%0 Articles
%T Developing spatial optimization in forest planning
%A Heinonen, Tero
%D 2007
%J Dissertationes Forestales
%V 2007
%N 34
%R doi:10.14214/df.34
%U http://dissertationesforestales.fi/article/1817
%X Forest management planning problems typically includes many objectives and several planning periods, and the planning area consists of numerous forest stands. The solution space can be enormous and numerical methods are needed to solve problems. When there are spatial management objectives, problem becomes combinatorial in nature. The aim of this thesis is to develop methods to improve the performance of heuristic optimization methods in spatial forest planning problems and to compare the ability of different heuristics to solve different problems. Another aim is also to develop improved methods to solve spatial forest planning problems.

Traditional local search heuristic, when applied to forest planning, consider one stand at the time and change its treatment if it improves the solution. In this thesis treatment was changed treatment simultaneously in two stands which enlarges the solution space. This clearly improved the spatial layout of desired features and led to better objective function values especially with simple heuristics. The performance of local search methods is highly dependent on the parameters controlling the search. In this thesis an automated procedure to look for optimal parameters was developed. The method of Hooke & Jeeves for nonlinear programming was adopted in the search process. The method was able to find logical and efficient parameters for local search methods when the search time was limited.

Stand borders are traditionally subjectively drawn and fixed, and individual stands are assumed homogeneous in terms of forest characteristics. This can restrict the efficient use of forest resources. The interest in the use of fine-grained forest data is increasing the prospects of obtaining reliable data with remote sensing tools. This thesis deals with the implications and possibilities of using raster cells in forest planning. By using spatial objectives these cells were aggregated into dynamic treatment units. Spatial optimization and raster data produced more old forest area with the same timber production level than the approach based on predefined stands.

The computational burden of large planning problems can be reduced using decentralized computing methods. Instead of controlling the whole system with one objective function, a cellular automaton and a spatial application of the reduced costs method were used for decentralized optimization. The decentralized approaches reduced the solution space into a small fraction of the solution space of local search heuristics and decreased the time consumption of spatial optimization. The quality of the solutions also improved.