Item

Abstract

For trees in tropical forests, competition for light is thought to be a central process that offers opportunities for niche
differentiation through light gradient partitioning. In previous studies, a canopy index based on three-dimensional canopy
census data has been shown to be a good predictor of species-specific demographic rates across the entire tree community
on Barro Colorado Island, Panama, and has allowed quantifying between-species variation in light response. However,
almost all other forest census plots lack data on the canopy structure. Hence, this study aims at assessing whether positionbased
neighborhood competition indices can replace information from canopy census data and produce similar estimates
of the interspecific variation of light responses. We used inventory data from the census plot at Barro Colorado Island and
calculated neighborhood competition indices with varying relative effects of the size and distance of neighboring trees.
Among these indices, we selected the one that was most strongly correlated with the canopy index. We then compared
outcomes of hierarchical Bayesian models for species-specific recruitment and growth rates including either the canopy
index or the selected neighborhood competition index as predictor. Mean posterior estimates of light response parameters
were highly correlated between models (r.0.85) and indicated that most species regenerate and grow better in higher
light. Both light estimation approaches consistently found that the interspecific variation of light response was larger for
recruitment than for growth rates. However, the classification of species into different groups of light response, e.g. weaker
than linear (decelerating) vs. stronger than linear (accelerating) differed between approaches. These results imply that while
the classification into light response groups might be biased when using neighborhood competition indices, they may be
useful for determining species rankings and between-species variation of light response and therefore enable large
comparative studies between different forest census plots.