Delineation of individual tree crowns for LiDAR tree and stand parameter estimation in scottish woodlands

There is an increasing need for accurate forest inventories to assist forest managers and decision makers in the planning of the forest resources. Airborne LiDAR methods enable the construction of Tree Canopy Models (TCM) at a fine resolution, which allow the delineation of individual tree crowns. This information can be useful for the prediction of forest parameters such as top height, basal area, standing volume and biomass. In this paper, we present a comparative analysis of the algorithms developed independently by Gougeon (1995), Popescu (2003) and Weinacker (2004a) for delineating individual tree crowns and as a means to extract forest parameters. The comparison was achieved as follow. Firstly, the algorithms were tested in their efficiency for delineating tree crowns. Secondly, single tree parameters were estimated using the crown delineation and finally, stand parameters were estimated by averaging single tree parameters. Results of the three algorithms were compared to each other and to field measurement for validation. The results showed that the algorithm by Popescu was the most suitable method to delineate crowns with 89% accuracy. However only 72% were linked with actual trees measured in the field. The algorithm by Popescu was the most suitable to estimate individual tree height with s RMSE (%) of 1.93 m (8.1%). The algorithm by Gougeon was the most suitable to estimate individual crown diameter and stem diameter with a RMSE (%) of 1.81 m (31.7%) and 7.05 cm (21.8%) respectively. The algorithm by Popescu was the most suitable to estimate top height with a RMSE (%) of 0.94 m (3.8%). Finally, the algorithm by Weinacker was the most suitable to estimate stand basal area and volume with a RMSE (%) of 9.10 m(2)/ha (24.3%) and 119.7 m(3)/ha (29.4%) respectively. All the methods underestimated the tree and stand parameters. However, it was shown that individual tree heights and stand top heights could be estimated with similar levels of accuracy than field methods.