Nutrient Cycling in Agroecosystems

, Volume 94, Issue 2, pp 123–160

Symbiotic dinitrogen fixation by trees: an underestimated resource in agroforestry systems?

Authors

    • Department of Forest Sciences
    • Finnish Society of Forest Science
  • María P. Fernández
    • Ecologie Microbienne, UMR5557, USC 1193Université Lyon1
  • Jean-Michel Harmand
    • CIRAD, UMR Eco&Sols
  • Humberto A. Leblanc
    • EARTH University
Review Article

DOI: 10.1007/s10705-012-9542-9

Cite this article as:
Nygren, P., Fernández, M.P., Harmand, J. et al. Nutr Cycl Agroecosyst (2012) 94: 123. doi:10.1007/s10705-012-9542-9
  • 858 Views

Abstract

We compiled quantitative estimates on symbiotic N2 fixation by trees in agroforestry systems (AFS) in order to evaluate the critical environmental and management factors that affect the benefit from N2 fixation to system N economy. The so-called “N2-fixing tree” is a tripartite symbiotic system composed of the plant, N2-fixing bacteria, and mycorrhizae-forming fungi. Almost 100 recognised rhizobial species associated with legumes do not form an evolutionary homologous clade and are functionally diverse. The global bacterial diversity is still unknown. Actinorrhizal symbioses in AFS remain almost unstudied. Dinitrogen fixation in AFS should be quantified using N isotopic methods or long-term system N balances. The general average ± standard deviation of tree dependency on N2 fixation (%Ndfa) in 38 cases using N isotopic analyses was 59 ± 16.6 %. Under humid and sub-humid conditions, the percentage was higher in young (69 ± 10.7 %) and periodically pruned trees (63 ± 11.8 %) than in free-growing trees (54 ± 11.7 %). High variability was observed in drylands (range 10–84 %) indicating need for careful species and provenance selection in these areas. Annual N2 fixation was the highest in improved fallow and protein bank systems, 300–650 kg [N] ha−1. General average for 16 very variable AFS was 246 kg [N] ha−1, which is enough for fulfilling crop N needs for sustained or increasing yield in low-input agriculture and reducing N-fertiliser use in large-scale agribusiness. Leaf litter and green mulch applications release N slowly to the soil and mostly benefit the crop through long-term soil improvement. Root and nodule turnover and N rhizodeposition from N2-fixing trees are sources of easily available N for the crop yet they have been largely ignored in agroforestry research. There is also increasing evidence on direct N transfer from N2-fixing trees to crops, e.g. via common mycelial networks of mycorrhizal fungi or absorption of tree root exudates by the crop. Research on the below-ground tree-crop-microbia interactions is needed for fully understanding and managing N2 fixation in AFS.

Keywords

15NActinorrhizal treesLegume treesManagement practicesNitrogen balanceRhizobial symbiosis

Introduction

Nitrogen is the first plant growth-limiting factor after water in most ecosystems. Agroecosystems may be even more N limited than natural ecosystems because of heavy N export in crop harvest (Nair et al. 1999). The atmospheric N2 is the biggest pool of N in the world but only some prokaryotic microbes are able to reduce it, thus playing a key role in both terrestrial and marine ecosystems. Crops that form symbiosis with N2-fixing microbes, most notably legumes (Fabaceae super family) with certain α- and β-Proteobacteria (rhizobia), are an alternative to cope with N deficiencies in agroecosystems.

Legume crops and forages respond only a part of human needs of plant food and fibre and, thus, vast agricultural areas depend on industrially-fixed N fertilizers. These are often unavailable for small-scale farmers in the developing world and may cause environmental problems such as contamination of water sources. Agroforestry systems (AFS) with “N2-fixing trees”1 provide alternatives to alleviate these problems if managed properly. These systems are diverse including but not restricted to cultivation of cereals and other crops between rows of periodically-pruned trees in alley cropping (Akinnifesi et al. 2010; Kang et al. 1981; Rowe et al. 1999), shade trees with perennial crops (Beer et al. 1998; Soto-Pinto et al. 2010), improvement of fallow phase with N2-fixing trees (Chikowo et al. 2004; Harmand et al. 2004; Ståhl et al. 2002, 2005), living supports for climbing crops (Salas et al. 2001), and simultaneous cultivation of fodder trees and grass (Blair et al. 1990; Dulormne et al. 2003).

Many tree species used in AFS provide multiple products including fuelwood, fodder, or several non-timber forest products. Dinitrogen-fixing trees are often preferred in comparison to other multi-purpose species because of the assumed benefit to the whole system N balance. Thus, symbiotic N2 fixation in AFS was enthusiastically studied in the 1970s and 1980s but it almost disappeared from research agenda in the 1990s because of studies suggesting little benefit of N2 fixation to system level N balance in AFS (Fassbender 1987; Garrity and Mercado 1994; van Kessel and Roskoski 1981) or sustainability (Kass 1995). Recent research based on N isotopic relations in whole plant (Leblanc et al. 2007; Peoples et al. 1996; Ståhl et al. 2002, 2005) or compilation of whole AFS N balance (Dulormne et al. 2003) indicate that symbiotic N2 fixation may have been underestimated as a N source for AFS.

The N2-fixing symbiosis is regulated by both intrinsic and environmental factors. The intrinsic physiological and morphological factors form the basis of the functional plant groups as they result in differences in resource requirements, seasonality of growth, and life history (Tilman et al. 1997), i.e. responses to the environment and interactions with other organisms. In spite of the recent advances in biotechnology, our ability to manage the intrinsic factors of plants and N2-fixing bacteria are still quite limited. In practical agroforestry, these intrinsic factors may be taken into account only by selecting suitable tree species (Aronson et al. 2002) and combinations of trees and bacterial strains (Acosta-Duran and Martínez-Romero 2002; Bala and Giller 2001; Bala et al. 2003).

The macroenvironment is out of human control, in spite of activities such as development of vegetation-based C sequestration to mitigate the global climate change (Soto-Pinto et al. 2010). On the other hand, agroforestry offers a wide variety of tools for managing the microenvironment for better sustainability and productivity, starting with simple techniques like optimisation of tree spacing in alley cropping for best N benefit and minimal crop shading (Akinnifesi et al. 2008). It is also important to understand the effects of the agroforestry management on the N2-fixing symbiosis; e.g., the green pruning of trees practiced for increasing nutrient recycling and reducing crop shading may also disturb nodulation (Nygren and Ramírez 1995).

Although several reviews on N2 fixation in AFS have been published (Bryan 2000; Giller 2001; Kass et al. 1997; Khanna 1998; Mafongoya et al. 2004; Sanginga et al. 1995), they are mostly descriptive compilations of relevant data. Thus, critical analysis on issues such as the methods used in N2 fixation research within the agroforestry context, functional importance of N2-fixing symbiosis for the legumes and actinorrhizal plants, and effects of the different AFS management practices on symbiotic N2 fixation seems timely. Recent research (André et al. 2005; Cardoso and Kuyper 2006; Duponnois and Plenchette 2003; Ingleby et al. 2001; Lesueur and Sarr 2008) also indicates the importance of the tripartite symbiosis between plants, N2-fixing bacteria, and mycorrhizae-forming fungi on N2 fixation.

Our aim is to analyse symbiotic N2 fixation by agroforestry trees as a part of their functions and indicate the critical environmental and management aspects needed to benefit from N2 fixation at AFS level. We also evaluate the research methodologies and compile quantitative data on N2 fixation in AFS based on the most reliable methods only. The specific objectives of the review are: revise current knowledge on N2-fixing microbia relevant for understanding the functioning of AFS; evaluate the methods for estimating N2 fixation suitable for use in AFS; evaluate the published data on symbiotic N2 fixation in AFS in the light of current knowledge on ecophysiology of legumes and actinorrhizal plants; and evaluate the importance of symbiotic N2 fixation for AFS functions and productivity.

Biological N2 fixation and N2-fixing organisms

Dinitrogen fixation process

All N2-fixing Prokaryota reduce atmospheric N2 using the nitrogenase enzyme:
$$ {\text{N}}_{2} + 8{\text{H}}^{ + } + 8{\text{e}}^{ - } + 16{\text{ATP}}\xrightarrow{{{\text{nitrogenase}}}}2{\text{NH}}_{3} + {\text{H}}_{2} + 16{\text{ADP}} + 16{\text{P}}_{{\text{i}}} $$
(1)

Thus, reducing a N2 molecule to two NH3 molecules requires 8 protons (H+) and electrons (e) and it releases a hydrogen molecule (H2). The energy for N2 fixation is released by breaking 16 adenosinetriphosphate (ATP) molecules to adenosinediphosphate (ADP) and inorganic phosphate (Pi). Details of the N2 fixation process can be found in most textbooks of plant physiology (e.g. Taiz and Zeiger 2006) or soil microbiology (e.g. Paul and Clark 1996) and they will not be repeated here, except for issues needed for better understanding the discussion on symbiotic N2 fixation in AFS.

Dinitrogen-fixing organisms

Dinitrogen-fixers have been reported among most of the taxonomic divisions of Prokaryota and the methanogenic Archae, thus presenting large genetic and physiological diversity. The N2-fixing bacteria are currently divided into symbiotic and free-living N2-fixers according to their capacity to form mutualistic association with eukaryotic higher organisms, either plants or animals, or their saprophytic life as components of environmental microflora. Most symbiotic N2-fixing bacteria also have a saprophytic stage. However, some symbiotic strains that remain non-isolated could represent obligate symbionts or intermediate stages in evolution towards a greater symbiotic dependence.

Dinitrogen-fixers associated with plants comprise 3 main groups: (1) Cyanobacteria that establish ectosymbiosis (non-intracellular location) with a large diversity of fungi and various plant groups, including mostly bryophytes and cycads but very few higher plants; (2) α- and β-Proteobacteria that form symbiosis in specialised structures, nodules, within roots or stems of legumes and Parasponia spp.; and (3) Frankiaceae (Actinobacteria) that nodulate plants belonging to 25 genera within eight Angiosperm families. Plants forming symbiosis with Frankiaceae are collectively called actinorrhizal plants. They are mostly trees or shrubs. (Vessey et al. 2004).

Cyanobacteria and other free-living diazotrophs

Taxonomy of Cyanobacteria is problematic (Castenholz 2001) and recent molecular studies indicate considerable biodiversity. Among plants symbiotically associated with Cyanobacteria, only the aquatic ferns of the genus Azolla have economic importance in farming systems (Herridge et al. 2008) and none is important in AFS.

Endophytic free-living and associative diazotrophic bacteria within the rhizosphere of Gramineae may make substantial contributions to N balance of an agroecosystem. The best-studied case is probably sugar cane (Saccharum spp.; Baldani et al. 1997), which is never used in AFS as far as we know. Some other plants known to harbour N2-fixing endophytes such as forage grasses (Herridge et al. 2008), coffee (Coffea arabica L.; Fuentes-Ramírez et al. 2001; Jiménez-Salgado et al. 1997), and banana (Musa spp.; Martínez et al. 2003) are often used in AFS. As far as we know, the role of the free-living N2-fixing bacteria in crop N supply has not been estimated in any AFS.

Symbiotic N2-fixing organisms

The nodules within which the symbiotic N2 fixation occurs are a plant organ. Nodule growth is stimulated by the symbiotic N2-fixing bacteria, and it contains a bacteroid space where the actual N2 fixation occurs (Minchin 1997). The symbiotic bacteria survive as free-living in the soil but many fix N2 only in symbiosis with a host plant. The endophytic bacteria, called bacteroids, lose part of their cellular organelles and, thus, become completely dependent on the host plant. Synthesis and catalytic activity of the nitrogenase enzyme are inhibited by oxygen. Consequently, the bacteroid space is anaerobic, protected by an O2 diffusion barrier in the inner cortex of the legume nodules (Minchin 1997) or by specialised structures, such as vesicles of Frankia spp. The bacteroids are, however, aerobic and the transport compound leghemoglobin supplies O2 to the bacteroids without damaging the nitrogenase. Ammonia produced by symbiotic N2 fixation is assimilated to amino acids or ureides in the root nodules and transferred in this form to the plant’s vascular system (Vessey et al. 2004).

Dinitrogen-fixing bacteria in agroforestry systems

Legume symbioses

The bacterial species associated with legumes described so far are diverse and do not form an evolutionary homologous clade. They are classified into 13 genera belonging to two distinct phylogenetic branches where rhizobia (i.e. bacteria forming N2-fixing nodules with legumes) are intermingled with many non-symbiotic bacteria. Earlier, rhizobia were assumed to belong to five genera of α-Proteobacteria recognized as Azorhizobium, Bradyrhizobium, Rhizobium, Mesorhizobium, and Ensifer (formerly Sinorhizobium) (Young and Haukka 1996; Zakhia and de Lajudie 2001).

Besides these genera, several other α-Proteobacteria (Allorhizobium, Devosia,Methylobacterium, Ochrobactrum, Phyllobacterium, and Shinella spp.) and β-Proteobacteria (Burkholderia, Cupriavidus (ex. Waustersia ex. Ralstonia), and Herbaspirillum spp.) have been isolated from legume nodules (reviewed by Balachandar et al. 2007). Several of these species (Blastobacter, Burkholderia, Cupriavidus,Devosia, Methylobacterium, Ochrobactrum, Phyllobacterium, and Shinella spp.) were recently shown to be rhizobia by the presence of nod and nif genes, which encode nodulation and N2 fixation. Some bacteria found in nodules, such as Agrobacterium spp., were shown to be devoid of nodulation and N2 fixation genes. These strains often form mixed populations with nodulating rhizobial strains in the nodules (reviewed by Balachandar et al. 2007). An up-to-date list of rhizobial taxa with recommended nomenclature is maintained by the “ICSP Subcommittee on the taxonomy of Rhizobium and Agrobacterium” (http://edzna.ccg.unam.mx/rhizobial-taxonomy/node/4) or the New Zealand rhizobia website (http://www.rhizobia.co.nz/taxonomy/rhizobia.html).

In July 2012, the New Zealand rhizobia website recognised 98 rhizobial species. Nevertheless, the global diversity is still unknown because less than 30 % of the 728 genera and the ca. 19,325 species of the Fabaceae have been tested for nodulation (Sprent 2009). Although considerable effort has been made during the last 20 years to describe rhizobial diversity from unstudied wild legumes, the symbionts have been identified to species level only in a few studies.

Many of the recently described “new rhizobia” have been isolated from tropical legume trees. They seem to be a source of large microbial diversity, particularly among the fast-growing rhizobia. In an extensive numerical analysis of 115 phenotypic characteristics of ca. 130 rhizobial strains, 12 clusters composed of tree rhizobia were found among the 19 clusters obtained (Zhang et al. 1991). An important diversity is found in the diversification centres of the legumes, most of which are tropical (Lie et al. 1987); e.g. in the sub-Saharan African centre of diversity of legume trees (The Sudan, Ethiopia, and Kenya), a wide phylogenetic bacterial diversity was found in a relatively low number of host species (Nick et al. 1999; Odee et al. 1997, 2002; Wolde-Meskel et al. 2005; Zhang et al. 1991). These studies suggest that studying symbionts of unexplored wild legumes from new biogeographical areas will reveal additional diversity.

Before the modern phylogenetic techniques became widely available, rhizobia were commonly classified to fast- and slow-growing strains; the latter were later classified to the Bradyrhizobium genus. It seems that often a given tree species recognises preferentially either slow- or fast-growing symbionts (Turk and Keyser 1992; Wolde-Meskel et al. 2005); e.g., strains isolated from Acacia senegal (L.) Willd. and Prosopis chilensis (Molina) Stuntz in the Sudan were all fast-growing whereas strains from Acacia mangium Willd. in Thailand were slow-growing (Zhang et al. 1991). The phenomenon has not been widely studied but Wolde-Meskel et al. (2005) noted that in a few cases a tree species was nodulated with both fast- and slow-growing rhizobia. However, the classification to slow- or fast-growing strains may not be genetically relevant; e.g., fast-growing strains from Inga edulis Mart. were genetically associated with slow-growing bradyrhizobia (Leblanc et al. 2005).

Because there is no specific selective medium for isolating rhizobia directly from soil, most of the biodiversity studies have been conducted sampling nodules from field-grown plants or from trap plant experiments. The natural rhizobial populations obtained from field-collected nodules seem to be more diverse than the trapped ones (Liu et al. 2005; Wang et al. 1999, 2002a). Indeed, the latter approach only allows studying compatible nodulating strains under the specific conditions used and estimates depend mostly of what and how many species are used for trapping, thus resulting in an underestimation of the global biodiversity.

In addition to taxonomic diversity, rhizobia that nodulate legume trees are functionally very diverse, i.e. with respect to their ecological, physiological, and biochemical properties (Zhang et al. 1991). Particularly, their cross-nodulation patterns vary remarkably (Table 1). Fast-growers seem to be more specific in comparison to slow-growers that are generally found promiscuous. Each rhizobial species has a defined host range, varying from very narrow to very broad (Duhoux and Dommergues 1985; Graham and Hubbell 1975; Trinick 1982); e.g. Calliandra calothyrsus Meisn. (sub-family Mimosoideae, tribe Ingeae), Leucaena leucocephala (Lam.) de Wit (Mimosoideae: Mimoseae), and Gliricidia sepium (Jacq.) Kunth ex Walp. (Papilionoideae: Robineae) were able to nodulate in most of the soils tested indicating that their symbiotic partners are widely distributed (Bala and Giller 2001). These phylogenetically diverse hosts also shared their symbionts, because same rhizobial species were isolated from their nodules (Bala and Giller 2001; Moreira et al. 1998; Oyaizu et al. 1993). Sesbania sesban (L.) Merr. (Papilionoideae: Sesbanieae) is a common species in AFS that failed to nodulate in most of the soils tested suggesting a higher symbiotic specificity (Bala and Giller 2001).
Table 1

Symbiotic N2-fixing bacteria found with legume and actinorrhizal trees commonly used in agroforestry systems

Host plant genus

Symbionts

References

Legume tree genera

 Acacia spp.

Sinorhizobium fredii

S. terangae

S. saheli

S. kostiense

S. americanum

S. arboris

Ensifer mexicanus

Allorhizobium undicola

Mesorhizobium plurifarium

M. huakuii

Bradyrhizobium sp.

Wolde-Meskel et al. (2005)

Boivin and Giraud (1999)

Lortet et al. (1996)

Nick et al. (1999)

Toledo et al. (2003)

Nick et al. (1999)

Lloret et al. (2007)

De Lajudie et al. (1998)

De Lajudie et al. (1998)

Sprent (2009)

Dupuy et al. (1994)

 Albizia spp.

M. albiziae

Wang FQ et al. (2007)

 Calliandra spp.

Rhizobium tropici

R. gallicum

R. mongolense

Martínez-Romero et al. (1991)

Zurdo-Piñeiro et al. (2004)

Wolde-Meskel et al. (2005)

 Erythrina spp.

Rhizobium sp.

B. liaoningense

Zhang et al. (1991)

Wolde-Meskel et al. (2005)

 Faidherbia spp.

Allorhizobium undicola

Ochrobactrum sp.

B. elkanii

De Lajudie et al. (1998)

Ngom et al. (2004)

Wolde-Meskel et al. (2005)

 Gliricidia spp.

R. tropici

R. etli

Sinorhizobium sp.

Hernández-Lucas et al. (1995)

Hernández-Lucas et al. (1995)

Acosta-Durán and Martínez-Romero (2002)

Inga spp.

B. japonicum

B. liaoningense

Leblanc et al. (2005)

Leblanc et al. (2005)

 Leucena spp.

R. tropici

R. etli biovar phaseoli

R. gallicum

R. giardinii

S. morelense

M. plurifarium

M. albiziae

Martínez-Romero et al. (1991)

Segovia et al. (1993)

Hernández-Lucas et al. (1995)

Amarger et al. (1997)

Wang ET et al. (2002b)

De Lajudie et al. (1998)

Wang FQ et al. (2007)

 Prosopis spp.

M. plurifarium

M. chacoense

S. kostiense

S. arboris

De Lajudie et al. (1998)

Velásquez et al. (2001)

Sprent (2009)

Sprent (2009)

 Robinia pseudoacacia

R. multihospitium

Han et al. (2008)

Sesbania spp.

R. huautlense

S. saheli

S. terangae

Ensifer adhaerens

Azorhizobium caulinodans

Azorhizobium johannense

S. meliloti

S. fredii

Wang ET et al. (1998),

De Lajudie et al. (1994)

De Lajudie et al. (1994)

Casida (1982)

Dreyfus et al. (1988)

Moreira et al. (2000)

Wolde-Meskel et al. (2005)

Wolde-Meskel et al. (2005)

 Tephrosia candida

R. hainanense

Chen et al. (1997)

Actinorrhizal tree genera

 Alnus spp.

3 genomic species

(3 to be confirmed)

An et al. (1985)

Fernández et al. (1989)

Akimov and Dobritsa (1992)

Shi and Ruan (1992)

 Casuarina spp.

1 genomic species

Fernández et al. (1989)

 Elaeagnus spp and Hippophae spp

5 genomic species

(5 to be confirmed)

Fernández et al. (1989)

Akimov and Dobritsa (1992)

Lumini et al. (1996)

An et al. (1985)

The specificity apparently varies depending on the level, at which it is analysed (nodulation, effectiveness, or both), and the applied methodology. Rhizobia may be promiscuous for nodulation and have high specificity for effectiveness as was demonstrated in Robinia pseudoacacia L., Acacia mearnsii De Wild. (Turk and Keyser 1992), and A. mangium (Galiana et al. 1990; Prin et al. 2003). Globally, there is no obvious correlation between phylogenies of rhizobia and the hosts, from which they were isolated. A given tropical legume tree may be nodulated by a wide diversity of rhizobia (Bala and Giller 2001; Odee et al. 1995, 1997) and several rhizobia isolated from tropical trees are also able to nodulate herbaceous legumes (Herrera et al. 1985; Zhang et al. 1991).

Actinorrhizal symbiosis

All N2-fixing actinorrhizal symbionts belong to the unique genus Frankia, which forms nodules in the roots of ca. 280 nodulating non-legume species that belong to 25 plant genera. This low number of species and genera compared with legumes does not imply low genetic diversity at the plant level because eight Angiosperm families are concerned (Dawson 2008). The main species used in tropical AFS belong to the genera Casuarina and Allocasuarina (Casuarinaceae), and Alnus (Betulaceae). Alnus spp. and Hippophae rhamnoides L. (Elaeagnaceae) are also used in some temperate AFS.

The first successful isolation of a Frankia sp. occurred only 30 years ago. Today at least 11 species have been reliably isolated and identified, and several others require verification. However, the majority of Frankia diversity remains to be described because almost half of the known Frankia strains have not been successfully grown ex-planta. Modern methods of molecular biology allow the characterisation of strains directly from the nodules, thus avoiding the limiting culture step. Due to the difficulties to cultivate Frankia spp. and lack of reproducibility of some classical taxonomical methods, most of the Frankia genomic species described are not yet named. The only recognised named species is F. alni (Table 1).

Frankia spp. are probably distributed in all continents except Antarctica under very diverse soil and environmental conditions, including areas devoid of actinorrhizal plants. The capacity of Frankia spp. to fix N2 saprophytically and sporulate may explain their survival out of the normal distribution of the host plants (Burleigh and Dawson 1994; Maunuksela et al. 2000, 2006; Paschke and Dawson 1992).

Consequently, many actinorrhizal plants are known for their high capacity to acclimate to environments different from their natural range and they are, thus, widely used as exotics, e.g. in plantation forestry. The nodulation out of the natural range is generally good in Alnus spp., Myricaceae, and Elaeagnaceae (Dawson 2008). Family Casuarinaceae that originates from Australia and New Zealand and has been introduced in Africa, the Americas, and Asia, is a good example on how distribution history can bias the knowledge about strain diversity and plant-strain specificity. The first studies indicated a strong genetic homogeneity of Casuarina-nodulating strains at both intra- and inter-specific level all around the world (Fernández et al. 1989; Nazaret et al. 1989; Rouvier et al. 1992). However, these studies concerned only the introduction areas and often the most widely distributed species Casuarina equisetifolia L. Further studies within the native range indicated genetic diversity of the strains and a high specificity between the plant species and the symbiont genotype (Rouvier et al. 1992). Globally, the actinorrhizal symbioses vary from very promiscuous to very specific associations even within a family, like in the case of Myricaceae (Huguet et al. 2005).

Estimation of N2 fixation

Need to quantify N2 fixation

When preparing this review, it became evident that symbiotic N2 fixation is often cited in agroforestry literature without actual estimates on N2 fixation rate or contribution of N2 fixation to plant or system N economy. In order to really evaluate the N benefit from N2-fixing trees to an AFS, N2 fixation must be estimated. Not all legume trees fix N2, e.g. Senna spp. that are common in AFS (Ladha et al. 1993; Ståhl et al. 2005). Non-nodulating species are most common among the Caesalpinoideae but exist also in Mimosoideae and Papilionoideae (Sprent 2009). Further, tree management may have important effects on nodulation and N2 fixation by trees reported as N2-fixers (Nygren and Ramírez 1995).

Several excellent reviews on the methods for quantifying N2 fixation have been published and they will be referred to in the subchapters dealing with relevant techniques. Unkovich et al. (2008), which is freely available in electronic form, is a good general starting point but the authors largely deal with crop legumes. Boddey et al. (2000) and Domenach (1995) provide reviews targeting N2-fixing trees. Before dealing with methods suitable for AFS, it is necessary to note limitations of some old methods.

Acetylene reduction assay (ARA) is based on the multiaffinity of nitrogenase to break triple bounds in gas molecules (see Giller 2001 or Hunt and Layzell 1993 for a review). Nodules are incubated in acetylene (C2H2) enriched atmosphere and the production of ethylene (C2H4) is measured. Theoretically, C2H4 production rate is 1/4 of N2 fixation rate (Giller 2001). However, the ratio seems to be highly variable in natural systems and ARA requires calibration with other methods in order to be quantitative (Hunt and Layzell 1993). Further, the incubation conditions may cause serious disturbance to nodule functions leading to underestimation of the actual N2 fixation rate (Giller 2001; Minchin et al. 1983; 1986). Thus, earlier reports on low N2 fixation rate by individual agroforestry trees (Lindblad and Russo 1986; Roskoski and Van Kessel 1985) and extrapolations to AFS level of ca. 40 kg ha−1 year−1 (Roskoski 1982) based on ARA should be dealt with caution. Unfortunately, they have been quite influential among agroforestry researchers in Latin America (Beer et al. 1998). While the ARA must be dismissed as a quantitative field method, it has undeniable value as a rapid method for checking if nitrogenase is active (Roggy et al. 1999b; Unkovich et al. 2008).

Another method unsuitable in agroforestry research is the xylem solute method (Peoples et al. 1996). It is used for ureide-transporting crop legumes such as soybean (Glycine max L.): xylem sap is collected and relative amounts of ureides, amino acids, and nitrate (NO3) out of total xylem N are calculated. The ureide proportion is used as an indicator of N2 fixation, which may be calibrated against another quantitative method (Unkovich et al. 2008). However, it has little value in agroforestry research because very few trees transport ureides; only two tree species of the legume tribe Desmodiae were found to be ureide transporters in an extensive review by Giller (2001). Thus, the often-repeated statement that temperate legumes transport fixed N in the form of amino acids and tropical legumes as ureides (Vessey et al. 2004) may be a false perception based on a few crop legumes.

Controlled and semi-controlled conditions

We call laboratory and greenhouse experiments where the growth conditions are mostly under the researchers’ control as controlled condition experiments. Semi-controlled conditions refer to field experiments where some environmental factors are restricted, typically root growing space by physical barriers (Kurppa et al. 2010; Leblanc et al. 2007; Ståhl et al. 2005). Under these conditions, using the stable heavy 15N isotope as a tracer may be the most adequate method. The method is based on the stable proportion of 15N out of atmospheric N, with mean 0.3663 % and standard deviation 0.0002 % (Mariotti 1983). If 15N enriched fertilizer is applied to the substrate, a N2-fixing plant becomes less enriched with 15N than a non-N2-fixing plant because the N taken up from the substrate is diluted with atmospheric N2 in the former. The symbiotic dependence of the N2-fixing plant (Chalk and Ladha 1999) is estimated as the percentage of N derived from atmosphere out of total plant N (%Ndfa):
$$ \% {\text{Ndfa}} = 1 - \left( {\%^{15} {\text{N}}_{\text{ex,fix}} / \%^{15} {\text{N}}_{\text{ex,ref}} } \right) $$
(2)
where %15Nex,fix and %15Nex,ref are 15N atom excess percentage with respect to atmosphere in the N2-fixing and non-N2-fixing reference plant, respectively. The %15Nex is calculated simply by subtracting the atmospheric 15N percentage from the measured 15N percentage in the sample, determined by an isotope ratio mass spectrometer.

The pros and cons of the 15N dilution method have been extensively revised by Unkovich et al. (2008). They also present different variants of the basic method for particular experimental conditions. The main advantage of the method is that it provides both yield-independent and time-integrated estimates of %Ndfa. Main difficulties are related to temporal and spatial non-uniformity of the distribution of the 15N label in the substrate. When small plants in small pots are studied, the 15N label may be efficiently mixed in the substrate (Unkovich et al. 2008). Studies on N2 fixation by seedlings or saplings are of little use for agroforestry research because juvenile trees may behave differently from mature trees in many aspects, as we will show in the chapter “Symbiotic N2 fixation in whole plant physiology”. The limits of complete mixing of 15N with substrate are soon met: when 200 l barrels buried in soil were used for controlling root growth and 15N distribution in the field, the top 0–15 cm of the barrel soil was significantly more enriched than the deeper layers (Kurppa et al. 2010).

Under semi-controlled field conditions, the only possibility to deal with the problems caused by the uneven distribution of the 15N tracer is to use an adequate reference plant. Unkovich et al. (2008) list characteristics, which the reference plant should have in common with the N2-fixing plant: (1) same life form (i.e. only trees should be used as references for trees); (2) adaptation to similar environmental conditions; (3) rooting zone and relative N uptake pattern within the root system; (4) mycorrhizal status; and (5) growth habit and phenology. Further, the N2-fixing and non-N2-fixing reference plant should prefer the same inorganic N form for uptake (Schimann et al. 2008). In natural ecosystems, plants tend to differentiate with respect to these factors in order to share soil resources within the same site, which results in niche differentiation (McKane et al. 2002). Ecologically, the N2-fixing and the reference plant should share the same niche except for N2 fixation. Following to that, the crop species associated with a N2-fixing tree in an AFS is seldom a good reference because species selection in agroforestry targets to complementary resource use (Ong et al. 2004) and, thus, niche differentiation.

Field studies

When N2 fixation is estimated under field conditions, the problems associated with the uniformity of 15N labelling and match between the characteristics of the N2-fixing and reference tree become manifold in comparison to controlled-environment studies. It is almost impossible to distribute the 15N tracer uniformly in a soil profile. Injecting the 15N tracer to different soil depths has been used for studying rooting depths and N uptake in AFS (Lehmann et al. 2001; Rowe et al. 1999; 2001). The injection is always punctual and extensive sampling is required for assuring a reliable recovery of the 15N signal in the plants (Rowe and Cadisch 2002). Although the deep injection labelling has some promise for agroforestry research, it has not yet been used for estimations of the contribution of N2 fixation to N economy of the N2-fixing trees or the whole AFS. It seems that selection of a “correct” reference tree species that uses the same form of N from same depth and horizontal area as the N2-fixing tree is the only method to overcome the problem of uneven distribution of the 15N tracer in soil (Chalk and Ladha 1999). However, often too little pre-information is available on these tree characteristics. Thus, using several reference species is recommended (Unkovich et al. 2008).

Given the problems of the 15N dilution method under field conditions, the 15N natural abundance method (Shearer and Kohl 1986) is often recommended for quantifying N2 fixation in AFS. Handley and Raven (1992), Högberg (1997), and Martinelli et al. (1999) reviewed the behaviour of 15N in plant metabolism, soil–plant systems, and natural forests, respectively. Boddey et al. (2000), Domenach (1995), and Gehring and Vlek (2004) reviewed the application of the 15N natural abundance method for estimating N2 fixation by trees. Here we revise the basic idea and some agroforestry applications.

In the basic application of the 15N natural abundance method, the deviation of the sample 15N content from that of atmosphere (δ15N) is calculated (Shearer and Kohl 1986):
$$ \delta {}^{15}N = \frac{{{}^{15}N/{}^{14}N_{\text{sa}} - {}^{15}N/{}^{15}N_{\text{at}} }}{{{}^{15}N/{}^{14}N_{\text{at}} }} \times { 1 , 0 0 0\,{\permille}} $$
(3)
where 15N/14N is the ratio of 15N to 14N and subscripts sa and at refer to the sample and atmosphere, respectively. In the 15N natural abundance method, %Ndfa is estimated by comparing the δ15N values in a N2-fixing plant growing in the field (δ15Nf), a non-N2-fixing reference plant growing in the same soil (δ15Nr), and the N2-fixing plant growing in N-free environment (δ15N0), i.e. depending only on N2 fixation for N supply (Shearer and Kohl 1986):
$$ \% Ndfa = \frac{{\delta^{15} N_{\text{r}} - \delta^{15} N_{\text{f}} }}{{\delta^{15} N_{\text{r}} - \delta^{15} N_{0} }} \times 100\% $$
(4)
While the 15N natural abundance in the atmosphere is very stable and uniform around the world (Mariotti 1983), it varies in the soils and plants because plant metabolism and several microbiological processes in the soil cause isotopic fractionation (Handley and Raven 1992). Most soil microbiological processes discriminate against 15N (Högberg 1997), thus, enriching soil with this isotope. This causes a slight difference in the δ15N of plants depending on soil N and plants partially fixing atmospheric N2. The difference can be detected with modern isotope ratio mass spectrometers but all work phases must be conducted with great care because the differences are only a few ‰-units.

The general trend of the soil becoming naturally 15N enriched does not hold in all soils. If the non-N2-fixing trees have a low soil-derived δ15N that cannot be distinguished from the δ15N of N2-fixing trees, application of the 15N natural abundance method is impossible (Gehring and Vlek 2004; Roggy et al. 1999a). In some cases, the soil may be so depleted of 15N and, consequently, the reference trees have so much lower δ15N than the N2-fixing trees that the 15N natural abundance method is applicable because active N2-fixers tend to have δ15N close to 0, in this case significantly higher than the non-N2-fixing reference (Augusto et al. 2005; Domenach 1995; Domenach et al. 1989).

Further, it was shown among non-N2-fixing caesalpinoids that Eperua falcata Aubl., which prefers NO3 as the inorganic N source, had significantly lower δ15N than Dicorynia guianensis Amshoff, which uses ammonium (NH4+) for N supply in the same soil (Schimann et al. 2008). This suggests that the N2-fixing tree and the reference tree should use the same inorganic N form. Most legumes form symbiosis, in addition to rhizobia, with arbuscular mycorrhizae-forming fungi (AMF) and some with ectomycorrhizae-forming fungi (ECM; Sprent and James 2007). According to a global analysis, AMF deplete the host plant with 15N and ECM even more (Craine et al. 2009) yet the difference between AMF and ECM plants decreases when they grow in the same site (Hobbie and Högberg 2012). It seems that the effect of mycorrhizae on plant δ15N depends on an interaction between fungal strain, environment, and host plant (Boddey et al. 2000). The mycorrhizal effects are further complicated by the fungal uptake of organic N (Boddey et al. 2000; Näsholm et al. 2009). However, before a better understanding on these effects arises, we recommend using reference trees that have the same mycorrhizal type and preference for soil N as the studied N2-fixing tree. As in the case of 15N dilution method, the reference plant should have similar rooting pattern and phenology as the N2-fixing plant.

The question whether to use the associated non-N2-fixing plant or an independent reference growing in similar soil but without contact with the N2-fixing plant is more important in the case of 15N natural abundance than the 15N dilution method. In the latter case, the selection is an issue of ecological and physiological similarity, which is seldom met between the trees and crops in AFS. In the field, a N2-fixing tree may slowly decrease the soil δ15N by recycling litter and root exudates depleted of 15N. Consequently, the δ15N of the associated plants also decreases over time (van Kessel et al. 1994; Sierra and Nygren 2006). Leucaena leucocephala altered so much the soil isotopic signature in six years that by the end of the study period, associated weeds had almost the same isotopic signature as the tree itself (van Kessel et al. 1994). The δ15N gradually increased in Dichanthium aristatum (Poir.) C.E. Hubb. grass brought from a field site with Gliricidia sepium trees to a greenhouse, probably because of removal of N transferred from the tree by successive grass cuttings (Sierra and Nygren 2006). Thus, the reference trees should grow without direct contact with the N2-fixing trees but in a similar soil.

Use of several reference tree species is recommendable also in 15N natural abundance method because often too little is known about the characteristics described above for selecting an ecologically and physiologically similar reference to a N2-fixing tree (Boddey et al. 2000; Unkovich et al. 2008). Several candidate references can also be screened in a pre-trial (Nygren and Leblanc 2009). This will allow the removal of apparent outliers; e.g. in a comparison of five non-N2-fixing tree species, Nygren and Leblanc (2009) found that one species had much lower δ15N than the four other species, which had little variation among themselves. Non-N2-fixing legume trees may be the best references for legume trees if available (Roggy et al. 1999b). Further, if the research involves comparisons over different sites, some of the reference species should be used in all sites if possible (Roggy et al. 1999b).

In some cases, it may also be possible to use a non-nodulating phenophase of the studied species as the reference (Salas et al. 2001). This method requires that the isotopic signature of the soil N is stable over time, sufficient variation in nodulation exists between the phenophases, and the temporal variation in N2 fixation of the nodules is small. The approach was successfully applied for estimating rainy season N2 fixation in Erythrina lanceolata Standl. using the dry season non-nodulating phenophase as the reference (Salas et al. 2001).

Sampling procedures

Whole tree harvesting is recommended for seedlings and saplings because the isotopic signal may vary in different parts of a tree (Kurppa et al. 2010; Leblanc et al. 2007). In the case of big trees, this is often not possible. Typically, both N2-fixing and non-N2-fixing trees tend to have the highest δ15N in leaves and the lowest in stem and coarse roots (Boddey et al. 2000). Leaves are often the preferred sink for recently fixed N (Domenach 1995) and they are, thus, the best choice for sampling if the objective is to follow seasonal or other variation in N2 fixation in time scale of weeks or few months. Large structural parts of a tree, stem and coarse roots, may be envisioned to integrate the isotopic signature of N sources over time. They may be the choice if the objective is to understand the long-term role of N2 fixation in trees’ N economy (Nygren and Leblanc 2009). Fine roots have seldom been sampled. In a study using 15N enrichment, higher proportion of 15N than total N was retained in the fine roots of Gliricidia sepium suggesting that soil N may be preferentially used for fine roots and fixed N is transported to the shoot. However, the root:shoot sharing of 15N and total N was the same in the legume tree Inga edulis and non-N2-fixing crop Theobroma cacao L. (Kurppa et al. 2010). Thus, dynamics of fine roots require further study for understanding the whole plant isotopic relationships.

Sanginga et al. (1995) claimed that the 15N natural abundances in different organs tend to be proportionally equal in N2-fixing and non-N2-fixing trees and, consequently, estimates of the %Ndfa are about the same if the same organ of both trees is used. Thus, they proposed that leaf sampling would result in reliable %Ndfa estimates. This hypothesis held between G. sepium and five different reference tree species in a cacao plantation but not with I. edulis (Nygren and Leblanc 2009). The estimates of %Ndfa based on 15N enrichment (Eq. 2) were about the same using leaves only and whole sapling harvesting for three agroforestry tree species, including I. edulis (Leblanc et al. 2007). Leaves were recommended for 15N sampling in trees but not in herbaceous legumes by Unkovich et al. (2008).

Isotopic fractionation within a tree resulting in differences in N isotopic relationships between tree organs may affect the usability of the 15N natural abundance method in some cases. Some nodulated Inga spp., in which ARA indicated nitrogenase activity, had high leaf δ15N values typical for non-N2-fixing trees in a rain forest (Roggy et al. 1999b) and a tropical fresh water swamp forest (Koponen et al. 2003) in French Guiana. In both cases, most Inga spp. had leaf δ15N values typical for N2-fixing trees. In a Costa Rican cacao plantation, I. edulis shade trees had a high δ15N in leaves but low stem and coarse root δ15N values were typical for N2-fixers (Nygren and Leblanc 2009). These data may indicate strong isotopic fractionation in the N transport and metabolism of some Inga spp. Similar observations have been made in Acacia mangium (Bouillet et al. 2008) and A. senegal (Isaac et al. 2011a). Although the strong within-tree isotopic fractionation may be a rare phenomenon, its possible occurrence should be taken into account when analysing 15N natural abundance data and nodulation must be revised in all trees sampled. The 15N dilution method does not seem to be affected by isotopic fractionation, indicating active N2 fixation in I. edulis (Kurppa et al. 2010; Leblanc et al. 2007).

Symbiotic N2 fixation in whole tree physiology

Conceptual model on nitrogen acquisition in the tripartite plant-bacteria-mycorrhizae symbiotic system

It may be assumed that legumes (Bethlenfalvay 1992; Bethlenfalvay and Newton 1991) and actinorrhizal plants (Gardner and Barrueco 1999) are mycorrhizal, except Lupinus spp. (Sprent and James 2007). Thus, ecophysiology of N2 fixation must be evaluated within the context of the tripartite symbiosis formed by plants, N2-fixing bacteria, and mycorrhizae in all natural and man-made ecosystems (Barea et al. 1992; Bethlenfalvay and Newton 1991; Cardoso and Kuyper 2006; Kuyper et al. 2004). Most legumes seem to form symbiosis with AMF (Bethlenfalvay 1992; Cardoso and Kuyper 2006; Kuyper et al. 2004; Sprent and James 2007). A minority of legume species, notably Acacia spp. and caesalpinoids, may form symbiosis with ECM or both AMF and ECM (Haselwandter and Bowen 1996; Sprent and James 2007). More than 40 actinorrhizal species have been recorded to support ECM, AMF, or both symbioses (Gardner and Barrueco 1999).

A simplified scheme of potential N acquisition strategies by the tripartite symbiosis is presented in Fig. 1. The soil N forms available for the mycorrhizal fungi vary: AMF may take up soil NH4+, NO3, and amino acids, while some ECM cannot use NO3 but they are able to use, in addition to amino acids, some other organic N forms in the soil such as simple proteins and oligopeptides (Lambers et al. 2008; Smith and Read 2008). Mycorrhizae also enhance plant N nutrition by largely increasing the soil volume available to uptake of N and other nutrients because of the extensive extraradical mycelium. Fungal hyphae also access smaller soil pores than plant roots (Smith and Read 2008). The N2 fixation process by symbiotic bacteria within the root nodules has been described above.
https://static-content.springer.com/image/art%3A10.1007%2Fs10705-012-9542-9/MediaObjects/10705_2012_9542_Fig1_HTML.gif
Fig. 1

A simplified scheme of nitrogen acquisition strategies of a plant, which forms symbiosis with N2-fixing bacteria and mycorrhiza-forming fungi. The potential N sources (clouds) are soil ammonium and nitrate, which the plant may take up directly via its own fine roots or via mycorrhizal symbiosis, soil amino acids (AA), which may be taken up by both arbuscular-mycorrhizal and ectomycorrhizal fungi, other soil organic N (No), which may be taken up by the ectomycorrhizal fungi, and atmospheric N2, which is fixed by the symbiotic bacteria in the root nodules. Flows of inorganic N are indicated by thin arrows, flows of organic N by thick arrows, flows of organic carbon (Co) by double arrows, feedback loops by dashed arrows, metabolic processes by valves, and NH4+ and NO3 transporters with ellipsoids. The metabolic processes are breakdown of AA to NH4+, NO3 reduction to NH4+, assimilation of NH4+ to AA, and biological N2 fixation. There is evidence on N transfer from the plant to mycorrhizal fungus as either NH4+ or AA, thus double-ended arrows. Details of leaf N metabolism and photochemistry that also affect plant N status are beyond the scope of this review

Soil inorganic N taken up by the mycorrhizal fungus is processed in the fungal hyphae (Smith and Read 2008). Nitrate is reduced to NH4+, which is further assimilated to amino acids together with NH4+ taken up from the soil. The plant may reduce NO3 in the roots or transport it to the leaves where NO3 is reduced using NADH and NADPH produced in photosynthesis (Oaks 1992; Zerihun et al. 1998). Free NH4+ is toxic to plants and must be assimilated in the site of reduction. Carbon transported in phloem from the leaves is needed for NH4+ assimilation in the roots or mycorrhizal mycelium. There is also evidence on N transfer from the plant to mycorrhizal fungus, especially from N-rich roots of N2-fixing plants (Arnebrant et al. 1993; He et al. 2003; Simard et al. 2002).

Parsons et al. (1993) suggested that nodulation and N2 fixation may be regulated by the concentration of reduced N in the phloem flow. Nitrogen is transported to leaves as NO3 or amino acids in the xylem, and amino acids may be transported from leaves back to roots in phloem (Fig. 1). Comparison of nodulation and canopy N flows in the agroforestry tree Erythrina poeppigiana (Walp.) O.F. Cook suggested that nodulation was regulated by N needs of the canopy with reduced nodulation when net N flow turned downward from canopy to roots (Nygren 1995). Accumulation of amino acids in the roots controls the NH4+ and NO3 transporters in all plants (Amtmann and Blatt 2009) and the blockage of the transporters by high root amino acid concentration may result in N exudation from roots to soil observed in many plants (Fustec et al. 2010; Wichern et al. 2008). Thus, root amino acid pool may be envisioned to play a key role in regulating the N acquisition of the whole tripartite symbiotic system (Fig. 1). It may be hypothesised that the root pool size is maintained by the balance between the influxes from the tripartite symbiosis and efflux to foliage. Thus, N acquisition may be ultimately regulated by N needs in the plant canopy. Not all evidence to support this hypothesis is available but we feel that it is a fruitful starting point for further research.

Energetics of N2 fixation

Symbiotic N2 fixation is quite an energy intensive strategy for acquiring N (Vance and Heichel 1991; Vitousek et al. 2002). Following the biochemically-based estimation method of Thornley and Johnson (1990), 4.29 g of glucose is needed to reduce 1 g of N to NH4+ from NO3 in the roots and 5.00 g of glucose for fixing 1 g of N from the atmosphere. Thus, the C cost of N2 fixation is only about 17 % higher than that of NO3 reduction in roots. However, glucose consumption per gram of protein produced is considerably lower when amino acids (on average 0.89 g glucose) or NH4+ (1.77 g) are available as the N source (Zerihun et al. 1998). At whole plant level, acquisition of soil N also requires C for growth and maintenance of the root system. Legumes probably evolved in the humid tropical forests (McKey 1994) where NH4+ is often a much more abundant inorganic N form than NO3 (Boddey et al. 2000). Under these conditions, N2-fixing symbiosis may have been energetically more appropriate N acquisition strategy than growing roots for competing for the scarce soil NO3 resources.

Carbon consumption to N2 fixation may be compensated by increased photosynthetic rate; photosynthetic rate of several nodulated herbaceous crop legumes was increased on average by 28 % in comparison to non-N2-fixing plants (Kaschuk et al. 2009). In most cases, increase in photosynthetic production was higher than the C allocation to the rhizobial symbiont (4–16 % of C fixed in photosynthesis). We are not aware of any respective studies on N2-fixing trees.

Phosphorus in the N2-fixing trees

Phosphorus has been considered a limiting nutrient for symbiotic N2 fixation process because of the high ATP requirements (Eq. 1). For evaluating the importance of the P on N2 fixation, we searched the CAB Abstracts database for articles published 2000–2009 on the effects of P fertilisation on nodulation and N2 fixation in trees. All studies dealing with tropical tree seedlings under greenhouse or nursery conditions (Binkley et al. 2003; Diouf et al. 2008; Pons et al. 2007; Uddin et al. 2008) indicated positive effects of P fertilisation on N2 fixation. All field studies were conducted on actinorrhizal trees in temperate forests. These studies indicated no direct P effect on nitrogenase activity (Uliassi and Ruess 2002) or N2 fixation rate (Augusto et al. 2005; Cavard et al. 2007; Gokkaya et al. 2006) but enhanced nodulation was observed in one case (Uliassi and Ruess 2002).

In a recent study, however, Isaac et al. (2011a) found that P supply rates did not markedly affect N2 fixation rates of Acacia senegal seedlings under non-limiting N supply, but higher P supply stimulated growth, which resulted in greater mineral N uptake from soil solution. On the other hand, the N2 fixation rate of A. senegal increased with increasing soil P availability in natural stands of the Rift Valley in Kenya (Isaac et al. 2011b). Application of rock phosphate did not significantly affect N2 fixation by Inga edulis in a P-poor soil in the humid Atlantic lowlands of Costa Rica (Leblanc 2004).

Severe P deficiency markedly impaired symbiotic N2 fixation in the tropical Casuarinaceae species (Sanginga et al. 1989) and other actinorrhizal plants (Gardner and Barrueco 1999). Several actinorrhizal genera such as Comptonia and Myrica (Myricaceae) develop cluster (proteoid-like) roots, which could enhance P uptake in P poor soils (Berliner and Torrey 1989).

Both plant roots and hyphae of mycorrhizal fungi secrete extracellular phosphatases to the rhizosphere (Grierson et al. 2004; Louche et al. 2010; Treseder and Vitousek 2001). Phosphatases hydrolyse the ester-phosphate bonds in soil organic P, thus releasing phosphate to soil solution where it may be taken by plant roots or hyphae of mycorrhizal fungi. Extracellular phosphatases in the rhizosphere assure the breakdown of organic P compounds in the proximity of roots. Phosphatases contain 8–32 % of N (Treseder and Vitousek 2001), and it has been proposed that N2-fixing plants have an advantage in producing these N-rich compounds (Wang et al. 2001). Houlton et al. (2008) presented a hypothesis that improved P nutrition because of a higher production rate of extracellular phosphatases may explain the abundance of legume trees in P-poor humid tropical forests. Higher phosphatase activity has been observed in the soil below both actinorrhizal (Giardina et al. 1995; Zou et al. 1995) and legume trees (Allison et al. 2006; Zou et al. 1995) than below non-N2-fixing trees. As far as we know, no studies on the phosphatase production in AFS exist. However, it may be an important yet at the moment unstudied positive interaction between N2-fixing trees and crops in AFS.

The fact that P fertilisation enhanced growth of tree seedlings in pot experiments (Diouf et al. 2008; Pons et al. 2007; Uddin et al. 2008) but results on mature trees in the field were more variable (Gokkaya et al. 2006; Uliassi and Ruess 2002) suggest that caution should be applied for extrapolating the pot results to the field. In one case, the same tree species, Falcataria moluccana (Miq.) Barnaby & J.W. Grimes, was studied both in a pot culture and in a forest. Phosphorus fertilisation enhanced N2 fixation by seedlings in the pot study (Binkley et al. 2003), while higher phosphatase activity was observed in soils under F. moluccana than non-N2-fixing trees in Hawaii (Allison et al. 2006). It is possible that mature trees and associated mycorrhizae produce more phosphatases than seedlings and, thus, better cope with soil P deficit.

Interaction between N2-fixing bacteria and mycorrhizae

Mycorrhizae may also alleviate the effects of soil P deficit on N2 fixation through enhanced P supply to the tripartite symbiotic system. Best growth and highest shoot nutrient concentrations were observed in P fertilised Acacia senegal seedlings inoculated with rhizobia and dual inoculation with rhizobia and AMF was the second best treatment (Colonna et al. 1991). In a sterile, low P substrate, AMF colonisation in roots of Calliandra calothyrsus seedlings decreased with increasing P addition (Ingleby et al. 2001). Biomass production was not affected by AMF colonisation or P level but higher nodulation was observed in the AMF-colonised plants. Rhizobia-ECM dual inoculated Acacia holosericea A. Cunn. ex. G. Don seedlings had the highest biomass production (André et al. 2005). Both AMF and ECM colonisation enhanced nodulation in the same tree species (Duponnois and Plenchette 2003). Inoculation with AMF enhanced nodulation of Leucaena leucocephala seedlings and dual inoculation was recommended (Araújo et al. 2001). Dual rhizobia-AMF inoculation also enhanced initial growth of C. calothyrsus under field conditions but statistically significant differences between various inoculation treatments disappeared after 24 months (Lesueur and Sarr 2008). It has also been shown that the percentage of roots infected by mycorrhizae is markedly higher in nodulated than non-nodulated actinorrhizal plants (Chatarpaul et al. 1989).

Based on their own observations and earlier work (Habte 1995; Habte and Turk 1991; Manjunath and Habte 1992), Ingleby et al. (2001) concluded that among the common agroforestry trees, N2-fixing C. calothyrsus, Gliricidia sepium, and L. leucocephala, and the non-N2-fixing Senna siamea (Lam.) H.S. Irwin & Barneby are highly responsive to mycorrhizae while the non-N2-fixing Senna reticulata (Willd.) H.S. Irwin & Barneby and N2-fixing Sesbania pachycarpa DC. are less responsive.

Most studies revised indicate beneficial effects of the tripartite symbiosis on plant development and N2 fixation. The beneficial effect of mycorrhizae seems to be most important under P limitation supporting Bethlenfalvay’s (1992) suggestion that the main function of the mycorrhizal symbiont is P supply to the tripartite system. The observations that AMF colonisation was more beneficial than P fertilisation (Araújo et al. 2001) and that dual inoculation enhanced biomass production independently of P level (Weber et al. 2005) suggest, however, that mycorrhizal colonisation may enhance the functioning of the tripartite symbiotic system also in ways other than just increased P supply. It should also be noted that mature trees might function differently as host than seedlings or saplings.

Estimates of N2 fixation in agroforestry systems

Field estimates on the dependence of agroforestry trees on symbiotic N2 fixation

When compiling quantitative data on N2 fixation in AFS, we excluded seedling studies, because we have shown in the previous chapter that seedlings may behave quite differently from mature trees. On the other hand, many legume trees used in AFS are fast-growing and some management practices, notably green pruning, maintain them in a physiological state resembling juvenile trees, e.g. by impeding flowering (Nygren et al. 2000). Thus, we accepted data from experiments with at least 1-year-old saplings. We also included data on tree plantations if the tree species is known to be used in agroforestry. Data were most abundant on the %Ndfa including 38 data points (Table 2). We classified the data according to management practices and climate. Climate was classified into three broad categories based on our interpretation of the data given in original articles: dry (water deficit prevails most of the year), subhumid (seasonal climate with water deficit for six months or less), and humid (short water deficit periods or none at all). Two irrigated experiments (Gauthier et al. 1985; Kadiata et al. 1997) were pooled with humid climate in the following analyses.
Table 2

Percentage of nitrogen derived from atmosphere out of total nitrogen (%Ndfa) in several tree species used in agroforestry

N2-fixing species

Associated species

Reference species

Climate

System

Management information

Estimation method

%Ndfa

References

Acacia angustisima

Zea mays

Hyparrhenia rufa

Subhumid

Improved fallow

Unpruned

15N natural abundance; whole tree

48–79a

Chikowo et al. (2004)

Acacia caven

None

Schinus polyganus and Fraxinus excelsior

Dry

Experimental plantation

Free growth

15N natural abundance; leaf biomass

50

Aronson et al. (2002)

Acacia mangium

Eucalyptus grandis

Eucalyptus grandis

Subhumid

Mixed plantation

Free growth

15N enrichment; whole tree biomass

59

Bouillet et al. (2008)

Acacia mangium

None

Eucalyptus urophylla

Humid

Experimental plantation

Free growth

15N natural abundance; sampling not reported

42–62b

Galiana et al. (1998)

Acacia mangium

None

Understorey weeds

Humid

Improved fallow, 12 years

Free growth

15N natural abundance; whole tree

57

Mercado et al. (2011)

Acacia senegal

None

Balanites aegyptiaca

Dry

Experimental arabic gum production

Free growth

15N natural abundance; leaf biomass

24–61c

Raddad et al. (2005)

Acacia senegal

Grass

Balanites aegyptiaca

Dry

Tree-grass savanna

Free growth

15N natural abundance; leaf biomass

33–39d

Isaac et al. (2011b)

Albizia lebbeck

None

Senna siamea

Irrigated

16-month greenhouse experiment

Variable pruning frequency

15N enrichment; shoot biomass

74–83e

Kadiata et al. (1997)

Calliandra calothyrsus

None

Eucalyptus deglupta and Grevillea robusta

Subhumid

Improved fallow

Free growth

15N enrichment; whole tree biomass

5–54f

Ståhl et al. (2002)

Calliandra calothyrsus

None

Senna spectabilis

Subhumid

Protein bank

7 prunings in 2 years

15N natural abundance; whole tree

24–84f

Peoples et al. (1996)

Calliandra calothyrsus

Khaya senegalensis, Ziziphus mauritaniana and Anacardium occidentale

Mean of associated tree species

Not reported (subhumid assumed)

Mixed experimental plantation; 2 years

Free growth

15N natural abundance; leaf biomass

30–60b

Lesueur and Sarr (2008)

Casuarina equisetifolia

Eucalyptus x robusta

Eucalyptus x robusta

Sudhumid-humid

Mixed tree plantation

Free growth

15N enrichment; whole tree biomass

59

Parrotta et al. (1996)

Casuarina equisetifolia

None

None

Irrigated

1-year experiment

Free growth

15N dilution; shoot biomass

55

Gauthier et al. (1985)

Chamaecytisus proliferus

None

Schinus polyganus and Fraxinus excelsior

Dry

Experimental plantation; 6 years

Free growth

15N natural abundance; leaf biomass

84

Aronson et al. (2002)

Chamaecytisus proliferus

Sequential Lupinus angustifolius and Avena sativa

Mean of Ptilotus polystachus and annual weeds

Dry

Experimental alley cropping, 4 years

Pruned

15N natural abundance; coppice biomass

83

Unkovich et al. (2000)

Codariocalyx gyroides

None

Senna spectabilis

Subhumid

Protein bank

7 prunings in 2 years

15N natural abundance; whole tree

48–86f

Peoples et al. (1996)

Erythrina fusca

None

Vochysia guatemalensis

Humid

14-month-old saplings

Free growth

15N enrichment; whole tree

64

Leblanc et al. (2007)

Erythrina lanceolata

Vanilla planifolia

Non-nodulated phenophase

Humid

Living support

Free growth

15N natural abundance; leaf biomass

53 g

Salas et al. (2001)

Erythrina poeppigiana

None

Vochysia guatemalensis

Humid

14-month-old saplings

Free growth

15N enrichment; whole tree

59

Leblanc et al. (2007)

Faidherbia albida

None

Parkia biglobosa

Dry

15-month-old saplings

Free growth

15N enrichment; whole tree

54

Gueye et al. (1997)

Gliricidia sepium

Dichanthium aristatum

Gmelina arborea

Subhumid

Cut-and-carry fodder production

100 % pruning every 6 months

15N natural abundance; shoot biomass

54–92f

Nygren et al. (2000)

Gliricidia sepium

Dichanthium aristatum

Gmelina arborea

Subhumid

Cut-and-carry fodder production; 8 years

50 % pruning every 2 months

15N natural abundance; shoot biomass

60–87f

Nygren et al. (2000)

Gliricidia sepium

None

Senna spectabilis

Subhumid

Protein bank

7 prunings in 2 years

15N natural abundance; whole tree

58–89f

Peoples et al. (1996)

Gliricidia sepium

Sequential Zea mays and Oryza sativa

Senna spectabilis

Humid

Alley cropping

4 prunings per year at 50 cm

15N natural abundance; shoot biomass

35–59f

Ladha et al. (1993)

Gliricidia sepium

Arachis pintoi (?)

Peltophorum dasyrrachis

Subhumid

Alley cropping

Complete pruning, frequency not reported

15N enrichment; prunings

50

Rowe et al. (1999)

Gliricidia sepium

Theobroma cacao

Theobroma cacao

Humid

14-month-old saplings

Free growth

15N enrichment; whole tree

85

Kurppa et al. (2010)

Gliricidia sepium

Theobroma cacao, Bactris gasipaes, non-leg. trees

Mean of 4 non-legume tree spp.

Humid

Shaded cacao plantation; 12 years

Free growth

15N natural abundance; leaf biomass

57–67f

Nygren and Leblanc (2009)

Gliricidia sepium

None

Senna siamea

Irrigated

16-month greenhouse experiment

Variable pruning frequency

15N enrichment; shoot biomass

69–75d

Kadiata et al. (1997)

Inga edulis

None

Vochysia guatemalensis

Humid

14-month-old saplings

Free growth

15N enrichment; whole tree

57

Leblanc et al. (2007)

Inga edulis

Theobroma cacao

Theobroma cacao

Humid

14-month-old saplings

Free growth

15N enrichment; whole tree

74

Kurppa et al. (2010)

Inga edulis

Theobroma cacao, Cordia alliodora

Cordia alliodora

Humid

Shaded cacao plantation; 15 years

Free growth

15N natural abundance; stem biomass

50–63f

Nygren and Leblanc (2009)

Leucaena leucocephala

Eucalyptus x robusta

Eucalyptus x robusta

Subhumid

Mixed tree plantation

Free growth

15N enrichment; whole tree biomass

38

Parrotta et al. (1996)

Leucaena leucocephala

None

Senna siamea

Irrigated

16-month greenhouse experiment

Variable pruning frequency

15N enrichment; shoot biomass

80–84e

Kadiata et al. (1997)

Prosopis alba

None

Schinus polyganus and Fraxinus excelsior

Dry

Experimental plantation; 6 years

Free growth

15N natural abundance; leaf biomass

10

Aronson et al. (2002)

Prosopis chilensis

None

Schinus polyganus and Fraxinus excelsior

Dry

Experimental plantation; 6 years

Free growth

15N natural abundance; leaf biomass

30

Aronson et al. (2002)

Sesbania sesban

Zea mays (sequential)

Hyparrhenia rufa

Subhumid

Improved fallow

Free growth

15N natural abundance; whole tree

42–73a

Chikowo et al. (2004)

Sesbania sesban

None

Eucalyptus deglupta and Grevillea robusta

Subhumid

Improved fallow

Free growth

15N enrichment; whole tree biomass

70–81f

Ståhl et al. (2002)

Sesbania sesban

None

Eucalyptus deglupta and Grevillea robusta

Subhumid

Improved fallow

Free growth

15N enrichment; whole tree biomass

61–71 h

Ståhl et al. (2005)

aDepending on organ

bDepending on rhizobial strain

cDepending on tree provenance

dDepending on tree age

eDepending on pruning frequency

fDepending on sampling time

gRainy season (non-nodulated in dry season)

hDepending on reference species

We pooled data on 15N dilution (Eq. 2) and natural abundance (Eq. 4) methods. The general mean of the %Ndfa of the compiled data (Table 2) was 59 % (Standard deviation, SD 16.6 %) and the range was from 10 to 85 %. We used the mean of minimum and maximum in the calculation of the general mean when a range was given in Table 2. Five studies were conducted in dry environments (Aronson et al. 2002; Gueye et al. 1997; Isaac et al. 2011b; Raddad et al. 2005; Unkovich et al. 2000) with the mean %Ndfa of 49 % (SD 25.3 %, N = 8). The data compiled also included ten cases of young (1–2 years) trees under humid or sub-humid conditions with mean %Ndfa of 69 % (SD 10.7 %). The remaining 20 studies were conducted with at least 2-year-old trees under humid or subhumid conditions with mean %Ndfa of 57 % (SD 12.1 %). However, only three of these studies were conducted in systems that had been subjected to the same management for several years, i.e. a 15-year-old shaded cacao plantation (Nygren and Leblanc 2009), a 12-year-old improved fallow of Acacia mangium (Mercado et al. 2011), and an 8-year-old cut-and-carry forage production system under partial pruning regime (Nygren et al. 2000). The studies on at least 2-year-old systems under humid or subhumid conditions were further divided into pruned and unpruned systems, with the mean %Ndfa of 63 % (SD 12.0 %, N = 7) and 54 % (SD 11.7 %, N = 13), respectively. The popular agroforestry tree Gliricidia sepium appeared eight times in Table 2 being the only species, for which we calculated the species mean for %Ndfa, 67 % (SD 13.0 %).
Table 3

Nitrogen fixation per hectare and year in some agroforestry systems with legume or actinorrhizal trees based on measurement of whole plant N isotopic relations or whole system N balance

N2-fixing species

Associated species

Reference species

Climate

System

Management information

Estimation method

N2 fixation kg ha−1 year−1

References

Acacia angustisima

Zea mays

Hyparrhenia rufa

Subhumid

Improved fallow

Free growth

15N natural abundance; whole tree

91a

Chikowo et al. (2004)

Acacia auriculiformis

None

Eucalyptus sp.

Humid

Improved fallow, 7 years

Free growth

Whole system N balance

140

Bernhard-Reversat (1996)

Acacia mangium

None

Understorey weeds

Humid

Improved fallow, 12 years

Free growth

15N natural abundance

129b

Mercado et al. (2011)

Acacia mangium

None

Eucalyptus sp.

Humid

Improved fallow, 7 years

Free growth

Whole system N balance

115

Bernhard-Reversat (1996)

Acacia polyacantha

None

Eucalyptus camaldulensis

Subhumid

Improved fallow, 7 years

Free growth

Whole system N balance

90

Harmand (1998)

Calliandra calothyrsus

None

Eucalyptus deglupta and Grevillea robusta

Subhumid

Improved fallow

Free growth

15N enrichment

122–171c

Ståhl et al. (2002)

Calliandra calothyrsus

None

Cassia spectabilis

Subhumid

Protein bank

7 prunings in 2 years

15N natural abundance

670d

Peoples et al. (1996)

Casuarina equisetifolia

Eucalyptus x robusta

Eucalyptus x robusta

Subhumid

Mixed tree plantation

Free growth

15N enrichment

73

Parrotta et al. (1996)

Chamaecytisus proliferus

Sequential Lupinus angustifolius and Avena sativa

Mean of Ptilotus polystachus and annual weeds

Dry

Experimental alley cropping, 4 years

Pruned

15N natural abundance

83a

Unkovich et al. (2000)

Chamaecytisus proliferus

None

Mean of Ptilotus polystachus and annual weeds

Dry

Experimental plantation, 4 years

Pruned

15N natural abundance

390a

Unkovich et al. (2000)

Gliricidia sepium

Dichanthium aristatum (grass)

None

Subhumid

Cut-and-carry fodder production

50 % pruning every 2-4 months

Whole system N balance

555e

Dulormne et al. (2003)

Gliricidia sepium

None

Cassia spectabilis

Subhumid

Protein bank

7 prunings in 2 years

15N natural abundance

675d

Peoples et al. (1996)

Leucaena leucocephala

Eucalyptus x robusta

Eucalyptus x robusta

Subhumid

Mixed tree plantation

Free growth

15N enrichment

74

Parrotta et al. (1996)

Sesbania sesban

Zea mays (sequential)

Hyparrhenia rufa

Subhumid

Improved fallow

Free growth

15N natural abundance; whole tree

56

Chikowo et al. (2004)

Sesbania sesban

None

Eucalyptus deglupta and Grevillea robusta

Subhumid

Improved fallow

Free growth

15N enrichment

282–363c

Ståhl et al. (2002)

Sesbania sesban

None

Eucalyptus deglupta and Grevillea robusta

Subhumid

Improved fallow

Free growth

15N enrichment

310–356f

Ståhl et al. (2005)

aIncluding fixed N in litter

bFixed N in roots estimated not measured

cDepending on sampling time

dThe authors did not show the whole tree values but provided data on N partitioning between shoot and root that we used for recalculating the estimate

eAll fixed N in the agroecosystem, including net transfer to the crop and soil accumulation

fDepending on reference species

All these means are higher than the general mean for legume trees in natural ecosystems according to the review by Andrews et al. (2011), 42 % (SD 5.4 %). Because competition with non-N2-fixing plants seems to increase the legume dependence on N2 fixation (Andrews et al. 2011) competition with crops may enhance N2 fixation by trees in AFS. Soil N depletion due to N export in crop harvest may also partially explain the higher legume dependence on N2 fixation in AFS than natural ecosystems.

Because of the high within-group variation and low number of studies under dry conditions, we compared statistically only the three distinct cases under humid or subhumid conditions: young trees (1–2 years) and older systems (>2 years) further divided into pruned and unpruned practices. Juvenile trees had a significantly higher %Ndfa than mature trees in unpruned systems while mature trees in pruned systems did not differ significantly from either juvenile or unpruned mature trees (analysis of variance followed by Duncan’s Multiple Range Test at 5 %). The ANOVA had a low r2 value, 0.258 with significant P value (0.0176). Thus, no strong conclusions may be drawn but the differences reveal some possible trends that may be fruitful for further research together with causes for the ranges observed in several cases included in Table 2. Especially a meta-analysis on these effects will be interesting when more data become available.

The ranges reported in Table 2 were caused by five factors. Differences because of reference species (Ståhl et al. 2005) or tree organ (Chikowo et al. 2004) are methodological issues. A range of N2 fixation estimates based on various reference species would be more reliable than single value because the range also gives an estimate of the reliability of the estimates. However, the range was reported only by Ståhl et al. (2005). Interestingly, the same reference species gave very similar N2 fixation estimates in another study and the seasonality had a much stronger effect on the N2 fixation estimates (Ståhl et al. 2002). An estimate based on the average δ15N of several non-N2-fixing references was reported in three other articles (Aronson et al. 2002; Lesueur and Sarr 2008; Nygren and Leblanc 2009). The estimates based on one or several herbaceous references (Chikowo et al. 2004; Mercado et al. 2011; Unkovich et al. 2000) should be dealt with caution; we agree with Unkovich et al. (2008) that the reference should be of the same life form as the N2-fixer. The wide range of estimates for different tree organs (Chikowo et al. 2004) may be caused by the use of the Hyparrhenia rufa (Nees) Stapf grass as the reference; thus, organs of the reference are not comparable with tree organs.

Sampling time (Ladha et al. 1993; Nygren and Leblanc 2009; Nygren et al. 2000; Peoples et al. 1996; Ståhl et al. 2002) refers to the seasonality of N2 fixation, which is mostly out of the control of agronomic management. The influence of tree provenance (Raddad et al. 2005) and rhizobial strain (Galiana et al. 1998; Lesueur and Sarr 2008) imply intrinsic factors that regulate N2 fixation. These influences may be managed by proper selection of tree sources, strains of N2-fixing bacteria, and their combinations. Finally, pruning frequency (Kadiata et al. 1997; Nygren et al. 2000) is essentially a management issue under the control of the farmers.

Dryland agroforestry

It is obvious from the data collected in Table 2 that dry environment poses challenges for the use of N2-fixing trees in AFS. Although the dry environment seems to reduce the average dependence on N2 fixation, the high within-group variation in the data indicates that some woody legumes may form N2-fixing symbiosis well adapted to dry habitats. In fact, the mean for dry environments was biased upwards because of high %Ndfa observed in Chamaecytisus proliferus (L. f.) Link in dry Central Chile (84 %; Aronson et al. 2002) and Australia (83 %; Unkovich et al. 2000). These numbers were comparable only with juvenile G. sepium under constantly humid conditions in Costa Rica (85 %; Kurppa et al. 2010). The adverse effect of dry environment on N2 fixation was also observed in a study of 11 Mimosoideae species in natural ecosystems of Namibia with %Ndfa typically varying 10–30 % (Schulze et al. 1991). They also observed considerable variation between species with %Ndfa of 49 % in two species, the maximum of 71 %, and the lowest indicating no N2 fixation. High variation (%Ndfa 24–61 %) was also observed between Acacia senegal provenances in the Sudan (Raddad et al. 2005).

Faidherbia albida (Delile) A. Chev., a tree species now strongly recommended for AFS in semi-arid Africa (Garrity et al. 2010), was cited as a N2-fixer in 39 references found in CAB Abstracts data base but only twice N2 fixation was estimated in the field. Active N2 fixation with %Ndfa 54 % at 15-month-age was observed in Senegal (Gueye et al. 1997), while no N2 fixation was observed in mature trees over a rainfall gradient 30–400 mm year−1 in Namibia (Schulze et al. 1991). The former result is based on a provenance that was a superior N2-fixer in a pot study and the trees were apparently very small, with total N accumulation of only 1.41 g [N] per seedling (Gueye et al. 1997). Functional nodules seem to be present in F. albida only during 2-3 months each year when the soil is superficially humid at the end of the rainy season and beginning of the dry season when tree foliage grows (Roupsard 1997). High variation has been observed in N2 fixation of F. albida seedlings in controlled environments (Gueye et al. 1997; Ndoye et al. 1995; Sellstedt et al. 1993). Thus, field estimates on N2 fixation in different environments and AFS are needed for this important tree species. Under field conditions, vigorous provenances from Burundi fixed more N2 than provenances from Niger and Burkina Faso (Roupsard 1997).

Among actinorrhizal plants, Casuarinaceae species, mostly used in AFS in arid and semi-arid regions, are known for their drought and temperature resistance and Frankia spp. isolates obtained from these species have high optimal growth temperatures (Dawson 2008).

These results indicate that it is possible to find tree species and provenances that are active N2-fixers in dry environments. Thus, we recommend screening trials of N2-fixing tree species and provenances for selecting suitable germplasm for dryland agroforestry. No potentially “N2-fixing tree” should be recommended for dryland agroforestry before its N2 fixation capacity is verified.

Dinitrogen-fixing trees in acidic soils

Acidic soils of the humid tropics are often cited as a major constraint to N2 fixation (Kass 1995) because most rhizobia and many Frankia strains do not tolerate high acidity and consequent solubilisation of aluminium. However, stains tolerant to acidity exist, especially among Bradyrhizobium spp. (Graham 1992) and N2-fixing trees are abundant in the humid tropical forests with acidic soils (McKey 1994; Roggy et al. 1999a, b). Although some Frankia strains survive in various acidic soils, a negative correlation was found between soil acidity and nodulation in Alnus spp. (Smolander and Sundman 1987) and Elaeagnus angustifolia L. (Jamann et al. 1992; Zitzer and Dawson 1992). We did not observe any trend related to soil acidity when compiling data for Table 2.

In a screening of N2-fixing potential of legume trees, Roggy et al. (1999b) classified 110 species as supposed N2-fixers and 33 species as supposed non-N2-fixers in a rain forest in French Guiana. In a selection trial of 25 legume tree species for forestry and agroforestry in Costa Rica, 18 species were inspected for nodulation and 9 for nitrogenase activity; all of them appeared to be N2-fixers, including Acacia, Albizia, Dahlbergia, Erythrina, Inga, and Pithecellobium spp. (Tilki and Fisher 1998). As is unfortunately common, no attempts to estimate N2 fixation were made in these studies but they indicate that a large collection of potential N2-fixers exists for acidic soils. Promising N2-fixers for AFS under these conditions include Acacia mangium, Codariocalyx gyroides (Roxb. ex Link) X.Y. Zhu, Erythrina fusca Lour., E. poeppigiana, Inga edulis, and G. sepium that has been introduced from seasonal to the humid tropics (Tables 2 and 3). Another introduction from seasonal climate, C. calothyrsus, has variable performance (Table 2; Peoples et al. 1996).

Host-bacteria interaction

Rhizobial strain may have a significant effect on N2 fixation by the symbiotic system (Lesueur and Sarr 2008). Many of the studies on the host × rhizobia interaction have been conducted with seedlings under controlled conditions (André et al. 2005; Bala and Giller 2001; Makatiani and Odee 2007; Weber et al. 2005). Although they clearly indicate the importance of the interaction, practical conclusions are hard to make beyond nursery production and early establishment in the field. Lesueur and Sarr (2008) observed that apparent growth differences between triple symbiotic systems of C. calothyrsus host and rhizobial and AMF strains disappeared after 5 months in the field. However, plants inoculated with AMF had higher foliar N, P and K content until 12 months after transplanting. According to the genetic analysis, one of the rhizobial strains persisted in the nodules of C. calothyrsus also in the field. Rhizobia-inoculated Acacia mangium grew better than non-inoculated trees up to 39 months after transplanting with detection of the efficient strains in nodules 42 months after transplanting (Galiana et al. 1998). The longest persisting effect of which we are aware, is the detection of rhizobial strains used for inoculating Leucaena leucocephala 10 years after introduction in a soil with only a few native rhizobia capable of nodulating it (Sanginga et al. 1994).

Table 1 indicates high diversity of rhizobia and Frankia spp. nodulating different tree genera. Further, both controlled-environment studies (Bala and Giller 2006) and genetic analyses of the rhizobia infecting legumes in the field (Diouf et al. 2007) indicate a high intra-specific genetic diversity among the rhizobia. Thus, the importance of rhizobial inoculation, at least with selected commercial strains, is questionable in small-holder agriculture and scarce resources may be better used for management improvements (Giller and Cadisch 1995). In large-scale nursery production, however, good legume-rhizobia combinations may improve the seedling establishment in the field (cf. Galiana et al. 1998; Lesueur and Sarr 2008). The observation that Rhizobium tropici type rhizobia are the most efficient symbionts in acidic soils, Mesorhizobium strains in intermediate, and Sinorhizobium strains in alkaline soils (Bala and Giller 2006) indicates that the host × rhizobia interactions may differ in different soils. Thus, only local trials may be valid for a particular environment.

Use of selected inoculants may be necessary when N2-fixing trees are introduced to new areas. When exotic legume or actinorrhizal trees have been planted in new locations where they did not occur previously, scarce nodulation or absence of nodulation has been observed (Diem and Dommergues 1990; Sanginga et al. 1994; Woomer et al. 1988). It has been postulated that in the absence of a compatible host, rhizobial or Frankia strains are confronted with competition from other soil bacteria and cannot maintain their population or can lose host specificity encoding genes (Barnet 1991).

Green pruning

Green pruning of N2-fixing trees is a common practice in many AFS for avoiding excessive shading of the crop (Kang et al. 1981; Akinnifesi et al. 2008), enhancing nutrient cycling (Beer et al. 1998; Kass et al. 1997), or harvesting tree fodder (Peoples et al. 1996). Symbiotic N2 fixation is often mentioned as an important benefit in these systems because it is assumed to contribute to increase in soil organic matter and N reserves (Beer et al. 1998; Haggar et al. 1993) or to produce protein-rich fodder (Peoples et al. 1996). Data in Table 2 indicate that pruned trees are often active N2-fixers with higher %Ndfa in pruned than unpruned systems. Because green pruning may result in partial rejuvenation of the trees (Nygren et al. 2000), the higher mean %Ndfa in pruned systems may be related to the higher dependence on N2 fixation in young trees also observable in Table 2. The placement of the pruned trees as an intermediate group between the juvenile and unpruned mature trees in the statistical analysis of Table 2 provides partial support for this “rejuvenation hypothesis”.

Not all trees fix actively N2 under a periodic pruning regime. Complete nodule turnover was observed in Erythrina poeppigiana in 2 weeks after complete pruning. Renodulation initiated at 10 weeks after pruning and nodulation was abundant at 14 weeks after pruning (Chesney and Nygren 2002; Nygren and Ramírez 1995). Leaving only 5 % of foliage in prunings twice-a-year was enough to retain low nodulation in 2-year-old trees and caused stabilisation of nodule growth rather than turnover in 8-year-old trees (Chesney and Nygren 2002). The nodule turnover after complete pruning of E. poeppigiana seemed to be a response to C starvation because of cessation of photosynthesis. Nodulation after foliage regrowth appeared to be regulated by canopy N needs with high nodulation when N flow was unidirectional from roots to foliage and reduction in nodule biomass when N flow to roots initiated (Nygren 1995), in line with the theoretical scheme on the regulation of N2 fixation presented in Fig. 1.

Even stronger negative response to green pruning was observed in Erythrina lanceolata, in which nodulation did not recover from complete pruning twice-a-year and partial pruning every three months caused a significant reduction in nodulation and %Ndfa in comparison to unpruned control (Salas et al. 2001). In contrast, pruning regime did not affect the %Ndfa in Gliricidia sepium but total N2 fixation in mass terms was reduced by bimonthly partial prunings in comparison to twice-a-year complete pruning (Nygren et al. 2000). Calliandra calothyrsus, Codariocalyx gyroides (Peoples et al. 1996), Albizia lebbeck (L.) Benth., and Leucaena leucocephala (Kadiata et al. 1997) also actively fixed N2 when pruned heavily. These data indicate that responses to green pruning are species-specific and N2 fixation should be estimated in all trials for selecting trees for AFS managed by pruning.

Whole tree N2 fixation

Much less data were available on the N2 fixation by agroforestry trees in terms of kg [Ndfa] ha−1 year−1, including fixed N in the root system (Table 3) than on the %Ndfa. Dulormne et al. (2003) reported the estimate based on all fixed N tracked in a cut-and-carry system of G. sepium and fodder grass Dichanthium aristatum, i.e. in the trees, grass, and soil over 12 years. The whole system N balance was also used to estimate N2 fixation by 7-year-old Acacia mangium, Acacia auriculiformis A. Cunn. ex. Benth. (Bernhard-Reversat, 1996), and Acacia polyacantha Willd. (Harmand 1998). Other data refer to fixed N in the trees and litter during a time range from 18 months (Ståhl et al. 2005) to 12 years (Mercado et al. 2011). All data were converted to correspond to the annual N2 fixation. Table 3 indicates a significant contribution of N2 fixation to the N economy of the trees or whole system. Trees pruned in AFS seem to fix more N than unpruned trees also in mass terms as the top three systems (Dulormne et al. 2003; Peoples et al. 1996; Unkovich et al. 2000) were managed by tree pruning.

Mafongoya et al. (2004) cite a “typical” range of N2 fixation by trees in AFS to be 70–200 kg [Ndfa] ha−1 year−1. Six out of the 16 cases compiled in Table 3 exceeded this range, probably because we accepted only estimates based on whole tree harvesting—including root system – or whole system N balance. Schroth et al. (1995) estimated that the N reserves in roots of 0–5 mm diameter of six legume tree species were 60–133 kg [N] ha−1 in 5-year-old fallows in a sub-humid area in Côte d’Ivoire. A similar amount, 71.5 kg [N] ha−1, was observed in fine roots of 0–2 mm diameter of Inga edulis shade trees in an organically-grown cacao plantation under humid conditions in Costa Rica (Gómez Luciano 2008). Higher amount, 190 [N] ha−1, was observed in fine roots of 0–2 mm in a 7-year-old Acacia polyacantha fallow in the subhumid zone of Cameroon (Harmand et al. 2004); whole root system of A. polyacantha contained 342 kg [N] ha−1. Giller (2001) refers to the range of 26–60 % of legume tree N to be below-ground, under variable field and experimental conditions.

Based on harvestable biomass only, it was estimated that G. sepium would fix only 147 kg [N] ha−1 year−1 (Nygren et al. 2000) in the system studied by Dulormne et al. (2003). However, N2 fixation by G. sepium is the only N input to this system and N balance taking into account N export in fodder harvest, N uptake by the companion grass, and soil N accumulation revealed the considerably higher estimate (Table 3). Thus, it is obvious that below-ground N as well as fixed N released by the trees to the environment should be taken into account when estimating the total amount of N2 fixation in mass terms. These data are still scarce in agroforestry literature.

Agronomic importance of N2 fixation by agroforestry trees

State-of-art in agroforestry systems

Our review on the estimates of N2 fixation by several tree species used in AFS shows that many of them are active N2-fixers (Table 2) and symbiotic N2 fixation may contribute annually tens or hundreds of kg [N] ha−1 to the farming system (Table 3). A critical review of the original literature in the light of current knowledge on the methods for quantifying the N2 fixation resulted in 16 estimates on the annual N2 fixation per hectare that we considered reliable enough for Table 3. Obviously, more data are needed at field- and farm-level. Even less data are available for evaluating the agronomic importance of N2 fixation in AFS at landscape, regional, or global level. Two recent reviews on the effects of “fertiliser trees” on crop productivity (Akinnifesi et al. 2010; Garrity et al. 2010) suggest that inclusion of N2-fixing trees into traditional cropping systems may significantly improve crop yields yet evaluation of the contribution of the symbiotic N2 fixation to the benefits was not evaluated.

The agronomic importance of N2 fixation by the trees depends on the function of the trees in an AFS. Nitrogen in soil and plant residues, including roots, after cutting an improved fallow is important for the succeeding crop (Chikowo et al. 2004; Mercado et al. 2011; Ståhl et al. 2002; 2005). Nitrogen recycling in the pruning residues potentially benefits the crop in alley cropping and other green manure systems (Haggar et al. 1993; Ladha et al. 1993; Unkovich et al. 2000), while N2 fixation by the trees reduces competition for soil N with the crop in unpruned shade tree systems and part of the fixed N is recycled to the crop in leaf and root litter of the trees (Escalante et al. 1984; Nygren and Leblanc 2009; Salas et al. 2001; Santana and Rosand 1985). Animal diet is enhanced by protein-rich tree foliage of the N2-fixing trees in fodder production systems, (Blair et al. 1990; Peoples et al. 1996).

All N2-fixing plants form the symbiosis for their own N supply and release N to the environment only when they have it in excess (Fig. 1). Nitrogen fixed by the trees may become available for the crop via indirect or direct pathways. Here, the indirect pathway refers to the complete microbial N cycle, including mineralisation of organic N in the litter or pruning residues to NH4+, partial immobilisation by soil microbiota and mobilisation because of microbial turnover, and nitrification of NH4+. While NH4+ is efficiently fixed by cation exchange on soil colloids, NO3 may be lost from the system via leaching to deep soil layers or denitrification, especially if it is available in excess with respect to plant N needs. Soil N cycle has been widely studied – also in the AFS with legume trees (Babbar and Zak 1994, 1995; Dulormne et al. 2003; Hergoualc’h et al. 2007, 2008; Kanmegne et al. 2006; Mafongoya et al. 1998)—and a discussion of the vast literature on the topic is out of our scope. Direct pathway refers here to the transfer of N from N2-fixing trees to crops either via a common mycelial network (CMN) of mycorrhizae-forming fungi colonising both trees and crops (He et al. 2003; Jalonen et al. 2009b) or absorption of tree root exudates by the crop before the complete decomposition cycle (Fustec et al. 2010; Jalonen et al. 2009a). The latter pathway requires that either crop plants absorb simple amino acids (Lambers et al. 2008; Näsholm et al. 2009) from the root exudates or the N2-fixing plants exude inorganic N (Fustec et al. 2010).

Fate of fixed nitrogen

In spite of the low C:N ratio (Mafongoya et al. 1998), legume tree mulches seem to have a relatively low efficiency as an immediate N source for a crop. Maize (Zea mays L.) gained 11 % of its N, ca. 10 kg [N] ha−1, from pruning residues of Gliricidia sepium and Erythrina poeppigiana in an alley cropping trial under humid tropical conditions (Haggar et al. 1993). In another experiment, maize used ca. 10 % of the N available in the pruning residues of Leucaena leucocephala (Vanlauwe et al. 1996). This apparently low efficiency depends on a cascade of factors. First, although amino acids and oligopeptides decompose fast, some proteins decompose slowly or may even be recalcitrant (Paul and Clark 1996). Thus, not all residual N becomes available for an annual crop during a cropping cycle.

Second, the N release is also restricted by the general litter quality: if the mulch C is not easily available for the decomposer microbes as an energy source, also N release from the residue slows down. Cellulose and hemicellulose are easily decomposable C compounds while lignin is recalcitrant (Paul and Clark 1996). Further, many legume mulches have high content of polyphenols that may be toxic to the decomposers. Nitrogen release rate is significantly higher from the legume mulches with a high N to polyphenols ratio than from residues with a low ratio (Barrios et al.1997, Ndufa et al. 2009). Following to these factors, 10–30 % of N in legume mulches is released in a month (Mafongoya et al. 1998) typically followed by an exponential decrease in the N release rate with time.

Third, soil microbiota are strong competitors for soil N with the plants; free N in soil solution is absorbed by microbes on average within 24 h after its release (Jones et al. 2005). Mobility of N in soil depends on soil and N type. In a loamy soil, the 24-h diffusion distance of NH4+, amino acids, and NO3 is ca. 1.7, 1.5, and 10.2 mm, respectively (Jones et al. 2005). Thus, N mineralisation must occur in the rhizosphere to be useful for a plant. Otherwise, it will be absorbed by microbiota. Based on the 24-h diffusion distance and data on fine root densities of cacao in a plantation with Inga edulis shade (Gómez Luciano 2008), we estimated that the cacao rhizosphere for NH4+ absorption was 45.6 % out of total soil volume in the 0–2 cm soil depth but only 16.4 % in the 2–10 cm depth (Nygren and Leblanc, unpublished). If the soil has a high nitrification rate, the more mobile NO3 is in better supply for plants but the high mobility makes it also subject to leaching to deeper soil layers, where it may become unavailable for plant roots (Babbar and Zak 1995; Harmand et al. 2007a).

The apparently low efficiency of new mulch as a crop N source has led some authors to argue that legume tree mulch is more important for long-term build-up of soil organic matter and N reserves than immediate crop N nutrition (Beer et al. 1998; Haggar et al. 1993; Kass et al. 1997). We partially share this opinion. However, the role of root litter in the N cycle of AFS has been neglected, although Nye and Greenland (1960) proposed more than 50 years ago that dead tree roots may be an important nutrient source for associated crops because they decompose in close proximity of the absorbing crop roots. Pot experiments with agroforestry combinations of maize with Paraserianthes falcataria (L.) I.C. Nielsen (Chintu and Zaharah 2003) and cacao with I. edulis (Kähkölä et al. 2012) suggest that root litter of the legume tree may be a more efficient N source for the crop than leaf mulch or litter. Further, ca. 300 kg ha−1 year−1 of N fixed by G. sepium was recycled below-ground to the soil and the associated fodder grass Dichanthium aristatum in a cut-and-carry fodder production system, where above-ground litter was eliminated by intensive fodder harvesting (Dulormne et al. 2003).

Direct N transfer between plants

Reviews of N transfer via CMN (He et al. 2003; Simard et al. 2002) and N exudation from legume roots (Fustec et al. 2010; Wichern et al. 2008) have been recently published but data on AFS are scarce. Direct N flow between plants is bidirectional and its net effect may be almost nil to both components (He et al. 2004; Johansen and Jensen 1996). Net flow seems to be more common from a N2-fixing to a non-N2-fixing plant; e.g. from soybean to maize (Bethlenfalvay et al. 1991), from Alnus subcordata C.A. Mey. and Elaeagnus angustifolia to Prunus avium L. (Roggy et al. 2004), and from soybean and peanut (Arachis hypogaea L.) to associated weeds (Moyer-Henry et al. 2006). In a pot study, direct N transfer from A. senegal to durum wheat (Triticum turgidum L.) was enhanced when crop N uptake was stimulated by high P availability and competition was low (Isaac et al. 2012).

In several coffee plantations in Burundi with different legume shade tree species, 6–22 % of N in coffee leaves was of atmospheric origin (Snoeck et al. 2000). In a cut-and-carry fodder production system, 27–35 % of N in D. aristatum grass (53–68 kg ha−1) was of atmospheric origin. Nitrogen isotopic data suggested that atmospheric N was directly transferred from the associated heavily pruned G. sepium (Sierra and Nygren 2006). In a cacao plantation with I. edulis shade, 8–25 % of N in cacao leaves was of atmospheric origin (Nygren and Leblanc 2009). Data on N isotopic relationships also suggest the possibility of direct N transfer from L. leucocephala to understorey weeds in a short-rotation plantation (van Kessel et al. 1994) and from Inga oerstediana Benth. ex Seem. to coffee and non-legume shade tree Liquidambar styraciflua L. in organic coffee farms in Chiapas, Mexico, (Grossman et al. 2006) although the authors themselves did not draw this conclusion.

Identification of the direct N transfer pathways is difficult under AFS field conditions. Root exudates of many legume trees have a remarkably low C:N ratio: 3.3–5.7 in G. sepium (Jalonen et al. 2009b), 4.7–6.8 in Robinia pseudoacacia (Uselman et al. 1999), and 5.2 in I. edulis (Nygren and Leblanc, unpublished data). Thus, they are potentially good N sources for an associated crop but tree and crop rhizospheres should overlap for the tree root exudates to be useful for the crop.

Formation of an effective CMN between trees and crops in an AFS requires that both share compatible strains of mycorrhizal fungi: anastomoses, functional connections that allow material transfer between mycelia of two fungi, have been observed only between fungi of the same population yet they may differ genetically (Croll et al. 2009). Glomus etunicatum Becker & Gerdemann and Gigaspora albida Schenck & Perez used for inoculating Calliandra calothyrsus spread their mycelia also to associated maize and common bean (Phaseolus vulgaris L.) (Ingleby et al. 2007). Millet (Pennisetum americanum (L.) Leeke) was effectively colonised by inocula from Acacia nilotica (L.) Willd. ex Delile, A. tortilis (Forssk.) Hayne, and Prosopis juliflora (Sw.) DC. (Diagne et al. 2001). Inga edulis positively responded to AMF originating from roots of cacao in a cross-inoculation trial indicating the potential of formation of a CMN between these species (Iglesias et al. 2011). In a pot study, N was transferred from G. sepium to D. aristatum both via the CMN and root exudates (Jalonen et al. 2009b). Under field conditions, both species were colonised by Rhizosphagus intraradices (ex. Glomus intraradices) and grass colonisation was more abundant between the tree rows than in an adjacent pure grass plot (Jalonen et al. 2012).

It seems obvious that the potential for direct N transfer from legume trees to crops in AFS exists and positive evidence on this interaction has been found in a few cases. However, the phenomenon requires further study before we well understand its occurrence and importance in different AFS.

Management of N2 fixation

Possibilities to manage N2 fixation and N recycling are quite limited in AFS with unpruned N2-fixing trees. According to the data in Table 2, sampling time was one of the factors affecting the %Ndfa. We interpreted this to reflect the seasonality of N2 fixation. If the climate is seasonal, the trees and crops often follow the same phenological cycle with higher metabolic activity and productivity during the same season. Because N2 fixation seems to be connected at least to some extend to the N needs of the canopy (Fig. 1; Nygren 1995), it is most active during the time of the highest general metabolic activity. Follow-up of annual biomass and N2 fixation dynamics of Erythrina lanceolata used as living support for vanilla (Vanilla planifolia Andrews) under a seasonal climate in Costa Rica indicated that main litterfall occurred during the dry season (Berninger and Salas 2003) when no N2 fixation was detected (Salas et al. 2001). Litterfall was close to nil during time of the most active foliage growth and N2 fixation. Thus, although the timing of N2 fixation and N recycling in unpruned multistrata AFS has not received much attention, we may expect the N2 fixation to occur at the time of the highest productivity of both trees and crops, thus, reducing competition between trees and crops.

An interesting case is the “reverse phenology” of Faidherbia albida in the African drylands (Garrity et al. 2010): the tree is leafy during the dry season. The low water consumption during the cropping season and the use of deep-water reserves out of cropping season reduce competition for water with the associated crops (Roupsard et al. 1999). However, N2 fixation is effective only 2–3 months during a year when foliage growth begins (Roupsard 1997). Because N2 fixation occurs out of the cropping season F. albida adds N for crop via litterfall at the beginning of cropping season. Green pruning may also reverse phenology; Central American cattle ranchers prune G. sepium at the beginning of dry season so that the trees remain leafy and provide dry season fodder (Hernández and Benavides 1994).

Green pruning is generally timed according to the crop phenology. Costa Rican coffee farmers traditionally prune shade trees leaving only a few branches (ca. 5–10 % of foliage) to promote coffee flowering at the end of the drier season and make a complete pruning for promoting the ripening of berries about half year later. The highest coffee N demand occurs 6–17 weeks after blossoming when the growing berries may consume 95 % of newly assimilated N (DaMatta et al. 2007). The popular shade tree E. poeppigiana remains unnodulated about 10 weeks following a complete pruning (Nygren and Ramírez 1995) but it retains reduced nodulation after a partial pruning (Chesney and Nygren 2002). Similar pattern has been observed in G. sepium but its renodulation rate is higher after a pruning (Nygren and Cruz 1998).

Thus, in coffee AFS we may envision partial synchrony between coffee N needs and N2 fixation and N recycling from the shade trees. The green mulch from the pruning to promote blossoming may release N for active coffee foliage growth that occurs at the time of coffee flowering (DaMatta et al. 2007) but trees fix only limited amount of N2 for their own foliage regrowth, thus, potentially competing for soil N and reabsorbing N released from the pruning residues. Nitrogen released from nodule (Escalante et al. 1984; Nygren and Ramírez 1995) and fine root turnover (Chesney and Nygren 2002) forms an additional N source, which may be more efficient than above-ground mulch (Chintu and Zaharah 2003; Kähkölä et al. 2012). During the high coffee N need for growing berries (DaMatta et al. 2007), the shade trees probably fix actively N2, which reduces competition between the trees and the crop. The coffee N needs are low during the second annual pruning for promoting berry ripening (DaMatta et al. 2007), and the residual N probably contributes to regrowth of the trees themselves, accumulation of soil organic N, or N losses through NO3 leaching.

In alley cropping, hedgerow trees are pruned before sowing of the annual crop and, if the crop requires and trees tolerate, in a later phase of cropping season (Akinnifesi et al. 2010; Haggar et al. 1993; Kang et al. 1981; Vanlauwe et al. 1996). The pruning often refers to complete defoliation. Thus, we may expect that a complete nodule turnover follows (Nygren and Cruz 1998; Nygren and Ramírez 1995) and trees do not fix N2 at the beginning of the cropping cycle. Nitrogen recycled in pruning residues is a strong input to the soil when the crop is relatively small. Many authors recommend a combination of rapidly and slowly decomposing tree mulches (“high-quality” and “low-quality” mulch) for enhancing the synchrony between the crop N demand and N release (Mafongoya et al. 1998; Nair et al. 1999; Palm 1995; van Noordwijk et al. 1996). These authors, however, ignored the effect of nodule turnover on trees’ N uptake. We argue that temporarily non-N2-fixing trees enforce the root safety-net (van Noordwijk et al. 1996) by at least partially reabsorbing N leaching from the mulch application. Further, the nodule and root turnover following the pruning may form a “high-quality mulch” that releases N at the beginning of the cropping cycle. The trees renodulate and fix N2 – unless too heavily pruned – during the crop flowering and seed production, thus, reducing competition with the crop.

Although at least partial synchrony between N2 fixation, N recycling, and crop N use is possible in AFS with both perennial and annual crops, it must be noted that precision agriculture is not possible in green manure systems. Thus, the main benefit of N2-fixing trees in AFS is probably the long-term accumulation of soil N reserves (Haggar et al. 1993) although below-ground N recycling and direct N transfer between plants probably provide N for more immediate crop N needs than foliage litter or mulch applied on soil surface (Chintu and Zaharah 2003; Dulormne et al. 2003; Jalonen et al. 2009b; Kähkölä et al. 2012). The long-term accumulation of organic matter and nutrients to soil is essential in improved fallow systems, in which cropping relies on the residual N after the fallow enhanced with N2-fixing trees (Chikowo et al. 2004; Mercado et al. 2011; Ståhl et al. 2002, 2005). Few attempts have been made for studying the fate of the high N release to soil after clearing the fallow for cropping. Nitrate leaching rates two year after clearing 7-year-old Acacia polyacantha, Eucalyptus camaldulensis Dehnh., and Senna siamea fallows were similar, ca. 10–15 kg [NO3–N] year−1 (Oliver et al. 2000).

Crop N needs and tree N2 fixation

Perennial crops coffee, cacao, and tea (Camellia sinensis (L.) Kuntze) cover globally ca. 21.6 Mha of agricultural land (Table 4). Perennial crops are often grown in AFS with N2-fixing shade trees (Beer et al. 1998; Kass et al. 1997) although global estimates are missing for other perennials than cacao; 7.8 Mha of cacao is cultivated in AFS (Zomer et al. 2009), which is 89 % out of the global cacao cultivation area of 8.73 Mha (Table 4). Central American, Caribbean, and Andean coffee is mostly shaded while full-sun cultivation prevails in large Brazilian plantations.
Table 4

Global cultivation area and production of perennial crops in 2009 (FAOStat 2011), and an estimate of annual N export in the harvest of these crops

Commodity and area

Total area (ha)

Total production (Mg)

Yield (Mg ha−1)

N export (kg ha−1 year−1)

Total N export (Mg year−1)

Cacao beans

     

 Sub-Saharan Africa

5,935,274

2,639,788

0.445

16.0

95,032

 Tropical Americas

1,585,728

548,360

0.346

12.4

19,741

 South and East Asia

1,068,476

837,766

0.784

28.2

30,160

Worlda

8,733,093

4,082,270

0.467

16.8

146,962

Green coffee

     

 Sub-Saharan Africa

2,060,527

1,006,318

0.488

14.7

30,190

 Tropical Americas

5,474,941

4,807,644

0.878

26.3

144,229

 South and East Asia

2,250,612

2,468,351

1.097

32.9

74,051

Worlda

9,841,317

8,342,636

0.848

25.4

250,279

Tea leaves

     

 Sub-Saharan Africa

283,161

530,992

1.875

106.9

30,267

 Tropical Americas

45,539

85,477

1.877

107.0

4,872

 South and East Asia

2,680,869

3,326,333

1.241

70.7

189,601

Worlda

3,014,909

3,950,047

1.310

74.7

225,153

aIncludes areas of low production not listed separately

We calculated the average yield of the perennial crops based on the 2009 cultivation area and production data (FAOStat 2011) and estimated the N export in crop harvest using the estimates on N needed for producing 1 Mg of harvest according to Bertsch (2003): 30, 36, and 57 kg Mg−1 for green coffee, cacao beans, and tea leaves, respectively. The average yields of all these three crops were relatively modest and N export in the harvest was, consequently, quite low (Table 4). In addition to the N export in the harvest, part of N is used for producing permanent biomass. We compiled data on N accumulation to permanent biomass of coffee, cacao, non-legume shade trees Cordia alliodora (Ruiz & Pav.) Oken and Eucalyptus deglupta Blume, and legume shade tree Erythrina poeppigiana in Table 5. Combining these few data with statistics in Table 4, we can roughly estimate that on average ca. 40–50 kg [N] ha−1 year−1 is exported in the average harvest or immobilised in permanent biomass in unshaded cacao and coffee plantations and ca. 60–80 kg [N] ha−1 year−1 in shaded plantations.
Table 5

Nitrogen accumulation to the permanent woody biomass in coffee (Coffea arabica L.) and cacao (Theobroma cacao L.) plantations

Species

Management

N accumulation (kg ha−1 year−1)

References

Coffea arabica

Shaded

11

Fassbender (1987)

Cordia alliodora

Free growth

25

 

System

 

36

 

Coffea arabica

Shaded

11

Fassbender (1987)

Erythrina poeppigiana

Pruned

11

 

System

 

22

 

Coffea arabica

Unshaded

15

Harmand et al. (2007a, b)

Coffea arabica

Shaded

13

Harmand et al. (2007a, b)

Eucalyptus deglupta

Free growth

7

 

System

 

20

 

Theobroma cacao

Shaded

9

Fassbender et al. 1988

Cordia alliodora

Free growth

28

 

System

 

37

 

Theobroma cacao

Shaded

9

Fassbender et al. 1988

Erythrina poeppigiana

Free growth

30

 

System

 

39

 

Table 3 shows that many trees commonly used in AFS may fix enough N for fulfilling the current or even increasing N needs of perennial crops. As green pruning seems to enhance N2 fixation both proportionally (Table 2) and absolutely (Table 3), it is a recommended practice for shade trees over perennial crops. Green pruning of legume trees Erythrina fusca, E. poeppigiana, and I. edulis under typical management potentially recycles 80, 70–115, and 100 kg ha−1 year−1, respectively, of N fixed from atmosphere to the soil (Leblanc et al. 2007), which is enough for compensating N accumulation to permanent crop biomass and harvest loss in1.5–3 Mg ha−1 yield.

Intensively managed large-scale perennial crop plantations may be over-fertilised; e.g. the typical coffee fertilisation rate in Costa Rica ranging from 150 (Harmand et al. 2007a) to 350 kg [N] ha−1 year−1 (Hergoualc’h et al. 2007) would be enough for compensating the N export in the harvest of 4–11 Mg ha−1 of green coffee yet the average yield in Costa Rica is 0.93 Mg ha−1 (calculated from cropping area and production data in FAOStat 2011). This results in NO3 leaching loss equivalent to N needed for producing 1–4 Mg of green coffee (Babbar and Zak 1995; Harmand et al. 2007a, b) and increased emissions of nitrous oxide (N2O) (Hergoualc’h et al. 2008), which is a greenhouse gas with 206 times higher atmospheric forcing potential than CO2. Including N2-fixing trees to highly fertilised coffee farms increases N2O emissions but the emissions peak after fertilisation in both shaded and unshaded systems (Hergoualc’h et al. 2008).

Perhaps the best known practice for combining N2-fixing trees with annual crops is alley cropping (Akinnifesi et al. 2008; Kang et al. 1981) that was intensively studied in the 1980s and 1990s but later abandoned. However, it showed some promise in certain humid and subhumid areas (Akinnifesi et al. 2008; Kass et al. 1997) and an improved version is now widely adopted by small-scale farmers in Southern Africa for maize cropping (Akinnifesi et al. 2008, 2010). The main improvement was achieved by modifying the spatial arrangement of the trees for reducing interference competition between trees and maize. Gliricidia sepium has been the most successful tree species in these systems (Akinnifesi et al. 2008), which have been adopted by over 120,000 Malawian farmers (Garrity et al. 2010).

Palm (1995) estimated that 40 kg of N is needed for 1 Mg of maize yield. Gliricidia sepium may recycle 70–126 kg ha−1 year−1 of N fixed from atmosphere (estimated from alley cropping data in Ladha et al. 1993 and Rowe et al. 1999), which is sufficient for harvesting ca. 1.7–3.1 Mg ha−1 year−1 of maize. Akinnifesi et al. (2010) report even higher sustained maize yields with G. sepium without fertiliser in Malawi (around 4 Mg ha−1 year−1, depending on the year vs. ca. 1.5 Mg ha−1 year−1 in pure maize with low N fertilisation). This may imply that also the tree root safety-net, which retains nutrients in the system out of the cropping season (van Noordwijk et al. 1996), is important for these systems. Further, N2 fixation is probably underestimated because the estimates of Ladha et al. (1993) and Rowe et al. (1999) are based on tree prunings only. Dulormne et al. (2003) found that 160 kg ha−1 year−1 of N fixed by G. sepium accumulated in soil. Direct N transfer to a non-N2-fixing crop may also occur (Sierra and Nygren 2006). As far as we know, the studies by Dulormne et al. (2003) and Bernhard-Reversat (1996) are the only ones that account for the N fixed from atmosphere in the soil and associated crop in addition to the trees. These kinds of studies would be most welcome for different AFS but the main constraint is the lack of sufficient time series for constructing the system N balance over years.

Concluding remarks

Symbiotic N2 fixation in AFS should not be studied as an isolated process because even the “N2-fixing component” in an AFS is a tripartite symbiotic system based on a plant capable of forming symbiosis with both N2-fixing bacteria and mycorrhizae-forming fungi. The bacterial partners of the symbiosis are highly diverse both taxonomically and functionally. So far, 98 rhizobial species forming N2-fixing symbioses with legumes have been identified with most of the diversity probably still remaining unrecognised. Actinorrhizal symbioses are much less studied than legume systems and they form an unexplored resource for AFS. Plants rely more on the N2-fixing symbiosis when the plant N needs exceed soil N supply. Thus, soil N supply reduced in AFS by the crop N use and N export in harvest may enhance the N2 fixation above levels observed in natural ecosystems. Mycorrhizae enhance P nutrition but they also appear to have other still poorly understood functions in the tripartite symbiosis. They also form CMN between trees’ and crops’ symbionts, which create a direct N transfer pathway.

Quantification of N2 fixation under field conditions is a challenging yet necessary task as controlled-environment studies with seedlings provide little information on functioning of mature trees in an AFS. Nitrogen isotopic analyses or detailed, long-term whole system N balances may provide the most reliable estimates on the N2 fixation by trees. The general average of %Ndfa for 19 tree species used in AFS was 59, with a higher proportion in juvenile and pruned trees and lower in unpruned trees. High variability was observed in drylands while N2 fixation was active in most of the AFS in the humid and sub-humid areas. In mass terms, N2-fixation may annually add from tens to hundreds of kilograms of N per hectare to an AFS. Few data were available on fixed N in system components other than the trees. The reports published so far indicate that estimates on annual N2 fixation based on trees only may hide considerable direct transfer of fixed N from trees to crops and an important rhizodeposition of tree N to the soil. It seems that symbiotic N2 fixation is indeed an underestimated resource in AFS.

Variation in the N2 fixation by trees in AFS depends on both intrinsic factors of the trees and their symbionts and the environment. Tree species selection and in some cases inoculation with compatible bacteria are the main options for managing the intrinsic factors. Species and provenance selection is also the only available response to the challenges caused by the macro environment, especially in dryland AFS, while several options exist for managing the microenvironment such as proper spacing of trees and crops. Green pruning that is practiced for reducing crop shading and enhancing nutrient cycling, seems to rejuvenate the trees and maintain high levels of N2 fixation. However, green pruning may temporarily impede N2 fixation by simultaneously reducing both the C flow to the roots and microsymbionts and tree N requirements. This may reduce the synchrony of N2 fixation with crop N needs. The N2-fixing trees mostly improve the crop N supply by long-term build-up of soil N reserves.

We may envision a role for N2-fixing trees in different kinds of AFS. In small-scale, low-input perennial cropping systems, N2-fixing trees may provide enough N for sustained or increasing yields. In large-scale, highly fertilised perennial cropping systems, they contribute to the reduction in the use of industrial fertilisers. Recent developments of legume trees mixed with annual crops show much promise for low-input farming. The highest N2 fixation activity was observed in improved fallows and intensive tree fodder production systems. Thus, we may envision significant contribution of N2 fixation to the production of N-rich mulch for soil improvement in the improved fallows and to the high yields of N-rich browse for domestic animals in protein banks and other tree fodder systems.

In order to manage the symbiotic N2 fixation by trees in AFS, basic research is needed on the functioning of the tripartite tree-rhizobia-mycorrhizae symbiosis; on the tree N rhizodeposition to soil via exudates and root turnover; on the direct N transfer from trees to crops; and if the trees really enhance P supply by excretion of extracellular phosphatases. Applied research is needed on tree and symbiont selection, especially for drylands, and on the effects of AFS management on N2 fixation.

Footnotes
1

In fact, no tree fixes atmospheric N2 because all organisms capable of N2 fixation are Bacteria or Archae. However, in order to avoid repeating the long correct expression “trees forming N2-fixing symbiosis with bacteria” we use the common though inaccurate term “N2-fixing trees” for referring to these trees as a group.

 

Acknowledgments

An early version of this review was presented in the 2nd World Congress of Agroforestry (Nairobi, August 2009). We thank Dr Anne-Marie Domenach for inspiring discussions and comments on a draft of this review. The contribution of PN was funded by the Academy of Finland (grant 129166).

Copyright information

© Springer Science+Business Media Dordrecht 2012