Talks and Poster Presentations (with Proceedings-Entry):
L. Eysn, M. Hollaus, K. Schadauer, A. Roncat:
"A novel method to calculate crown coverage based on ALS data";
Talk: 1st EARSeL SIG Forestry workshop: Operational remote sensing in forest management,
Prague, Czech Republic;
- 2011-06-03; in: "1st Forestry Workshop: Operational Remote Sensing in Forest Management",
The delineation as well as the classification of forests has a long tradition in Remote Sensing. Considering different forest definitions (e.g. Austrian forest law, FAO) forest land can for example be composed of tree crowns, forest gaps, forest streets or harvested areas. This complex land cover class "forest" is often difficult to derive from remotely sensed data with high accuracy on a reliable and comprehensible way. One significant parameter of forest definitions is the crown coverage (CC), which defines the percentage of the ground covered by tree crowns. Furthermore, the CC is an important parameter to describe for example the forest structure or the compactness of a forests canopy. However, for the calculation of the CC it is necessary to define a reference ground area i.e. their size and shape. Unfortunately, the size and shape of the reference area is not clearly defined in most of the forest definitions which makes the CC often to a doubtful criterion.
Airborne laser scanning (ALS), as an active remote sensing technique, is not influences by shadowing effect or different sun illumination conditions and is able to deliver reliable information even for small forest gaps. The normalized digital surface model (nDSM), calculated by subtracting the terrain model from the surface model provides an excellent data source for calculating the CC. Using the nDSM a height threshold can be applied to decide whether a pixel is covered by tree crowns or not. In a next step the CC can be calculated by dividing the reference area by the tree crown covered area. As reference area forest stands or moving windows with user defined circular or squared kernel shapes are commonly in use. Due to the lack of precise geometric descriptions of the CC the derived results are often not comparable.
In this contribution a geometrically unambiguously defined approach to calculate CC from ALS data is presented. The basic assumption is to define the CC as a relation between the sum of the crown areas of three neighboring trees at a time and the area of their convex hull. The crown diameters are assessed using empirical functions describing the relation between the tree height and the crown diameter. These functions are calibrated based on the national forest inventory data. The tree heights are extracted from the nDSM whereas the tree positions are detected with a local maxima filter. A Delaunay triangulation is used to find the tree triples for calculating the CC. The approaches are applied for different test sites in Tyrol, Austria and are compared with forest masks generated by the local foresters considering the forest definition of the Austrian forest law. Furthermore, the derived results are compared with forest masks that are generated from moving window algorithms using different kernel shapes and sizes.
LiDAR, forest law, forest definitions, forest area, forest border delineation
Electronic version of the publication:
Created from the Publication Database of the Vienna University of Technology.