Abstract: The current trends in forestry in Europe include the increased use of continuous cover forestry (CCF) and the increased availability of tree-level forest inventory data. Accordingly, recent literature suggests methodologies for optimizing the harvest decisions at the tree level. Using tree-level optimization for all trees of the stand is computationally demanding. This study proposed a two-level optimization method for CCF where the harvest prescriptions are optimized at the tree level for only a part of the trees or the first cuttings. The higher-level algorithm optimizes the cutting years and the harvest rates of those diameter classes for which tree-level optimization is not used. The lower-level algorithm allocates the individually optimized trees to different cutting events. The most detailed problem formulations, employing much tree-level optimization, resulted in the highest net present value and longest optimization time. However, restricting tree-level optimization to the largest trees and first cuttings did not significantly alter the time, intensity, or type of first cutting. Computing times could also be shortened by applying accumulated knowledge from previous optimizations, implementing learning aspects in heuristic search, and optimizing the search algorithms for short computing time and good-quality solutions.

Key words: Management optimization, Forest planning, Differential evolution, Simulated annealing