Growth modeling of forests at the individual tree and stand levels is
a highly refined procedure for many forest types. A method to incorpor
ate predictions from such models into a forest inventory system is dev
eloped. Variance components from the actual measurements and from the
predicted measurements are used to estimate the variance of the combin
ed predicted value. The only assumption required to justify this metho
d is that the model estimate has a bias that does not change from one
time period to the next. The estimation procedure proposed here can al
so incorporate remotely sensed information via a regression estimator.