LIDAR-based geometric reconstruction of boreal type forest stands at single tree level for forest and wildland fire management

Abstract

Vegetation structure is an important parameter in fire risk assessment and fire behavior modeling. We present a new approach deriving the structure of the upper canopy by segmenting single trees from small footprint LIDAR data and deducing their geometric properties. The accuracy of the LIDAR data is evaluated using six geometric reference targets, with the standard deviation of the LIDAR returns on the targets being as low as 0.06 m. The segmentation is carried out by using cluster analysis on the LIDAR raw data in all three coordinate dimensions. From the segmented clusters, tree position, tree height, and crown diameter are derived and compared with field measurements. A robust linear regression of 917 tree height measurements yields a slope of 0.96 with an offset of 1 m and the adjusted R 2 resulting at 0.92. However, crown diameter is not well matched by the field measurements, with R 2 being as low as 0.2, which is most certainly due to random errors in the field measurements. Finally, a geometric reconstruction of the forest scene using a paraboloid model is carried out using values of tree position, tree height, crown diameter, and crown base height.