Precision Forestry: Using Machine Learning and LiDAR to Inform Thinning in Pinus taeda Plantations – A Case Study

Abstract

We demonstrate that the inputs and conceptual foundation needed for individual-tree-level precision thinning optimization algorithms are already available. As more resource managers adopt LiDAR-based inventories, precision thinning can become a value-added outcome of collecting these data. We use LiDAR to assess individual-tree stem volumes in Pinus taeda L. plantations in the southeast US. Rather than arbitrarily selecting starting rows in row thinning operations, we use field- and LiDAR-derived stem volume data to inform row selection. Among all three study sites, row-to-row tree volume variability was present, indicating that selecting rows to be removed deliberately could improve thinning outcomes. A machine learning model based on LiDAR-derived metrics was also accurate in estimating individual stem volume in the primary study site and LiDAR was accurate in measuring pre- and post-thinning stem counts, the data that would be needed to audit thins.