Annualized diameter and height growth equations for Pacific Northwest plantation-grown Douglas-fir, western hemlock, and red alder

Simulating the influence of intensive management and annual weather fluctuations on tree growth requires a shorter time step than currently employed by most regional growth models. High-quality data sets are available for several plantation species in the Pacific Northwest region of the United States, but the growth periods ranged from 2 to 12 years in length. Measurement periods of varying length complicate efforts to fit growth models because observed growth rates must be interpolated to a common length growth period or those growth periods longer or shorter than the desired model time step must be discarded. A variation of the iterative technique suggested by Cao [Cao, Q.V., 2000. Prediction of annual diameter growth and survival for individual trees from periodic measurements. Forest Sci. 46, 127–131] was applied to estimate annualized diameter and height growth equations for pure plantations of Douglas-fir, western hemlock, and red alder. Using this technique, fits were significantly improved for all three species by embedding a multi-level nonlinear mixed-effects framework (likelihood ratio test: p < 0.0001). The final models were consistent with expected biological behavior of diameter and height growth over tree, stand, and site variables. The random effects showed some correlation with key physiographic variables such as slope and aspect for Douglas-fir and red alder, but these relationships were not observed for western hemlock. Further, the random effects were more correlated with physiographic variables than actual climate or soils information. Long-term simulations (12–16 years) on an independent dataset using these annualized equations showed that the multi-level mixed effects models were more accurate and precise than those fitted without random effects as mean square error (MSE) was reduced by 13 and 21% for diameter and height growth prediction, respectively. The level of prediction error was also smaller than an existing similar growth model with a longer time step (ORGANON v8) as the annualized equations reduced MSE by 17 and 38% for diameter and height growth prediction, respectively. These models will prove to be quite useful for understanding the interaction of weather and silviculture in the Pacific Northwest and refining the precision of future growth model projections.