Estimating canopy cover from Eucalypt dominant tropical savanna using the extraction of tree crowns from very high resolution imagery

Very high spatial resolution satellite imagery provides data that enables spatially detailed analysis of landscapes. The identification and extraction of information about tree crowns is one such use. Tree crown or canopy cover is one parameter of vegetation structural classification. The estimation of canopy cover has a wide range if uses related to management and policies. Tree crown extraction of Eucalypts is not without its challenges. The inherent characteristics of Eucalypt crowns include open spaces within the crowns, verti-cally angled leaves, and the irregular crown shapes provide a number of challenges to remote sensing. This paper proposes an object-based method for extracting tree crowns from Eucalypt dominant savanna in the wet/dry tropics of northern Australia. A two level multi-resolution segmentation was undertaken upon QuickBird data (multispectral, panchromatic and derivatives) covering the area under investigation. The first broader segmentation allowed the differentiation of Eucalypt dominant communities from other vegeta-tion types. The second finer segmentation produced segments smaller than the tree crowns. A rule-set containing a series of classifi-cation and object growing algorithms were then used to firstly identify objects within a crown and then to expand the objects to cover the entire crown. Results indicate the potential of this method for delineating tree crowns from Eucalypt savanna and the use of this information to estimate canopy cover. The approach used here offers a method of tree crown delineation where the availability of other forms of data such as hyperspectral and laser scanning imagery may not be available.