SEARCH

SEARCH BY CITATION

Keywords:

  • Amazon;
  • community assembly;
  • leaf chemistry;
  • leaf traits;
  • Peru;
  • remote sensing;
  • spectranomics;
  • tropical forest

Summary

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information
  • Canopy chemistry and spectroscopy offer insight into community assembly and ecosystem processes in high-diversity tropical forests, but phylogenetic and environmental factors controlling chemical traits underpinning spectral signatures remain poorly understood.
  • We measured 21 leaf chemical traits and spectroscopic signatures of 594 canopy individuals on high-fertility Inceptisols and low-fertility Ultisols in a lowland Amazonian forest. The spectranomics approach, which explicitly connects phylogenetic, chemical and spectral patterns in tropical canopies, provided the basis for analysis.
  • Intracrown and intraspecific variation in chemical traits varied from 1.4 to 36.7% (median 9.3%), depending upon the chemical constituent. Principal components analysis showed that 14 orthogonal combinations were required to explain 95% of the variation among 21 traits, indicating the high dimensionality of canopy chemical signatures among taxa. Inceptisols and lianas were associated with high leaf nutrient concentrations and low concentrations of defense compounds. Independent of soils or plant habit, an average 70% (maximum 89%) of chemical trait variation was explained by taxonomy. At least 10 traits were quantitatively linked to remotely sensed signatures, which provided highly accurate species classification.
  • The results suggest that taxa found on fertile soils carry chemical portfolios with a deep evolutionary history, whereas taxa found on low-fertility soils have undergone trait evolution at the species level. Spectranomics provides a new connection between remote sensing and community assembly theory in high-diversity tropical canopies.

Introduction

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

High species diversity in tropical forest canopies results from environmental filtering, biotic interactions, and neutral processes, played out over timescales of evolution and biogeographic migration. However, the functional variation among tropical forest canopies is often expressed in plant traits (Wright et al., 2006), which may or may not track patterns in species composition and diversity (Kraft et al., 2009). Among these traits, leaf chemicals regulate light capture, growth, respiration, longevity and defense. Leaf chemicals are also key mediators of ecosystem processes such as decomposition and nutrient cycling (Vitousek, 1984). But the relationship between phylogeny and leaf chemistry is only starting to be explored in humid tropical forests (Fyllas et al., 2009), and the relevance of high canopy diversity to canopy chemistry or ecosystem processes remains unclear (Townsend et al., 2008).

Leaf chemical variation occurs among many important elements and molecular compounds. Nitrogen (N), phosphorus (P), base cations (calcium (Ca), potassium (K), and magnesium (Mg)), and micronutrients (manganese (Mn), zinc (Zn), boron (B), and iron (Fe)) vary in concentration based on investments in processes ranging from CO2 fixation to protection against toxic metals (Johnson & Todd, 1983; Aber & Melillo, 1991). Polyphenols including tannins play a lead role in the defense against herbivores and other pests (Coley et al., 1993; Rothstein et al., 2004). Chlorophylls and carotenoids regulate light harvesting, while lightweight, soluble carbon fractions form the high-energy products of photosynthesis (Evans et al., 1988). Larger and heavier secondary metabolites including cellulose and lignin require greater energetic investment by plants, but yield increased leaf toughness, longevity, and defense capability (Hikosaka, 2004). This chemical portfolio is maintained in a structure of varying leaf mass per unit area (LMA) (Poorter et al., 2009).

Traditionally, studies of leaf chemical variation have focused on soil fertility and climate controls, which differentially influence concentrations of N, P, base cations, and other foliar constituents (e.g. Vitousek & Sanford, 1986; Raich et al., 1996; Aerts, 1997). Others have focused on chemical differences among plant functional types (PFTs) as a means to generalize patterns at large biogeographic scales (Bonan et al., 2002; Wright et al., 2004). However, recent studies have highlighted the potential importance of species-level diversity and taxonomic organization of several leaf chemical properties in tropical forest canopies. Townsend et al. (2007) reported that species exert a dominating influence on variation in leaf N and P concentrations in Costa Rican and Brazilian lowland forests. Hattenschwiler et al. (2008) found pronounced inter-specific variation in leaf and litter chemical properties in Guyana, noting that such high chemical diversity weakens the role of general PFT-based rules in predicting canopy function or ecosystem processes. Fyllas et al. (2009) and Asner et al. (2009) documented taxonomic organization among leaf chemicals, against a backdrop of varying climate and soils, in Amazonian and Australian tropical forest canopies, respectively. In a lowland Borneo rainforest, Paoli (2006) found a differential effect of environment and phylogeny on leaf chemical variation: within the genus Shorea, variation in leaf P and specific leaf area (LMA−1) was influenced more by soil fertility than was leaf N, which more closely tracked phylogeny. These and other studies give us a sense that leaf chemical attributes are indeed strongly influenced by species composition. However, the role of soil fertility and climate in mediating the connection between phylogeny and leaf chemical traits is not well understood.

A major barrier to linking canopy diversity and chemistry, and to understanding the role of this linkage to ecosystem processes, rests in measurement and tracking of the taxa at geographic scales commensurate with community dynamics and demographic change. Field measurements cannot easily resolve changes in forest canopy composition because the pertinent demographic dynamics occur at scales larger than most plots. In humid tropical forests, this limitation is evidenced by the fact that the spatial co-occurrence of species or even congeners is often relatively low and many singletons exist in a given plot (Condit et al., 2005). To understand the ecological importance of varying canopy composition, we need a way to observe and quantify plant traits that may indicate the presence of species and their functional role over relatively large areas. The potential observables are a challenge to identify, yet the regional perspective is proving critical to understanding ecological change for conservation and management decision-making.

Recent work demonstrates that remotely sensed optical spectroscopy provides a window into the composition and diversity of tropical forests. In Hawaiian forests, Asner et al. (2008) developed airborne spectroscopic signatures to identify native and invasive species, while Carlson et al. (2007) employed the concept of spectral variance to map canopy species richness. Castro-Esau et al. (2004) and Sanchez-Azofeifa et al. (2009) used leaf-level spectroscopy to delineate liana and tree species in Panamanian forests. Clark et al. (2005) used airborne spectroscopy to classify several canopy tree species in a Costa Rican forest. These and other studies provide novel links between spectral data and species information, but none have developed the general approach required to broadly understand the interconnection between canopy composition and spectroscopy. We believe that this interconnection can be made robustly and generally via the chemical properties of the canopies.

Asner & Martin (2009) introduced the concept of spectranomics to link a specific type of remote sensing – high-fidelity spectroscopy – of foliage to canopy taxa via their detailed chemical signatures. Variation in spectroscopic properties of canopies is determined by multiple molecular compounds ranging from pigments to secondary metabolites, along with variation in leaf area and volume, and canopy architecture (Curran, 1989; Asner, 1998). In highly foliated canopies of the humid tropics, leaf chemical traits are primary determinants over high-fidelity spectra (Asner, 2008; Asner & Martin, 2008). The spectranomics approach suggests that a spectral–chemical link would allow taxonomic analysis of tropical forest canopies from aircraft, yet community composition may be disconnected from the spectral-to-chemical linkages needed to indicate species presence and functional status. This disconnect may occur if intraspecific variation in chemical attributes is high and/or if phenotypic plasticity trumps phylogenetic patterns among chemical traits. Even if spectral measurements yield quantitative information on the chemical signatures of canopy foliage, closely related species may have similar chemical portfolios (phylogenetically conserved), making it difficult to differentiate taxa. To our knowledge, the link between community composition and spectral properties through chemical traits has not been broadly demonstrated.

We sought to integrate phylogenetic, chemical and spectral properties of canopies in a lowland tropical forest spanning low- and high-fertility soils in the Peruvian Amazon. Our goals were to determine the degree to which canopy chemical traits, and combinations of traits termed ‘chemical signatures’, are phylogenetically organized within and across contrasting soil types, and to assess which chemical constituents can be quantitatively linked to canopy spectroscopy. This study is the first major test of the spectranomics concept, carried out in a very high diversity forest. Here we present data on 21 leaf chemical traits and high-fidelity spectroscopic signatures of 594 individuals of tree, palm, vine and liana growth habits. We carefully controlled for full-sunlight, upper canopy position to ensure that the light environment was relatively constant, thereby avoiding unwanted variation in chemical traits resulting from shade, and because upper canopy foliage plays a dominant role in determining the spectroscopic remote sensing signatures of tropical forests (Asner, 2008).

Materials and Methods

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

Site description

The study was conducted in terra firme and floodplain forests along the Tambopata River in the southern Peruvian Amazon basin. Terra firme forests are located on flat to undulating Pleistocene surfaces c. 15–20 m above the floodplain forests and are classified as weathered, lower fertility haplic Ultisols (or Alisols in the Food and Agriculture Organization (FAO) World Reference Base) (Quesada et al., 2009). Floodplain forests are characterized by high-fertility alluvium originating in the Andes, and deposited throughout the Holocene, forming humic Inceptisols (or Cambisols) that do not currently experience frequent or extensive flooding. These soil types are immediately adjacent to one another and support large-statured forest canopies reaching 40 m in height. Mean annual precipitation and temperature are 2600 mm yr−1 and 24.0°C, respectively. The Holdridge Life Zone classification is moist lowland tropical forest.

Canopy collections were undertaken over a total forest area of c. 1600 ha, with each soil type covering about half of this area. Our sampling was designed to capture the diversity of sunlit canopies throughout the site, while also maintaining statistical power for replication at the family, genus, species and branch (within crown) levels. In total, 594 individual canopies were randomly selected from 328 unique species spanning the two soil types and a range of growth habits. Full triplicate replication at the branch level within crowns was carried out for 450 of these individuals (= 450 × 3 branches = 1350 samples) to assess the intracrown variability. In addition, 126 species (65 on Inceptisols and 61 on Ultisols) were selected for replication with two or more representatives. Of these 126, a total of 48 had three or more replicates. These different levels of replication were based on the requirements of the various statistical analyses employed in the study, described below and in the Methods S1 section of the online Supporting Information. Tree species comprised the majority of the individuals with 531 representatives, followed by lianas (40), hemi-epiphytes (11), palms (nine) and vines (three). The distribution was only slightly different on the two soil types. On the Inceptisols, the 294 individuals selected were comprised of 202 unique species with the following habits: tree (257), liana (20), hemi-epiphyte (eight), palm (seven), and vine (two). On the Ultisols, 300 individuals were selected consisting of 193 unique species (29 with three or more representatives; 103 individuals) with the following habits: tree (274), liana (20), hemi-epiphyte (three), palm (two), and vine (one). Taxonomically the 328 species were partitioned into 177 genera and 55 families.

Samples were selected to control for full-sunlight canopies, providing comparable canopy position and illumination conditions among samples. Leaf collections were conducted using tree-climbing techniques to ensure that full-sunlight samples were taken. Samples were cryocooled (−80°C) and dried, with additional pre-processing, in the field and then transported to the laboratory for multi-chemical assays. Spectroscopic measurements were made in the field using fresh samples with a high-fidelity 400–2500-nm custom-built spectroradiometer, integrating sphere and light source. Statistical analyses included intraspecific coefficients of variation (with at least two replicates per species), general linear modeling (at least three replicates), nested random-effects modeling (at least three replicates), principal components analysis (all data used), and stepwise linear discriminant analysis (at least two replicates). Detailed methodologies are provided in the online Supporting Information Methods S1, and laboratory protocols are downloadable from the Carnegie spectranomics website (http://spectranomics.ciw.edu).

Results

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

Soil and habit effects

Most leaf properties showed enormous variation on both soil types (Table 1). Nonetheless, chlorophylls, cellulose, lignin, phenols, and tannins were 7–23% lower in concentration on Inceptisols (Fig. 1). By contrast, P, B, Fe, and base cation contents were 10–52% higher on Inceptisols. Zn and Mn are 18 and 39% higher on Ultisols, respectively, but soluble C and hemi-cellulose fractions were lower in canopies found on these soils. LMA and N, water and carotenoid concentrations did not differ between Ultisols and Inceptisols.

Table 1.   Descriptive statistics for canopy leaf samples collected on high-fertility floodplain Inceptisols and low-fertility upland Ultisols
 InceptisolsUltisols
MSDCVMinMaxMSDCVMinMax
  1. t-tests, performed on the loge-transformed data, indicated significant differences (< 0.01) for all leaf properties except for leaf mass per unit area (LMA), N, water, and carotenoids (Car) as marked by ‘ns’.

  2. M, mean chemical concentration; SD, standard deviation; CV, coefficient of variation; Min, minimum; Max, maximum.

LMA (ns)105.331.630.034.4194.9101.628.327.938.3215.9
Water (ns)57.86.711.644.379.057.77.012.140.783.6
Chla5.221.7733.951.6013.265.571.6930.421.6411.44
Chlb1.960.7035.980.565.122.080.6832.850.644.73
Carotenoids (ns)1.560.4830.630.533.671.620.4829.670.514.32
Phenols73.248.366.00.7225.393.551.154.60.0244.7
Tannins36.020.958.10.0115.846.522.848.90.0119.2
N (ns)2.190.6830.841.044.662.200.5826.411.174.62
C46.83.37.034.753.549.03.46.934.056.0
P0.170.0846.550.060.510.130.0538.020.060.36
Ca1.290.9170.380.034.400.850.6981.660.063.97
K1.110.5953.280.204.400.880.4348.930.292.75
Mg0.310.1343.200.090.850.280.1551.920.090.91
B28.721.675.54.0173.022.815.266.74.484.7
Fe75.443.257.328.3391.261.633.754.819.7296.1
Mn289.2664.5230.410.66594.3467.9719.3153.710.04843.5
Zn17.712.570.34.688.0322.013.159.34.8118.9
Soluble C45.810.422.718.771.843.111.025.622.976.4
Hemi-cellulose15.65.132.33.335.212.34.435.31.934.7
Cellulose18.14.927.56.541.819.95.427.47.735.7
Lignin20.48.642.16.950.824.58.835.95.348.9
image

Figure 1.  Percentage differences in the ratio of each canopy chemical trait found on Inceptisols (I) and Ultisols (U), I : U (black bars); and in the ratio of each chemical trait for liana (L) and tree (T) habits, L : T (gray bars). LMA, leaf mass per unit area.

Download figure to PowerPoint

We summarize differences between trees and lianas in Table 2, and provide additional data for less common growth habits in Table S2. All means (loge-transformed data) were statistically different with the exception of P, Fe, soluble C, and cellulose. Lianas had much higher concentrations of key growth-related attributes including chlorophylls, carotenoids, N, and base cations (Fig. 1). By contrast, trees had 20–40% higher concentrations of lignin, phenols and tannins. Finally, two-way ANOVA tests indicated no significant interaction between soil and growth habit for any leaf property, with the exception of Mn (= 3.9, = 0.05).

Table 2.   Descriptive statistics for canopy leaf samples collected from species with tree and liana growth habits
 TreeLiana
MSDCVMinMaxMSDCVMinMax
  1. t-tests, performed on the loge-transformed data, indicated significant differences (< 0.01) for all leaf properties except for P, Fe, soluble C and cellulose, as marked by ‘ns’. LMA, leaf mass per unit area; M, mean chemical concentration; SD, standard deviation; CV, coefficient of variation; Min, minimum; Max, maximum. Statistics for less common growth habits including hemi-epiphytes (= 11), palms (= 9) and nonwoody vines (= 3) are provided in Supporting Information Table S2.

LMA104.329.228.038.3215.978.323.429.934.4140.9
Water57.46.611.540.779.061.68.413.644.683.6
Chla5.301.6831.611.611.447.021.9527.794.1113.26
Chlb1.970.6733.770.564.732.700.7828.761.665.12
Carotenoids1.570.4729.760.514.322.030.524.851.213.67
Phenols87.250.557.90.0244.753.447.188.30.7184.4
Tannins42.822.252.00.0119.230.123.277.10.087.8
N2.190.6228.531.044.662.520.5923.381.484.14
C48.13.67.434.056.046.73.26.939.552.3
P (ns)0.150.0746.620.060.510.160.0637.410.080.38
Ca1.050.8278.020.034.851.471.0974.020.134.18
K0.970.5253.70.24.41.210.6352.050.332.79
Mg0.290.1344.580.090.80.410.2356.750.10.91
B25.819.375.04.0173.029.115.753.86.876.7
Fe (ns)67.238.557.319.7391.275.639.051.738.4213.3
Mn352.7646.7183.510.06594.3834.11199.4143.811.45034.3
Zn19.411.659.84.681.227.623.886.37.9118.9
Soluble C (ns)44.610.824.122.976.445.69.921.825.268.3
Hemi-cellulose13.74.936.11.934.717.15.230.48.128.2
Cellulose (ns)18.75.026.86.535.718.84.926.29.731.9
Lignin22.89.139.85.350.818.46.535.47.833.0

Within-crown and intraspecific variation

Within-crown variation in leaf properties varied from medians of only 1.4 and 1.7% for C and water, respectively, to 27.7% for phenols (Fig. 2a). Median variation in macro- and micronutrients within crowns ranged from 4.0 to 9.8%, which was similar to the range measured for C fractions including lignin, cellulose, hemi-cellulose and soluble C (4.0–9.3%). Median variation in photosynthetic pigment concentrations ranged from 10.3 to 12.3%. Phenols and tannins showed occasional very high values for intracrown variation, reaching an absolute maximum of 123% in one case. Nonetheless, the upper quartile value for these leaf constituents was only 45 and 31% for phenols and tannins, respectively.

image

Figure 2.  Coefficients of variation (CVs) for 20 chemical traits and leaf mass per unit area (LMA) for (a) samples collected on different branches within crowns and (b) samples collected from different crowns within species. Chemical data are mass-based; = 126 species with two or more replicates per species.

Download figure to PowerPoint

Intraspecific variation among leaf properties exceeded that of within-crown variation in most cases (Fig. 2b). Total C had the minimum median variation of 1.8%, and Mn had the maximum of 36.7%. Maximum intraspecific variation was about twice that of the maximum intracrown variation. We also analyzed intraspecific variation by soil type (Fig. S1), and this indicated no general differences in the variability of most chemicals within species. A few notable exceptions included 44–60% greater variation in Mn and Zn, and 10–17% less variation in photosynthetic pigments, in species found on Inceptisols.

Phylogenetic patterns

Above, we have reported wide-ranging values for all chemicals and LMA, independent of soil type or growth habit. However, we also showed that Inceptisols or lianas were associated with higher concentrations of nutrients and lower concentrations of defense-related compounds (lignin and phenols) compared with Ultisols or the tree growth habit. Yet despite these underlying substrate and habit effects, and the degree of intraspecific variation measured, an average 70% of the variation in the elemental and molecular composition of canopy foliage can be explained by phylogenetic groupings that incorporate families, genera, and species (Fig. 3a). Here we use the term ‘taxonomy’ as an expression of phylogeny; phylogeny resolves more detailed evolutionary relationships among taxa, which we did not assess. Factors other than taxonomy, including micro-environment, growth habit, sample selection and analytical error, contributed to the average remaining 30% residual in the nested ANOVA analyses. Family, genus within family, and species within genus within family groupings accounted for an average of 25, 15, and 30% of the chemical taxonomy, respectively. Chemical attributes with the highest degree of taxonomic organization were lignin (86%), total C (85%), phenols (84%), and N (82%). Taxonomy accounted for just 38–52% of the variation in chlorophylls and carotenoids, suggesting a more universal approach to light harvesting by tropical canopies.

image

Figure 3.  Nested random-effects ANOVA results for canopy taxa from (a) combined sites, (b) Inceptisols and (c) Ultisols in the Tambopata National Forest, southern Peruvian Amazon basin. Results show the percentage variance of each chemical attributable to plant families, genera within families, species within genera within families, and the residual. The residual includes factors such as intraspecific variation, micro-site variability, canopy selection, and analytical error.

Download figure to PowerPoint

Soil type imparted few effects on the degree and pattern of taxonomic organization for most chemicals (Fig. 3b,c). A few exceptions were the photosynthetic pigments: Chla, Chlb, and carotenoids showed 36, 39, and 18% less overall taxonomic organization on the higher fertility Inceptisols. Moreover, family- and genus-level organization of these pigments was 76–152% greater on Inceptisols, although species-level control was > 100% higher on the Ultisols. More generally, although families accounted for 39–423% more variation in all leaf chemicals on Inceptisols than on Ultisols, this heightened degree of family-level control was balanced by increased genus- and/or species-level control on the Ultisols. N is a good example: families accounted for 43% more variation in N on Inceptisols, whereas genus and species accounted for 15 and 34% more variation on Ultisols, respectively. This led to no net difference in overall phylogenetic control of leaf N on either soil.

Regression analyses supported results from the nested ANOVA tests (Table S3). Family, genus, and species accounted for an average 25, 32, and 68% (adjusted r2; < 0.01), respectively, of the variation in leaf chemicals and LMA. Maximum species-level regression (adjusted r2) values were 0.88 for C, 0.87 for lignin, 0.82 for phenols and 0.81 for N. The weakest species-level regressions were for chlorophylls and carotenoids (0.34–0.49). Families on Inceptisols had an average 60% higher adjusted r2 for leaf chemicals, and were nearly three times stronger determinants of Ca concentrations. By contrast, photosynthetic pigments showed 67–81% higher regression coefficients at the genus level on Inceptisols, but 48–64% higher at the species level on Ultisols. This is similar to the compensating effects at family, genus and species levels described earlier for N in the nested ANOVA analyses.

Inter-relationships among leaf properties

The diversity of chemical signatures among species is a function of stoichiometric relationships between constituents. Correlation among chemicals reduces the dimensionality of would-be chemical signatures, whereas orthogonal properties add to the chemical diversity of the signatures. Principal components analysis (PCA) showed that a single linear combination of traits (PC1) explained just 31% of the variation among 21 leaf properties, and 14 orthogonal combinations were required to explain 95% of the variation (Table 3). There was no effect of soil type on this pattern (data not shown). We also considered other leaf chemical trait combinations, starting with the photosynthetic pigments alone, for which PC1 accounted for 97% of the variation between chlorophylls and carotenoids. We analyzed three traits contributing to the growth-related leaf ‘economics spectrum’ (Wright et al., 2004) as a group, and PC1 accounted for 61% of the variation among N, P and LMA. Finally, the first PC accounted for 59% of the variation between C fractions.

Table 3.   Principal components analysis (PCA) results for different combinations of leaf properties
 Leaf properties% of variance explained (raw data)% of variance explained (loge-transformed data)
  1. The percentage variance explained by the first principal component is shown for raw and loge-transformed data. Data in parentheses are results for the tree habit only.

  2. 1Ten leaf properties selected for remote sensing analysis including water, C, N, leaf mass per unit area (LMA), chlorophylls, carotenoids, cellulose, soluble C, phenols and hemi-cellulose (see Table 5).

(i) Chla, Chlb, and carotenoids96.5 (96.2)97.7 (96.5)
(ii)LMA, N, and P59.1 (58.7)61.2 (60.8)
(iii)N, P, Ca, K, and Mg43.3 (49.1)49.3 (52.4)
(iv)Soluble C, cellulose, hemi-cellulose, and lignin58.0 (59.0)58.9 (59.6)
(iv)i + ii+ iii + iv above37.9 (36.7)38.5 (37.4)
(v)Remote sensing properties137.9 (37.1)38.2 (37.6)
(vi)All leaf properties30.5 (30.1)31.6 (31.2)

Correlation analyses further supported results from the PCA work, indicating relatively weak correlation (e.g. < 0.4 for 86% of trait pairs) among most chemicals (Table S4). Exceptions again included the photosynthetic pigments, for which inter-correlations are well understood (Sims & Gamon, 2002), but all other chemical relationships left 31–99% (mode = 80%) of the information uncorrelated. There was a highly variable effect of soil type on leaf chemical correlations (Tables S5, S6). About 68% of the trait pairs showed some level of increased correlation on the Inceptisols, with the majority of these increases observed among nutrients. About 32% of the trait pairs showed no difference or slightly lower correlation on Inceptisols than on Ultisols.

Taxonomic classification using chemistry

Stepwise linear discriminant analysis (LDA) indicated that 96.9% of the species were correctly classified using their full chemical signatures (Fig. 4). A combination of lignin, N, C and phenols alone classified 65.6% of the species correctly (Table 4); adding other elements and compounds yielded 1–10% more power per chemical to the classification until the signature was comprised of the full 21 constituents. The results of family- and genus-level analyses generally mirrored the species results (Fig. 4), although with lower overall classification accuracy because of the increased variation among leaf traits at these lower taxonomic levels. Together, these results indicate that the complete chemical signature, as defined in our study, accounts for nearly all of the species-level variation among canopies measured on both soils.

image

Figure 4.  The increase in the accuracy of chemically based taxonomic classification is shown here using stepwise linear discriminant analysis (LDA). Chemical signature composition steps (1–21) map to specific chemical elements and molecular compounds listed for family (blue), genus (red) and species (green) in Table 4.

Download figure to PowerPoint

Table 4.   Results of stepwise linear discriminant analysis (LDA) showing the percentage of species correctly classified using an increasing number of leaf traits (1–21, cumulatively)
StepSpeciesGeneraFamilies
  1. The analysis was repeated at genus and family levels. Results are shown in Fig. 3. Chl, chlorophyll; LMA, leaf mass per unit area.

1LigninLigninC
2NNPhenols
3CCMg
4PhenolsPhenolsN
5MnWaterLignin
6Hemi-celluloseBB
7MgMgWater
8CelluloseMnK
9Soluble CKP
10WaterLMACellulose
11KCelluloseZn
12BPSoluble C
13LMAZnLMA
14CaCaMn
15ZnSoluble CCa
16PHemi-celluloseCarotenoids
17TanninsFeChlb
18FeChlbChla
19CarotenoidsCarotenoidsFe
20ChlbChlaHemi-cellulose
21ChlaTanninsTannins

We tested for soil effects on taxonomic classifications using LDA with the 21-trait signatures (Fig. 5). Soils exerted relatively little influence on the classification of taxa, with c. 10% effect at the family and genus levels. On Inceptisols, species-level classification reached 100% accuracy using just nine chemical traits, with the initial five most important attributes being lignin (13.6%), Mn (33.3%), hemi-cellulose (16.4%), cellulose (17.7%) and water (10.9%) (Table S7). These initial five attributes accounted for 91.8% of classification accuracy on Inceptisols. On Ultisols, a similar set of five initial components (C, lignin, hemi-cellulose, Mn, and N) were needed to obtain 87.9% accuracy, and 15 of 21 chemical constituents were required to approach a maximum classification accuracy of 99%. Substrate-related differences in LDA performance were smaller for genus and family classifications (Fig. 5).

image

Figure 5.  The effect of soil type (Ultisols, dashed line; Inceptisols, dotted line; combined, solid line) on linear discriminant analyses of (a) families, (b) genera and (c) species. The 21 leaf properties are provided in Supporting Information Table S7.

Download figure to PowerPoint

Spectroscopy of canopy chemicals

Variation among the spectral reflectance signatures of 594 individuals is shown in Fig. 6(a). Although the greatest reflectance range was observed in the near-infrared (750–1300 nm), the highest coefficients of variation (CVs) averaged 29% in the shortwave-infrared (1500–2500 nm) followed by the visible spectral region (400–750 nm) (Fig. 6b). This confirmed that, although the reflectance range was low in the visible and shortwave-infrared, the variation among canopies was most pronounced in these spectral regions (Asner, 2008). Intraspecific variation in canopy reflectance varied by wavelength (Fig. 6c), with variation in the visible (400–700 nm), near-infrared (700–1300 nm), and shortwave-infrared (1300–2500 nm) regions averaging only 7, 5, and 8%, respectively. There was no clear differential effect of soil or habit on spectral reflectance signature variability (Fig. 6d). However, spectral variation was 2–4 times greater by soil than within any given habit (tree and liana shown in Fig. 6(d), but true for all habits). This indicates the relatively weak influence of plant habits on spectral variation among canopies found on either soil.

image

Figure 6.  (a) Minimum and maximum values of spectral reflectance for canopies in the Tambopata National Forest, southern Peruvian Amazon basin. (b) Spectral coefficients of variation (CVs) for all canopies (= 594). (c) Mean intraspecific CV in canopy spectral reflectance. (d) Spectral CV by soil type and by tree and liana growth habits.

Download figure to PowerPoint

The brightness-normalized Partial Least Squares (bnPLSR; Feilhauer et al. 2010) analyses indicated that many of the 21 leaf properties can be estimated using high-fidelity spectroscopy at the canopy scale. Validation r2 values ranged from a low of 0.16 for Mn to a high of 0.86 for water (Table 5). Root mean squared error (RMSE) values expressed as a percentage of the mean for each chemical trait varied from a low of 3.9% for water to a high of 234.4% for Mn. Chemical traits found to be critical to taxonomic classification (Table 4) such as C, N, cellulose and phenols showed %RMSE values of 3.2, 12.8, 14.6, and 31.1%, respectively (Table 5). Taking a cut-off value of RMSE ≤ 30% and r 0.50, we found that 11 of 21 leaf properties might be considered best for taxonomic classification based on remotely sensed spectroscopy (Table 5). Here we used only Chla because it is highly correlated with Chlb, resulting in 10 chemical properties for LDA analysis. We then tested for inter-correlation among these remotely sensible traits using PCA, and found that only c. 37% of the trait correlation was expressed in the first axes of variation (Table 3). Furthermore, seven components were required to account for 95% of the variance among these 10 chemical traits. Finally, the LDA analysis showed that 91% of species could be correctly classified, with C, N, water, and several molecular compounds accounting for 84% of the process (Fig. 7). Genera and families were classified to 64 and 43%, respectively.

Table 5.   Spectroscopic estimation of leaf chemical properties and leaf mass per unit area (LMA) using canopy reflectance signatures (Fig. 6a)
Leaf propertyRMSE%RMSEr2RS use index
  1. RMSE, root mean square error in units of the original chemical assays (Supporting Information Table S1); %RMSE, RMSE expressed as a percentage of the mean value of the leaf trait (Table 1); r2, regression coefficient for validation data used during the bnPLS analyses; remote sensing (RS) use index, %RMSE × (1 − r2).

Water2.263.90.860.55
C1.513.20.661.10
LMA11.9011.30.822.05
Chlb0.2814.50.793.06
Chla0.7514.40.783.10
N0.2812.80.693.98
Carotenoids0.2415.20.744.02
Cellulose2.6514.60.645.26
Soluble C6.5014.20.516.97
Phenols22.7830.10.767.56
Hemi-cellulose3.1019.90.509.99
Tannins11.9533.20.6511.58
K0.3228.50.5313.49
Lignin5.9229.00.4914.71
Mg0.0930.40.4516.77
P0.0529.10.4017.52
Fe30.2540.10.4621.65
Ca0.6449.50.4427.76
B15.2253.00.2937.52
Zn11.8366.90.3146.46
Mn677.84234.40.16197.18
image

Figure 7.  Stepwise linear discriminant analyses of families (blue), genera (red) and species (green) based on the chemical traits with the highest degree of remote sensing precision and accuracy (Table 5). LMA, leaf mass per unit area.

Download figure to PowerPoint

Discussion

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

Chemical diversity

Leaf properties showed enormous variation on both soil types (Table 1), often matching or exceeding the variation reported for larger ecoregions and biomes (McGroddy et al., 2004; Reich & Oleskyn, 2004; Wright et al., 2004; Townsend et al., 2007; Poorter et al., 2009). For example, our LMA range was 34–216 g m−2, approaching the range of 30–299 g m−2 for a regional sampling of Amazonian forests (Fyllas et al., 2009). The range of N and P variation among all terrestrial ecosystems was previously reported as 0.7–4.1% and 0.04–0.23%, respectively (Wright et al., 2004), while our measured range was 1.0–4.6% for N and 0.06–0.51% for P in a single forest. We also found an average 8500% variation in large molecular compounds including lignin, hemi-cellulose and phenols among coexisting canopy species on both soil types. Total C showed the least amount of variation, yet still a substantial 54% on Inceptisols and 65% on Ultisols. Overall, the range of chemical concentrations and LMA found in the Tambopata forest canopy is enormous, providing a basis upon which phylogenetic patterns in leaf chemistry might be expressed.

Environmental filtering

We found major differences in the chemical composition of canopies on adjacent high- and low-fertility soils (Fig. 1). P and all base cations were at higher concentrations in canopies on the Inceptisols, with Ca showing a maximum difference of > 50% compared with canopies sampled on Ultisols. By contrast, we found no difference in leaf N concentration by soil type. The leaf N : P ratio was 14.2 (SD = 5.3) on Inceptisols, which was significantly lower than the mean value of 18.1 (SD = 5.4) in canopies on the Ultisols (t-test, < 0.05). This suggests P limitation on Ultisols, and neither N nor P limitation on the Inceptisols (Hedin, 2004; Reich & Oleskyn, 2004). Compared with P, relatively little work has been done on the ratio of Ca to other nutrients, in the context of identifying elements that might affect growth and/or demographic patterns. Ca : P ratios were 7.6 and 6.5 on Inceptisols and Ultisols, respectively – a 17% difference, suggesting that Ca may be limiting on the Ultisols.

Our results support the hypothesis that soil fertility affects spatial patterns of canopy composition in high-diversity tropical forests (John et al., 2007; Fyllas et al., 2009). For example, a comparison of family-level composition by soil type (34 families containing species with at least two replicates) indicated that only 34% of families were found on both soils (Table 6). Moreover, only 16 and 9% of genera and species tested were found on both soils, with the remainder specializing on either Inceptisols or Ultisols. While our sampling was not specifically designed to quantify compositional differences between soil types, we did attempt a comprehensive sampling of the most common taxa found on each soil, and we have a good working knowledge of canopy composition in these forests. Our family-level partitioning also parallels the findings of Phillips et al. (2003), who reported that c. 25% of families tend to be found on both soil types in neighboring forest areas. We therefore believe that substrate-based differences in leaf properties are an expression of differences in community composition. This landscape-scale variation in composition is the broadest environmental filter recognizable in our data set.

Table 6.   Number of families, genera and species with two or more replicates used in the nested random-effects ANOVA shown in Fig. 3
 Both soilsInceptisolsUltisolsTotal
  1. Counts and percentage of total are shown for taxa found on both soils, only Inceptisols and only Ultisols.

Family12 (35%)13 (38%)9 (26%)34
AnnonaceaeAnacardiaceaeCannabaceae
ApocynaceaeArecaceaeChrysobalanaceae
BignoniaceaeCaricaceaeClusiaceae
BurseraceaeCaryocaraceaeEuphorbiaceae
FabaceaeCombretaceaeLauraceae
LecythidaceaeCucurbitaceaeMelastomataceae
MeliaceaeEbenaceaeMenispermaceae
MoraceaeElaeocarpaceaeRubiaceae
MyristicaceaeMalpighiaceaeSabiaceae
SapotaceaeMalvaceae 
SimaroubaceaePhytolaccaceae 
UrticaceaeRutaceae 
 Sapindaceae 
Genera12 (16%)34 (46%)28 (38%)74
Species10 (9%)51 (46%)50 (45%)111

Secondary metabolites such as lignin, phenols and tannins were found in greater leaf concentrations on the Ultisols (Fig. 1). This pattern suggests that chemical allocation to defense and longevity is greater on low-fertility soils, probably expressing strategies to negotiate lower nutrient conditions by building tougher, pest-resistant foliage. Such patterns in leaf chemicals may also be an indirect response to differences in nutrient availability affecting leaf turnover and growth strategies. Support for this interpretation may be found in the concentrations of Zn and Mn, which were much higher on Ultisols. Zn is key to the chelation process involved in protecting leaves from the toxic metals typically found in weathered soils (Broadley et al., 2007). We also think that Mn is more abundant in foliage on Ultisols in response to the prevailing conditions of metabolic stress, as has been simulated in laboratory analyses (Adams et al., 1999).

Beyond the environmental filtering imparted by soil fertility, variation in chemical traits also depends upon the degree of intracrown and intraspecific plasticity. Variation within and among crowns of a species could also produce seemingly random taxonomic patterns in chemical traits. However, we found that most chemicals varied by < 10% within individual crowns, although phenols and tannins showed higher levels of variability among branches (22–23%; Fig. 2). This relatively high variance in phenols and tannins was probably a result of variations in leaf age (Coley et al., 1993), as well as random error in the wet chemistry method used for these particular assays (Ainsworth & Gillespie, 2007). Nonetheless, intracrown variation was quite low, suggesting that chemical variation at this scale is not sufficient to cause large swings in the concentrations of most chemicals throughout the forest canopy.

Chemicals with the lowest intracrown and intraspecific variation, such as soluble C, lignin, LMA, N and water (Fig. 2), were the most consistent taxonomically (Fig. 3), suggesting that microsite resource variation has little effect on these properties. Chemicals with low intracrown but relatively high intraspecific variation, such as pigments and several micronutrients, indicate plasticity within species, probably in response to micro-environmental variation in illumination and soil conditions. Chemicals with relatively high intracrown and intraspecific variation, such as phenols and tannins, indicate adjustments made at the scale of leaves and branches, in addition to the scale of whole crowns and individuals. While the overall intraspecific variation among chemical constituents was low, the significance of this variation for subsequent taxonomic classifications depends upon these absolute chemical differences within and between crowns.

Phylogenetic patterns

Despite clear substrate effects and some intraspecific variation among foliar chemicals, phylogeny remained the most evident source of chemical variation throughout the forest. We found that lianas maintained much higher concentrations of photosynthetic pigments, N, P, cations and micronutrients, and much higher LMA, than trees. In contrast, concentrations of lignin, phenols and tannins were 20–40% higher in trees. Although it was not possible to know if lianas differed in chemical concentrations as a result of soils or chemical evolution, these habit-based differences did not impart a detectable bias on the taxonomic organization of chemical properties (Fig. 3, Table S3). Instead, we found that nesting of species within genera within families accounted for a low of 39% (Chla) to a high of 86% (lignin) of the chemical variance among canopies. Using nested ANOVA, we reanalyzed the data with plant habit as a fourth factor, but did not find a significant contribution of habit to any given chemical (data not shown). It is also clear that pigments showed the least amount of taxonomic organization, which would be expected given localized variation in self-shading conditions and leaf age in the upper canopy. However, secondary metabolites such as lignin and phenols showed marked taxonomic organization – a clear sign that phylogeny matters when it comes to understanding spatial and biological patterns in defense and leaf longevity throughout the forest. Families accounted for 5–45% of the variation in leaf chemicals. Elements and compounds such as C, N, lignin, phenols and Mg displaying strong family-level organization may have a deeper evolutionary history than those chemicals showing little family-level structure, such as photosynthetic pigments.

Chemical signatures among taxa

The role of chemical signatures – combinations of chemical properties within the leaf – in determining linkages between phylogeny and ecosystem function is only beginning to be addressed in humid tropical forests. Yet an evolutionary history under a regime of intense biotic interaction (competition, mutualism, and pest defense) and generally good growth conditions (e.g. high net primary production) might result in highly organized inter-specific and, more generally, taxonomic organization of canopy chemical signatures. A key prerequisite involves determining which chemical properties are inter-correlated as a means to understand stoichiometry in the context of life strategies. Our PCA and correlation analyses indicated a high degree of correspondence among pigments (light capture) and traits closely associated with light harvesting (N and LMA) (Tables 3, S3). This result, like that of many others, supports the hypothesis that these traits display coordinated variation to maximize growth (Reich et al., 1997; Wright et al., 2004). However, we also found combinations of additional leaf properties that remained largely uncorrelated, and it is these chemicals that add dimensionality to leaf chemical signatures. In turn, this added dimensionality expresses a wide range of strategies that support growth and maintenance via numerous metabolic processes, and that support leaf survival via chemical and physical defense. A clear expression of this can be found in the LDA analyses of canopy chemical signatures: although we did find intraspecific variation among individual elements and molecular compounds, when combined the multi-chemical signatures were largely unique among species found on each soil type and throughout the site as a whole (Figs 4, 5). In combination, our results uncover a multitude of chemical signatures among species within an Amazonian canopy, and suggest a wide range of strategies played out on the forest landscape to negotiate changing niche conditions and biotic interactions over time.

Canopy chemical assembly

This study provides new insights into patterns of forest canopy chemical variation, from which processes of chemical trait evolution and community assembly can be inferred. At the landscape scale, canopy chemical responses to soil fertility cannot be easily disentangled from changes in species composition. Despite the clear environmental filtering of chemical concentrations and species composition, we uncovered a strong phylogenetic component in the distribution of individual chemicals and of multi-chemical signatures among canopies. Several traits displaying clear taxonomic organization – namely N, C, lignin, phenols, and water (Fig. 3) – are also the most important contributors to classifications of families, genera and species based on chemical signatures (Table 5). These results, along with the observed low intraspecific variation of leaf chemicals, suggest that species maintain tightly regulated chemical signatures which would predispose them to one soil type or the other. However, linear discriminant analyses of multi-chemical signatures also demonstrated that family-level organization is important, with 21 traits correctly classifying > 60% of individuals at the family level (Fig. 4).

The nested ANOVA analyses provide further insights into the effects of soils on species composition and chemical concentrations in canopies. Families accounted for much more variation in leaf chemicals on the Inceptisols than on Ultisols, but this pattern was reversed in canopies on Ultisols, which showed stronger organization at the species level. Taxa found on more fertile soils maintain chemical portfolios with an older evolutionary history expressed at the family level, possibly because selection is relaxed where resources are abundant and traits are free to drift at lower taxonomic levels. By contrast, taxa predominantly found in low-fertility soils may have undergone trait evolution under a regime of resource-dependent niche differentiation and biotic interactions. To our knowledge, these are new patterns uncovered in an Amazonian forest, suggesting that canopies carry chemical portfolios that vary based on different evolutionary histories and modes of natural selection.

Our results emphasize a strong phylogenetic component of chemical trait variation, but they do not indicate whether the traits are phylogenetically conserved or overdispersed. Additional study is planned using DNA barcodes to develop phylogenetic trees against which chemical trait patterns can be assessed. Independent of phylogenetic signals, our results indicate a strong taxonomic organization of canopy chemical signatures representing variation in functional strategies related to growth, metabolism, longevity, and defense.

Canopy spectranomics

We have drawn a quantitative connection between canopy phylogeny, chemistry and spectroscopy for hundreds of species in a high-diversity tropical forest. Of the 21 leaf traits analyzed, all displayed some level of taxonomic organization (Fig. 3), and at least 10 showed a strong connection to canopy spectroscopic signatures (Fig. 6, Table 5). Using those 10 leaf traits, we were able to classify taxa to a high degree of accuracy – over 90% at the species level and 40% at the family level (Fig. 7). Combined, these results indicate that canopy spectroscopy can quantify leaf chemical signatures, which subsequently can be used to classify hundreds of taxa. This constitutes the spectranomics approach, demonstrated for the first time here in a high-diversity canopy, and greatly extending the original concept of Asner & Martin (2009). Application of this approach to airborne spectroscopic data depends on additional challenges including sensor performance, data fusion techniques and biophysical scaling methods (Asner et al., 2007; Asner & Martin, 2008).

Past work has focused on the use of leaf or canopy spectra to identify species using various spectral bands or spectral features (e.g. Roberts et al., 1998; Cochrane, 2000; Clark et al., 2005; Castro-Esau et al., 2006; Zhang et al., 2006). Spectranomics differs from these and similar studies by first focusing on the relationship between phylogeny and chemical traits, and then making the link between chemical traits and the spectroscopic signatures of the taxa. This approach provides advantages over methods that attempt to directly connect species to spectra. First, much of the variation among canopy spectra results from factors having little to do with species identity and their chemical traits. These factors include variation in canopy structure (e.g. leaf area index), illumination conditions, and viewing geometry (Kennedy et al., 1997; Asner, 1998). To address this problem, we convert canopy spectra to chemical signatures using analytical techniques that dampen the effect of these other factors (Asner & Martin, 2008; Feilhauer et al., 2010). Next, the chemical traits derived from the spectra can be used to assess community composition, while the chemistries can also be analyzed for intraspecific variation and inter-correlation, thereby refining the remote sensing measurements on leaf constituents that will best delineate the taxa. Finally, the chemical trait data can be used to explore functional diversity and phylogenetic patterns in both a spatial and an evolutionary context, providing a novel link between remote sensing, ecosystem function, and community assembly theory. Spectranomics provides a new connection between these disparate fields of study, with results that touch on the ecology and evolution of humid tropical forests.

Acknowledgements

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

We thank G. Powell, R. Tupayachi, R. Emerson, D. Knapp, F. Sinca, P. Martinez, N. Jaramillo, L. Carranza, C. Anderson, M. Houcheime, M. Papes, P. Weiss, K. Smith, K. Ledesma, and the many assistants who helped with logistics, field work and laboratory analyses. We thank A. Austin and three anonymous reviewers for critiquing the manuscript. The Carnegie Spectranomics Project (http://spectranomics.ciw.edu) is funded by the John D. and Catherine T. MacArthur Foundation.

References

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information
  • Aber JD, Melillo JM. 1991. Terrestrial ecosystems. Philadephia, PA, USA: Saunders College Publishing.
  • Adams ML, Philpot WD, Norvell WA. 1999. Yellowness index: an application of spectral second derivatives to estimate chlorosis of leaves in stressed vegetation. International Journal of Remote Sensing 20: 36633675.
  • Aerts R. 1997. Climate, leaf litter chemistry and leaf litter decomposition in terrestrial ecosystems: a triangular relationship. Oikos 79: 439449.
  • Ainsworth EA, Gillespie KM. 2007. Estimation of total phenolic content and other oxidation substrates in plant tissues using Folin Coicalteau reagent. Nature Protocols 2: 875877.
  • Asner GP. 1998. Biophysical and biochemical sources of variability in canopy reflectance. Remote Sensing of Environment 64: 134153.
  • Asner GP. 2008. Hyperspectral remote sensing of canopy chemistry, physiology and diversity in tropical rainforests. In: KalacskaM, Sanchez-AzofeifaGA, eds. Hyperspectral remote sening of tropical and subtropical forests. Boca Raton, FL, USA: Taylor and Francis Group, 261296.
  • Asner GP, Jones MO, Martin RE, Knapp DE, Hughes RF. 2008. Remote sensing of native and invasive species in Hawaiian forests. Remote Sensing of Environment 112: 19121926.
  • Asner GP, Knapp DE, Kennedy-Bowdoin T, Jones MO, Martin RE, Boardman J, Field CB. 2007. Carnegie Airborne Observatory: in-flight fusion of hyperspectral imaging and waveform light detection and ranging (LiDAR) for three-dimensional studies of ecosystems. Journal of Applied Remote Sensing 1: paper 013536, doi:10.1117/1111.2794018.
  • Asner GP, Martin RE. 2008. Spectral and chemical analysis of tropical forests: scaling from leaf to canopy levels. Remote Sensing of Environment 112: 39583970.
  • Asner GP, Martin RE. 2009. Airborne spectranomics: mapping canopy chemical and taxonomic diversity in tropical forests. Frontiers in Ecology and the Environment 7: 269276.
  • Asner GP, Martin RE, Ford AJ, Metcalfe DJ, Liddell MJ. 2009. Leaf chemical and spectral diversity of Australian tropical forests. Ecological Applications 19: 236253.
  • Bonan GB, Levis S, Kergoat L, Oleson KW. 2002. Landscapes as patches of plant functional types: an integrating concept for climate and ecosystem models. Global Biogeochemical Cycles 16: 101029101059.
  • Broadley MR, White PJ, Hammond JP, Zelko I, Lux A. 2007. Zinc in plants. New Phytologist 173: 677702.
  • Carlson KM, Asner GP, Hughes FR, Ostertag R, Martin RE. 2007. Hyperspectral remote sensing of canopy biodiversity in Hawaiian lowland rainforests. Ecosystems 10: 536549.
  • Castro-Esau KL, Sanchez-Azofeifa GA, Caelli T. 2004. Discrimination of lianas and trees with leaf-level hyperspectral data. Remote Sensing of Environment 90: 353372.
  • Castro-Esau KL, Sanchez-Azofeifa GA, Rivard B, Wright SJ, Quesada M. 2006. Variability in leaf optical properties of Mesoamerican trees and the potential for species classification. American Journal of Botany 93: 517530.
  • Clark ML, Roberts DA, Clark DB. 2005. Hyperspectral discrimination of tropical rain forest tree species at leaf to crown scales. Remote Sensing of Environment 96: 375398.
  • Cochrane MA. 2000. Using vegetation reflectance variability for species level classification of hyperspectral data. International Journal of Remote Sensing 21: 20752087.
  • Coley PD, Kursar TA, Machado J-L. 1993. Colonization of tropical rain forest leaves by epiphylls: effects of site and host-plant leaf lifetime. Ecology 74: 619623.
  • Condit R, Ashton P, Balslev H, Brokaw N, Bunyavejchewin S, Chuyong G, Co L, Dattaraja HS, Davies S, Esufali S et al. 2005. Tropical tree alpha-diversity: results from a worldwide network of large plots. Science 55: 565582.
  • Curran PJ. 1989. Remote sensing of foliar chemistry. Remote Sensing of Environment 30: 271278.
  • EvansJR, CaemmererSvon, AdamsWWIII, eds. 1988. Ecology of photosynthesis in sun and shade. Melbourne, Australia: CSIRO.
  • Feilhauer H, Asner GP, Martin RE, Schmidtlein S. 2010. Brigthness-normalized partial least squares regression for hyperspectral data. Journal of Quantitative Spectroscopy and Radiative Transfer 111: 19471957.
  • Fyllas N, Patiño S, Baker T, Bielefeld Nardoto G, Martinelli L, Quesada C, Paiva R, Schwarz M, Horna V, Mercado L. 2009. Basin-wide variations in foliar properties of Amazonian forest: phylogeny, soils and climate. Biogeosciences 6: 26772708.
  • Hattenschwiler S, Aeschlimann B, Couteaux M, Roy J, Bonal D. 2008. High variation in foliage and leaf litter chemistry among 45 tree species of a neotropical rainforest community. New Phytologist 179: 165175.
  • Hedin LO. 2004. Global organization of terrestrial plant–nutrient interactions. Proceedings of the National Academy of Sciences, USA 101: 1084910850.
  • Hikosaka K. 2004. Interspecific difference in the photosynthesis–nitrogen relationship: patterns, physiological causes, and ecological importance. Journal of Plant Research 117: 481494.
  • John R, Dalling JW, Harms KE, Yavitt JB, Stallard RF, Mirabello M, Hubbell SP, Valencia R, Navarrete H, Vallejo M et al. 2007. Soil nutrients influence spatial distributions of tropical tree species. Proceedings of the National Academy of Sciences, USA 104: 864869.
  • Johnson DW, Todd DE. 1983. Relatioships among iron, aluminum, carbon, and sulfate in a variety of forest soils. Soil Science Society of America Journal 47: 792800.
  • Kennedy RE, Cohen WB, Takao G. 1997. Empirical methods to compensate for a view-angle-dependent brightness gradient in AVIRIS imagery. Remote Sensing of Environment 62: 277291.
  • Kraft NJB, Valencia R, Ackerly DD. 2009. Functional traits and niche-based tree community assembly in an Amazonian forest. Science 322: 580582.
  • McGroddy ME, Daufresne T, Hedin LO. 2004. Scaling of C:N:P stoichiometry in forests worldwide: implications of terrestrial Redfield-type ratios. Ecology 85: 23902401.
  • Paoli GD. 2006. Divergent leaf traits among congeneric tropical trees with contrasting habitat associations in Borneo. Journal of Tropical Ecology 22: 397408.
  • Phillips OL, Vargas PN, Lorenzo A, Cruz AP, Chuspe M-e, Sánchez WG, Yli-halla M, Rose SAM. 2003. Habitat association among Amazonian tree species: a landscape-scale approach. Journal of Ecology 91: 757775.
  • Poorter H, Niinemets U, Poorter L, Wright IJ, Villar R. 2009. Causes and consequences of variation in leaf mass per area (LMA): a meta-analysis. New Phytologist 182: 565588.
  • Quesada C, Lloyd J, Schwarz M, Baker T, Phillips O, Patiño S, Czimczik C, Hodnett M, Herrera R, Arneth A. 2009. Regional and large-scale patterns in Amazon forest structure and function are mediated by variations in soil physical and chemical properties. Biogeosciences 6: 39934057.
  • Raich JW, Russell AE, Crews TE, Farrington H, Vitousek PM. 1996. Both nitrogen and phosphorus limit plant production on young Hawaiian lava flows. Biogeochemistry (Dordrecht) 32: 114.
  • Reich PB, Oleskyn J. 2004. Global patterns of plant leaf N and P in relation to temperature and latitude. Proceedings of the National Academy of Sciences, USA 101: 1100111006.
  • Reich PB, Walters MB, Ellsworth DS. 1997. From tropics to tundra: global convergence in plant functioning. Proceedings of the National Academy of Sciences, USA 94: 1373013734.
  • Roberts DA, Gardner M, Church R, Ustin S, Scheer G, Green RO. 1998. Mapping chaparral in the Santa Monica Mountains using multiple endmember spectral mixture models. Remote Sensing of Environment 65: 267279.
  • Rothstein DE, Vitousek PM, Simmons BL. 2004. An exotic tree alters decomposition and nutrient cycling in a Hawaiian montane forest. Ecosystems 7: 805814.
  • Sanchez-Azofeifa GA, Castro K, Wright SJ, Gamon J, Kalacska M, Rivard B, Schnitzer SA, Feng JL. 2009. Differences in leaf traits, leaf internal structure, and spectral reflectance between two communities of lianas and trees: implications for remote sensing in tropical environments. Remote Sensing of Environment 113: 20762088.
  • Sims DA, Gamon JA. 2002. Relationships between leaf pigment content and spectral reflectance across a wide range of species, leaf structures and developmental stages. Remote Sensing of Environment 81: 337354.
  • Townsend AR, Asner GP, Cleveland CC. 2008. The biogeochemical heterogeneity of tropical forests. Trends in Ecology and the Environment 23: 424431.
  • Townsend AR, Cleveland CC, Asner GP, Bustamante MMC. 2007. Controls over foliar N:P ratios in tropical rain forests. Ecology 88: 107118.
  • Vitousek PM. 1984. Litterfall, nutrient cycling, and nutrient limitation in tropical forests. Ecology 65: 285298.
  • Vitousek PM, Sanford RL Jr. 1986. Nutrient cycling in moist tropical forest. Annual Review of Ecology and Systematics 17: 137167.
  • Wright IJ, Ackerly DD, Bongers F, Harms KE, Ibarra-Manriquez G, Martinez-Ramos M, Mazer SJ, Muller-Landau HC, Paz H, Pitman NCA et al. 2006. Relationships among ecologically important dimensions of plant trait variation in seven Neotropical forests. Annals of Botany 99: 10031015.
  • Wright IJ, Reich PB, Westoby M, Ackerly DD, Baruch Z, Bongers F, Cavender-Bares J, Chapin T, Cornelissen JHC, Diemer M et al. 2004. The worldwide leaf economics spectrum. Nature 428: 821827.
  • Zhang J, Rivard B, Sanchez-Azofeifa GA, Castro-Esau KL. 2006. Intra- and inter-class spectral variability of tropical tree species at La Selva, Costa Rica: implications for species identification using HYDICE imagery. Remote Sensing of Environment 105: 129141.

Supporting Information

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

Fig. S1 Coefficients of variation (CVs) within species for leaf chemical properties and leaf mass per unit area (LMA) by soil type.

Methods S1 Leaf collections and mobile processing, chemical assays, leaf and canopy spectroscopy, and statistical analyses.

Table S1 List of laboratory assays conducted to develop the chemical portfolio for each sample

Table S2 Descriptive statistics for canopy leaf samples collected from less common growth habits

Table S3 Adjusted r2 values relating leaf properties to taxonomic family, genus and species on Ultisols, Inceptisols, and the two substrates combined

Table S4 Pearson correlation coefficients among leaf properties of all individuals

Table S5 Pearson correlation coefficients among leaf properties of individuals located on Inceptisols

Table S6 Pearson correlation coefficients among leaf properties of individuals located on Ultisols

Table S7 Chemical variables associated with stepwise linear discriminant analysis with taxa containing two or more representatives

Please note: Wiley-Blackwell are not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing material) should be directed to the New Phytologist Central Office.

FilenameFormatSizeDescription
NPH_3549_sm_MethodsS1.doc77KSupporting info item
NPH_3549_sm_TablesS1-S7-FigS1.doc379KSupporting info item