High variation in foliage and leaf litter chemistry among 45 tree species of a neotropical rainforest community


  • Stephan Hättenschwiler,

    1. Centre d’Ecologie Fonctionnelle et Evolutive (CEFE), CNRS, 1919, route de Mende, F–34293 Montpellier cedex 5, France;
    Search for more papers by this author
  • Beat Aeschlimann,

    1. Centre d’Ecologie Fonctionnelle et Evolutive (CEFE), CNRS, 1919, route de Mende, F–34293 Montpellier cedex 5, France;
    2. Botanisches Institut, Universität Basel, Schönbeinstrasse 6, CH–4056 Basel, Switzerland;
    Search for more papers by this author
  • Marie-Madeleine Coûteaux,

    1. Centre d’Ecologie Fonctionnelle et Evolutive (CEFE), CNRS, 1919, route de Mende, F–34293 Montpellier cedex 5, France;
    Search for more papers by this author
  • Jacques Roy,

    1. Centre d’Ecologie Fonctionnelle et Evolutive (CEFE), CNRS, 1919, route de Mende, F–34293 Montpellier cedex 5, France;
    Search for more papers by this author
  • Damien Bonal

    1. UMR Ecologie des Forêts de Guyane, INRA Campus Agronomique, BP 709, F–97387 Kourou cedex, French Guiana
    Search for more papers by this author

Author for correspondence: Stephan Hättenschwiler Tel: +33 467 61 22 36 Fax: +33 467 41 21 38 Email: stephan.hattenschwiler@cefe.cnrs.fr


  • • Distinct ecosystem level carbon : nitrogen : phosphorus (C : N : P) stoichiometries in forest foliage have been suggested to reflect ecosystem-scale selection for physiological strategies in plant nutrient use. Here, this hypothesis was explored in a nutrient-poor lowland rainforest in French Guiana.
  • • Variation in C, N and P concentrations was evaluated in leaf litter and foliage from neighbour trees of 45 different species, and the litter concentrations of major C fractions were also measured.
  • • Litter C ranged from 45.3 to 52.4%, litter N varied threefold (0.68–2.01%), and litter P varied seven-fold (0.009–0.062%) among species. Compared with foliage, mean litter N and P concentrations decreased by 30% and 65%, respectively. Accordingly, the range in mass-based N : P shifted from 14 to 55 in foliage to 26 to 105 in litter. Resorption proficiencies indicated maximum P withdrawal in most species, but with a substantial increase in variation in litter P compared with foliage.
  • • These data suggest that constrained ecosystem-level C : N : P ratios do not preclude the evolution of highly diversified strategies of nutrient use and conservation among tropical rainforest tree species. The resulting large variation in litter quality will influence stoichiometric constraints within the decomposer food web, with potentially far-ranging consequences on nutrient dynamics and plant–soil feedbacks.


Plant leaves vary widely in morphological and physiological traits, despite their shared key functional purpose of photosynthetic carbon assimilation and transpiration (Aerts & Chapin, 2000; Reich & Oleksyn, 2004; Wright et al., 2004, 2005). At a global scale, ratios between leaf nitrogen (N) and phosphorus (P) concentrations, two key foliar traits, were reported to increase with decreasing latitude (McGroddy et al., 2004; Güsewell, 2004; Hedin, 2004; Reich & Oleksyn, 2004). Relatively high foliar N and low P concentrations in tropical ecosystems are in accord with the widespread belief that tropical forests are limited by P rather than N (Vitousek, 1984; Hedin, 2004; Paoli et al., 2005), and reflect the fact that many tropical forests grow on highly weathered and P impoverished soils (Irion, 1978; Uehara & Gillman, 1981; Vitousek & Sanford, 1986). The relationship between resource availability and the balance of chemical elements in primary producers, and its consequences for trophic interactions and biogeochemical cycles have intensively been studied in aquatic ecosystems (Sterner & Elser, 2002) and to a lesser extent in terrestrial ecosystems (Reiners, 1986; Elser et al., 2000a). Although carbon (C) : N : P stoichiometry in terrestrial plant leaves is usually more variable compared with the well-constrained C : N : P ratios in planktonic biomass (Redfield, 1958; Sterner & Elser, 2002), it has recently been suggested that terrestrial C : N : P ratios are equally well constrained at the level of ecosystems as are marine ratios (McGroddy et al., 2004). Based on their analyses from a literature survey, these authors further proposed ecosystem-scale selection for nutrient use strategies at the individual plant level (McGroddy et al., 2004).

However, leaf nutrients can vary substantially within an ecosystem at small spatial scales. In a temperate forest community in south-eastern Ontario, for example, Ricklefs & Matthew (1982) observed ranges in foliage N concentration from 1.35 to 2.95% among 34 broadleaf deciduous tree species. In a neotropical rainforest, N concentration of sunlit leaves of canopy trees have been reported to vary between 0.6% and 3.0% (Roggy et al., 1999; Bonal et al., 2000). Such large differences in green leaf quality among neighbour trees of the same community may imply similar interspecific variation in leaf litter chemistry, with important consequences for ecosystem processes such as decomposition and nutrient cycling (Hättenschwiler et al., 2005; Wardle et al., 2006). Reports on simultaneous measurements of green leaf and litter chemistry from species-rich plant communities are rare, particularly for tropical forests, but are strongly needed for a closer evaluation of plant nutrient use strategies, potentially underlying evolutionary processes, and plant-soil feedbacks on nutrient cycling.

Here we analysed green leaf and litter quality from 45 different tropical canopy tree species from several harvests, with a focus on C, N and P concentrations, but also measuring water-soluble compounds, nonstructural carbohydrates, cellulose, lignin and phenolics composition of leaf litter. We addressed the questions of how variable leaf chemistry is among co-occurring tree species from a lowland rainforest community in French Guiana, and how litter chemistry relates to foliage chemistry. The relationship between green leaf and litter quality is complex because leaf senescence is a physiological process involving nutrient resorption (Killingbeck, 1996; Aerts, 1996) that can vary among species (Aerts, 1996; Kobe et al., 2005). Interspecific variability in litter quality may thus diverge, converge or may remain the same compared with foliage quality. Withdrawal of nutrients before leaf death and abscission can be of key importance for the plant nutrient budget (Aerts, 1996), especially in nutrient-poor ecosystems (Vitousek, 1982) such as the tropical forest studied here. Under the assumption that general physiological constraints, and the biochemical and/or biophysical minimum threshold for nutrient resorption are similar among species (Killingbeck, 1996), we would predict that interspecific differences in foliage nutrient concentrations should converge in litter. Consequently, and in line with the hypothesis of ecosystem-scale selection for plant nutrient use strategies, we might expect narrowing variance in litter C : N : P ratios compared with foliage as a mechanism to maximize nutrient retention across species.

Materials and Methods

Study site

Green leaves and leaf litter were collected at the Guyaflux experimental site located near Sinnamary, French Guiana (5°18′N, 52°55′W). This site was set up in 2003 to study water and carbon balance of a neotropical forest. Mean annual temperatures in 2004, 2005 and 2006 were 25.7°C, 25.9°C and 25.5°C, and total annual rainfall was 2756 mm, 3072 mm, and 3510 mm, respectively. Temperature varies only slightly during the course of the year, but relatively large intra-annual variations in rainfall are observed. This results in two distinct rainy seasons, a moderate one from December to February and a stronger one from April to July. Soils in the study area are acrisol (FAO-ISRIC-ISSS 1998) developed over a Precambrian metamorphic formation called the Bonidoro series. The soil is nutrient-poor with 24% clay, 7% silt and 69% sand, and a pH (water extract) of 4.7 in the top 0.2 m (V. Freycon & L. Soucémarianadin, pers. comm.). Soil C : N is 15.3 with a total N of 1.12 g kg−1 soil, total P of 0.3 g kg−1 soil and plant available P (Olson P) of 0.0045 g kg−1 soil (all top 0.2 m of soil).

Plant material

Our analyses include three different complementary samplings of foliage and/or leaf litter. First, we sampled leaf litter from three individuals of nine different species to evaluate local intraspecific variation in litter chemistry. Second, we sampled leaf litter from entire stands of four species to compare population level litter chemistry with that at the individual tree level. Third, our main sampling effort concentrated on 45 different tree species, all within a 0.98 ha experimental forest plot, to evaluate interspecific variation in green leaf and litter chemistry.

Intraspecific variation in litter chemistry  From nine relatively abundant species, that were also included in the main sampling (Table 1), we selected a total of three fully sunlit canopy trees, each from a different 0.98 ha experimental forest plot in close vicinity of each other at the Guyaflux site. Litter traps (1 × 2 m) were installed 1.5 m above the ground underneath each tree, and fresh fallen leaf litter was collected several times per month in February, May, and September 2006 in order to obtain a representative sample for each month. Litter traps were emptied every 3–4 d to prevent leaching and initial decomposition. Only fresh fallen litter was selected, while leaves with obvious signs of herbivory, galls, fungal attacks, leaves that were clearly very thin (shaded canopy position), leaves that were still green or greenish, or having an otherwise atypical coloration were excluded. Excluded leaves typically represented > 15% of total collected leaf litter per species. All litter collected in the traps was sorted and only litter from the target species was kept and dried at 65°C to constant weight and kept dry until analyses.

Table 1.  List of species with their taxonomic grouping
  1. Names in bold type indicate species that have been replicated across different forest plots (see Fig. 1) and POP at the beginning of a species name designates species for which litter was additionally sampled at the population level in a forest plantation.

EricalesLecythidaceaeEschweilera coriacea (A.L. De Candolle) S.A. Mori
 Lecythis persistens Sagot
 Lecythis poiteaui O.C. Berg
 Lecythis spp.
SapotaceaeChrysophyllum pomiferum (Eyma) Pennington
 Chrysophyllum prieurii A.L. De Candolle
 Chrysophyllum sanguinolentum (Pierre) Baehni
 Manilkara bidentata (A.L. De Candolle) A.J. Chevalier
 Manilkara huberi (Ducke) A.J. Chevalier
 Micropholis obscura Pennington
 Pouteria gonggrijpii Eyma
 Pouteria guianensis J.B. Aublet
 Pouteria singularis Pennington
 Pouteria spp.
 Pradosia cochlearia (Lecomte) Pennington
FabalesCaesalpiniaceaePOP Dicorynia guianensis G.J. Amshoff
 POP Eperua falcata J.B. Aublet
 Tachigali paraensis (Huber) Barneby
 Vouacapoua americana J.B. Aublet
MimosaceaeInga alba (O.P. Swartz) C.L. Willdenow
 Inga jenmanii Sandwith
PapilionaceaeDiplotropis purpurea (L.C. Richard) G.J. Amshoff
GentianalesApocynaceaeCouma guianensis J.B. Aublet
RubiaceaeDuroia genipoides J.D. Hooker ex K. Schumann
LamialesBignoniaceaeJacaranda copaia (J.B. Aublet) D. Don
LauralesLauraceaeSextonia rubra (Mez) van der Werff
MagnolialesMyristicaceaeIryanthera sagotiana (Bentham) Warburg
MalpighialesCaryocaraceaePOP Caryocar glabrum (J.B. Aublet) Persoon
ChrysobalanaceaeLicania alba (Bernoulli) Cuatrecasas
 Licania densiflora Kleinhoonte
ClusiaceaePOP Platonia insignis Martius
 Symphonia globulifera Linnaeus f.
 Tovomita ssp.
EuphorbiaceaeGlycydendron amazonicum Ducke
HumiriaceaeHumiriastrum subcrenatum (Bentham) Cuatrecasas
MalvalesBombacaceaeEriotheca longitubulosa A. Robyns
SterculiaceaeSterculia pruriens K. Schumann
MyrtalesCombretaceaeBuchenavia grandis Ducke
OxalidalesElaeocarpaceaeSloanea ssp.
RosalesCecropiaceaeCecropia obtusa Trécul
 Coussapoa angustifolia J.B. Aublet
 Pourouma guianensis J.B. Aublet
SapindalesBurseraceaeProtium opacum Swart
 Protium subserratum (Engler) Engler

Population vs individual tree litter chemistry  From four species, which were also represented in the main sampling (Table 1), we collected fresh fallen leaf litter within a tree plantation located 10 km from the Guyaflux study site. Litter was collected regularly from June 2004 to February 2005 and pooled across sampling dates. Seedlings of local seed sources from 19 tree species (four of which occurred in our studied natural forest at the Guyaflux study site) have been planted as monocultures by CIRAD in 1983 and 1984. The natural forest was cleared and the soil ploughed before the planting of 49 individuals within 20 × 20 m plots per species. The somewhat more than 20-yr-old tree stands have a fully closed canopy with a leaf area index (LAI) of c. 5 (Roy et al., 2005).

Interspecific variation in foliage and litter chemistry  Green leaves and leaf litter were collected from the same individual of fully sunlit canopy trees of each of 45 species within one of the three 0.98-ha experimental plots at the Guyaflux site mentioned earlier. The 45 species belong to 21 different families and 12 orders (Table 1). Seven species belong to the Leguminosae, but only four of them are considered to be N fixing (Diplotropis purpurea, Inga alba, Inga jenmanii and Tachigali paraense; Roggy et al., 1999). Litter was collected several times per month in January, May, July and November 2004 in the same way as described earlier. Fully sunlit mature green leaves, with no signs of herbivory or disease, were harvested from the top canopy in November 2003 and November 2004 using a shotgun. Litter and green leaves were dried at 65°C to constant weight and kept dry until analyses.

Leaf chemistry

All samples were ground using a centrifugal mill (Cyclotec Sample Mill; Tecator, Höganäs, Sweden) to obtain a uniform particle size of > 1 mm. The petioles were cut off before grinding to avoid differences in species-specific contribution of this predominantly woody leaf part.

Carbon, N, P, water soluble compounds (WSC), hemicelluloses, cellulose, lignin, sugar, starch and polyphenols were analysed using standard methods for each component. Carbon and N concentrations were measured using a CN elemental analyser (Flash EA1112 Series; ThermoFinnigan, Milan, Italy). For P measurements, 2 ml of 36 n H2SO4 and 3 ml of H2O2 were added to a 20 mg sample of leaf material and heated at 360°C for 4 h. After this mineralization step, P concentration was measured colorimetrically with an autoanalyser (Evolution II; Alliance Instruments, Frépillon, France) using the molybdenum blue method (Grimshaw et al., 1989).

Sugar and starch were determined from a 10 mg sample that was mixed with 3 ml of water and heated in a pressure cooker for 30 min. An aliquot of the dissolved sample was digested by invertase to transform sucrose into fructose and glucose, followed by complete conversion into glucose using isomerase. An aliquot of the initial sample was treated with clarase to break up starch into glucose. Final glucose concentration was measured colometrically at a wavelength of 340 nm by adding hexokinase and dehydrogenase (Hoch et al., 2002).

Water-soluble compounds, hemicelluloses, cellulose and lignin were determined according to the van Soest extraction protocol (van Soest & Wine, 1967) using a fibre analyser (Fibersac 24; Ankom, Macedon, NJ, USA). Soluble polyphenols were measured from 1 g of the sample to which 60 ml of water was added and shaken for 2 h. After filtering, the filtrate was diluted and polyphenol concentration was measured colorimetrically with Folin–Ciocalteau reagent following the method of Marigo (1973) using gallic acid as a standard. Total polyphenol concentration was measured using methanol (50%) instead of water as solvent followed by the same procedure already described.

Data analysis

Litter quality for the subset of nine species for which three individuals have been sampled, was analysed with analyses of variance to test for effects of species identity (df = 8), date of harvest (df = 2) and their interaction (df = 16) with a replication of three individual trees. Individual-based and population-based data (from the tree plantation with 31–46 individuals per species) were compared using one-factorial analyses of variance with location (natural forest vs plantation) as factor. Green leaf and leaf litter quality parameters from the main sampling were analysed with analyses of variance to test for effects of species identity (df = 44) and date of harvest (df = 3 for litter chemistry and df = 1 for foliage chemistry). Variation among the four sampling dates was not significant for any of the variables tested, indicating that there was no consistent seasonal variation in litter quality parameters at the community level. However, most species varied among sampling dates. Because of our limited sampling of one individual per species, the species effect must be interpreted with caution. In our analysis the variability within species actually represents the variability within individuals representing a given species, with the four different sampling dates as the replicates. So, strictly speaking, the reported differences among species are differences among individuals, which, however, have the same implications for processes at the ecosystem level in this highly diverse community of mostly one individual per species and per hectare. Moreover, the nine species for which litter chemistry was compared among different individuals, and the comparison with population level data for four species, provide strong evidence for well-conserved species-specific leaf chemistry at this small local scale.

Relationships between different litter quality parameters, and relationships between litter quality and green leaf quality, were tested with simple linear regression analyses. All data expressed in percent of leaf dry mass were arcsin square-root transformed to improve the homogeneity of variance and normality before analyses. Analyses were performed with General Linear Model (GLM) procedure using SYSTAT, version 5.2.1 (Systat, Inc., Evanston, IL, USA).


Variation in litter chemistry within species

The nine species sampled in three different experimental forest plots differed significantly in their litter C concentration (P < 0.001, data not shown). Litter C was not different among sampling dates (P = 0.35); neither were species differently affected by date of sampling (P = 0.68 for the species × date interaction). The nine species varied significantly in their litter N (P < 0.001, data not shown) and P concentrations (P < 0.001, Fig. 1). Litter N (P = 0.02) and P (P < 0.001) also differed among dates of sampling. Dates of sampling affected litter P distinctly among species (P < 0.01 for the species × date interaction), but not litter N (P = 0.17 for the species × date interaction). Differences among the three individuals were not significant for any of the three litter parameters tested in any of the nine species (Fig. 1).

Figure 1.

Mean litter phosphorus (P) concentrations of each of three individuals (Individual 1, open bars; Individual 2, tinted bars; Individual 3, closed bars) from nine species (mean ± SD, n = 3 sampling dates) of Amazonian rainforest trees. Differences among individuals were not significant in any of the nine species (P > 0.15, sampling dates as replicates). Different letters indicate significant differences among species (Tukey post hoc contrasts).

Individual-based data from the natural forest were compared with population-based data from a nearby plantation where some of the species included in our study were present, to test whether or not the individual-based measurements are representative for a local population. Individual-based values from the natural forest matched those at the population level of planted trees (Fig. 2). Also, species-specific rankings were maintained between the two groups of data for all compounds. The differences in absolute values between individuals from the natural forest and those from the plantation were significant only for N (P = 0.03) and WSC of C fraction analyses (P = 0.01). All other parameters measured were not statistically different between natural forest trees and plantation trees.

Figure 2.

Litter concentrations of (a) carbon (C), (b) nitrogen (N), and (c) phosphorus (P) in individual trees from a diverse community as a function of litter C, N and P at the population-level of the same four species measured in a nearby plantation. Tree populations were represented by 31, 38, 43 and 46 individuals of Caryocar, Dicorynia, Eperua and Platonia, respectively. The broken line indicates the 1 : 1 line along which elemental concentrations in individual tree leaf litter and leaf litter at the population level of the respective species are equal. Full lines and r2 values are given for simple linear regressions.

Interspecific variation in litter chemistry

Based on the data mentioned earlier, which suggested little intraspecific variation at a small spatial scale, the single individual approach to assess the range of local variation in foliage and litter chemistry among 45 neighbour tree species, was considered as rather robust. Carbon concentration in fresh fallen leaf litter varied among species individuals (P < 0.001) between 45.3 and 52.4% (Fig. 3), with a mean (± SD) of 49.2 ± 1.9% (Table 2). The Licania alba tree showed the lowest C concentrations, and Pouteria singularis and Lecythis poiteaui trees were at the higher end of the range. Litter N concentration ranged between 0.68% and 2.01% (Fig. 3, significant species effect, P < 0.001), with an average (± SD) of 1.14 ± 0.29% (Table 2). Individuals of the genus Manilkara (M. bidentata and M. huberi) had the lowest and individuals of Inga (I. alba and I. jenmanii) the highest N concentrations. From all litter quality traits measured, P concentration showed the largest differences among species (Fig. 3, P < 0.001) with values ranging from 0.009 to 0.062% (mean ± SD of 0.024 ± 0.011%, Table 2). Trees of Humiriastrum subcrenatum and T. paraense had the lowest P concentrations, and Tovomita ssp. and I. alba trees were at the high end of the P gradient.

Figure 3.

Litter phosphorus (P, circles) and nitrogen (N, squares) concentrations as a function of litter carbon concentration from 45 co-occurring Amazonian rainforest trees. Each symbol represents the mean value of a single species (see Table 1 for species identity). The double-arrow lines indicate the range in P (the highest value differs by a factor of 7 from the lowest value) and N (the highest value differs by a factor of 3 from the lowest value). Note that P concentrations are multiplied by a factor of 10 for better presentation.

Table 2.  Green leaf and leaf litter concentrations of carbon, nitrogen, phosphorus, water-soluble compounds (WSC), hemicelluloses, cellulose, lignin and polyphenols across all 45 species with maximum and minimum values
Leaf parameterGreen leavesLeaf litter
Mean ± SDMaximumMinimumMean ± SDMaximumMinimum
  • Values are means (± SD).

  • *

    Because of limited amounts of material polyphenols have only been measured for a total of 12 species for green foliage and for a total of 17 species for leaf litter. Comparisons between green foliage vs leaf litter are thus approximate and should be interpreted carefully.

Carbon (% DM)48.8 ± 1.652.544.549.2 ± 1.952.445.3
Nitrogen (% DM)1.6 ± 0.42.6 1.01.1 ± 0.32.0 0.7
Phosphorus (% DM)0.07 ± 0.020.12 0.040.02 ± 0.010.06 0.01
WSC (% DM)33 ± 846 831 ± 84812
Hemicelluloses (% DM)16 ± 745 414 ± 639 4
Cellulose (% DM)20 ± 4271319 ± 32610
Lignin (% DM)31 ± 11611337 ± 95719
Soluble polyphenols* (% DM)4.0 ± 2.78.6 0.72.9 ± 2.48.4 0.5
Total polyphenols* (% DM)8.4 ± 4.921.5 2.69.5 ± 4.917.7 2.2

While N resorption proficiency (the lowest measured concentration in litter, cf. Killingbeck, 1996) of only four trees (Buchenavia grandis, Coussapoa angustifolia, M. bidentata and M. huberi) was below 0.7%, and thus, according to Killingbeck (1996), can be considered as complete, 42 species individuals showed P concentrations below 0.04%, as the supposed threshold value for maximum P resorption.

Water-soluble compounds (WSC) were significantly different among species (P < 0.001), and ranged from 12% to 48% of total litter dry mass with a mean of 30 ± 8% (Table 2). The WSC correlated positively with total C concentration (r2 = 0.18, P < 0.01, Fig. 4). A small part of the water-soluble fraction consisted of nonstructural carbohydrates. Twenty-three out of the 45 study species had detectable amounts of sugar in their leaf litter. Sugar concentration varied between 0.1% and 2.1% with an average of 0.7 ± 0.5% (data not shown). Only seven species (Chrysophyllum prieurii, Eschweilera coriacea, Lecythis ssp., Lecythis persistens, Platonia insignis, Pradosia cochlearia, Symphonia globulifera) contained > 1% sugar in their leaf litter. In 13 species there was also some starch left in leaf litter, but the concentrations were very low, with an average of only 0.3 ± 0.2%, (data not shown).

Figure 4.

Concentrations of water-soluble compounds (diamonds) and lignin (triangles) as a function of litter carbon (C) concentration in 45 rainforest tree species (a), and average concentrations of different carbon compounds across all study species (mean ± SD) determined by the van Soest method (b). Lines and corresponding r2 and P values are shown for simple linear regressions.

van Soest lignin showed a high variation among species (P < 0.001), with concentrations between 19 and 57%, and a mean (± SD) of 37 ± 9% (Table 2). Surprisingly, in contrast to WSC, lignin concentration showed no correlation with total C concentration (P = 0.88, Fig. 4). Cellulose and hemicellulose concentrations were generally lower than WSC and lignin, but together accounted roughly for one-third of total dry mass (Fig. 4). Concentrations of hemicellulose and cellulose were highly variable among species (P < 0.001). Neither the cellulose nor the hemicellulose fraction correlated with total litter C. Because of limited amounts of litter material polyphenols in litter from only 17 species were analysed. Soluble and total polyphenols in this limited set of species varied substantially (Table 2).

Relationship between foliage and leaf litter chemistry

A comparison between mean values of the different sampling dates of green leaves and leaf litter of the same tree individuals revealed a relatively tight correlation for C concentration across all species (r2 = 0.745, P < 0.001, Fig. 5) with no significant difference between green and senescent leaves (P = 0.47). Similarly, N concentration in green leaves and leaf litter correlated well (r2 = 0.554, P < 0.001, Fig. 5), but there was significantly less N in litter than in green leaves (P < 0.0001). Averaged across all species, N concentration was 30% lower in litter than in green leaves. By contrast, P concentration correlated only weakly between green leaves and leaf litter (r2 = 0.084, P = 0.047, Fig. 5), but concentrations in litter were substantially lower than in green leaves (on average 65% less in litter than in green leaves, P < 0.0001). Concentrations of WSC (P = 0.33) and cellulose (P = 0.57) were not significantly different between green leaves and leaf litter. However, green leaves had higher hemicellulose concentrations (P = 0.039), and lower lignin concentrations (P = 0.012).

Figure 5.

Concentrations of (a) carbon (C), (b) nitrogen (N), and (c) phosphorus (P) in leaf litter from 45 rainforest tree species as a function of C, N and P concentrations in green foliage of the same tree individuals. The broken line indicates the 1 : 1 line along which elemental concentrations in leaf litter and foliage are identical. Full lines and corresponding r2 values are shown for simple linear regressions.

Variation among species remained largely the same in litter compared with green leaves for most of the parameters measured (Fig. 6). Based on our initial hypothesis smaller variation in litter nutrients compared with green leaf nutrients was expected. However, our data showed similar variation for litter N and green leaf N, and much higher variation for litter P compared with green leaf P (Fig. 6). The substantially higher variation in P concentration in litter than in green leaves with a doubling of the coefficient of variation (CV) was particularly surprising. It is important to note that the higher CV for litter P is not just driven by some extreme values. When the three lowest and the three highest values of measured P concentrations are removed, the litter CV of 36% is still substantially higher than that calculated for foliage (19%).

Figure 6.

Comparison of coefficients of variation (CV) for the different leaf quality parameters between green leaves (open bars) and leaf litter (closed bars) across all 45 tree species (17 and 12 species for polyphenols in leaf litter and green foliage, respectively). WSC, water-soluble compounds.

Foliage and litter stoichiometry

Green leaf C : N ratios (mass basis) ranged from 18.1 to 51.7 with a mean (± SD) of 32.1 ± 7.8 (Fig. 7). Green leaf C : P ratios ranged from 406 to 1372 (mean of 767 ± 219) and green leaf N : P ratios varied between 14.3 and 54.8 (mean of 24.5 ± 6.6, Fig. 7) across the 45 species studied. According to the proposed leaf N : P ratio < 14, indicative of N limitation and that of > 16, indicative of P limitation (Koerselman & Meuleman, 1996; Aerts & Chapin, 2000), these results suggest P rather than N limitation for tree growth at our study site. Compared with green leaves, stoichiometric relationships in litter were substantially wider and varied more for C : P and N : P ratios (Fig. 7). Mass-based litter C : N ratios ranged between 24.8 and 76.7 (mean ± SD of 45.7 ± 11.7). Litter C : P ratios ranged from 801 to 5367 (mean of 2409 ± 1126) and litter N : P ratios from 26.5 to 105 (mean of 52.3 ± 19.1).

Figure 7.

Leaf litter stoichiometry from 45 rainforest tree species as a function of stoichiometry in green foliage of the same tree individuals. The broken line indicates the 1 : 1 line along which elemental ratios in leaf litter and foliage are identical. The double-arrow lines indicate the range in elemental ratios (vertical for litter and horizontal for green leaves) along with the mean value and the coefficient of variance (CV).


Variation in leaf chemistry within a tropical forest community

Substantial variation in foliage and litter chemistry was observed among 45 co-occurring neotropical rainforest tree species. Since all trees were sampled within a well-defined, homogeneous 0.98 ha area, environmental conditions such as climate and general soil characteristics can largely be ruled out as driving factors for the observed variation in leaf quality. Rather, the reported differences appear to reflect inherent species-specific characteristics.

Species identity-driven variation in leaf traits is supported by the small intraspecific variation observed in the subset of nine species from different experimental forest plots (Fig. 1), and the well conserved species-specific differences at the population level of four of the studied species (Fig. 2). This is in agreement with a study by Ricklefs & Matthew (1982) on temperate deciduous trees, where they concluded that the small variation among individuals within species can be ignored for among-species comparisons within a site. Variation within species is arguably less important than variation among species in highly species rich communities such as most tropical forests, with typically between 100 and 200 different tree species per hectare, and often only one individual canopy tree for the majority of species (Ter Steege et al., 2000). In forest ecosystems dominated by only one or a few tree species, however, intraspecific variation in litter traits can be significant for ecosystem-level processes (Madritch & Hunter, 2002; Schweitzer et al., 2004).

Functional groups of plant species sharing similar leaf traits, such as N-fixing legumes or conifers, can account for a large amount of variation in leaf chemistry (Aerts, 1996; Cornelissen et al., 1997; Perez-Harguindeguy et al., 2000; Quested et al., 2003). In our study, however, the majority of species examined belong to the rather well-defined functional group of evergreen trees with long-lived, leathery leaves. Although some of the highest N and P concentrations were measured for species from the Fabales, only four (I. alba, I. jenmanii, D. purpurea and T. paraense) out of the seven Fabales species, are actually believed to fix N (Roggy et al., 1999), and two of them (Diplotropis purpurea and Tachigali paraense) are not among the 10 N and P richest species. This suggests that N fixation contributes little to a functional explanation of the reported range in leaf chemistry in the forest ecosystem studied. The local variation in litter quality observed here is similar to that reported across different plant functional groups and along a wide climatic gradient in central Argentina (Perez-Harguindeguy et al., 2000). For example, litter C : N ratio that correlated well with decomposition in the central Argentina study, varied by a factor of 3.2 (between 16 and 51) among the functional groups of ‘woody deciduous’, ‘woody evergreen’, ‘aphyllous’, ‘bromeliads’, ‘succulents’, ‘graminoids’ and ‘herbaceous dicots’, including a total of 52 plant species (Perez-Harguindeguy et al., 2000). This compares to the difference in C : N ratio by a factor of 3.1 (between 25 and 77) in our study. A somewhat lower range in C : N ratio between 25 and 66, varying by a factor of 2.6, was reported for 22 mostly deciduous tree species from a relatively infertile dry tropical forest (Lal et al., 2001).

In conclusion, the widely different leaf traits reported here for a rainforest tree community dominated by broadleaf evergreen tree species does not readily fit the conventional concepts of plant functional groups and their segregation within a leaf trait matrix. A finer grained resolution of ‘functional groups’ based, for example, on a more detailed assessment of leaf life-spans within the group of ‘evergreens’, could perhaps contribute to a functional interpretation of the observed variation in foliage and litter chemistry.

C : N : P stoichiometry and plant nutrient use strategies

Mean foliage C : N, C : P, and N : P ratios of 37.5, 1982 and 54.3 (on a molar basis) reported here, compare well with the average molar ratios of 35.5, 2457 and 43.4 across several tropical forest studies summarized in McGroddy et al. (2004). Although we did not account for the relative abundance of species in the calculation of these data, they are approximately representative for the community studied, characterized by a low abundance of most species. Mean foliage stoichiometry is in accord with previous findings of relatively constant C : N ratios, and distinctly higher C : P and N : P ratios in tropical forests compared with other forest ecosystems at higher latitudes (Güsewell, 2004; McGroddy et al., 2004; Reich & Oleksyn, 2004). These broad global patterns across biomes indicate large-scale environmental constraints (e.g. P limitation in tropical ecosystems vs N limitation in temperate ecosystems; McGroddy et al., 2004; Reich & Oleksyn, 2004) and/or temperature related physiological processes leading to broadly different growth rates, and thus, to different N and P demands across climatic gradients (Elser et al., 2000b; Kerkhoff et al., 2005). However, such large-scale considerations might be too coarse-grained to understand evolutionary processes in nutrient use strategies. The large variance in foliage stoichiometry among canopy trees in a rather homogeneous environment at a small local scale reported here, suggest different adaptive responses to selection pressures related to nutrient limitation, and does not support the view of well-constrained terrestrial C : N : P ratios of autotrophs as was suggested by McGroddy et al. (2004). In line with our findings, Townsend et al. (2007) recently reported a strikingly large interspecific variation in green leaf N : P ratios at different tropical forest sites, and concluded that this diversity needs to be accounted for in order to understand ecosystem processes in species rich tropical forests.

Compared with green leaves, C : N : P ratios widened considerably in litter from the same individuals because of nutrient resorption, and C : P and N : P in litter showed a larger variance than in foliage (Fig. 7). Nitrogen and P concentrations were on average 30% and 65% lower in litter than in green leaves. This corresponds to an estimated mean N resorption efficiency of 40 ± 13% and a mean P resorption efficiency of 70 ± 13% across our studied species (calculated on a unit-lignin basis, data not shown). In agreement with our study, higher mean P than N withdrawal in 73 species from eastern Australian ecosystems (Wright & Westoby, 2003), and in the dominant tree species Metrosideros polymorpha in Hawaiian tropical montane forest ecosystems (Vitousek, 1998), was associated with P-poor soils suggesting higher resorption of the more limiting nutrient. Similar to the Australian study (Wright & Westoby, 2003), we reported lower litter P concentrations than most values previously published (Killingbeck, 1996), and lower than what has been observed in other tropical rainforests (Proctor, 1984; Scott et al., 1992; Chuyong et al., 2000).

The large variation in litter C : P and N : P ratios among species was driven by the considerable increase in variation of litter P concentrations (Fig. 6). The much higher variability in litter P than in green leaf P concentration contrasts with the initial hypotheses of converging interspecific differences in litter compared with foliage nutrient concentrations, and concomitantly decreasing variances in litter C : N : P compared with foliage. Phosphorus withdrawal during senescence seems to vary among species independently of their foliage P concentration (Fig. 5). If a constant proportion of foliage P was recycled from senescing leaves, as could be expected when the relative amount of mobile P is a function of total leaf P, we would expect a close correlation between foliage and litter P. Conversely, if the absolute amount of immobile P within a leaf was determined by species-independent general biochemical and/or biophysical constraints, we would expect a higher proportion of total P being recycled from comparatively high P leaves. According to our hypothesis, this would then result in converging litter P concentrations towards possible minimum values in all species and maximum P conservation under the supposedly P limiting conditions of the studied forest. Why did our data not support either one of these two predictions?

Several nonexclusive explanations might exist which would need further testing. The degree of P limitation might vary among species, which then could exhibit distinct P use strategies, including P resorption efficiencies. For example, a fertilization experiment in a secondary tropical forest indicated species-specific limitations by both N and P (Davidson et al., 2004). A common minimum P value across species might actually not exist because of multiple constraints in biochemical and biophysical processes during leaf ageing, nutrient transformation, phloem loading, and nutrient retranslocation, processes that are, at present, not well understood. Independently of any physiological processes during senescence, the amount of remaining P in fresh fallen litter may also be determined by leaf structural and morphological characteristics (e.g. thickness of cuticula). These might more or less protect P compounds from leaching during leaf senescence, thus masking potential patterns of P resorption across species. Possible differences in leaching losses from senescing leaves among species have not been assessed here; however, they are commonly not considered as a major nutrient flux (Aerts, 1996; Killingbeck, 1996).

In conclusion, the tremendous variation in nutrient proficiencies and leaf stoichiometries among neighbour trees within a single site, suggests a wide range of adaptive strategies among tree species at the level of green leaf functioning, plant nutrient acquisition, and nutrient resorption physiology, to apparent common environmental constraints. This is in contradiction to simple generalizations on plant nutrient economies based upon broad functional groups, and indicates that ecosystem-scale selection is of minor relevance for the evolution of plant nutrient use strategies.

Implications for ecosystem nutrient dynamics

The observed high variation in litter chemistry and stoichiometry among trees and the low abundance of individual species within the canopy implies a qualitatively highly heterogeneous litter input to the soil at small spatial scales. The wide range of C : N : P stoichiometry in litter, but also differences in C quality, will affect higher trophic levels of the decomposer community with varying constraints depending on the local species composition of the litter. This might have important consequences for nutrient dynamics during litter decomposition and subsequent nutrient availability, and could, thus, contribute to the larger than expected range of plant stoichiometry. Spatially separated resources can affect competitive interactions between microbes and plants (Jingguo & Bakken, 1997) and microsite dynamics in soils are fundamentally important for the understanding of whole soil processes (Jackson et al., 1989; Ettema & Wardle, 2002; Schimel & Bennett, 2004). Recent advances in microsite-based mineralization theory were mainly achieved on nitrogen dynamics in primarily N-limited northern ecosystems (Schimel & Bennett, 2004), whereas the role of functionally distinct microsites for P cycling in P-limited ecosystems is not well understood. Recent studies showed that experimental P additions to P-limited ecosystems can increase decomposition rates (Hobbie & Vitousek, 2000), as well as the mineralization of dissolved organic matter derived from litter (Cleveland et al., 2002, 2006). Whether the wide variation in litter P concentration among species observed in our study translates into spatially heterogeneous soil P availability with feedback effects on trees and their leaf nutrient concentrations remains to be tested in further studies.


We thank Jean-Yves Goret and Chris Baraloto for their invaluable help in tree species determination of litter material, Bruno Buatois and Laurette Sonié for laboratory support, Audin Patient for help in litter collections, and Anne-Marie and Vincent Domenach for their hospitality during several stays in Kourou. The comments of three anonymous reviewers improved a previous version of the manuscript. This research was funded through an ATIP research grant from CNRS (SDV) to S.H.