Variance propagation in growth and yield projections

This paper reports the results of a study on the propagated variance associated with stand estimates in a forest growth and yield model. A cumulative variance as a result of input measurement and regression estimation errors is propagated in a growth and yield model using the method of statistical differentials. To provide an assessment of relative performance, these variance estimates are compared with a Monte Carlo simulation estimate of propagated error for increasing levels of sampling intensity. The method of statistical differentials is used to estimate the propagated variance through five 10-year growth projections. The results indicate growth projection estimates may have substantial error components that are not readily apparent from model calibration statistics or bias assessment procedures.