- Two key challenges emerge when initiating individual-tree growth models for bare ground or very young plantations. Both involve the need for a list of individual trees with known or predicted size and known expansion factor (number of trees per unit of area represented by each tree). In the case of bare ground or young plantations with no tree measurements, the tree list must be generated based on some assumptions about the size distribution. Even when a measured tree list is available, simulations resulting from input of the very young tree lists do not produce future stands with observed levels of tree size diversity because initial tree lists do have not differentiated to a degree that allows deterministic growth predictions to drive observed growth differences. Tree growth, particularly in the presence of initial size differences and competition, is a deviation amplifying process. Therefore, initial height and DBH distributions are critical for accurately simulating tree size differentiation. The main objectives of this work are to understand and model how differentiation emerges and evolves in relatively homogeneous newly planted seedlings, how early silvicultural practices affect tree size distributions, how well our current growth models can represent the process of size differentiation, and how to identify features of simulation models to improve simulation of young stand development under varying site conditions and silvicultural practices.
Estimating the progression of tree size differentiation during the first decade or two of Douglas-fir plantation development is challenging due to the multiplicity of factors that influence seedling growth and are not typically measured or knowable in forest operations. From a modeling perspective, these factors contribute an important stochastic component to deterministic growth estimates during the first years after planting. The increased variability in growth that results from the addition of these stochastic components drives size differentiation among trees. However, as trees grow and start competing with each other, a greater portion of the growth variability among trees in the stand can be accounted for by predictors such as its initial size, crown ratio, and relative social position. Thus, at these older stages, the deterministic component of the models can amplify the size differentiation to produce realistic future structure with less need for introducing stochasticity.
The overall goal of this dissertation research was to develop and compare methodologies to generate tree lists for young plantations and to project growth of both generated and measured tree lists of Douglas-fir newly planted forest in a manner that produces stand structures or size distributions that resemble observed stands of later ages that started with similar initial conditions.
Weibull distributions were first fitted to height distributions of trees on individual plots in a large dataset compiled from research plots designed to test growth responses to early silvicultural treatments. Different approaches to introducing varying degrees of variability on fitted smooth Weibull distributions were explored with the objective of representing as closely as possible the variation of the observed tree heights around the smoothed distributions. The hypothesis tested was that different degrees of initial size variations can cause a significant impact on simulated stand structures 15-20 years later. Simulations demonstrated that the CIPSANON growth model for intensively managed Douglas-fir plantations was quite unresponsive to different initial size variation in out planted nursery seedlings that had not been through their first growing season. However, initiating CIPSANON with tree lists from older plantations (>5 years), the simulations produced stand structure at plantation ages 15-20-yrs that represented the observed tree size differentiation with sufficient accuracy. The poorest performance of the model resulted from initiating the model with generated or measured trees lists for plantations with age < 5yrs, suggesting that additional stochastic variation was required in annual height growth predictions during the first five years of plantation development. Three alternatives variance-covariance structures were used to represent the residual variability around fitted height growth equation. The parameters for these alternative variance-covariance structures were used to incorporate stochastic variation into deterministic CIPSANON height growth predictions during this initial five-year period after planting. The alternatives included: 1) white noise (based on a random component with constant variance around predicted height growth); 2) proportion noise based on a random component proportional to initial tree height; and 3) noise proportional to height as in alternative 2 plus autocorrelated noise with 1-yr lag (AR(1)) based on the observed covariance among successive residuals on the same tree). The three alternative simulation approaches produced significantly more accurate tree lists for 5-yr-old plantations than purely deterministic simulations. The white noise approach produced unrealistically greater height differentiation than the observed in 2-yr-old plantations, leading to overprediction of height variability in 15-yr-old plantations. Results from the other two approaches were similar. The inclusion of stochasticity during these first five years also reproduced observed patterns in height rank changes over time, in DBH size structures, and in amelioration of multimodality identified as an artifact of the DBH assignment procedure (each tree assigned a DBH in the year that each tree exceeded breast height or 1.37 m). Other model impacts included reductions of 6.8 to 5.1% in cumulative net volume growth at plantation age 30 years compared to purely deterministic predictions.
Finally, the effects of different vegetation regimes on tree size distributions during the first two decades after planting were analyzed. Vegetation regimes were grouped by the number of spring releases received. Treatment regimes impacted DBH proportionally more than height size distributions, and the effects of different treatment regimes on the size distributions of both variables were not the same. Spring releases produced an upward shift in the whole DBH distribution, but with respect to height the shorter trees were more positively affected than taller trees. This differential effect on different portions of the height distribution lasted for the first decade, but gradually disappeared over the second decade, responses to SRs treatments tending to converge with the untreated control in absolute cumulative height growth.