| This thesis has developed a metaheuristic model, i.e. methodology, based on tabu search principles. To develop this model, a forest landscape management design problem was characterized by a set of six aspatial and spatial objectives that included gross timber volume, periodic volume flow, minimum and maximum harvest block size limits, harvest adjacency delay, age class composition and patch-size distribution. All objectives were addressed over an 80-year planning horizon. Unique to this model is a partial move-evaluation scheme guided by dynamic memory structures and an oscillating objective weighting scheme.; The model was evaluated with two case studies. The first involved four hypothetical forests, each comprised of 987 stands (19,740 ha) and representing a unique combination of age class structure and their spatial arrangements. Achievement levels of the six objectives varied depending upon initial forest conditions and harvest level, ranging from close to 100% to as low as 50% of the specified targets. These results, however, were reasonable and explained. The first case study helped validate the model's outcomes, while also showing the influence of initial spatial forest structure on objective achievement. In a second case study, a real forest of 11,000 stands (52,000 ha) was used. In this forest, the model was able to maintain all objectives close to 100%, except for the gross volume and periodic flow objectives that were achieved at only 70% levels. Given the forest's initial age class structure and spatial pattern of operable areas, the results were reasonable and explained. The second case study demonstrated the model's performance on a realistic problem size. For this case, the model took 14–20 hours of CPU time on a Pentium III 733 MHz computer to complete a single run of 5,000 iterations while examining a maximum of 30–50 solutions per iteration.; Both case studies showed that a well-designed tabu search metaheuristic can provide good solutions to complex forest landscape management design problems. While the model in the thesis generated solutions that were reasonable and explainable, it did not guarantee an optimal solution. Nonetheless, the model is flexible and easily extensible to a wide range of problems in forest landscape management design. |
