Abstract
The long-term planning of sustainable forest treatment at the landscape level is an increasingly more complex task. Local treatment schedules, pertaining to homogeneous sub-areas called stands, must be developed over a time horizon of a few centuries. Thousands of local schedules must be coordinated to satisfy hard constraints, and balance soft constraints and optimization criteria. Constraints and objectives are defined in terms of economical, recreational, and environmental effect. The aim of the forest treatment schedule is twofold. Over the near time horizon, it must provide clear instructions for forest treatment. In addition, sustainability over the full horizon must be demonstrated. In this context, sustainability means balancing growth and yield in the long term, the preservation of bio-diversity, and catering for human recreational and cultural value. Conventional OR based approaches have failed to give satisfactory results for this type of problem. We describe a method built on explicit constraint descriptions and a memory-based local search procedure for solving rich models of the long-term forest treatment scheduling problem. We also describe a configurable decision support system, called Ecoplan, where the scheduling kernel is based on our method. The system relies heavily on close interaction with a stand simulator, which must provide forestry knowledge necessary to guide the scheduling process, including the definition of abstract forest treatment actions. Ecoplan also provides facilities for user interaction in the planning process, functionality for locking specific parts of a plan, and flexibility to alter key factors in the plan such as active constraints and objective criteria. In this way, the system supports the definition and exploration of "what-if" scenarios. The Ecoplan system has been built on the initiative of the major Norwegian forest owners, addressing a problem area that is becoming increasingly more complex to handle and more critical to society.
