Processing network models for forest management

The Forest Management Problem of determining when and how much to plant and harvest, to convert tree species, to spray pesticides, and to take other silvicultural actions is a large scale optimization problem that is often formulated and solved as a linear program. Unfortunately, the algebraic representation of the model is often unintelligible to the practicing forest managers who ought to be able to manipulate the model in order to set up and examine various ‘what-if’ scenarios. This paper proposes a solution to this communication problem: the time-expanded processing network. The goal of the network diagram is to clearly convey the underlying structure, limitations, and assumptions to both mathematical programming experts and nontechnical forest managers. The linear programming formulation then follows directly from the diagram. The paper illustrates the model-building procedure by constructing processing network models for common forest behaviours including growth, fire, and pest infestation, and for common management actions including conversion harvests, pest spraying, and nondeclining yield constraints.