Taxonomic analysis of airborne laser scanning data for the recognition of stand vertical structure and its spatial variability

The aim of the research was to develop and verify the usefulness of the algorithm for grouping multi−feature objects known as Ward’s taxonomic method for distinguishing homogeneous spatial structures in a forest stands. Scots pine forest stand (20.137720 E, 51.754974 N) was divided into 10 zones differing in the share of undergrowth and up−growth of deciduous species, and the presence of single oak trees growing to the upper layer. The LiDAR data obtained in the first days of May 2007 was used. 459 sample plots of a radius of R=11.28 m located in a 20×20 m grid were set up in the study area. Histograms of spatial distribution of airborne laser scanning pulses were made for them. The histograms are described using eight variables: (1) height above ground level of the highest point of reflection of the laser pulse, (2) height above the upper maximum of the histogram, (3) height above ground level of the lower maximum of the histogram, (4) share of pulses reflected in the upper maximum of the histogram in the total number of pulses, (5) share of pulses in the area of the lower maximum of the histogram in the total number of pulses, (6) share of pulses of the first histogram layer in the total number of pulses, (7) share of pulses in the lower layers of the histogram (from layer no. 2 to no. 11, i.e. from 0.5 to 5.5 m) in the total number of pulses in layers 1−11, i.e. from 0 to 5.5 m, as well as (8) average height above ground of the reflection places of all pulses. The measure of the similarity of the extracted structures was the Euclidean distance of the points representing the histograms in the multidimensional feature space. The results of the division (segmentation) of forest space using the Ward’s method into 2 to 9 groups were analysed. It was found that the obtained divisions are related to the tree stand zones distinguished during the inventory carried out in the field. The proposed taxonomic method of LiDAR data analysis allows for the identification of homogeneous spatial structures in forest stands