Diagnosing structure and composition typologies in uneven-aged broad-leaved forests: a comparison of classification methods

Structure and composition of forest stands are crucial factors for forest planning and biodiversity management. In Belgium, typologies of structure and composition exist to support planning in uneven-aged broadleaved forests (typically dominated by oak and beech). The principle of these typologies is to classify irregular stands with the percentage of small, medium, large, and very large trees (regarding dbh), and the percentage of basal area of oak and beech. This paper investigates the potential of LiDAR data processed with classification methods (k-nn, K-Means, CART, etc.) to allocate a forest structure and composition type. For this purpose several supervised and unsupervised classification methods are compared, as well as the impact of leaf-on (summer) and leaf-off (winter) data to discriminate the forest types