SEARCH

SEARCH BY CITATION

Keywords:

  • dominance;
  • functional trait diversity;
  • functional traits;
  • mixed-species plantations;
  • Panama;
  • species diversity;
  • tree carbon storage;
  • tropical forests

Summary

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  • Linking tree diversity to carbon storage can provide further motivation to conserve tropical forests and to design carbon-enriched plantations. Here, we examine the role of tree diversity and functional traits in determining carbon storage in a mixed-species plantation and in a natural tropical forest in Panama.
  • We used species richness, functional trait diversity, species dominance and functional trait dominance to predict tree carbon storage across these two forests. Then we compared the species ranking based on wood density, maximum diameter, maximum height, and leaf mass per area (LMA) between sites to reveal how these values changed between different forests.
  • Increased species richness, a higher proportion of nitrogen fixers and species with low LMA increased carbon storage in the mixed-species plantation, while a higher proportion of large trees and species with high LMA increased tree carbon storage in the natural forest. Furthermore, we found that tree species varied greatly in their absolute and relative values between study sites.
  • Different results in different forests mean that we cannot easily predict carbon storage capacity in natural forests using data from experimental plantations. Managers should be cautious when applying functional traits measured in natural populations in the design of carbon-enriched plantations.

Introduction

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

To fully understand the effect of species loss within tropical ecosystems, we need to link measures of biodiversity to ecosystem functions, such as carbon sequestration. The relationship between diversity and ecosystem function has been most intensively studied in temperate grasslands, where a decline in species number has negative effects on plant productivity (Tilman et al., 1996; Schwartz et al., 2000; Spehn et al., 2005). Similarly, tropical mixed-species plantations have greater total biomass and higher annual rates of carbon sequestration than monocultures (Erskine et al., 2006; Healy et al., 2008; Piotto, 2008). Moreover, the dominant species has been shown to partly determine carbon storage in tropical plantations and natural forests where, depending on its functional characteristics it can increase or decrease carbon storage (Balvanera et al., 2005; Kirby & Potvin, 2007; Ruiz-Jaen & Potvin, 2010).

There have been attempts to link tree diversity to carbon storage in tropical ecosystems using experimental plantations (Erskine et al., 2006; Healy et al., 2008) because they can allow for testing the mechanisms responsible for this link, but do not contain the natural variability that occurs in natural forests where species richness can vary by an order of magnitude or more (Scherer-Lorenzen et al., 2005). Thus, we do not know if results found in mixed-species plantations can be extrapolated to predict what is found in natural forests. For example, the most diverse tropical plantations have an average of eight species in a quarter of a hectare (Scherer-Lorenzen et al., 2005; Erskine et al., 2006; Piotto, 2008), while natural tropical forests have an average of 26 species in a similar area but with an upper limit of up to 300 or more species (Condit et al., 2005). If we can find a similar relationship between tree diversity and carbon storage in natural forests and experimental plantations, we can generalize about the importance for conserving biodiversity in tropical forests.

In the present study, we examined the relationship between tree carbon storage, species diversity, and functional diversity in a well-studied natural tropical forest in central Panama and compared results with a mixed-species plantation nearby. We are interested in understanding what explains variation in tree carbon storage to identify carbon sinks and help to decrease atmospheric CO2 concentrations (Malhi & Phillips, 2004; Houghton, 2005; Houghton et al., 2009). Tropical forests store more than one-quarter of the terrestrial carbon (Bonan, 2008; Houghton et al., 2009) and trees represent 90% of this (Kirby & Potvin, 2007). This carbon storage can be determined in part by species diversity (Balvanera et al., 2005; Spehn et al., 2005; Vila et al., 2007). Functional traits have also been proposed as an improved way to understand forest dynamics in hyperdiverse tropical forests (Wright et al., in press), because functional traits considered the redundancy in function of species. This function can group species according to their resource use and life history strategies (Grime, 2002). To link functional traits to a specific function, we need to select traits that are related to the ecosystem function of interest, in this case, carbon storage. Therefore, we chose wood density, maximum diameter at breast height (DBH), maximum height, leaf mass per area (LMA) and nitrogen fixers as functional traits related to tree carbon storage (Ishida et al., 2008). Wood density partly determines aboveground biomass (Baker et al., 2004) and correlates with growth rates and tree mortality (King et al., 2005; Chave et al., 2009), with high wood density being correlated with slow growth and longer-lived tree species. Maximum diameter is an indirect measure of the maximum aboveground biomass that can be attained by a species (Nelson et al., 1999; Chave et al., 2003) and a measure of light competition among species (Sheil et al., 2006). Maximum height (Hmax) is also a measure of light competition among species (Poorter et al., 2005; Moles et al., 2009). Leaf mass per area (LMA) is leaf biomass invested to produce the light-capturing foliar area and it is negatively correlated to photosynthetic rates (Thomas & Winner, 2002; Rozendaal et al., 2006; Poorter et al., 2009). The presence of root nodules in some trees to fix atmospheric nitrogen can increase the overall carbon uptake by plants (Cornelissen et al., 2003b; Hedin et al., 2009).

With these data available, we asked the following question. Does species richness, functional trait diversity, species dominance and functional trait dominance similarly explain proportions of tree carbon storage in a mixed-species plantation and in a natural forest? We are also interested in exploring how functional traits vary between sites within and among species. If functional traits can order species similarly in natural forest and mixed-species plantations, we can use the available information on functional traits from different natural forests to design plantations for carbon sequestration (Baroloto et al., 2010). Therefore, our second question was: how are functional traits differently ranked within in tropical forests and in plantations? Plantations can act as carbon sinks in areas where deforestation is a threat for preserving native species and could provide an alternative source of timber (Canadell & Raupach, 2008; Paquette & Messier, 2010).

Materials and Methods

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

Study sites

This study was conducted in natural forest on Barro Colorado Island (BCI) and a mixed-species plantation in Sardinilla < 20 km apart with similar precipitation regimes (both locations are in Panama). We used the same sampling protocols to collect wood density, maximum diameter, maximum height and LMA for all the species present at both sites. In Sardinilla we used a high-diversity experimental plantation established in a pastureland (9°18′N, 79°38′W) in 2003; > 800 trees were planted in 24 plots of 18 × 18 m with treatments of 6, 9 or 18 species replicated 8 times (Scherer-Lorenzen et al., 2005). These species were randomly drawn from a pool of 28 species found in the nearby natural forest of BCI based on how common they were in midstory to canopy layer (i.e. > 15 m of maximum height at maturity), their shade tolerance (i.e. light demanding, light intermediate and shade tolerant), and their timber value (Delagrange et al., 2008). The maximum number of species planted in the experimental plantation was based on the average number of species larger than 10 cm DBH in 20 × 20 m found in BCI (Scherer-Lorenzen et al., 2005). In this plantation, tree height, basal diameter at 10 cm from the ground and diameter at breast height of all individuals have been recorded yearly since its establishment. For statistical analysis, we used the data of the census in 2009. The topography of Sardinilla is almost uniformly flat, with a total elevation difference of 5 m (Potvin et al., 2004). This plantation receives a mean annual precipitation of 2350 mm with very low rainfall reported during the months of January to March and with an annual mean temperature of 25°C (Scherer-Lorenzen, 2007). The soil is an Alfisols dominated by clay over a Tertiary limestone and other sedimentary rocks (Potvin et al., 2004).

For natural forest, we used the nearby tropical moist forest of BCI that has a 50 ha permanent dynamic plot (9°09′N, 79°51′W). This site is a seasonal forest with a mean annual rainfall of 2600 mm with very dry periods during January–April (Leigh, 1999). Soils are weathered kaolinitic Oxisols composed mostly of red light clay (John et al., 2007; Barthold et al., 2008). In the 50 ha permanent plot, every individual > 1 cm DBH has been mapped and identified to species and has been censused every 5 yr since 1985 (Hubbell & Foster, 1987). For statistical analysis, we used data of the last available census on BCI, 2005.

To compare BCI with Sardinilla, we examined equivalent plot and tree sizes. For the plot size, we subdivided the 50-ha forest plot into plots of 20 × 20 m and excluded the ones along streams and with steep slopes (see Harms et al., 2001 for habitat classification). The small plot size helps to control for the effect of habitat heterogeneity, even though it increases carbon storage variation among plots (Chave et al., 2004). Moreover, the chosen plot size reflects the scale of individual tree competition, as neighbor effects are detected in radius of < 20 m (Hubbell et al., 2001; Wiegand et al., 2007). For tree size, we restricted our analysis to species with a maximum height at maturity of > 15 m, because the mixed-species plantation is composed only of midstory to canopy trees.

Tree carbon storage

We calculated the aboveground biomass (AGB) for individual trees using the following allometric regression for moist forest (Chave et al., 2005):

  • image

(DBH is the diameter at breast height (1.30 m) in cm; WD is the wood density for each species in g cm−3). To determine the WD values in the experimental plantation, we collected wood cores of five individuals per species and followed the standardized protocols in Cornelissen et al. (2003b). The WD values for the natural forest were provided S.J. Wright (Wright et al., in press). For those species with no WD values, we used the average wood density, 0.54 g cm−3 for moist forests in Panama (Chave et al., 2003). For individuals with multiple stems, we calculated the AGB of each stem and summed them. The estimation of AGB in the natural forests was done using the biomass function in the CTFS package in r (Hall, 2006). Then, we estimated tree carbon (C) storage per plot using: C (Mg C ha−1) = AGB × 0.46/plot area (Elias & Potvin, 2003).

Species and functional traits

We examined how species richness, functional diversity, and species dominance determined carbon storage in natural forest and mixed-species plantation. Species richness was estimated using rarefaction curves (Hurlbert, 1971), with the minimum number of stems in the mixed-species plantation (19 individuals) as the sample size. For functional diversity, we calculated the dispersion for wood density, maximum DBH, Hmax, LMA and potential for nitrogen fixers (NF) based on the trait value of the species present at each plot. Functional dispersion for each functional trait comes from a species–species distance matrix, where it calculates the species average distance from the centroid of each plot. We weighted species distance by the basal area of the species within each plot (Laliberté & Legendre, 2010). We estimated the functional diversity using the FD package in r using the function dbFD (Laliberté, 2009).

Wood density is the oven-dry mass divided by its fresh volume (Cornelissen et al., 2003b) and was calculated from wood cores of five individuals of each species (Wright et al., in press). In Sardinilla (25 species), we chose the largest individuals of each species within the study site, while for BCI (c. 300 species) we measured individuals close to the study site. Maximum DBH and Hmax is the maximum size a species can reach at maturity. To calculate maximum DBH and Hmax, we selected the six largest individuals of each species in Sardinilla and in the 50-ha plot of BCI based on previous census in 2009 and in 2005, respectively. The Hmax was measured using a telescopic measuring pole for trees < 15 m and a laser rangefinder and a clinometer for trees > 15 m (Wright et al., in press).

The LMA is the leaf oven-dry weight divided by its fresh area (Cornelissen et al., 2003b) and was calculated from two leaves of five individuals per species in the mixed-species plantation and in the natural forests. Most of the leaves collected were exposed to full sunlight, as suggested by Cornelissen et al. (2003b). However, we collected leaves in the shade when a species did not have individuals that were fully exposed to the sun (for details see Wright et al., in press). For nitrogen fixers, we assigned a one to species that had the potential to nodulate (i.e. observed having a symbiotic interaction with nitrogen fixing bacteria) based on field observation or existing literature. In the mixed-species plantation, we searched for active root-nodules from five individuals of each species identified as nitrogen fixers in previous studies (Cornelissen et al., 2003b). In the natural forest, we classified nitrogen fixers based their potential for nodulation according to existing literature (De Faria et al., 1984, 2010; Sprent, 2005).

For species dominance, we calculated the relative basal area (BA) for all the species in a plot to account for the contribution to the total basal area of the species dominating the BA in the plot. Previous studies have shown that only a few species accounted for > 90% of total carbon storage (Walker et al., 1999; Balvanera et al., 2005). For functional dominance, we calculated the community weight mean (CWM) for wood density, Hmax, LMA and nitrogen fixing. CWM is the mean of each species trait value weighted by the relative basal area of the species at each plot (Lavorel et al., 2008). We estimated the CWM using the FD package in r (Laliberté, 2009).

Statistical analysis

Before the statistical analysis, we controlled for the variation in stem density and light availability among plots. Stem density has been related to tree carbon storage because stem density is a trade-off between stem size classes, where a high density of trees is related to smaller average stem size (Clark & Clark, 2000; DeWalt & Chave, 2004). Light availability in a plot can indicate the presence of a gap or can help to determine stand age. Different light levels can affect species traits (Rijkers et al., 2000) and can enhance tree growth (King et al., 2005). Light availability in the mixed-species plantation was measured taking hemispheric photographs at 1 m above the ground at the center of each plot and using the software, gap light analyzer 2.0 (http://www.ecostudies.org/gla/), to obtain light availability. For the natural forest, we used existing data for light availability (Ruger et al., 2009) that was calculated as a function of the vertical distribution of canopy density at six height intervals. We regressed tree carbon storage against stem density and light availability and used the model residuals as the dependent variable. To compare the mixed-species plantation in Sardinilla with the natural forest of BCI, we partitioned the variation of the dependent variable among: species richness, functional diversity, species dominance and functional dominance (Borcard et al., 1992; Legendre, 2008). To estimate the variation explained by each independent matrix and assess model fit, we used adjusted R2, which controls for sample size and number of predictors (Peres-Neto et al., 2006). We performed these analyses using the vegan package in r (Oksanen et al., 2009). We selected independent variables using stepwise procedures to avoid multicollinearity and over-fitting the model (Crawley, 2007). We also used the variance inflator factor to select independent variables and excluded the ones with values > 10 (Oksanen et al., 2009). After these variable selection procedures, we excluded diversity and dominance of wood density and maximum diameter and retained the relative basal area of two dominant species. To estimate functional trait variability between study sites, we compared species ranking based on four traits (wood density, maximum diameter, maximum height and LMA) for species shared in Sardinilla and BCI. According to Garnier et al. (2001), no change in species ranking among sites implies that there is no trait plasticity in different environments (e.g. plantation vs natural forests), or that all species responded similarly to the environment.

Results

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

Determinants of carbon storage

Tree carbon storage was highly variable at both study sites (Table 1). It ranged from 0.83 to 10.75 Mg C ha−1 in Sardinilla and from 12.91 to 856.40 Mg C ha−1 in BCI. Not surprisingly, carbon storage was much higher in BCI because of the young age of Sardinilla. It is interesting to note, that after only 6 yr of growth some of the plantation plots had approached the minimum biomass observed in the forest plots. Nitrogen fixers represented 32% of the species in Sardinilla (from a total of 25 species) but 14% of the species in BCI (from a total of 157 species ≥ 15 cm DBH). Functional diversity and the CWM of Hmax, LMA, and N fixers had similar coefficient of variation at the plot level in both sites (Table 1).

Table 1.   Mean and coefficient of variation of tree carbon storage, stem density, light availability, species richness, functional diversity (standard deviation of maximum height (Hmax), leaf mass per area (LMA), and nitrogen fixers (NF)), species dominance (relative basal area (BA) of Anacardium excelsum and Tabebuia rosea), and functional dominance (community weight mean of Hmax, LMA and NF weighted by the BA) in the mixed-species plantation of Sardinilla and the natural forest of Barro Colorado Island (BCI) in Central Panama
VariablesSardinilla (n = 24)BCI (n = 815)
  1. 1Plots removed where these species were absent.

  2. CWM, community weight mean.

Carbon storage (Mg C ha−1)5.44 (0.47)146.80 (0.79)
Stem density (trees ha−1)700.88 (0.18)2371.73 (0.23)
Light availability (%)0.10 (0.38)0.01 (0.62)
Species richness7.18 (0.28)13.33 (0.14)
Functional diversity
 Hmax (m)0.76 (0.22)0.66 (0.28)
 LMA (g m−2)0.51 (0.29)0.58 (0.42)
 NF0.40 (0.47)0.28 (0.23)
Species dominance1
 Anacardium excelsum0.10 (0.92)0.48 (0.64)
 Tabebuia rosea0.23 (0.38)0.04 (2.04)
Functional dominance
 CWM Hmax (m)28.26 (0.13)28.82 (0.11)
 CWM LMA (g m−2)69.64 (0.07)72.81 (0.14)
 CWM NF0.29 (0.68)0.09 (1.63)

Species richness, functional diversity, species dominance and functional dominance together explained 67% of the variation of carbon storage in Sardinilla and 49% for BCI, and their contribution to carbon storage variation differed between the two sites. Species richness explained most of the tree carbon storage variation in Sardinilla but did not contribute significantly to carbon in BCI (Fig. 1). By contrast, functional dominance explained most of the variation in the natural forest (Fig. 1) with CWM of Hmax as the variable most responsible for this pattern (Table 2). Another important source of variation was the interaction between species dominance and functional dominance; it explained 8% of carbon storage variation in Sardinilla and 12% in BCI. Functional diversity was only important for BCI, which explained 6% of carbon storage variation (Fig. 1; Table 2).

image

Figure 1.  Proportion of tree carbon storage variation explained after controlling for stem density and light availability in the mixed-species plantation of Sardinilla (black bars) and the natural forest of Barro Colorado Island (grey bars) located in Central Panama. Independent matrices were: species richness (SpDiv), functional diversity (FunDiv), species dominance (SpDom) and functional dominance (FunDom).

Download figure to PowerPoint

Table 2.   Pearson correlations (r) between tree carbon storage controlled by stem density and light availability with species richness, functional diversity, species dominance and functional dominance in the mixed-species plantation of Sardinilla and the natural forest of Barro Colorado Island (BCI) in Central Panama
 Carbon storage (Mg C ha−1)
Sardinilla (n = 24)BCI (n = 751)
  1. Abbreviations: Hmax, maximum height (m); LMA, leaf mass per area (g m−2); NF, nitrogen fixers; CWM, community weight mean weighted by basal area.

  2. Significance level: ns, nonsignificant; *, < 0.05; **, < 0.01; ***, < 0.001.

Species richness0.544**−0.084*
Functional diversity
 Hmax0.053ns−0.315***
 LMA0.490*−0.004ns
 NF0.343*−0.001ns
Species dominance
 Anacardium excelsum−0.485*0.171***
 Tabebuia rosea−0.494*0.008ns
Functional dominance
 CWM Hmax−0.523**0.628***
 CWM LMA−0.450*0.130***
 CWM NF0.397*0.098**

Species dominance explained the higher proportion of tree carbon storage variation in Sardinilla (12%) than in BCI (2%). The two species chosen after forward selection were nonpioneer species and allocated carbon differently (Tables 2 and 3). Anacardium excelsum has higher investment producing more branches than Tabebuia rosea, which invests in vertical growth (Table 3). In BCI, T. rosea is more abundant (0.23% of the stems) than A. excelsum (0.02% of the stems), but achieved smaller DBH sizes (10.5 cm and 105.6 cm average DBH, respectively). In Sardinilla, these two species along with five others contributed more than half of the basal area (Fig. 2a). For BCI, A. excelsum, along with 12 other species, contributed half of the basal area and T. rosea was ranked as the 49th largest contributor to basal area (Fig. 2b).

Table 3.   Description of the two tree species used to explain the variation of carbon storage in the mixed-species plantation, Sardinilla, and the natural forest of Barro Colorado Island (BCI) located in Central Panama
DescriptionAnacardium excelsumTabebuia rosea
SardinillaBCISardinillaBCI
  1. Traits values represented the average of the six largest individuals and values in parenthesis are the coefficient of variation (Wright et al., in press). LMA, leaf mass per area.

  2. 1Information taken from Delagrange et al. (2008).

FamilyAnacardiaceaeBignoniaceae
Wood density (g cm−3)0.31 (0.24)0.42 (0.06)0.39 (0.21)0.49 (0.14)
Maximum diameter at breast height (mm)171.82 (0.14)1775.67 (0.08)149.62 (0.17)752.20 (0.13)
Maximum height (m)10.45 (0.02)35.11 (0.09)10.20 (0.04)40.42 (0.07)
LMA (g m−2)82.82 (0.12)98.26 (0.14)116.06 (0.09)101.15 (0.22)
Economic use1Timber valueHigh timber value and ornamental
image

Figure 2.  Species ranked by their dominance in the relative basal area in the mixed-species plantation of Sardinilla (a) and the natural forest of Barro Colorado Island (BCI; b), Central Panama. Values are average per cent of basal area (BA) by species in each plot. Lines are 95% confidence intervals. Note that only the first 100 species of the natural forests are included, the rest are rare species that contribute very little to total BA in each plot. AE, Anacardium excelsum; TR, Tabebuia rosea.

Download figure to PowerPoint

Species trait ranking

None of the traits gives a consistent ranking of species between sites. Wood density and Hmax values were consistently lower in Sardinilla than BCI (Fig. 3a,c). For maximum DBH and LMA, trait values increased or decreased depending on the tree species (Fig. 3b,d). There was a large range of LMA values in BCI; by contrast, values were low to medium in Sardinilla (Fig. 3d).

image

Figure 3.  Species ranking comparison based on four functional traits of the species shared across the mixed-species plantation (Sardinilla) and the natural forest (Barro Colorado Island). Traits are: wood density (WD; a), log maximum diameter at breast height (DBH; b), maximum height (c) and leaf mass per area (LMA; d). Species abbreviations: AG, Astronium graveolens; CO, Cedrela odorata; CG, Colubrina glandulosa; DO, Dipteryx oleifera; GU, Guazuma ulmifolia; HC, Hura crepitans; IP, Inga punctata ; LS, Luehea seemannii; OM, Ormosia macrocalyx; TR, Tabebuia rosea; TA, Terminalia amazonia.

Download figure to PowerPoint

Based on species ranking by traits, we can identify species that perform very differently in the mixed-species plantation of Sardinilla and in the natural forest of BCI. For example, Astronium graveolens (AG), Dipteryx oleifera (DO) and Ormosia macrocalyx (OM) were consistently lower in the species ranking from BCI to Sardinilla (Fig. 3a–d). By contrast, Colubrina glandulosa (CG), Guazuma ulmifolia (GU), Inga punctata (IP), and Terminalia amazonia (TA) improved their species ranking from BCI to Sardinilla.

Discussion

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

Sources of variation for carbon storage

Tree carbon storage in the mixed-species plantation was mainly explained by species richness and in the natural forests by functional trait diversity. For example, species richness only mattered in Sardinilla, where increasing the number of species also increased carbon storage. This relationship was expected because trees were planted according to different species functions, including shade tolerance and nutrient acquisition ability (Scherer-Lorenzen et al., 2005). By contrast, we found no relationship between species richness and carbon storage in the BCI forest. One explanation of the lack of relationship is that carbon storage in the natural forest may have reached saturation in species richness, an effect that can be found in high species richness treatments in experimental grasslands (Silver et al., 1996; Wright, 1996; Hooper et al., 2005). In those systems, < 15 species are needed to reach the highest values of plant productivity (Hector et al., 2001; Balvanera et al., 2005). Saturation between species number and carbon storage can vary among sites and will depend on the niche overlap among species. Previous studies on BCI have found that most species have overlapping ecological niches along an environmental gradient within the same forest type (Wright, 1996; Harms et al., 2001; Cardinale et al., 2007).

Functional dominance explained most of the variation of tree carbon storage on BCI; more specifically, we found that the more a canopy tree dominates (i.e. higher CWM of Hmax) a plot, the greater the increase in carbon storage. Canopy trees have a higher probability of survival and can grow faster owing to asymmetry in the size relative to their neighbors’ (Hubbell et al., 2001; Potvin & Dutilleul, 2009). Plots with low Hmax diversity, where most trees have similar Hmax, had higher tree carbon storage. By contrast, we can expect that sites with more variation in Hmax are associated with younger forests and forest gaps but this could also be a result of greater liana loads (Schnitzer & Carson, 2001), which reduces carbon storage.

The diversity and the proportion of nitrogen fixers enhanced carbon storage only in the mixed-species plantation in Sardinilla. Nitrogen content in the soil was lower and less variable in Sardinilla (15.56 mg kg−1 N; Zeugin et al., 2010) than in BCI (25.92 mg kg−1 N; John et al., 2007). In Sardinilla, the previous presence of pasturelands has altered the nitrogen cycle by reducing nitrogen content in the soil after forest cutting (Jordan, 1985; Silver et al., 2005). Thus, an increase in the proportion and diversity of nitrogen fixers in Sardinilla has increased carbon storage through soil improvement in these nutrient poor sites. Nitrogen fixers have been found to be facultative, where nitrogen fixation occurs in disturbed sites such as in Sardinilla or in forest gaps (Barron et al., 2010). The high nitrogen soil-content BCI could suppress nitrogen fixation by trees, by reducing the relative advantage of symbiotic fixation over direct uptake (Barron et al., 2010; Hedin et al., 2009). Moreover, preliminary data along a forest chronosequence have shown that younger stands, such as Sardinilla, have a higher occurrence of nodulation than old-growth forest (S. Batterman et al., unpublished).

In both sites, the effect of LMA diversity on carbon storage was generally lower than that found for Hmax and nitrogen fixers. In the experimental plantation, we found that plots dominated by trees with low LMA have higher carbon storage. In general, species that have lower LMA have higher growth rates because of their fast resource acquisition (Reich et al., 2007; Poorter et al., 2009), these are also the species that have greater aboveground biomass in the experimental plantations. Thus, when they dominate the plot, there is more tree carbon storage. We found the opposite relationship in the natural forests, as the old-growth forest canopy is mostly dominated by late successional species with high LMA. Similarly, LMA values have been shown to increase with tree height (Rijkers et al., 2000; Poorter et al., 2009).

The dominance of a given species alone is not a strong determinant for tree carbon storage. However, we still found significant effect of a species and this effect depends on the basal area each species occupied. For example, the positive effect of A. excelsum on carbon storage in the natural forest results mostly from its capacity to reach large DBH even though it has a low wood density. In natural forests, aboveground biomass in trees has been mostly determined by their diameter size and not by their wood density (Chave et al., 2004).

Species trait ranking

There is little interspecific trait consistency among species present at both study sites. Others studies have found that their high intraspecific trait variability provides no clear pattern (Rozendaal et al., 2006; Albert et al., 2010). Changes in interspecific trait ranking have been observed from seedlings grown in a glasshouse with plants found in natural environments (Cornelissen et al., 2003a). Species ranking based on single functional traits can be attributed to ontogenetic changes between study sites and to different responses of species to the environment (Garnier et al., 2001; Cornelissen et al., 2003a). For example, there are changes in wood density and LMA at different ontogenetic stages in other tropical forests (Rozendaal et al., 2006; Ollivier et al., 2007), where low wood density at early stages is associated with a higher proportion of sapwood relative to heartwood (Chave et al., 2009). We would have expected to observe lower values for most of the functional traits in Sardinilla than in BCI due to ontogeny because this plantation has only been established for 6 yr. However, we were surprised to see no clear pattern in the species ranking by all functional traits between study sites.

The convergence in the LMA values observed in Sardinilla can be explained by the higher light availability for each individual compared with the forest of BCI (Table 1). Similarly, some species in Sardinilla invested more in growing horizontally (i.e. achieved higher values for maximum DBH than Hmax) than vertically. Investing in tree height is a response of light competition in tropical forests (Poorter et al., 2005; Kooyman & Westoby, 2009). Thus, if light is not a limiting resource, tree species will not invest in height, but in growing branches.

Species that reached larger sizes in the mixed-species plantation of Sardinilla showed no clear similarities among their functional traits. However, we found that these species had low LMA and low wood density in natural forest. These characteristics are associated with pioneer species, which have a trade-off between high growth and low carbohydrate storage (Poorter & Kitajima, 2007; Chave et al., 2009). By contrast, the species that could be considered good carbon storers (e.g. D. oleifera and A. graveolens), based on their high wood density and Hmax in the natural forest, had low carbon storage in Sardinilla, even though these species have high growth rates relative to their wood density (i.e. growth rate is negatively correlated to wood density) and are light-demanding species (Condit et al., 2006). Wishnie et al. (2007) found that these two species grow poorly in other mixed-species plantations in Panama.

Can we use plant functional traits to design carbon-enriched plantations? We have to be cautious if we want to use information on functional traits from natural forest plots to design experimental high-carbon plantations. However, some species have been identified as having potential for mixed-species plantations, as they can grow to larger sizes, are good for timber (Kirby & Potvin, 2007) and can store carbon for long periods of time (Chambers et al., 1998). Despite the poor grouping of species based on functional traits, we can recommend the use of C. glandulosa, G. ulmifolia, Hura crepitans, T. rosea and T. amazonia. Most of these have been reported to grow well in both diameter and height in mixed-species plantations (Wishnie et al., 2007) and, more specifically, T. amazonia and I. punctata have been used to reduce degraded areas dominated by grasslands (Kim et al., 2008; Craven et al., 2009).

Acknowledgements

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

Authors were supported by the Panamanian Government through the IFARHU-SENACYT (M.C.R-J.) and the Canadian Government with the Discovery Grant from NSERC (C.P.). The FH Levinson Fund supported plant trait measurements for Barro Colorado Island. National Science and McArthur Foundations supported the 50-ha plot censuses. We thank Andy Jones, Richard Norby and three anonymous reviewers for insightful comments on previous versions of the manuscript. We thank Cristina Salvador, David Brassfield, Sebastian Bernal, Paulino Villarreal, Eduardo Medina, Rolando Perez, Salomon Aguilar and Javier Ballesteros for collecting traits for BCI data. Jose Monteza, Lady Mancilla and Jurgis Sapijanskas collected trait and light availability data in Sardinilla.

References

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References