Development and evaluation of models of tree and forest growth and climat

There are numerous models of tree and forest growth that have been developed for different purposes, and for different scales. In this thesis, various growth models are systematically compared in relation to forest stress by comparing model assumptions and formulations, and by comparing model performance at two specific sites. Incorporation of climate in forest models is reviewed and techniques for generating climate data are evaluated. A stochastic model was developed to generate daily weather variables for use with forest models. A preliminary successional model for miombo woodlands of dry tropical Africa was developed.;None of the forest models evaluated performed well when predictions were compared with seven years of observed growth data, making site-specific predictions impossible without local calibrations. In models that make assumptions about growth of unstressed trees, stress factors were evaluated directly by altering model functions and coefficients to represent the means by which a particular stress factor affects growth. Physiologically based models are the most easily linked to controlled experiments. Management models remain preferable for indirect, empirical evaluation of historical forest growth information and for projecting the economic consequences of a given level of forest growth reduction.;In an analysis of the JABOWA maximum diameter growth model, the requirement that two-thirds of maximum diameter is produced at half maximum age was shown to be inconsistent with the fixed curve form of the JABOWA model. We generalized the model to have variable shapes to accommodate different patterns of growth in different species.;Many modelers have used different methods to generate weather data, without much thought into the properties of observed weather data or the data being generated, and subsequent effects on model results. Stochastic models offer the best method for preserving the characteristics of weather from a particular station when actual measurements cannot be used directly. A stochastic daily weather model was developed that recognizes and assures observed serial and cross-correlation relationships among the simulated weather variables.;Development of successional models for tropical forests remains difficult due to lack of suitable data, and the large variety of species involved.