Date: 08 Feb 2014

Optimization in Natural Resources Conservation

* Final gross prices may vary according to local VAT.

Get Access

Abstract

The previous three chapters of this book have been devoted to specific components of informed decision processes: objectives, potential actions, model(s) predicting system change and response to potential actions, and monitoring to provide estimates of system status. The final component of an informed decision process is a solution algorithm, providing a means for deciding which potential action to take. Optimization algorithms provide an objective and transparent approach to select the action that will do the best job of meeting objectives. Static optimization provides a solution to decision problems that are not iterative, and we provide examples for one or more decision variables (variables that are components of potential actions). Many decision problems in natural resource management are best viewed as dynamic, in that they are iterative and require decisions that are repeated through time. In dynamic decision problems, decisions made at one point in time are expected to influence system state of the next time step, thus influencing the state-dependent decision at that time. For any specific decision, dynamic optimization algorithms must thus consider all subsequent time steps for the time horizon of the decision problem. In addition to being dynamic, most decision problems in natural resource management are characterized by substantial uncertainty, and dynamic optimization algorithms have been extended to deal with several sources of uncertainty. An important source is uncertainty about how the system responds to management actions, and we may develop multiple models to characterize this uncertainty. Adaptive dynamic optimization algorithms provide solutions that deal not only with objectives, but with the anticipated reduction in uncertainty that will characterize future decisions. The output of an optimization algorithm is frequently a graph or table of recommended actions for specific values of system state variables. Decision thresholds are thus defined by the optimization algorithm and are simply locations in state space where a small change in the value of a state variable produces a change in the optimal or recommended management action.