterized by shadows (low DN values), exist within and between individual deciduous crowns. Large, 'deep' shadows distributed along the edge of trees provide a skeleton outline from which crown segmentation can begin. An automated route selection method using a least cost path based on image DN values was developed and tested in two deciduous forest types, a Liriodendron tulipifera - Prunus serotina dominated mesophytic stand and a nearby Quercus alba - Q. rubra dominated mesophytic stand of known species composition, crown location, and crown juxtaposition. Delineation of individuals within the Quercus dominated stand was more accurate than in the Liriodendron - Prunus stand. Liriodendron and Prunus form densely packed patches, are more likely to be of stump sprout origin (appear to be two individual crowns as imaged from above the canopy), and subsequently offer few consistent shadow patterns between individual crowns. Because of their relatively isolated, individually discrete crowns and their relative lack of deep shadows within crowns, Quercus spp. are more likely to be accurately delineated by the automated least cost path method.

Keywords: remote sensing, canopy delineation, mesophytic forest, West Virginia

This abstract is being presented at: 10:30 AM in session:
REMOTE SENSING