Graduation date: 2007Forest growth models in the Pacific Northwest are predominantly empirical. Predictions of yield under alternative silvicultural regimes cannot rely completely on field trials; yet empirical growth models are often inadequate for extrapolating untested regimes and genotypes. The limitations of current models include (1) long time-steps (e.g. 5-10 years); (2) insufficient detail for characterizing crowns; and (3) inability to capture physiological mechanisms. The overall goal of this dissertation was to test the ability of a hybrid model (empirical + process-based) to predict the growth of intensively managed plantations. The first step of model development was to refine current characterizations of Douglas-fir crown structure across several silvicultural treatments. The effects of fertilization, thinning, precommercial thinning, vegetation control, and disease intensity (Swiss needle cast) were found to influence important structural attributes of the crown. Among the crown attributes affected, maximum branch size and total- and nonfoliated-crown profile were the most dynamic attributes. Conversely, treatments had no effect on the number of branches or on branch angle. Equations based solely on bole and crown variables predicted crown structural attributes reasonably well across these varied stand conditions. Annualized empirical equations for individual tree diameter and height growth were developed next and found to outperform similar models with a longer time-step. The parameters of these empirical equations showed very few meaningful relationships with physiography, soil, or climate, suggesting that representation of key physiological processes was a necessary next step. Individual branch growth and mortality were significantly influenced by fertilization, thinning, precommercial thinning, vegetation control, and Swiss needle cast. Dynamic equations developed from this dataset significantly improved predictions of crown recession, compared to a traditional empirical approach. The improvements, however, had a relatively minor impact on short-term stand volume growth. The combination of these equations into a hybrid framework showed improvements in leaf area index and periodic annual increment when compared to other stand-level hybrid models. At the individual tree-level, the use of both empirical and mechanistic components was necessary to achieve a level of bias slightly better than that of a purely empirical approach. Beyond growth predictions, this hybrid model offers many other uses.