Validation of forest height-age models

Forest height-age models are used in forest management to estimate height and (or) site index. It is useful to know the bias and precision of these models in order to evaluate their applicability. Methods are available for validating the models; however, many problems exist with the methods because of a lack of independence in the data and nonconstant error variance across a range of ages. A validation procedure is presented that overcomes these problems by using a multivariate technique (random coefficients) to model the structure of the errors associated with the models. Confidence intervals for bias and precision can then be constructed based on the error structure. This method of validation was demonstrated on the white spruce (Picea glauca (Moench) Voss) height-age model for British Columbia, Canada. The preliminary validation showed the model to be unbiased for estimating both height and site index; however, its precision was poor.