A comparison of forest growth and yield models for inventory updating

Abstract

A system of loblolly pine growth and yield models was developed which used database files as input. Using database files as direct input created a compatible link between the growth and yield models and a Geographic Information System (GIS). Since growth and yield models can be used to update forest inventories and a GIS is a common method for maintaining forest inventory data, this compatibility provided a method to rapidly update past forest inventory records.

This system was used to evaluate four different loblolly pine growth and yield models. The growth and yield models examined were: a whole stand, a diameter distribution - parameter prediction, a diameter distribution - parameter prediction, and an individual tree. Three different validation approaches were used to create fitting and validation data sets from permanent plot remeasurement data, and evaluate each of the four growth and yield models at varying projection periods. The periods used were zero, three, six, and nine years. Evaluations were based solely on the capability of each to model to predict merchantable volume.

In terms of root mean square error of prediction, the individual tree and whole stand models performed superior than the diameter distribution models. At shorter projection periods the individual tree model performed better than the whole stand model, but the whole stand was superior at the nine year period. The parameter recovery models performed better for shorter periods than the parameter prediction model, but this difference diminished with longer periods.