Parameter uncertainty analysis of forest carbon sequestration

Uncertainty analysis plays an important role in policy decisions by providing appropriate information about descriptions of uncertainty as well as the causes of such uncertainty. Even though sensitivity analysis allows the analysis of the uncertainty of a model, it is unable to express precisely the results of model uncertainty due to the voluntary and discrete definition of the range of uncertainty relative to input parameters. Furthermore, it cannot provide explanations about the relative relationships that exist among the possible combinations of input parameters since it analyzes the effects of change based on a single parameter for model results when the remaining parameters are fixed. This paper analyzes the effect of parameter uncertainties in land-use change between agriculture and forestland as well as the uncertainty of forest yield on forest carbon sequestration. The Monte Carlo technique is used to turn a global forest model with no probability (Sohngen and Mendelsohn, 2003; 2007) into a proper probability model through Latin Hypercube sampling. With scarce information of parameters, we apply a multi-level regression model to estimate the parameters of yield function since data in the forest are structured hierarchically, and fitting regression models that ignore this hierarchical structure can lead to false inferences. However, when analyzing the influences of parameters on model uncertainty, it is critical to apply the common criteria for ranges of uncertainty because the wider the range of uncertainty, the greater the influence it has on mode