Low-cost phone-based LiDAR scanning technology provides sub-centimeter accuracy when measuring the main dimensions of motor-manual tree felling cuts

Motor-manual tree felling in one of the most important technical alternatives in timber harvesting. Commonly, it requires a sequence of cuts and may cause higher volume losses compared to mechanized tree felling. The wood lost depends also on the tree size, but accurate quantifications and their dependencies are difficult to establish since there is a limited ability to manually measure key dimensions and assimilate the tree bottom to parametric geometric features. Short-range LiDAR technology integrated in affordable mobile platforms has already been proved to produce reliable estimates on objects located in a limited space, and point cloud processing algorithms have been developed to compare two instances of the same object, potentially enabling the quantification of tree-level wood loss. However, volume loss estimates based on LiDAR scanning of tree bottoms are very sensitive to the accuracy of scans, while low-cost platforms may lack the capability to produce point clouds of an acceptable density at a reasonable distance. This study was setup to check to what extent proximal scanning by low-cost LiDAR platforms can provide accurate data to support the extraction and computation of volume loss by motor-manual tree felling. Eleven Norway spruce logs having a length of one meter and mid-diameters between 22 and 43 cm were used to make 63 notches by a chainsaw. The main dimensions of each notch were then measured manually to the nearest millimeter by a tape and used as reference data. These were the hinge width, depth of the bottom cut, depth of the top (inclined) cut and the notch height. Close-range (up to 50 cm) scans were taken on each notch by an iPhone 13 Pro Max platform using the freely available software 3D Scanner App. The resulted point clouds were imported to Cloud Compare software, where the same measurements were taken digitally and used as data for comparison. By the commonly used error metrics such as the bias (−0.73–0.10), mean absolute error (0.51–0.78) and root mean squared error (0.68–0.92), the differences between the two were in the sub-centimeter domain. Taken individually, all the measurements agreed well in a ± 2 cm range, with an obvious dominance in much lower ranges and no evident trends in variance related to the measurement size. These results are promising for the forest operations science and practice because they provide evidence on the fact that by low-cost close-range LiDAR scanning and point cloud processing one can get accurate estimations notch volume losses. In turn, this will provide the basis for leveraging the losses by considering the operational conditions, the used procedures and experience of the workers while supporting the attempts of quantifying and relating the losses to the mentioned factors.