A cellulose price prediction model was adjusted using time and lagged price as covariates. From the model parameter estimates, 48 possible trends were proposed for future cellulose price. Following, three simulation methods were used to predict the future values defined by the various trends: M1<FONT FACE=Symbol>Þ</FONT> Pcel.f = µ; M2 <FONT FACE=Symbol>Þ</FONT> Pcel.f = µ + epsilonf, e M3 µf + epsilonf, where m is the systematic part and e f is the stochastic component. The Monte Carlo method and a triangular distribution were used for the simulation. To compare the values simulated by the methods and the future values of the various trends, the Average Relative Difference was used. In case of no trend, M1 and M2 were satisfactory, although M2 included disturbances around the mean. In the case of a real trend, M3 had the best performance, though it was influenced by the accuracy in the predicted trend.
Simulation; risk analysis; forest planning
