Multicriteria forest decisionmaking under risk with goal-programming Markov decision process models

Abstract

As multiple risks pervade forest decisionmaking, Markov decision process (MDP) models offer an analytically tractable approach to seek optimal policies that are straightforward for implementation in practice. By incorporating goal programming (GP), this study extended MDP models with both average and discounted criteria to deal with multiple, often noncommensurable and conflicting, objectives. This method (GPMDP) was applied to the management of mixed loblolly pine-hardwood forests in the southern United States. The decision criteria were the values of harvests, carbon sequestered by trees, diversity of tree species and sizes, and fraction of old-growth stands in the forested landscape. For the case study, the results showed that given equal weights for normalized criteria, with both average and discounted GPMDPs, minimum deviations from the highest diversity of tree size and species were achieved at the cost of, on average, one-third of the decline of other criteria from their maximum levels.