Supporting information belonging to the article: Can traits predict individual growth performance? A test in a hyperdiverse tropical forest

Summary The functional trait approach has, as a central tenet, that plant traits are functional and shape individual performance, but this has rarely been tested in the field. Here, we tested the individual‐based trait approach in a hyperdiverse Amazonian tropical rainforest and evaluated intraspecific variation in trait values, plant strategies at the individual level, and whether traits are functional and predict individual performance. We evaluated > 1300 tree saplings belonging to > 383 species, measured 25 traits related to growth and defense, and evaluated the effects of environmental conditions, plant size, and traits on stem growth. A total of 44% of the trait variation was observed within species, indicating a strong potential for acclimation. Individuals showed two strategy spectra, related to tissue toughness and organ size vs leaf display. In this nutrient‐ and light‐limited forest, traits measured at the individual level were surprisingly poor predictors of individual growth performance because of convergence of traits and growth rates. Functional trait approaches based on individuals or species are conceptually fundamentally different: the species‐based approach focuses on the potential and the individual‐based approach on the realized traits and growth rates. Counterintuitively, the individual approach leads to a poor prediction of individual performance, although it provides a more realistic view on community dynamics.


Notes S2
Relationship between traits and growth rates using an individual-based approach (that uses traits and growth rates of 370 saplings) and a species-based approach (that uses mean trait values per species for 36 species). Pearson's correlations (r) and coefficients of determination (R 2 ) and shown. Significant correlations (P<0.05) are shown in bold. The last column indicates the difference in explanatory power between the individual-based approach (R 2 individuals) and the speciesbased approach (R 2 species). For trait abbreviations see Table 1.
Notes S3. Discussion why some traits are more plastic than others We hypothesized that traits related to organ size and leaf display would be strongly related to tree sapling size and light, and would vary therefore mostly within species, whereas traits related to tissue quality and toughness would be phylogenetically more conserved and therefore vary also little within species. We indeed found high intraspecific trait variation, ranging from 17-78% across traits. As expected, traits related to tissue toughness (e.g., dry matter content and density) showed the least intraspecific variation, as they tend to be phylogenetically conserved (Chave et al., 2006). Surprisingly, several leaf display traits (SLA and LAR) showed relatively little intraspecific variation, which strongly contrasts with controlled studies, which have shown that leaf display traits change strongly with the light environment, to enhance light capture and carbon gain (Poorter, 1999;Poorter et al., 2009;Sterck et al., 2013). Perhaps in our study intraspecific variation in SLA and LAR was smaller than expected, because they did not respond to our quantitative index of crown exposure. We predicted that size-related traits should be very plastic in response to tree sapling size, but found instead that they show little intraspecific variation. For example, leaf size was the trait that showed the second-lowest variation within species (Fig. 1). Alternatively, intraspecific trait variation is simply the consequence of ontogenetic drift, tissue ageing, and wear and tear, which raises the question whether this large trait variation is functional, and whether it also implies a large trait acclimation.

Notes S4 Comparison between the mean and variation in trait values of trees in tropical dry, moist and wet forest in South America. Mean and variation (coefficient of variation, CV in %)
in trait values of branch wood density (WD), leaf dry matter content (LDMC), specific leaf area (SLA) and individual leaf size (LS) are shown. In the Bolivian studies trait data have been sampled with a different sampling design, so they are not strictly comparable with Ducke, but they are included here to illustrate the point how these forests might differ. In Bolivia traits have been sampled for shaded trees, but larger trees (between 10-20 cm cbh) compared to Ducke (1-5 cm DBH). In Bolivia, traits were sampled for 37-38 abundant species, differing in shade tolerance, 5 trees per species. Because of this design, not the whole community was sampled, and trait values may potentially show a larger spread because species were deliberately selected that differed in their strategy. To correct for this, we weighted the Bolivian trait values for the tree abundance of these species in 32-48 1-ha plots, where all trees >10 cm dbh were measured (Peña-Claros et al. 2012). Weighting was only done, based on the relative stem abundance of those 36 species for which the traits were measured. For Ducke each measured stem weighted as 1.
The difference in sampling design may have affected the results in two ways. Within species, taller trees tend to have more conservative trait values (i.e., higher WD, LDMC, LS and lower SLA), which may make the trait means of the Bolivian forests more similar to that of the conservative forest in Ducke, thus underestimating mean trait differences. Because in the Bolivian forest different strategies were sampled we may have overestimated their trait variation (CV) even when correcting for stem abundances. Because in Bolivia not the whole community and all plants in different micro environmental conditions were measured, we may have underestimated their trait variation (CV).
In terms of average trait values, saplings in the wet and nutrient poor Ducke forest have, compared to the other forests high LDMC and low SLA , which enhance leaf longevity