A hybrid method for tree-level forest planning

Abstract Forest inventory is increasingly producing information on the locations and sizes of individual trees. This information can be acquired by airborne or terrestrial laser scanning or analyzing photogrammetric data. However, all trees are seldom detected, especially in young, dense, or multi-layered stands. On the other hand, the complete size distributions of trees can be predicted with various methods, for instance, kNN data imputation in an area-based LiDAR inventory, predicting the parameters of a distribution function from remote sensing data, field sampling, or using histogram matching and calibration methods. The predicted distribution can be used to estimate the number and sizes of the non-detected trees. The study’s objective was to develop a method for forest planning that efficiently uses the available tree-level data in management optimization. The study developed a two-stage hierarchical method for tree-level management optimization for cases where only part of the trees is detected or measured individually. Cutting years and harvest rate curves for the non-detected trees are optimized at the higher level, and the cutting events of the detected trees are optimized at the lower level. The study used differential evolution at the higher level and simulated annealing at the lower level. The method was tested and demonstrated in even-aged Larix olgensis plantations in the Heilongjiang province of China. The optimizations showed that optimizing the harvest decisions at the tree level improves the profitability of management compared to optimizations in which only the dependence of thinning intensity on tree diameter is optimized. The approach demonstrated in this study provides feasible options for tree-level forest planning based on LiDAR inventories. The method is immediately applicable to forestry practice, especially in plantations.