Introduction

Versatile legume (Fabaceae) trees may be used for various purposes in tropical agroforestry systems (AFS) like shade trees for perennial crops, production of green manure and fuel wood in alley cropping, soil recovery in enriched fallow systems, fodder, or support trees for climbing crops (Nair 1993). The use of legume trees is motivated by the assumed benefit of symbiotic fixation of atmospheric dinitrogen by the trees for the whole AFS. Both traditional knowledge and research results indicate that the presence of legume trees has beneficial effects on soil fertility in many tropical AFS (Kass et al. 1997). These benefits include increased soil and microbial C and N content in comparison to annual cropping (Mazzarino et al. 1993; Sierra et al. 2002), increased residual soil N content after legume-enriched fallows in comparison to continuous cropping (Ståhl et al. 2002), and long-term accumulation of C and N in soil in simultaneous AFS (Dulormne et al. 2003; Haggar et al. 1993; Sierra and Nygren 2005). Some legume benefits like increased N mineralization and nitrification rate (Babbar and Zak 1994) may be caused by interaction of increased N availability from symbiotic N2 fixation and beneficial modification of microclimate below tree canopy.

The actual contribution of symbiotic N2 fixation to the observed benefits remains debatable. In a review, Mafongoya et al. (2004) cite a range of 14–92% of legume tree N originating from symbiotic N2 fixation, with large majority of values being between 40% and 80%. Nodulation and symbiotic N2 fixation rate depend heavily on environmental conditions (Salas et al. 2001) and tree management (Nygren and Ramírez 1995; Nygren et al. 2000). Further, estimating the N contribution by legume trees to N economy of an AFS under field conditions is a methodologically challenging task. It is nowadays generally accepted that the acetylene reduction assay (ARA) may only be used for showing if the plants fix N or not but not for quantification of the absolute N2 fixation rate (Hunt and Layzell 1993). Constructing a whole-system N balance may be a reliable method (Dulormne et al. 2003) but it requires several measurements on all system components over time, thus requiring a lot of human and financial resources, and long-term experiments.

Methods based on the relationships between the heavy, stable 15N isotope and the common 14N isotope are considered the best yet not perfect way to estimate the percentage of N fixed from atmosphere out of total plant N—an integrating estimate of N2 fixation capacity by legumes. The main problem of the 15N enrichment methods under field conditions is achieving uniform isotopic labelling of the soil both spatially and temporarily, and selecting a suitable non-N2-fixing reference species (Chalk and Ladha 1999). This problem is especially important in the case of trees that may scavenge nutrients from a horizontally extensive area and from several soil layers.

The 15N natural abundance method (Shearer and Kohl 1986) overcomes the problems of applying the 15N label but it is even more sensitive to selecting a proper reference species (Domenach 1995). Further, the 15N natural abundance in plants may vary for a number of factors other than N2 fixation. Soil biological activity alters the isotopic signature of soil (Högberg 1997); plant metabolism tends to discriminate against the heavy 15N isotope (Handley and Raven 1992) resulting in differences between plant organs and species; and different types of mycorrhizae have different effects on plant isotopic signature (Handley et al. 1999; Högberg 1997). Plant metabolism and mycorrhizae affect especially the isotopic signature of non-N2-fixing plants but may also alter the isotopic relationships in N2-fixing plants (Roggy et al. 1999; Wheeler et al. 2000). It is recommended that the reference species should grow close to the N2-fixing plants (Domenach 1995; Roggy et al. 1999). Direct transfer of N fixed from atmosphere from the legume trees to associated crops has recently been observed in some AFS (Sierra and Nygren 2006; Snoeck et al. 2000). Thus, the isotopic signature of the non-N2-fixing reference near the N2-fixer may have been directly affected by the N transfer from the latter.

Even considering the constraints listed above, the 15N natural abundance method may be the most practical way to obtain field estimates on N2 fixation in AFS, if conducted carefully, because it does not require expensive 15N labelled fertilizers and is not sensitive to short-term dilution of the 15N label in the soil (Boddey et al. 2000). In this contribution, we report results of 15N natural abundance sampling in two cacao (Theobroma cacao) plantations with legume tree (Inga edulis) and mixed-species shade canopy under humid tropical conditions. Reference sampling was also conducted in a tree hedgerow at one side of the cacao plantations. The objectives of the study were to (i) measure the 15N natural abundance in two legume and several non-N2-fixing tree species in the area during the driest and rainiest season of a year; (ii) identify suitable non-N2-fixing reference tree species for estimating the percentage of N fixed from atmosphere out of total N (%Nf) in the N2-fixers; (iii) estimate the %Nf in the assumed N2-fixers; and (iv) evaluate possible evidence of direct N transfer from the legume trees to non-N2-fixing species.

Materials and methods

Field sampling

The study was conducted in two cacao plantations located near to each other in the EARTH University academic farm located in the Caribbean coastal plain of Costa Rica (10°10′ N, 83°37′ W, 64 m a.s.l.), and a tree hedgerow that separates the cacao plantations from an adjacent pasture (Fig. 1). The climatic zone is premontane wet forest basal belt transition (Bolaños and Watson 1993). Average annual rainfall in 1996–2005 was 3,713 mm (Fig. 2). The study year 2006 was somewhat drier than the average with a rainfall of 2,911 mm. The rainfall is quite evenly distributed throughout a year with relatively less rain in March and September. Monthly variation in the rainiest months is higher than in the driest months. The high rainfall and variation in May are strongly affected by three very rainy years (1997, 2002 and 2004) when the May rainfall was ca. 1,000 mm. The annual mean temperature is 25.1°C with very little monthly variation. The soil of the study area is an andic humitropept (Sancho et al. 1989).

Fig. 1
figure 1

Map of the study site indicating the relative positions of cacao Plantations I (with Inga edulis as shade tree) and II (with several shade tree species), and the adjacent tree hedgerow separating the cacao plantations from a pasture. Approximate positions of Theobroma cacao, Gliricidia sepium, and Hieronyma alchorneoides samplings in Plantation II are shown. Samplings in Plantation I and Hedgerow were distributed throughout the whole area. See Table 1 for a complete listing of tree species sampled. The inset shows the geographic position of the study area

Full size image
Fig. 2
figure 2

Average monthly rainfall in the humid tropical study site in Guácimo, Costa Rica, from 1996 to 2005 and during the sampling year 2006. Error bars for the monthly means indicate standard deviation

Full size image

One of the cacao plantations (hereafter Plantation I; Fig. 1) was established in 1992 using various T. cacao clones. It has been cultivated organically since 1997. The plantation size is about 60 × 60 m with T. cacao planted in 3 × 3-m spacing. A few scattered Cordia alliodora trees were left on the site while establishing the plantation. Inga edulis was planted as the main shade tree in 9 × 9-m spacing. Each I. edulis replaced a T. cacao. By October 2005, the I. edulis canopy had become closed and too dense for good cacao production, and the shade trees were thinned to approximate 18 × 18-m spacing, leaving all C. alliodoras untouched. Theobroma cacao trees were pruned in December 2005, after the major annual harvest of cacao pods. All organic fertilisation was discontinued in 2005 because the isotopic signature of the fertilizer could not be controlled between preparations of different fertilizer lots.

The other cacao plantation (hereafter Plantation II; Fig. 1) was established in 1995 using matina cultivar of T. cacao and industrial fertilizers were applied. The original plantation was partially destroyed by escaped cattle and it was renewed using various T. cacao clones in 2002. The clonal trees were planted between the old T. cacao rows. The plantation has been managed organically since re-establishment. Old T. cacao trees that survived the browsing and various planted banana varieties served as temporal shade, and Peach palm (Bactris gasipaes Kunth) was planted as permanent shade. At the time of 15N natural abundance sampling, the palms did not provide any significant shade. Plantation II also contained a mixture of scattered, naturally-established shade trees (Table 1). The 15N natural abundance samples were collected from the old (matina) T. cacao trees that were situated at least 15 m from the nearest N2-fixing shade tree. All organic fertilization was discontinued in 2005.

Table 1 Tree species studied and location of sample trees in the humid tropical study site in Guácimo, Costa Rica
Full size table

Two N2-fixing and five non-N2-fixing tree species were sampled (Table 1). Theobroma cacaos from Plantations I and II were dealt as separate pseudo-species in all analyses. Plant organs sampled were:

  • Young leaves: youngest leaf that had reached full size but still had lighter green colour than mature leaves. Depending on species, these leaves were 1–2 weeks old.

  • Mature leaves: fully developed, deep-green leaves from lower position of a branch, without marked insect or disease damage.

  • Branches: small woody branches that carried leaves.

  • Stem: a core sample taken with an increment borer to stem centre at breast height (1.3 m; top of stem in T. cacao).

  • Coarse roots: a sample from a large structural root at 1–2 m from root collar. Depending on tree size, the sample was taken as a cylinder of whole root or as a core sample passing the whole root diameter.

The sampling was conducted in March and July. These months correspond to the driest and rainiest period of a year in the study site (Fig. 2). Cordia alliodora did not have any young leaves in March; otherwise all sample types could be collected from all species in both March and July.

15N natural abundance analyses and calculations

The N isotopic analyses were conducted by isotope ratio mass spectrometry (IRMS) in the Dating Laboratory at the University of Helsinki. The samples were weighted in tin capsules, which were fed by an auto sampler in He–O2 flow to an element analyzer (Carlo Erba NC2500, Italy) for volatilisation. The volatilised nitrogen oxides were reduced to N2 by Cu. After reduction, the samples were directed through CO2 and H2O traps and a ConFlo III interface (Thermo Scientific, Bremen, Germany) to the IRMS (Delta Advantages, Thermo Scientific). The laboratory references used in the analyses were calibrated against international references IAEA-N-1 (δ15N = 0.4 ± 0.2‰) and IAEA-N-2 (δ15N = 20.3 ± 0.2‰). Total N content of the samples was determined in the Department of Forest Ecology at the University of Helsinki by dry combustion (Leco CNS-1000, Leco Corp., St. Joseph, MI, USA).

The deviation of the sample 15N proportion from that of atmosphere (δ15N) was calculated (Shearer and Kohl 1986):

$$ \delta ^{15} {\text{N}} = \frac{{{{{}^{15}{\text{N}}} \mathord{\left/ {\vphantom {{{}^{15}{\text{N}}} {^{14} {\text{N}}_{\text{s}} - {{{}^{15}{\text{N}}} \mathord{\left/ {\vphantom {{{}^{15}{\text{N}}} {{}^{14}{\text{N}}_{{\text{atm}}} }}} \right. \kern-\nulldelimiterspace} {{}^{14}{\text{N}}_{{\text{atm}}} }}}}} \right. \kern-\nulldelimiterspace} {^{14} {\text{N}}_{\text{s}} - {{{}^{15}{\text{N}}} \mathord{\left/ {\vphantom {{{}^{15}{\text{N}}} {{}^{14}{\text{N}}_{{\text{atm}}} }}} \right. \kern-\nulldelimiterspace} {{}^{14}{\text{N}}_{{\text{atm}}} }}}}}}{{{{{}^{15}{\text{N}}} \mathord{\left/ {\vphantom {{{}^{15}{\text{N}}} {{}^{14}{\text{N}}_{{\text{atm}}} }}} \right. \kern-\nulldelimiterspace} {{}^{14}{\text{N}}_{{\text{atm}}} }}}} \times 1000\permille $$
(1)

where 15N/14Ns and 15N/14Natm are 15N/14N ratios in the sample and atmosphere, respectively. The percentage of N fixed from atmosphere out of total N in the sample (%Nf) was calculated (Shearer and Kohl 1986):

$$ \% {\text{N}}_{\text{f}} = \frac{{\delta ^{15} {\text{N}}_{\text{r}} - \delta ^{15} {\text{N}}_{\text{f}} }}{{\delta ^{15} {\text{N}}_{\text{r}} - \delta ^{15} {\text{N}}_0 }} \times 100\% $$
(2)

where subscript f refers to the N2-fixing species growing in the field, r to a non-N2-fixing reference species growing in the same field as the N2-fixer, and 0 to the N2-fixing species growing in a N-free medium. In practice, the δ15N0 must be determined from potted plants under controlled conditions. We performed the %Nf calculations using mean δ15N values for each component of Eq. 2. In that case, the standard error of the %Nf estimate was calculated (Shearer and Kohl 1986):

$$ \% S_{\text{f}} = \left[ {\frac{{\left( {\delta ^{15} {\text{N}}_{\text{f}} - \delta ^{15} {\text{N}}_0 } \right)^2 }}{{\left( {\delta ^{15} {\text{N}}_{\text{r}} - \delta ^{15} {\text{N}}_0 } \right)^4 }} \times S_{\text{r}}^ 2+ \frac{{S_{\text{f}}^2 }}{{\left( {\delta ^{15} {\text{N}}_{\text{r}} - \delta ^{15} {\text{N}}_0 } \right)^2 }} + \frac{{\left( {\delta ^{15} {\text{N}}_{\text{r}} - \delta ^{15} {\text{N}}_{\text{f}} } \right)^2 }}{{\left( {\delta ^{15} {\text{N}}_{\text{r}} - \delta ^{15} {\text{N}}_0 } \right)^4 }} \times S_0^2 } \right]^{\frac{1}{2}} $$
(3)

whereS r, S f, and S 0 refer to the standard error of δ15Nr, δ15Nf, and δ15N0, respectively.

A preliminary estimate on the possible transfer of N fixed by I. edulis to T. cacao in Plantation I was calculated by modifying Eq. 2 according to Snoeck et al. (2000): δ15Nr denoted to T. cacao without contact with N2-fixers (Plantation II) and δ15Nf denoted to T. cacao in contact with N2-fixers (Plantation I). This modification of Eq. 2 results in the percentage of N of atmospheric origin out of total N (%Na) in the receiver plant (Sierra and Nygren 2006), not the percentage of all N potentially transferred from I. edulis to T. cacao.

Inga edulis in N-free medium

Because no information on the δ15N0 value for I. edulis was found in the literature, it was determined from potted seedlings grown in N-free medium. Seeds of I. edulis were collected from trees at the EARTH University campus. After removing fresh seed pulp, the seeds were immersed for 12 h in water containing macerated root nodules taken from mature I. edulis trees for inoculating the seeds with N2-fixing bacteria. The inoculated seeds were planted in 5 l plastic pots of autoclaved vermiculite mixed with N-free fertilizer. The seedlings were grown in a greenhouse for 3 month.

At the time of harvest, the seedlings were divided into leaves, woody above-ground parts, coarse roots (diameter ≥2 mm), fine roots (<2 mm), and nodules. Only leaves without damage or senescence indications were used for the isotopic analyses. The above-ground woody parts included stem and woody branches because they did not differ in the seedlings. Green soft tissue of branches was excluded from the isotopic analyses because the respective branch parts were not sampled in the field, where they formed a negligible proportion of total branch biomass. The samples were analysed by IRMS in the Dating Laboratory at the University of Helsinki.

Results

15N natural abundance

Rainfall during the 2 weeks preceding the March sampling was 67 mm and during 4 weeks 97 mm, which is relatively little in the humid tropical study site (cf. Fig. 2). Rainfall during 2 and 4 weeks prior to the July sampling was 259 and 530 mm, respectively. These numbers indicate that the March and July sampling corresponded to the driest and rainiest season of the year, respectively.

The δ15N values varied by species and organ sampled (Fig. 3; Table 2). Gliricidia sepium had the lowest δ15N value, typical for a N2-fixer, in all organs and both seasons. Inga edulis had the second lowest δ15N value in stem and coarse roots during both seasons but δ15N values in young and mature leaves were high like in non-N2-fixing species. The relative differences between organs within a species appeared to be similar in March and July except in H. alchorneoides (Fig. 3). Hieronyma alchorneoides had lower δ15N values than the other non-N2-fixers. Student’s t-test performed by species and organ for testing the differences in δ15N values between seasons indicated that statistically significant differences occurred in woody tissue only in stem and coarse roots of H. alchorneoides and T. cacao in Plantation I (both lower in March). Statistically significant differences between seasons were detected in young leaves of G. sepium and H. alchorneoides (higher in March); young leaves of I. edulis and T. cacao in Plantation I (lower in March); and in mature leaves of G. sepium and C. alliodora (higher in March).

Fig. 3
figure 3

Mean δ15N values in different organs of trees of two cacao plantations and adjacent hedgerow fence under humid tropical conditions in Guácimo, Costa Rica, during the driest (a; March) and rainiest (b; August) period of a year

Full size image
Table 2 Mean of δ15N values in the sampled organs of trees of two cacao plantations and adjacent hedgerow fence under humid tropical conditions in Guácimo, Costa Rica
Full size table

Because of the relatively few cases of season differences observed, the data for both seasons were pooled for testing the significance of the differences in δ15N values between species within an organ and between organs within a species using analysis of variance followed by Duncan’s Multiple Range Test (Table 2). Statistically significant differences in δ15N values between young and mature leaves were detected only in C. alliodora but it should be noted that this species did not have young leaves in March due to natural phenological variation. In general, all species had higher δ15N in leaves than in woody tissue. Gliricidia sepium had the significantly lowest δ15N in all organs. Inga edulis had the second lowest δ15N in stem and coarse roots but it did not separate from non-legumes according to the δ15N in leaves and branches. Hieronyma alchorneoides had significantly lower δ15N than the other non-legume species in all organs but higher than in G. sepium (Table 2).

The δ15N in different organs of I. edulis seedlings grown in N-free medium (Table 3) followed the same pattern as in the mature field-grown trees: the leaf sample, which corresponded to a mixture of young and mature leaves from the field, had a relatively high δ15N, followed by coarse roots and stem (cf. Fig. 3). However, all these δ15N values in N-free medium were negative, typical to a N2-fixing tree. Fine roots that were not sampled in the field had a δ15N, which did not differ significantly from leaf value, and nodules had the highest δ15N. The nodule sample of the smallest seedling that had δ15N 1.9‰-units higher than the second highest (6.48 vs. 4.58) was excluded from the mean in Table 3.

Table 3 Mean ± SE of δ15N values in Inga edulis seedlings grown for 3 month in N-free vermiculite
Full size table

When the δ15N in mature leaves and stem was plotted against total N concentration in the same organ and season (Fig. 4), three species—I. edulis, G. sepium, and C. alliodora—formed a group with high total N concentration but varying δ15N. Gliricidia sepium had the lowest δ15N in all cases and C. alliodora the highest, while δ15N of I. edulis was intermediate in mature leaves and close to that of G. sepium in stem. The other five species seemed to be quite similar but H. alchorneoides had the significantly lowest δ15N of this group, except in stem in July (Duncan’s MRT at 5%). The δ15N of C. alliodora did not differ significantly from other non-N2-fixing species than H. alchorneoides. Following the criteria proposed by Roggy et al. (1999), I. edulis would be classified as a non-N2-fixer according to mature leaf characteristics but a N2-fixer according to stem characteristics. All non-legumes appeared to be non-N2-fixers.

Fig. 4
figure 4

Mean δ15N value plotted against mean total N concentration in mature leaves during the driest month, March (a), and the rainiest month, July (c), and stem in March (b) and July (d) of trees in two cacao plantations and adjacent hedgerow fence under humid tropical conditions in Guácimo, Costa Rica. The error bars indicate standard error of mean

Full size image

Estimation of N2 fixation by I. edulis and G. sepium and transfer of fixed N to T. cacao

Estimates of %Nf were calculated for I. edulis and G. sepium based on mature leaf and stem δ15N values. Based on Fig. 4, a non-N2-fixing reference group was formed using the average δ15N of C. alliodora, T. cacao in Plantations I and II, R. pittieri, and P. latifolia for %Nf estimations. Hieronyma alchorneoides was excluded from %Nf estimations because its low δ15N values may indicate that it uses a different soil N pool than the other non-legumes (Roggy et al. 1999). Thus, it may be an unsuitable reference species for estimating %Nf in the legumes (Domenach 1995; Högberg 1997; Shearer and Kohl 1986). The %Nf estimates for I. edulis were calculated using this group and the non-N2-fixers of Plantation I, C. alliodora and T. cacao separately as reference (Table 4). The %Nf estimates of I. edulis based on stem δ15N in July were almost the same using any of the references, while C. alliodora produced higher estimates than T. cacao or the non-N2-fixing group in March. The %Nf estimates based on mature leaves were low, except in March using C. alliodora as the reference species. However, even in this case, the %Nf estimate based on mature leaves was about half of the estimate based on stem δ15N (Table 4). The stem-based %Nf estimates suggest active N2 fixation by I. edulis while leaf-based estimates suggest that it would not fix N2.

Table 4 Mean (±SE) percentage of N fixed from atmosphere out of total N in Inga edulis growing as shade tree in the cacao Plantation I calculated on the basis of δ15N values in mature leaves and stem
Full size table

Estimates of %Nf were calculated for G. sepium using T. cacao growing in the same Plantation (II) and the non-N2-fixing group as reference. The δ15N0 in leaves of G. sepium, −2.07 (Nygren et al. 2000), was also used for stem-based estimates because no information on δ15N0 for other organs was available and between-organ differences in δ15N of G. sepium appeared to be small (Fig. 3). The results based on both references and organs were quite similar (Table 5). The leaf-based %Nf estimate was about 10%-units higher in July than in March against both references while season differences were small in stem-based estimates. All %Nf values clearly indicated N2 fixation (Table 5).

Table 5 Mean (±SE) percentage of N fixed from atmosphere out of total N in Gliricidia sepium in the cacao Plantation II calculated on the basis of δ15N values in mature leaves and stem
Full size table

The effect of organ used for %Nf estimates was studied by calculating the %Nf estimate of G. sepium against the non-N2-fixing reference group for all organs separately (Fig. 5). Most organs produced roughly similar %Nf (ca. 60–65% with extremes of 56% and 74%). The estimates in July were higher for both leaf types and coarse roots and lower for branch- and stem-based calculations (Fig. 5). The δ15N of the reference group was about the same in March and July in all organs, and the seasonal differences in %Nf were mainly caused by variation of δ15N in G. sepium.

Fig. 5
figure 5

Mean percentage of N fixed from atmosphere out of total N in different organs of Gliricidia sepium growing as a shade tree in a cacao plantation with mixed-species shade stratum (Plantation II) under humid tropical conditions in Guácimo, Costa Rica. The mean δ15N of four non-fixing tree species (Cordia alliodora, Rollinia pittieri, Posoqueria latifolia, and Theobroma cacao) was used as the reference value for each organ. The error bars indicate standard error of mean

Full size image

Because T. cacao had significantly lower δ15N values in Plantation I with I. edulis shade trees than in Plantation II without contact with legume trees (Table 2) preliminary estimates on the possible transfer of N fixed by I. edulis to T. cacao in Plantation I were calculated using δ15N in mature leaves and stem (Table 6). The leaf-based estimates indicated potential transfer of N fixed by I. edulis to T. cacao in both March and July. Stem-based estimates suggested that N transfer would not occur.

Table 6 Mean (±SE) percentage of N of atmospheric origin out of total N in Theobroma cacao growing below Inga edulis shade trees in the cacao Plantation I in March (the driest season) and July (the rainiest season)
Full size table

Discussion

Natural abundance of 15N, measured as δ15N, did not vary much between the non-N2-fixing trees C. alliodora, P. latifolia, R. pittieri, and T. cacao (Fig. 3; Table 2). The δ15N values were typical to tropical forests (Martinelli et al. 1999). Differences between organs followed the same pattern in this group, i.e. the highest δ15N in leaves and slightly lower values in woody tissue. The δ15N in all organs of G. sepium was typical for a N2-fixing tree (Roggy et al. 1999; Yoneyama et al. 1993). The ratio of total N concentration to δ15N in P. latifolia, R. pittieri, and T. cacao (Fig. 4) was also typical to non-N2-fixing trees and the relationship in G. sepium was typical to a N2-fixer (Roggy et al. 1999).

Two of the non-legume trees differed from the general pattern of the non-N2-fixing group: C. alliodora had a high leaf total N concentration (Fig. 4) and H. alchorneoides had significantly lower δ15N than the other non-N2-fixing trees (Table 2). Cordia alliodora also had higher yet not significantly different δ15N than the other non-N2-fixers. Differences in δ15N may reflect variations between N sources available for different species and/or distinct mycorrhizal symbionts (Boddey et al. 2000; Handley et al. 1999; Högberg 1997). Nitrate reduction discriminates against 15N (Handley and Raven 1992) causing 15N depletion in plants depending heavily on NO3 − for N supply. Ectomycorrhizae tend to deplete host plants of 15N (Högberg 1997) while arbuscular mycorrhizae (AM) tend to slightly enrich their hosts by 15N (Handley et al. 1999).

Roots of I. edulis and T. cacao were colonised by AM in the study site (Iglesias et al. 2007). Soil fungal community in C. alliodora and H. alchorneoides plantations in Sarapiquí, ca. 50 km NW from the study site under similar climatic conditions, imply AM symbiosis although actual root colonisation was not analysed (Lovelock and Ewel 2005). Gliricidia sepium forms AM symbiosis (Okon et al. 1996). No information on mycorrhizal symbioses of any Rollinia or Posoqueria sp. could be found. Thus, mycorrhizal differences probably do not explain the observed lower δ15N in H. alchorneoides.

Both I. edulis and T. cacao had superficial rooting pattern in the study site (Nygren et al. 2007) and G. sepium seems to root superficially under humid tropical conditions (Rowe et al. 2001; Salas et al. 2004). Root system of C. alliodora was observed to be superficial in Sarapiquí in a plantation with abundant understorey vegetation while root system of H. alchorneoides was deeper (Haggar and Ewel 1997). Deep rooting pattern of H. alchorneoides was also observed in a natural forest and in a pure plantation in Sarapiquí (Arnáez and Moreira 2006). Thus, we assume that H. alchorneoides had access to a different soil N pool than the other studied species, which affected its δ15N. This hypothesis could not be fully verified because information was not available on rooting patterns of all species but it was considered strong enough to exclude H. alchorneoides from the non-N2-fixing reference group used for estimating %Nf. Cordia alliodora was considered a valid reference species because it appeared to form AM symbiosis and have similar rooting pattern as the legume trees I. edulis and G. sepium.

The legume tree I. edulis had a strange δ15N profile with high values, typical to non-N2-fixing trees, in leaves and branches and almost as low δ15N as G. sepium in stem and coarse roots (Fig. 3). High foliar δ15N of I. edulis has also been reported earlier (Roggy et al. 1999; Yoneyama et al. 1993). In French Guiana, I. edulis with high foliar δ15N was nodulated and the nodules showed nitrogenase activity in the acetylene reduction assay (Roggy et al. 1999). Inga edulis has also been shown to fix N2 using 15N enrichment method under greenhouse (Leblanc et al. 2005) and semi-controlled field conditions close to our study site (Leblanc et al. 2007); however, all studied individuals did not fix N2 in the field. The survival of I. edulis in the N-free medium of this study also indicated that it is a N2-fixer (Table 3).

Thus, I. edulis may be considered a confirmed N2-fixer and its δ15N profile is probably caused by dissimilar 15N fractionation processes in different organs. The δ15N between organs of the seedlings grown in N-free medium followed the same pattern as in the Plantation I (Table 3; Fig. 3), i.e. the lowest values in stem and coarse roots. The same profile has also observed in other N2-fixing trees (reviewed by Boddey et al. 2000) and in both G. sepium and non-N2-fixing trees in this study (Fig. 3). Nodules had the highest δ15N in the I. edulis seedlings grown in N-free medium (Table 3), which seems to be a typical feature of N2-fixing trees (Boddey et al. 2000). Although the tendency of lower δ15N in stems than leaves was observed in other N2-fixing and non-N2-fixing trees, the difference between leaves and woody tissue in I. edulis was the most pronounced. In the 15N enrichment experiment, stem of I. edulis had lower 15N atom excess than leaves but the differences were proportionally equal to differences observed in the non-N2-fixing reference tree Vochysia guatemalensis Donn. Sm. (Leblanc et al. 2007) indicating that strong isotopic enrichment of the growth medium surpasses the effects of fractionating processes. High foliar δ15N was also observed in some other, but not all, nodulating Inga spp. (Koponen et al. 2003; Roggy et al. 1999), which suggests that this characteristic may be associated to some subgroups of the large genus.

Theobroma cacao had significantly lower δ15N in Plantation I with I. edulis shade trees than in Plantation II without contact with legume trees (Table 2; Fig. 3). The estimation of %Na using the modified Eq. 2 is based on the assumption that the isotopic signature of the N fixed from atmosphere by I. edulis does not change if it is directly transferred to T. cacao, e.g. via common mycorrhizal network or root exudates of I. edulis absorbed by T. cacao (He et al. 2003). The isotopic signature of N mineralised from decomposing leaves and fine roots is strongly altered by microbiological processes of the soil (Boddey et al. 2000; Handley and Raven 1992; Högberg 1997). Thus, N released by I. edulis to soil through the complete mineralization process and reabsorbed by T. cacao would have the general soil isotopic signature. Leaf-based estimates of %Na in T. cacao in Plantation I suggested small but significant direct N transfer from I. edulis in both seasons (Table 6). However, the results are still inconclusive because of differences between seasons and organs. Direct N transfer via common mycorrhizal networks in the studied plantation may be envisioned because both species were colonised by the same AM morphospecies (Iglesias et al. 2007). Thus, the question requires further research, in which the apparent 15N fractionation processes will be carefully considered.

Statistically significant differences in δ15N between seasons were observed mainly in young leaves (Fig. 3); higher values were measured in July for I. edulis and T. cacao in Plantation I. Opposite was observed in G. sepium and H. alchorneoides. This may reflect variation in plant-available N pools between the driest and rainiest season of a year and higher N2 fixation rate of G. sepium under humid conditions. Leaf δ15N tends to respond most rapidly to these changes (Domenach 1995). However, significant differences in stems and coarse roots of T. cacao in Plantation I and H. alchorneoides were somewhat unexpected because the δ15N in these organs integrates effects of long-term changes in the δ15N of N sources. The stem of cultivated T. cacao is relatively small—most above-ground woody tissue is in morphological branches—and the H. alchorneoides individuals sampled were quite young. Thus, it is possible that the small stem of T. cacao reacts rapidly to changes in the available N sources. The stems of H. alchorneoides were almost entirely of sapwood, and it may be envisioned that large stems are stabilised by the inactive heartwood rather than the physiologically active sapwood.

The estimates of %Nf for I. edulis calculated on the basis of stem δ15N (Table 4) were about the same as the estimates based on 15N enrichment of growth medium, 57% (Leblanc et al. 2007). Leaf δ15N did not seem to be a reliable basis for estimating %Nf in I. edulis when the 15N natural abundance method was used. The stem-based estimates were quite similar with various reference trees. Leaf- and stem-based estimates of %Nf for G. sepium were close to each other’s independently of reference plant used but seasonal variation appeared to be higher in leaf-based estimates (Table 5). It may be assumed that leaf-based estimates better reflect seasonal variation in N2 fixation because leaf δ15N tends to respond more rapidly to changes in N2 fixation than other organs (Domenach 1995); the δ15N in stem integrates effects of long-term N2 fixation. Estimates of %Nf calculated on the basis of different organs of G. sepium were quite close to others (Fig. 5), which indicates that relatively reliable %Nf estimates can be made for this species by collecting leaves only. The same was observed in 15N enrichment-based estimates for I. edulis and two Erythrina L. spp. (Leblanc et al. 2007). The %Nf estimates for G. sepium were close to values found in other studies (e.g. Nygren et al. 2000; Peoples et al. 1996).

Concluding remarks

General mean of δ15N in the non-N2-fixing trees in the studied cacao plantations, 3.67‰, was very close to the average value of various tropical forests, 3.7‰ (Martinelli et al. 1999). Little variation was observed between the non-N2-fixing trees C. alliodora, P. latifolia, R. pittieri, and T. cacao. Variation between organs within a species was more pronounced than general variation between these species. Hieronyma alchorneoides had significantly lower δ15N than the other non-N2-fixers, which probably reflected use of a different soil N pool; it had the deepest rooting pattern of the species, for which information was available. Mycorrhizal differences were assumed to be unimportant in the study site; all species, for which information was available, formed AM symbiosis. The %Nf of G. sepium varied between 56% and 74%, depending on season and organ used for estimation. It may be assumed that coarse roots and stem δ15N of I. edulis reflected the ratio of N2 fixation to soil N uptake and the high leaf values were altered by fractionation due to plant metabolism. Some transfer of fixed N from I. edulis to T. cacao may occur but results so far are inconclusive. Further research on N transfer in this AFS should carefully consider the effects of 15N fractionation within trees. The δ15N appeared to vary according to ecophysiological characteristics of the trees in the cacao plantations.