Comparing strategies for representing individual-tree secondary growth in mixed-species stands in the Acadian Forest region

Tree diameter increment (ΔDBH) is a key component of a forest growth and yield model as predictions are passed to other submodels and tree-level estimates are scaled up to represent plot- and stand-level measures. A common problem faced in mixed-species stands is that ΔDBH needs to be characterized for numerous species, each with varying growth rates, shade tolerances, and competitive abilities. In addition, a variety of approaches have been used to model ΔDBH with unclear implications for general suitability for each species and overall prediction accuracy. This analysis used remeasurement data comprising 2,656,354 observations from 16,204 permanent sample plots across the Acadian Forest region of North America to develop and compare alternative approaches to estimating ΔDBH as well as stem basal area increment (ΔBA). Sixty-one species or genera including several with N < 10 were represented where observed mean growth rates ranged from 0 to 1.08 cm yr⁻¹, depending on species. Analyzing several modeling approaches to project DBH of the 15 most abundant species using various evaluation statistics to quantify prediction performance, this study showed that i) modeling ΔDBH was generally superior compared to approaches that estimated ΔBA, ii) a two-stage modeling procedure predicting potential growth and a corresponding multiplicative modifier to derive ultimate increment was mostly inferior compared to strategies predicting realized ΔDBH or ΔBA in a unified model form, and iii) species-specific, realized increment models exhibited similar behavior and accuracy compared to models fitted with modeling species as random effect. These key findings became even more evident when projection lengths increased (≥ 30 years). Our study thus showed the efficiency and flexibility of diameter predictions by including tree species as a random effect to account for ΔDBH differences of trees in mixed-species stands, including infrequent species. However, curves for rare species derived with the mixed effects modeling approach still need to be evaluated for biological plausibility as unbalanced or biased data can lead to uncharacteristic and potentially illogical behavior. Overall, the study highlights the challenges of accurately predicting ΔDBH across a range of species and conditions, but offers a general framework for future analyses in mixed species forests.