Deep learning techniques have been widely applied to classify tree species and segment tree structures. However, most recent studies have focused on the canopy and trunk segmentation, neglecting the branch segmentation. In this study, we proposed a new approach involving the use of the PointNet++ model for segmenting the canopy, trunk, and branches of trees. We introduced a preprocessing method for training LiDAR point cloud data specific to trees and identified an optimal learning environment for the PointNet++ model. We created two learning environments with varying numbers of representative points (between 2048 and 8192) for the PointNet++ model. To validate the performance of our approach, we empirically evaluated the model using LiDAR point cloud data obtained from 435 tree samples scanned by terrestrial LiDAR. These tree samples comprised Korean red pine, Korean pine, and Japanese larch species. When segmenting the canopy, trunk, and branches using the PointNet++ model, we found that resampling 25,000–30,000 points was suitable. The best performance was achieved when the number of representative points was set to 4096.

In Republic of Korea, the digital transformation of forest data has emerged as a critical priority at the governmental level. To support this effort, numerous case studies have been conducted to collect and analyze forest data. This study evaluated the accuracy of forest resource assessment methods using terrestrial laser scanning (TLS) and backpack personal laser scanning (BPLS) under Leaf-on and Leaf-off conditions in the Gari Mountain Forest Management Complex, Hongcheon, Republic of Korea. The research was conducted across six sample plots representing low, medium, and high stand densities, dominated by Larix kaempferi and Pinus koraiensis. Conventional field survey methods and LiDAR technologies were used to compare key forest attributes such as tree height and volume. The results revealed that Leaf-off LiDAR data exhibited higher accuracy in capturing tree height and canopy structures, particularly in high-density plots. In contrast, during the Leaf-on season, measurements of understory vegetation and lower canopy were hindered by foliage obstruction, reducing precision. Seasonal differences significantly impacted LiDAR measurement accuracy, with Leaf-off data providing a clearer and more reliable representation of forest structures. This study underscores the necessity of considering seasonal conditions to improve the accuracy of LiDAR-derived metrics. It offers valuable insights for enhancing forest inventory practices and advancing the application of remote sensing technologies in forest management.

The aim of this study was to select the optimal deep learning model for land cover classification through hyperparameter adjustment. A U-Net model with encoder and decoder structures was used as the deep learning model, and RapidEye satellite images and a sub-divided land cover map provided by the Ministry of Environment were used as the training dataset and label images, respectively. According to different combinations of hyperparameters, including the size of the input image, the configuration of convolutional layers, the kernel size, and the number of pooling and up-convolutional layers, 90 deep learning models were built, and the model performance was evaluated through the training accuracy and loss, as well as the validation accuracy and loss values. The evaluation results showed that the accuracy was higher with a smaller image size and a smaller kernel size, and was more dependent on the convolutional layer configuration and number of layers than the kernel size. The loss tended to be lower as the convolutional layer composition and number of layers increased, regardless of the image size or kernel size. The deep learning model with the best performance recorded a validation loss of 0.11 with an image size of 64 × 64, a convolutional layer configuration of C→C→C→P, a kernel size of 5 × 5, and five layers. Regarding the classification accuracy of the land cover map constructed using this model, the overall accuracy and kappa coefficient for three study cities showed high agreement at approximately 82.9% and 66.3%, respectively.

This study provides the methodology for the development of sustainable forest management activities and systematic strategies using national spatial data, satellite imagery, and a random forest machine learning classifier. This study conducts a regional province-scale approach that can be used to analyze forest clear-cutting in South Korea; we focused on the Chungcheongnam-do region. Based on spatial information from digital forestry data, Sentinel-2 satellite imagery, random forest (RF) classifier, and digital forest-type maps (DFTMs), we detected and analyzed the characteristics of clear-cut areas. We identified forest clear-cut areas (accounting for 2.48% of the total forest area). The methodology integrates various vegetation indices and the RF classifier to ensure the effective detection of clear-cut areas at the provincial level with an accuracy of 92.8%. Specific leaf area vegetation index (SLAVI) was determined as the most important factor for accurately detecting clear-cut areas. Moreover, using a DFTM, we analyzed clear-cutting characteristics in different forest types (including private, national, natural, and planted forests), along with age class and diameter-at-breast-height class. Our method can serve as a basis for forest management and monitoring by analyzing tree-cutting trends in countries with forest areas, such as Republic of Korea.

In order to address the enhanced transparency framework in the Paris Agreement, the completeness of the National GHG Inventory should be ensured. Up-to-date information on changes in both land-use categories and area by land-use conversion must be included in the Land-Use Change Matrix (LUCM), which contains critical activity data for the LULUCF (Land Use, Land-Use Change and Forestry) sector. This study was conducted to determine the optimal sampling intensity and develop the LUCM through a comparison of different sampling intensities. In this study, an empirical analysis was based on relative standard error and relative efficiency in uncertainty assessment for seven sampling intensities from 2 to 8 km. The findings indicate no clear difference in the uncertainty by sampling intensity concerning forest land and cropland, which account for large areas of Gangwon-do and Jeollanam-do. In contrast, the uncertainties for grassland and wetlands, each of which covers less than 10% of the total area, were decreased with increasing sampling intensity, but no statistical significance was found. To improve the LUCM at the national level, it is therefore necessary to select an efficient sampling intensity because there are trade-offs between statistical accuracy and cost regarding the application of different sampling approaches.

This study aims to segment individual tree structures (stem, crown, and ground) from terrestrial LiDAR-derived point cloud data (PCD) and to compare the segmentation accuracy between two models: XGBoost (machine learning) and PointNet++ (deep learning). A total of 17 input features were categorized into spatial coordinates and normals, geometric structure features, and local distribution features. These were combined into four input configurations and evaluated under three downsampling conditions (2048, 4096, and 8192 points), resulting in 12 experimental setups. XGBoost achieved the highest stem segmentation F₁-score of 87.8% using all features with 8192 points, whereas Point-Net++ reached 92.1% using only spatial coordinates and normals with 4096 points. The analysis of missegmentation patterns showed that XGBoost frequently confused structures near stem-to-ground boundaries and around branch junctions, while PointNet++ occasionally missegmented complex regions between stems and crowns. Regarding processing time, XGBoost required 10 to 47 min across all conditions, whereas Point-Net++ required 49 min for the 2048-point condition and up to 168 min for 8192 points. Overall, XGBoost provided advantages in computational efficiency and in generating feature-importance scores, while PointNet++ outperformed XGBoost in segmentation accuracy and the recognition of structurally complex regions.