Evaluating the potential of a harvester-mounted LiDAR system in monitoring biodiversity indicators such as low vegetation during forest harvesting could enhance sustainable forest management and habitat conservation including dense forest areas for game. However, there is a lack of understanding on the capabilities and limitations of these systems to detect low vegetation characteristics. To address this knowledge gap, this study investigated the performance of a harvester-mounted LiDAR system for measuring low vegetation (height <5 m) attributes in a boreal forest in Finland, by comparing it with handheld mobile laser scanning (HMLS) and drone laser scanning (DLS) systems. LiDAR point cloud data was collected in September 2023 to quantify the low vegetation height (maximum, mean, and percentiles), volume (voxel-based and mean height-based) and cover (grid method). Depending on the system, LiDAR point cloud data was collected either before (HMLS and DLS), during (harvester LiDAR) or after (HMLS and DLS) harvesting operations. A total of 46 fixed-sized (5 m × 5 m) grid cells were studied and analyzed. Results showed harvester-mounted LiDAR provided consistent estimates with HMLS and DLS for maximum height, 99th height percentile, and volume across various grids (5 cm, 10 cm, 20 cm) and voxel (20 cm) sizes. High correlation was observed between the systems used for these attributes. This study demonstrated that harvester-mounted LiDAR is comparable to HMLS and DLS for assessing low vegetation height and volume. The findings could assist forest harvester operators in identifying potential low vegetation and dense areas for conservation and game management.
A new method for the co-registration of single tree data in forest stands and forest plots applicable to static as well as dynamic data capture is presented. This method consists of a stem diameter weighted linking algorithm that improves the linking accuracy when operating on diverse diameter stands with stem position errors in the single tree detectors. A co-registration quality metric threshold, QT, is also introduced which makes it possible to discriminate between correct and incorrect stem map co-registrations with high probability (>99%). These two features are combined to a simultaneous location and mapping-based co-registration method that operates with high linking accuracy and that can handle sensors with drifting errors and signal bias. A test with simulated data shows that the method has an 89.35% detection rate. The statistics of different settings in a simulation study are presented, where the effect of stem density and position errors were investigated. A test case with real sensor data from a forest stand shows that the average nearest neighbor distances decreased from 1.90 m to 0.51 m, which indicates the feasibility of this method.