A tabu search method for a tactical forest planning problem

At the tactical level of forest planning, decision support systems must deal explicitly with spatial and temporal restrictions on clearcuts and the design of road access to the forest. Formulating and solving optimization models which provide useful decision choices in this context is a serious challenge.;This thesis advances a new model formulation for the harvest scheduling and road building optimization problem. The tactical planning problem is treated in the context of a hierarchical planning system, where a strategic planning process has been first executed. The strategic plan goals of sustainable harvest are inputs to the tactical planning process, where the object is to produce a spatially and temporally explicit schedule of harvesting and road building. The model is designed to produce harvest schedules which minimize both biological productivity losses due to sub-optimal-timing of harvests, and the costs of road construction. The objective function to be minimized is a weighted sum of these two opposing cost factors. Solving the optimization problem for a range of weightings produces a spectrum of solutions, from which non-dominated solutions are selected to produce an efficient frontier of roading cost versus lost productivity.;The model uses forest stands as spatial decision units for harvest scheduling, thus avoiding the reduction of the solution space which occurs when stands are pre-blocked. The model produces schedules which are compliant with maximum opening size and adjacency delay requirements, without restricting adjacency delay to one planning period. The configuration of the road system is not restricted to be a tree structure, in that multiple access points and cycles in the proposed network are permitted. As a result, this model can address a wider range of practical road network designs than other models in the literature. An heuristic algorithm was developed to solve the multi-period road network optimization problem. This algorithm was shown to provide solutions which consistently fall within a small percentage of the optimum solution, and which are optimal in the majority of cases.;Topographical adjacency relationships between stands are represented by a general undirected graph. Openings which are created by harvesting are connected sub-graphs of this forest graph. These sub-graphs are dynamically created throughout the solution process. Spatial feasibility of a proposed schedule is determined by utilizing depth first search and articulation point searches on these sub-graphs.;The combined harvest and road building problem is solved using a tabu search metaheuristic. The search algorithm has a dynamic feedback mechanism to set tabu tenure, and a strategic oscillation strategy to smooth transitions throughout the feasible region. These features of the algorithm essential to ensure solution quality over many problem instances. Computational studies were carried out on a forested region in Cumberland County, Nova Scotia. The study results demonstrate consistent results over the original dataset and five additional datasets. Tradeoff curves which are produced from these solutions provide valuable information, clearly showing the relationship between road building budgets and the range of harvest timing choices available to the decision maker. Thus, this model and solution algorithm represent a significant contribution to the tactical forest management problem.