A distance-independent tree growth model for radiata pine in New Zealand [microform]

With the evolution of more intensive forest management practices in New Zealand a higher level of resolution in growth and yield forecasts is needed. A review of growth modeling revealed that a distance-independent tree model would provide the detailed information required. Such a model was developed for radiata pine in Kaingaroa Forest, New Zealand. Preliminary work suggested that diameter increment was a better dependent variable than basal area increment or the use of a logarithmic transform. Different forms of diameter increment models were examined, ranging from purely empirical models to models based on theoretical considerations of tree growth. An initial comparison of the predictive ability of three diameter growth functions indicated that, while all performed reasonably, the theoretically-based model did not allow for the range of increment observed within a particular stand while the form of site in the empirical linear regression model did not appear realistic.

The selected diameter increment function has site acting as a multiplier on a linear combination of stand basal area, relative spacing, relative tree diameter and top height. Mortality was modeled using the logistic function. Stand basal area and relative tree diameter are the predictors of tree mortality. Simulation of an independent set of 22 plots showed good agreement between actual and predicted stand parameters with the exception of some high stocking plots. Such plots show considerable variability in the onset of serious competitive mortality. Further validation of the model against such stands is required. Otherwise the model performed well, particularly for the heavy early thinning regimes that dominate current management practice in New Zealand.

The model was compared with a stand growth model and a distance-dependent tree model previously developed for Kaingaroa Forest. This comparison proved favourable for the distance-independent tree model developed in this study as it generally predicted stand parameters better than the distance-dependent tree model and at least as well as the stand model.