Abstract
Optimization of forest stand management is an important basic technique to plan cutting. Convergence for a global solution is the most important point when measuring confidence. In this paper, we compared three methods, dynamic programming, random search, and full search, and examined their effectiveness. We took larch in Nagano as the object of optimization, and applied a stand density diagram as a growth model. Profits were calculated by yield rate of mass, but the value of the trees shifted with DBH. Net present value of yield was adopted as the criterion for optimizing. Separation of thinning trees was fixed every 5 per hectares in dynamic programming and random search, and every 5% in full search. Interval of thinning available age was 5 or 10 to plan length. Then we optimized those models with each method. As a result, the criterion of the full search was the highest, next was that of the random search and that of dynamic programming was the lowest. When the model in which the value of trees increased and decreased with DBH was adopted, the annual criterion of the solution for dynamic programming differed from that of the other two methods. The number of calculations for one forest stand in dynamic programming was the fewest; next was random search, and that of full search was much larger than for the other methods. So considering for number of calculations and confidence in the solution, random search was the best method under this profit model.
