ABSTRACT
In recent years the term "precise forestry" has been used more and more often, referring to a modern and sustainable model of forest management. Functioning of such management of wood biomass resources is based, among others, on precisely defined and log-term monitored selected forest taxation parameters of single trees and whole forest stands based on modern geoinformation technologies, including Airborne Laser Scanning (ALS) and digital photogrammetry. The purpose of the work was the analysis of the usefulness of the CHM (Canopy Height Model) generated from the image-based point cloud or ALS technology to define the number of trees using the method of the segmentation of single Scots pine (Pinus sylvestris L.) crowns. The study was carried out in the Scots pine stands located in the Bory Tucholskie National Park (Poland). Due to the intentional lack of certain silviculture treatments, over the recent decades, these forest stands have been characterized by relatively high tree density, compared to managed forests. The CHM was generated from digital airborne photos (CIR composition; GSD 0.15 m) and on the other hand - from the ALS point clouds (4 points/m2; ISOK project). To generate point clouds from airborne photos using stereomatching method, the PhotoScan Professional (Agisoft) software was applied. The CHM coming from the Image-Based Point Cloud (CHM_IPC; GSD: 0.30 m) and ALS data (CHM_ALS; GSD: 0.75 m) were generated using FUSION (USDA Forest Service) software. The segmentation of tree crowns was carried out in eCognition Developer (TRIMBLE GeoSpatial) software. Apart from height models, also spectral information was used (so-called true CIR orthophotomaps; GSD: 0.3 and 0.75 m). To assess the accuracy of the obtained results, the ground truth data from 248 reference areas were used. The carried out analyses showed that in forest stands of younger age classes (< 120 years) better results were achieved applying the method of image matching (CHM_IPC), while in the case of older stands (> 120 years) the accuracy of the detection rate of tree crowns was the highest when CHM_ALS model was applied. The mean percentage error (defined by the number of trees, based on the detection of single pine crowns), calculated based on 248 ground truth areas was 0.89%, which shows a great potential of digital photogrammetry (IPC) and GEOBIA. In case of almost full nationwide cover in Poland of airborne digital images (present IPC models) and ALS point clouds (DTM and DSM), at almost 71% forest stands in the Polish State Forests National Forest Holding (PGL LP), one can assume wide application of geodata (available free of charge) in precise modelling of selected tree stand parameters all over Poland.