Asymptotic allometry and transition to the canopy inAbies balsamea

Summary There is a lack of consensus in theoretical and empirical literature on whether height–diameter (H:D) relationships of canopy trees are asymptotic. To investigate H:D allometry, particularly in the transition to the canopy, we focused on shade-tolerant Abies balsamea, across steep physical gradients associated with elevation, and correlated biotically generated gradients of stem density, canopy height and canopy species composition. We addressed these questions: A. What is the relation between H:D allometric form and emergence into the canopy? B. Is H:D allometry asymptotic? C. What aspects of the H:D relationship are most sensitive to environmental conditions? Using maximum likelihood, we compared the performance of alternative H:D forms (linear, power function, non-sigmoid asymptotic and sigmoid asymptotic), incorporating elevation and local canopy height as covariates. Abies balsamea H:D allometric form was clearly sigmoid asymptotic, by a wide AIC margin, across all gradients, to an asymptote slightly above canopy height. The commonly used power function had essentially no empirical support (ΔAIC >> 10). Canopy height decreased with elevation, but with much variation, so elevation was a poor predictor of the asymptote. Across environmental gradients, H:D allometries overlapped for trees far below the canopy, but diverged strongly for trees approaching different canopy heights. Changes in resource allocation and modification of growth form that could generate asymptotic H:D allometries are consistent with known physical forces and diverse research findings in tree developmental biology and physiology. Synthesis. The H:D asymptote and its tight relation to canopy height, unapparent in aggregated H:D data, became clear only when canopy height and elevation were included explicitly in allometric models. Asymptotic H:D allometries may be the norm, though often undetected, in unmanaged forests without recent catastrophic disturbance. Appropriate asymptotic H:D models may reduce error in forest carbon and biomass estimation and promote theoretical and empirical integration in the ecology and evolution of tree allometry.