Does increasing canopy liana density decrease the tropical forest carbon sink?

The ongoing decline in the American tropical forest carbon sink has serious ramifications for atmospheric carbon levels and global climate change. Increasing liana abundance may explain the decaying carbon sink because lianas reduce canopy tree growth and survival, which limits forest carbon storage. However, canopy lianas, not solely understory lianas, would have to be increasing for this hypothesis to be credible because canopy lianas compete especially intensely with canopy trees. Data in this archive were used to examine the change in canopy lianas over 10 years on Barro Colorado Island (BCI), Panama to test three hypotheses. 1) Canopy lianas are increasing on BCI. 2) Increasing canopy lianas decrease aboveground canopy tree and forest carbon storage. 3) Lianas are the numerically dominant and most diverse woody plant growth form in the canopy. Our findings indicated that lianas are the numerically dominant and diverse woody plant group in the BCI canopy, and this dominance is increasing, reducing forest-level carbon storage and possibly explaining the decaying American tropical forest carbon sink.

Methods
We conducted this study in the BCI 50-ha forest dynamics plot in central Panama. The BCI 50-ha plot is one km in length and one-half km in width and is divided into 1250 20 x 20 m quadrats. A single hectare in the north-eastern part of the plot may have been disturbed ~125 years ago during the construction of the Panama Canal; however, the rest of the plot appears to be intact old-growth forest with no previous agriculture or burning.

We tested the effect of canopy liana change on canopy tree carbon change (H2) using Bayesian linear regression (Eq. 1 in the associated article) to model the mean change in aboveground canopy tree carbon (∆AGC) in each quadrat () as a Gaussian response to the change in liana density, basal area, and mean stem diameter. We confirmed that the relationship was linear by comparing linear and non-linear regression techniques (Appendix S1: Fig. S2). We controlled for spatial autocorrelation among adjacent quadrats using a quadrat-specific spatial random effect. We tested the numerical dominance and diversity of canopy lianas (H3) using Kruskal-Wallis H tests to determine whether median per-quadrat canopy liana density and species richness were greater than canopy tree density and species richness.