Functional trait divergence of juveniles and adults of nine Inga species with contrasting soil preference in a tropical rain forest

Authors


Correspondence author. E-mail: dpalow@ufl.edu

Summary

1. Community-wide studies have shown that functional traits of tropical trees vary with soil-type preference, but few have examined trait diversification among closely related taxa. In this study, we asked how functional traits of adults and saplings within a speciose genus, Inga, differ in relation to their soil-type preferences.

2. We quantified soil-type preference and functional traits of nine Inga species (Fabaceae) in the wet tropical forest at La Selva Biological Station, Costa Rica, where rich alluvial soil and less fertile residual soil of volcanic origin occur in close proximity.

3. From the extensive trail network, we randomly selected 104 trail segments, each 50 m long, along which all Inga individuals >50 cm tall were counted and identified to species. Five and three common species showed significant bias to residual or alluvial soil, respectively, while the remaining one showed no significant bias. For these nine species, we quantified five leaf traits for saplings and adults, as well as wood-specific gravity (WSG) and three seed traits.

4. For both adults and saplings, leaf mass per area (LMA) and lamina density were higher for residual-soil specialists than fertile alluvial-soil specialists. The opposite trend was found for nitrogen (N) and phosphorus (P) concentrations in leaves. Lamina thickness did not differ between soil-type preference groups. All leaf trait values except N concentration increased from saplings to adults, with adults showing larger interspecific variations in trait values than saplings.

5. Saplings and adults of the single species found on both soils had trait values intermediate between residual and alluvial-soil specialists.

6. Seed mass and seed N, as well as WSG of adults, were higher in residual-soil specialists than alluvial-soil specialists. Higher seed N indicates higher maternal investment of species growing on less fertile soil.

7. These differences in traits among closely related species across continuous soils suggest that habitat filtering has contributed to the evolution of seed, juvenile and adult traits within a speciose tropical tree genus.

Introduction

Contrasting soil resource niches should favour establishment, growth and reproduction of individuals with certain combinations of functional traits, and hence non-random spatial distribution of trait values in plant communities (Kraft, Valencia & Ackerly 2008). Such habitat filtering is a potential mechanism that explains non-random distribution of tropical tree species depicted in community-wide comparisons (Russo et al. 2005; John et al. 2007) and within-family comparisons conducted for Lauraceae, Burseraceae, Malvaceae (Sterculiaceae), Myristicaceae and Fabaceae (Sri-Ngernyuang et al. 2003; Fine et al. 2005; Yamada et al. 2006; Baraloto et al. 2007; Queenborough et al. 2007). To link the ecological significance of individual traits with habitat preference of species, the most direct evidence could be sought among closest relatives that share phylogenetic backgrounds and hence non-focal traits (e.g. within genus Quercus by Cavender-Bares, Kitajima & Bazzaz 2004). But, functional trait diversification within a genus has rarely been quantified for tropical trees (Queenborough et al. 2007), even though soil-habitat specialization is known for species-rich genera, such as Inga (Harms et al. 2001; Dexter, Pennington & Cunningham 2010; Endara & Jaramillo 2011) and Shorea (Russo et al. 2005) that contribute significantly to local tree diversity. Furthermore, juveniles and adults may differ in functional trait divergence with soil-type preference, but little is known about such ontogenetic shifts. In this study, we aim to ask how functional traits of adults and saplings within a speciose genus, Inga, differ in relation to their soil-type preferences.

Functional traits are traits that lead to growth and survival differences, as well as differences in resource niches and ecological strategies, among coexisting species (Cornelissen et al. 2003; Cavender-Bares, Kitajima & Bazzaz 2004; Kitajima & Poorter 2008; Wright et al. 2010). In particular, leaf mass per area (LMA), leaf nitrogen (N) concentration, leaf size, seed mass and wood-specific gravity (WSG) have been linked to growth-survival trade-offs and resource niche specialization (Baraloto et al. 2007). Low LMA and high leaf N are typical of fast-growing species common in resource-rich environments such as treefall gaps and fertile soils (Wright et al. 2004; Poorter & Bongers 2006; Kraft, Valencia & Ackerly 2008; Ordoñez et al. 2009). Dry mass density (dry mass per area) of leaves and stems are associated with each other and growth-survival trade-offs; high values of WSG and lamina density are associated with high survival, slow growth and specialization to shaded or infertile habitats (Poorter et al. 2008; Russo et al. 2008; Chave et al. 2009; Kitajima & Poorter 2010).

The degree of soil-habitat specialization increases from juveniles to adults (Russo et al. 2005), but it is not well understood whether environmental filters acting upon functional traits operate at either or both juvenile and adult stages to create the observed patterns of trait-environmental associations. Information on how adult and juvenile traits differ in relation to ecological niche specialization of species will contribute to mechanistic understanding of the demographic processes that lead to habitat filtering (Lusk & Warton 2007; Poorter 2007; Kitajima & Poorter 2010). Seedling establishment represents the earliest ontogenetic stage in which trait-based habitat filtering may operate, for which functional traits of seeds are relevant. Larger seeds initially produce larger seedlings that can tolerate various hazards including burial by litter and physical disturbance (Moles et al. 2004) and enhance establishment in nutrient-poor soils by enabling construction of larger initial root system (Jurado & Westoby 1992; Hanley et al. 2007). Furthermore, higher concentration of a mineral element may extend the time during which seed reserves support seedling demands for that nutrient (Hanley & Fenner 1997; Kitajima 2002). However, the existing data are equivocal as to the relationship of seed size and mineral nutrient concentration with specialization to fertile vs. infertile soils; tropical species found on more infertile soils have smaller seeds with higher N concentrations (Grubb & Coomes 1997), and Proteaceae species from infertile sandstone soils have larger seeds with higher phosphorus (P) concentrations than those from less infertile soils (Esler et al. 1989; Milberg & Lamont 1997; Groom & Lamont 2010). Russo et al. (2007) found no association between seed size and soil-habitat specialization for Malaysian tropical tree species, but comparison within a narrow phylogenetic group may be more conclusive because phylogenetic constraints are particularly strong for seed size.

The objective of this study is to examine how habitat association and ontogeny affect leaf, wood and seed traits of species in the genus Inga in Fabaceae (legumes). Fabaceae are a species-rich family that represents approximately 16% of the woody species in Neotropical forests (Burnham & Johnson 2004). Inga is a large neotropical genus consisting of approximately 300 species that rapidly diversified within the last 2–10 million years (Richardson et al. 2001). Within the wet tropical forest of La Selva Biological Station (LS) in Costa Rica, expert observations suggest that several Inga species appear to grow either on rich alluvial soil or less fertile residual soil that occur in close proximity (Zamora & Pennington 2001, N. Zamora pers. comm.).

We asked the following questions using common Inga species at LS: (i) What is the degree of habitat association of Inga species according to soil type, residual or alluvial, at LS? (ii) Do leaf and wood functional traits of adults and juveniles differ with the observed soil-type preferences? (iii) How do juvenile leaf functional traits correlate with adult leaf functional traits? (iv) Do residual-soil specialists differ in seed mass and nutrient content from alluvial specialists? We predicted that both adults and juveniles of species associated with less fertile residual soils show trait syndromes associated with slower growth, such as high values of LMA, WSG and seed size, but low concentrations of N and P compared to alluvial-soil specialists. On the basis of other studies that examined juvenile and adult leaf traits (e.g. Poorter 2007; Kitajima & Poorter 2010), we expected that a given leaf trait exhibit positive correlation between juveniles and adults.

Materials and methods

Site, Species and Spatial Distribution

La Selva Biological Station is a tropical wet forest in the central lowlands of Costa Rica located adjacent to Braulio Carrillo National Park (10° 26′N, 83° 59′W). The two major soil types at LS, ‘residual’ soils on terra firma and alluvial (old alluvium) soils on river terraces, both belong to Oxisols (Kleber et al. 2007) with similar total N, pH and water contents, but total nutrient stocks in 0–75 cm depth (kg ha−1) are much lower in residual soils than alluvial soils for phosphorus (44% lower) and potassium (39% lower) (Espeleta & Clark 2007). Distribution of several species of trees and palms at LS exhibit significant associations with either of these soil types (Clark, Palmer & Clark 1999). Zamora & Pennington (2001) classify species within Inga to those with no soil-type preference (generalists), alluvial-soil specialists, residual-soil specialists and riparian species, and describe their growth habits as ranging from subcanopy trees to canopy emergents. Twenty species of Inga are found at LS, and many are considered secondary forest species (Zamora & Pennington 2001). According to the best molecular phylogeny available, there was no obvious phylogenetic bias in soil-type preferences among Inga species at La Selva (Fig. S1, K. Dexter, unpublished data).

To confirm these expert observations, we surveyed the distribution of all Inga individuals along 52 transects, each 50 m long, per major soil type (residual or alluvial) along the existing trails at LS. The trail segments surveyed were randomly chosen from all trail segments within old growth, second growth and selectively logged forest. We recorded all individuals of Inga that were >50 cm tall and occurred within 2 m of the trail edge, identifying them to species and recording the location of individuals >2 m tall. Voucher specimens of each species were collected and identified by Nelson Zamora at the National Biodiversity Institute of Costa Rica. A selection of these vouchers was left at the National Herbarium of Costa Rica.

Functional Traits Measurements

For the nine most common species of Inga encountered in our survey, we determined species mean values for leaf traits (LMA, lamina thickness, lamina density, lamina N, lamina P) for adults and juveniles, WSG of adults, seed biomass, seed N and seed P. While most individuals sampled occurred within 2 m of an established trail, some individuals were sampled up to 20 m away from an established trail. All were collected from areas where the soil type has been mapped. Leaves and wood were primarily collected from individuals growing on their preferred soil type (as identified by the species distribution survey). We attempted to collect samples from individuals occurring on both their preferred and non-preferred soil types to assess how trait values may show plastic responses to in situ soil types where individuals were sampled. Most species, however, rarely if ever occurred on both soil types, and we could collect an adequate number of samples only for three species. For the three species sampled from the two soil types, the traits of individuals did not differ significantly (see Table S1). Leaves from adults (a minimum of five per individual from the preferred soil type) were collected from the canopy using a crossbow to break off small branches from the tree. For saplings (defined as 1–3 m tall), we collected three leaves per individual. Leaves were collected from the highest position of the tree crown, wrapped in wet paper towel along with branches, sealed in plastic bags and kept cool in a refrigerator overnight until they were processed the following day. Light environment of each tree was categorized using a crown illumination index (CII, see Clark & Clark 1992); the amount of the crown exposed to light and the direction of light were estimated. Trait measurements generally followed Cornelissen et al. (2003). Leaf processing included wiping them clean of epiphylls with a paper towel, followed by measurement of leaf size (including rachis and petiole) with a leaf area metre (LI-3100; LICOR, Lincoln, NE, USA), lamina thickness with a micrometre (at three positions avoiding veins as much as possible) and leaf fresh mass.

WSG sampling followed the protocol described in Chave et al. (2006); all individuals were between 10 and 36 cm diameter at breast height (130 cm). Wood cores were collected from five individuals per species (the same individuals from which leaves were collected whenever possible), except I. acuminata for which samples from only three individuals were collected. Seeds were collected during March 2007, May–June 2007, February–July 2008, January–August 2009 and May 2010 directly from multiple parent trees or from the ground near fruiting adults. Seeds were removed from pods, rinsed with de-ionized water, towel-dried and weighed for fresh mass.

Leaves and seeds were dried at 60 °C for 3–4 days, and their dry mass determined, then ground for analysis of N with an elemental analyzer ((ECS 4010 CHNSO Analyzer; Costech Analytical Technologies, Valencia, CA)) and P with an ash-digest ascorbate assay (Murphy & Riley 1962). Seeds including paper-thin seed coat for each parent tree were pooled as a replicate before grinding.

Statistical Analysis

The significant bias of individual distribution in relation to soil types was tested with the Fisher's exact chi-squared test under the null hypothesis that equal numbers of individuals should grow on both soil types. Furthermore, species were ranked according to the per cent of individuals found on residual soil. Changes in CII among species and life stages were also analyzed using Fisher's exact chi-squared test. For leaf functional traits, we examined the effects of soil preference (alluvial vs. residual), species nested within soil-preference type and ontogenetic stage (sapling vs. adults) and interaction between ontogenetic stage and species nested within soil-preference type. As sapling data were not collected for I. acuminata, it was excluded from this analysis. The models were tested with and without the only species without soil bias, I. leiocalycina. Seed and wood data were also tested for soil preference and species nested within soil preference. Whenever the effect of species was significant, a Tukey's post hoc test was conducted across all nine species.

Results

Species Distribution

In 97 of the 104 survey locations at least one Inga individual > 50 cm tall was encountered. In total, 813 individuals >50 cm tall were identified to species. We found 16 Inga species; the nine species with a total of at least ten individuals were used in further analyses (Table 1). Individuals of these nine species comprised 96% of all Inga individuals found in the survey. These nine species were classified according to their soil-preference types based on the number of individuals occurring on residual and alluvial soils. Eight species showed distribution biases to either residual (5 spp.) or alluvial (3 spp.) soil, while one occurred on both soil types (Table 1). The CII experienced by individuals was not different among species for saplings (χ2 = 36·7, P = 0·38); however, it was different among species for adults (χ2 = 79·3, P < 0·001). Additionally, adults had higher CII than saplings (χ2 = 75·3, P < 0·001).

Table 1. Distribution of Inga species along 104–50-m-long trail segments at La Selva Biological Station for which at least a total of 10 individuals taller than 50 cm were encountered. Chi-squared and P-values indicate the results of Fisher's exact test for soil-type preference for residual soil, alluvial soil or without significant bias. The species were ranked from greater to lesser degree of bias to the residual soil based on the per cent of individuals found on residual soils
SpeciesRankNo. on residual soilNo. on alluvial soilSoil-type preferenceχ2 P
Inga cocleensis 1111Residual14·8<0·001
Inga acuminata 2101Residual13·2<0·001
Inga thibaudiana 311714Residual154·0<0·001
Inga alba 4203Residual25·1<0·001
Inga pezizifera 57023Residual59·6<0·001
Inga leiocalycina 62537None0·20·25
Inga oerstediana 721139Alluvial41·5<0·001
Inga sapindoides 89192Alluvial95·1<0·001
Inga marginata 9373Alluvial37·1<0·001

Functional Traits

All of the leaf functional traits examined, except for leaf lamina thickness, were significantly different between soil-type preference groups regardless of whether the only unbiased species (Ileiocalycina) was excluded from the analysis (Tables 2 and 3) or included (results not shown). Residual specialists had higher values of LMA and leaf density than alluvial specialists, while the unbiased species (I. leiocalycina) had intermediate LMA, low lamina thickness and high leaf density as adults (Fig. 1). Nutrient concentrations in leaf laminas were higher for alluvial specialists than residual specialists per unit mass (%N and %P, Table 2), but not per unit area (Table S2). Residual specialists also had higher WSG than alluvial specialists, although there was a large overlap between the two groups (Table 2).

Figure 1.

Three leaf traits (mean + SE) of saplings (left side) and adults (right side) of nine Inga species, listed in the order of strong to lesser bias to residual soil (Tables 1 and 2). (a) Leaf mass per area (LMA, g m−2), (b) lamina thickness (mm) and (c) lamina tissue density (g cm−3). Open bars for residual-soil specialists, dotted bars for unbiased species and hatched bars for the alluvial-soil specialists. No data available for sapling leaves of species 2.

Table 2. Species means for seed mass, N and P concentrations and wood-specific gravity (WSG) of nine Inga species (See Table 1 for species identity, Table 3 for anova results). Also shown at the bottom are means for soil-type preference to residual (R) and alluvial (A) soils. For leaf traits, n = 10 individuals per species except n = 7 for Sp. 2. For WSG, which was determined only for adult trees, n = 5 except n = 3 for Sp. 2. Letters that follow mean values indicate the results of post hoc Tukey–Kramer pairwise comparisons. ND, no data
Species rankSoilSeed mass (g)N (%)P (%)WSG (g cm−3)
SeedSaplingAdultSeedSaplingAdultAdult
1RNDND3·23d2·77dND0·11de0·10d0·59a
2R0·30bcd3·67bcND3·53b0·15ND0·14c0·61a
3R0·13d3·31bcde3·13d3·06cd0·190·10e0·11d0·44b
4R0·30bcd2·03de3·34cd3·73b0·140·11de0·16bc0·52ab
5R0·73a4·99a3·67b3·44b0·200·14cd0·14c0·44b
6N0·71a3·38bcde3·66b3·43bc0·190·15c0·15c0·53ab
7A0·27c3·26bd3·54bc3·56b0·180·15bc0·17bc0·47ab
8A0·36b2·94bcde4·34a4·28a0·180·19ab0·19ab0·49ab
9A0·15d2·48ce3·19d3·53b0·160·20a0·22a0·41b
Soil pref.R0·494·033·343·290·180·120·130·51
MeansA0·302·903·693·790·170·180·190·46
Table 3. The results of nested two-way anova testing the effects of soil preference (alluvial or residual), ontogenetic stage (sapling or adult), species (sp., nested within soil-preference type) and their interactions on five leaf traits. LMA, leaf mass per area. Shown are P-values with degree of freedom (d.f.) at the top. Leaf trait analyses excluded Inga leiocalycina (with no soil preference) and Inga acuminata (with no sapling data), and seed trait analyses excluded Inga leiocalycina and Inga cocleensis
d.f.Soil prefStageSp (soil pref)Stage*Sp (soil pref)Stage*soil pref
11551
Leaf traits (determine for saplings and adults) 
LMA<0·001<0·001<0·0010·0030·06
Leaf thickness0·16<0·001<0·001<0·0010·09
Leaf density<0·001<0·001<0·0010·0010·72
N (%)<0·0010·87<0·001<0·0010·014
P (%)<0·0010·011<0·001<0·0010·75
Seed traits and wood-specific gravity (WSG)
Seed N (%)0·002 <0·001  
Seed P (%)0·95 0·10  
Seed mass<0·001 <0·001  
WSG0·001 0·001  

LMA, lamina thickness and lamina density tended to be higher for adults than for saplings, although the magnitude of change observed between the two ontogenetic stages differed among species (Fig. 1, Tables 2 and 3). However, the degree of ontogenetic plasticity was unrelated to soil-type preference, because there were no significant interactions between stage and habitat preference type. The only exception was lamina %N, which showed only marginally significant interaction of ontogenetic stage with soil preference. Lamina %N did not differ significantly between saplings and adults (Table 2), but N per unit area was higher for adults than saplings because of greater LMA in the former (Table S2). Lamina P was higher for adults in only some species (Table 2).

Seed mass also differed significantly among species, and residual specialists had greater seed mass than alluvial specialists (Tables 2, 3). Seed N was significantly higher for residual specialists than alluvial specialists, but seed P did not differ among species. The one unbiased species, I. leiocalycina, had values intermediate between the average values for the two specialist groups.

Discussion

Habitat Association

Our trail-based survey demonstrated that distribution of eight of the nine most common species of Inga exhibited significant edaphic biases, confirming the soil-type associations observed by experts (Zamora & Pennington 2001). Inga species studied in other neotropical forests also show non-random distributions in relation to slope (Harms et al. 2001), soil water content and pH (Endara & Jaramillo 2011), and soil nutrient availability and geographic distance (Dexter, Pennington & Cunningham 2010). Soil-type preference of each species was not absolute; some individuals were found on their non-preferred soil type, a pattern often observed in other studies of soil-type preferences (e.g. Clark, Clark & Read 1998; Russo et al. 2005; Malaysia, Queenborough et al. 2007; Ecuador). This is not surprising, as seeds can easily disperse to the other soil type. Stochasticity and additional dimensions in ecological niches, such as light availability, water availability and forest successional stage may promote the establishment of juveniles on non-preferred soil types (Baraloto et al. 2007). After initial establishment, ecological sorting continues throughout juvenile-to-adult transition, as found by Russo et al. (2005) in a Malaysian forest plot and Clark, Clark & Read (1998) in the same forest for non-Inga species. This study focused on functional traits associated with soil-type preference, and the study design would not allow us to examine temporal changes in species distribution patterns with ontogeny. Such data collected with appropriate spatial sampling would provide greater insights on the process of trait-based habitat filtering.

Owing to recent and rapid diversification, phylogenetic relationships within Inga are difficult to resolve (Dexter, Pennington & Cunningham 2010). Although this may have influenced the statistical power, according to the best available phylogeny within Inga, there was no obvious phylogenetic bias in soil-preference type among the nine Inga species in the study (Fig. S1). Perhaps, soil-type preference is an evolutionarily labile character as found by Russo et al. (2007) for a broad range of Malaysian tropical tree species. Alternatively, trait diversification because of recent niche sorting within a speciose genus can show weak phylogenetic signals, as found for chemical defence traits of 37 Inga species in Panama and Ecuador (Kursar et al. 2009), and for Quercus spp. that distribute along soil fertility gradients (Cavender-Bares, Kitajima & Bazzaz 2004).

Soil-type preference and functional traits

All traits of both adults and saplings, except for leaf lamina thickness, N per unit area, P in seed, showed significant difference in relation to the soil-type preference of the species (Tables 3 and S2). For the six species that exhibited a significant bias to one or the other soil type, a sufficient number of individuals could be sampled only in the preferred soil type. It is possible that plastic responses to soil nutrient availability may contribute to the observed differences among species, but results for three species sampled from both soil types suggest that phenotypic plastic responses to in situ soil types appeared small relative to inherent differences among species (see Table S1). The relative contribution of genetic basis and plasticity, however, differ among traits, and it is not possible to know whether adaptation or acclimation is responsible for observed variations when plants are sampled only in their typical habitats (Endara & Coley 2011). For example, Fine et al. (2006) found that habitat preference of species (to infertile white sand vs. richer clay soil), but not in situ soil type, affect leaf protein concentration, but in situ soil type affects leaf toughness.

The results supported our a priori expectation that LMA should be higher for infertile residual-soil specialists than alluvial-soil specialists, a result also found for 79 species in Australia (Wright & Westoby 2003) and other studies that compared resource-rich vs. resource-poor habitats (Reich, Ellsworth & Uhl 1995; Wright et al. 2004; Poorter et al. 2008). Change in LMA may be achieved either by increasing lamina thickness, lamina density (dry mass per volume) or both (Onoda et al. 2011; Westbrook et al. 2011). We found that only leaf density, but not lamina thickness, increased with specialization to less fertile soil. Parallel to this, across 19 Bolivian tree species, LMA and lamina density, but not lamina thickness, were associated with slow-growth strategies of shade-adapted species (Kitajima & Poorter 2010). Thus, lamina density, rather than lamina thickness, is the general correlate of slow-growth syndromes in resource-poor habitats, just as high WSG shows general association with slow growth of low-resource specialists across a broad range of species (Poorter et al. 2008, 2009; Chave et al. 2009). This is a broadly convergent pattern despite significant phylogenetic signals across broad phylogenetic ranges for both WSG and lamina density (Chave et al. 2009; Westbrook et al. 2011). The divergence in these traits within a single species-rich genus in relation to soil-type preference could indicate the significance of high tissue density in leaves and stems as part of an adaptive trait-syndrome to resource-poor environments, and such a conclusion would be more strongly supported with a better-resolved phylogeny.

Leaf N was significantly lower for residual-soil specialists than alluvial-soil specialists (adult leaf mean N = 3·29 and 3·79%, respectively, Table 2), although there was no significant difference for N per unit lamina area (Table S2), which suggests that leaf N reflects allocation strategies for optimal light and nitrogen use (Hikosaka & Hirose 2000). The observed %N values were high relative to other tropical trees in Fabaceace (mean %N = 2·5, Townsend et al. 2007). Similarly, leaf P concentrations (adult leaf mean = 0·13 and 0·19%, for residual and alluvial-soil specialists, respectively, Table 2) were high compared with other tropical species (mean = 0·09%, Townsend et al. 2007). Inga species at LS had high values of leaf N and P, possibly because even residual soil at LS is relatively fertile compared with many other tropical sites (Porder, Clark & Vitousek 2006). But, leaf N concentrations were high even in comparison with other legume species at LS (e.g. Pentaclethra macroloba, Porder, Clark & Vitousek 2006). We recognize that ecological niche dimensions other than soil fertility may also influence functional trait evolution. For example, of the five residual-soil specialists in our study, I. acuminata had the highest WSG (Table 2) and the lowest LMA (Fig. 1a). Inga acuminata is also the only species in the group classified as a subcanopy tree, and thus, other variables, such as light availability, may contribute to the observed differences in trait values.

Ontogenetic Changes in Leaf Traits

All traits examined in this study differed between saplings and adults, but species varied widely in the degree of ontogenetic change within, rather than between, soil-type preference groups (Table 2). These changes may reflect changes in height per se or changes in microenvironmental factors with height growth. Crown illumination index (CII) was higher for adults than for saplings, but hydraulic limitation associated with tree height might also be important in explaining height-associated variation in LMA (Thomas & Winner 2002; Cavaleri et al. 2010). Both components of LMA, lamina thickness and lamina density, increased from saplings to adults (Fig. 1), although such increases in lamina thickness may not be universal (Kitajima & Poorter 2010). Leaf N concentration did not differ between saplings and adults, but adults had greater N per unit area, which is an adaptive plasticity response to increased light availability experienced by adult leaves. CII was not different among species at the sapling stage; however, it was at the adult stage (data not shown). The species differences in degrees of CII change from saplings to adults may be responsible for species differences in degrees of ontogenetic changes in leaf traits.

Seed Traits

Particularly interesting among our results was seed N, which was higher for infertile residual-soil specialists. This was the opposite of the trend for leaves, and also the opposite of the trend that would be expected if seed N reserves merely reflect soil N availability. Grubb & Coomes (1997) found that tall trees in the infertile soils of caatinga forests have seeds with higher N compared with tree species occurring in adjacent more fertile soils. But our results differed from Grubb & Coomes (1997) for seed size; seed size tended to be greater for Inga spp. that prefer less fertile soil, suggesting that both higher nutrient concentrations and larger seeds may be selected in infertile soil, because the larger N capital is advantageous for seed reserves to support seedling nutrient demands (Kitajima 2002; Groom & Lamont 2010). While the advantage of seed size per se may be equivocal in adaptation to infertile soil (Leishman et al. 2000; Russo et al. 2007), higher total N in seeds may be favoured for in infertile soil as a strategy to prolong the duration of seed-reserve dependency necessary for initial seedling establishment (Kitajima 2002). Even though Inga seedlings form N-fixing nodules and the two soil types did not differ in total soil N, higher initial N availability from seeds may be adaptive for initial root development and establishment of mycorrhizal association. Following the same logic, we hypothesized that seed P should be higher for residual-soil specialists, but this expectation was not supported (Table 2). It may be that, regardless of soil fertility, the initial seed P concentration must be high enough to meet the seedling P demands until the time or ontogenetic stage when seedlings establish symbiotic associations with mycorrhizal fungi (Janos 1980).

Conclusions

Inga is one of the hyper-diverse tropical tree genera that have gone through rapid and recent diversification. Eight of the nine common Inga species in this lowland tropical wet forest exhibited clear preference to either rich alluvial soil or less fertile residual soil of volcanic origin. The patterns of leaf functional traits of adults and juveniles, as well as WSG, were consistent with the expected trait differences as a result of habitat filtering in resource-rich vs. less rich environments. Species that become successfully recruited as adults on less fertile soils have high LMA, high tissue densities of leaves and wood, and low N and P per unit mass in leaves, observations that are typically associated with slow growth and high survival as a life-history strategy. In addition, N concentrations in seeds were higher for residual specialists. This suggests that nutrient allocation strategy to seeds could reflect selection for different seedling establishment strategies in fertile vs. less fertile soils. We conclude that community assembly is likely to reflect trait-mediated niche sorting at multiple life stages.

Acknowledgements

We thank Orlando Vargas and Nelson Zamora for their assistance with species identification, Deborah Clark and David Clark for advice on experimental design, Julia Reiskind and Grace Crummer for the chemical analyses, the staff at La Selva Biological Station for logistic support and Steven Oberbauer, John Ewel, Martijn Slot and Jared Westbrook for constructive comments. DP was funded by fellowships from the South East Alliance for Graduate Education and the Professoriate and the Organization for Tropical Studies, and KN was funded by Research Experience for Undergraduates Program of the National Science Foundation of the United States.

Ancillary