Open access publications in the SLU publication database

Abstract

It is well known that decision making problem solving in forest management involves risks from different sources. In many cases, the incorporation of risk will change the optimal decisions in different ways. Usually, stochastic optimization models are much more complicated than the corresponding deterministic ones, and greater computational efforts are required for identification of the optimal solutions. Thus, simplification of the decision making problems is often made in one way or another, the cost being deviations from reality. In this thesis, a new optimization approach has been designed for decision problems concerning harvesting under price risk, both for single species stands (paper I) and for mixed species stands (paper II). The intention with this approach has been to keep the model simple with respect to the computational burden, without imposing very restrictive assumptions on the choice of feasible decision alternatives. Optimal stocking level functions, which are modeled as functions of stand age and stumpage price, are applied to guide the thinning decisions. A reservation price function, which is formulated using stand age and standing timber stock as independent variables, and market price observations are used to determine the final harvest time. The coefficients of these functions are calculated by simulation. Flexibility is valuable in decision making under risk. Recently, the risk of damage to forests by moose in Sweden has been widely recognized. This problem will most likely exist also in the future, since moose hunting is very popular in the country. Study of this problem is highly relevant to forestry. In Paper III, the value of flexibility for a mixed species stand is analyzed when risks in timber price and moose damage are incorporated. An adaptive optimization model has been developed to determine the initial mix of species which maximizes the expected present value. The results show that mixed-species stands of Scots pine and Norway spruce are preferable to pure Scots pine stands when the risks of moose damage and timber price are taken into account. The efficiency of the model also depends on the optimization technique applied. Paper IV works on a hybrid heuristic algorithm, based on a genetic algorithm and a traditional non-linear optimization method: the Hooke and Jeeves method. It is shown that the performance of the hybrid algorithm is significantly better than that of the Hooke and Jeeves method, and of the Powell search method.