There is a limited understanding of how forest structure affects the performance of methods based on terrestrial laser scanning (TLS) in characterizing trees and forest environments. We aim to improve this understanding by studying how different forest management activities that shape tree size distributions affect the TLS-based forest characterization accuracy in managed Scots pine (Pinus sylvestris L.) stands. For that purpose, we investigated 27 sample plots consisting of three different thinning types, two thinning intensities as well as control plots without any treatments. Multi-scan TLS point clouds were collected from the sample plots, and a point cloud processing algorithm was used to segment individual trees and classify the segmented point clouds into stem and crown points. The classified point clouds were further used to estimate tree and forest structural attributes. With the TLS-based forest characterization, almost 100% completeness in tree detection, 0.7 cm (3.4%) root-mean-square- error (RMSE) in diameter-at-breast-height measurements, 0.9–1.4 m (4.5–7.3%) RMSE in tree height measure-ments, and <6% relative RMSE in the estimates of forest structural attributes (i.e. mean basal area, number of trees per hectare, mean volume, basal area-weighted mean diameter and height) were obtained depending on the applied thinning type. Thinnings decreased variation in horizontal and vertical forest structure, which especially favoured the TLS-based tree detection and tree height measurements, enabling reliable estimates for forest structural attributes. A considerably lower performance was recorded for the control plots. Thinning intensity was noticed to affect more on the accuracy of TLS-based forest characterization than thinning type. The number of trees per hectare and the proportion of suppressed trees were recognized as the main factors affecting the accuracy of TLS-based forest characterization. The more variation there was in the tree size distribution, the more challenging it was for the TLS-based method to capture all the trees and derive the tree and forest structural attributes. In general, consistent accuracy and reliability in the estimates of tree and forest attributes can be expected when using TLS for characterizing managed boreal forests.
|Tidskrift||ISPRS Journal of Photogrammetry and Remote Sensing|
|Status||Publicerad - okt 2020|
- 4112 Skogsvetenskap