U 11 certainty and the Intertemporal Forest Manageme 11 t of Natural Resources : An Empirical Application to the Stanislaus National Forest

The empirical problem of natural resource management is typically the intertemporal allocation of product flows and resource stocks under uncertainty. National forest harvest scheduling is conceptualized in this study as a stochastic optimal control problem. In theory, optimal solutions to most stochastic control problems exist, but required computer costs are excessive even for problems of moderate dimensions. Hence, approximate solution techniques are required, and this study employs one such approach called the Linear-Quadratic-Gaussian (LQG) control method. Given a set of desired or target levels for stocks and flow variables, the LQG optimization criterion is to keep the actual evolution of the system close to the target levels. Optimal harvest levels are given as the solution to a simplified model of the actual problem where the model is characterized by quadratic preferences and linear system dynamics. Imperfect observation of timber stocks is also modeled in the LQG approach, and part of the solution is a recursive estimator of timber levels that is the optimal estimator given management objectives. The relative costs of uncertainty are also calculated in the LQG solution. Empirical results isolate. those sources of uncertainty critical to management actions and indicate superiority of recursive estimation over static estimation.