Comparing tree attributes derived from quantitative structure models based on drone and mobile laser scanning point clouds across varying canopy cover conditions

Abstract

Complex canopy cover conditions often challenge the accurate measurement of many individual tree attributes that are pivotal to the sustainable management of forest resources. Advances in drone laser scanning (DLS) and mobile laser scanning (MLS) have enabled the acquisition of high-density point clouds with the potential to better resolve detailed tree structures. Yet, the quality of DLS and MLS data can be limited by occlusions and environmental complexities. To quantify the impacts of canopy cover on the tree attribute estimation, this study investigated the utility of DLS and MLS data both individually and combined. Considering the scanning characteristics, we examined direct fusion and a new strategy using a relative weighting scheme based on the probability density of vertical point distribution. We compared the accuracy of seven tree attributes derived from quantitative structure models (QSMs) based on (1) DLS, (2) MLS, (3) fused, and (4) weighted point clouds under low, moderate, and high canopy cover levels. We found that the weighted data improved the modelling efficiency of QSMs by ∼ 20% on average, compared to fused and MLS data. Across canopy cover levels, the fused and weighted data achieved comparable results and outperformed DLS/MLS data in estimating tree attributes. Specifically, diameter at breast height and crown base height were accurately extracted from the fused, weighted, and MLS data under low canopy cover with the concordance correlation coefficient (CCC) > 0.80. As canopy cover increased, they were best estimated using the fused data (CCC > 0.90, RRMSE < 22%). Height was accurate regardless of canopy cover, which was independent of data collection platforms (CCC > 0.80, RRMSE < 16%). The crown diameter was also well estimated by fused, weighted, and MLS data across canopy cover levels (CCC > 0.82, RRMSE < 19%). The total, stem, and branch volumes could be best modelled by the fused data with increasing canopy cover. Overall, the fusion of DLS and MLS point clouds allowed the retrieval of comprehensive tree-level information. However, forestry practitioners still need to evaluate the trade-offs in selecting the most appropriate platform for laser scanning data based on their needs. Future studies should also enhance the modelling of trees with complex branching structures to strengthen the extraction of diverse attributes.