Environmental change and carbon limitation in trees: a biochemical, ecophysiological and ecosystem appraisal


Author for correspondence: Peter Millard Tel: +44 1224318611 Fax: +44 1224311556 Email: p.millard@macaulay.ac.uk

I. Introduction

As a consequence of burning fossil fuels and land use change, atmospheric concentrations of CO2 are rising at a rate unprecedented in the recent history of the planet. The current CO2 concentration of 382 µmol mol−1 (2006) is 38% above the preindustrial figure of 270 µmol mol−1 (which had been stable for thousands of years) and nearly twice as high as it was towards the end of the last ice age (Neftel et al., 1988). In the face of rapidly rising atmospheric CO2 concentrations, there is considerable interest in whether forests will grow faster and so sequester more C, thereby helping to mitigate C emissions and their consequences for global warming. Forests cover over 45% of the land surface of the Earth and account for some 75% of terrestrial net primary production (Melillo et al., 1993). Therefore, as C3 photosynthesis is not yet CO2-saturated, forests potentially offer the possibility of enhanced growth and C sequestration as atmospheric CO2 concentrations continue to rise. At a physiological level, elevation of CO2 results in a direct increase in net photosynthesis, because of both an increased velocity of carboxylation and an inhibition of photorespiration (Long et al., 2004). However, at an ecosystem scale, increased photosynthetic rates are not always translated into faster tree growth (Oren et al., 2001) or C sequestration (Luo et al., 2004) and initial stimulation in above-ground growth often declines with time (Körner, 2006). So is tree growth C-limited?

The origin of the concept of limitation is Liebig's law of the minimum (von Liebig, 1840) and was first applied to investigate the limitation on the productivity of individual plants or crops imposed by an inadequate supply of individual nutrients from managed soils. Since then, element limitation has been used to interpret differences in productivity of natural ecosystems (Chapin, 1980; Vitousek & Howarth, 1991) and, even though still focusing on the measure of plant productivity, the original concept has been developed further in attempts to overcome some of its inherent constraints in explaining the behaviour of integrated systems. The realization that not one but several resources can constrain productivity simultaneously (Chapin & Shaver, 1985) led to the concept of multielement limitation (Bloom et al., 1985) and the resource optimization paradigm (Rastetter & Shaver, 1992). The concept of stoichiometry has provided a wider angle on factors controlling the flow of elements in biological systems, including plant production, in that it stresses the importance of element ratios, particularly C, N and phosphorus (P), at key interaction points of the element cycles (Hessen et al., 2004). The resulting requirement to focus on the interaction between element cycles (Chapin et al., 1987) and their individual time lags in responding to a change of plant resource investments, particularly when addressing links between atmospheric CO2 and soil resources (such as N), has triggered concepts like progressive nitrogen limitation (Luo et al., 2004). Hence, C limitation needs to be assessed in the context of the use of other resources by the tree.

Resource use by plants in relation to their environment is often considered using C as the basic currency. There are several reasons for this common approach. First, C is the major component of the dry matter of plants. Second, physiologists have implicitly assumed that because photosynthesis is limited by CO2, plant functioning can therefore be considered in terms of the C ‘cost’. Third, because of concerns over global environmental change, the C cycle has become a major research theme in its own right, particularly understanding the processes governing C sequestration in ecosystems. As a consequence, our knowledge of the C physiology of trees is far more advanced than that of other nutrients, and so the response of trees to environmental change is often interpreted solely in terms of C. The apparent assumption underlying this approach is that the ability of trees to assimilate and allocate C ultimately regulates their capture and use of other resources and hence growth.

Three other issues in this carbon-centric approach need to be considered. First, many of the physiological experiments have used ‘optimal’ conditions, rather than using conditions that impose constraints of light, nutrients or temperature. Yet these conditions of constraint are more biologically meaningful (Warren & Adams, 2004). Second, the difference between a deficiency and a limitation to tree growth needs to be stressed. Under experimental conditions where nutrient supply is carefully controlled, a nutrient deficiency (which has visible symptoms such as foliage discoloration) is easy to induce. However, outside of intensive forestry plantations such symptoms are very rare for trees growing in soils. Instead, there are fast-growing trees on fertile sites or slower-growing plants where nutrient availability is lower. However, the absence of deficiency symptoms does not in any way preclude a limitation to growth. Finally, most ecological experiments that have manipulated resource availability to trees (e.g. by elevated CO2) have been of a relatively short duration. As a consequence, it has not been possible to consider the effect of potentially complex feedback processes that operate at the ecosystem scale and which can occur over a timescale of years to decades (Fig. 1).

Figure 1.

A schematic representation of the carbon (C) flux through forest ecosystems, highlighting the different scales at which the fluxes in and out are controlled. Three different scales are considered: the leaf, the whole tree (including roots, mycorrhizal fungi and their extrametrical mycelial network in the soil and bacteria in the rhizosphere of roots) and the forest soil. C inputs into the system (→) are represented by the following: 1, C assimilation by Rubisco; 2, C export from leaves to woody tissues; 3, C transfers from the whole tree to soil as litter returns, turnover of roots and mycorrhizal hyphae and exudation. C losses from the system (inline image) are represented by the following: 4, leaching of dissolved organic C from the soil; 5, heterotrophic soil respiration resulting from the turnover of soil organic matter; 6, autotrophic soil respiration and tree respiration (excluding photorespiration); 7, leaf photorespiration. Together, 5–7 determine ecosystem respiration. 8, Within each scale, C inputs and outputs are regulated by the internal status of the compartment involved, that is, leaf and whole-tree physiological status (inline image), and amount of soil organic matter and biogeochemical cycling (inline image), with feedback controls of the C status of the compartment acting upon the inputs and outputs (inline image). Two main feedback controls operate between scales (inline image): between the leaf and whole-tree scale through physiological processes operating over a timescale of seconds to days (A); and between the soil and whole tree, through processes operating at a timescale of hours to decades (B).

To overcome some of these potential problems in interpreting experimental data, an alternative approach is to consider evidence over a range of temporal or spatial scales. Körner (2003) published a study which purportedly showed that tree growth in four climatic zones (boreal and mountain, temperate, Mediterranean and tropical) was not limited by the availability of C. He showed that, with only one exception (dry mid-summer in the Mediterranean), nonstructural carbohydrate pools (as a measure of C shortage or surplus for growth) were maximal when tree growth was reduced or zero, irrespective of the time of year. He interpreted these results as meaning that sink activity and its direct control by the environment, or developmental constraints, restricted growth, not C availability (Körner, 2003). The aim of this paper is to extend this approach by reviewing the evidence for a C limitation to tree growth and challenge the ‘carbon-centric’ approach by considering the response of trees to environmental change. Evidence for C limitation will be considered at three different scales (Fig. 1). First, photosynthetic responses at the leaf scale will be considered in relation to the evolution of the carboxylation activity of Rubisco and the role of the protein in N storage. How Rubisco is affected by elevated CO2 will also be considered. Second, at the whole-tree scale, C storage and allocation to roots and associated microbes in order to gain nutrients will be considered and it is argued that any C limitation is only through a limitation to soil nutrient cycling. Finally, we will consider how forest ecosystems are responding to rising atmospheric CO2 concentrations and the soil feedback processes involved.

II. The biochemical scale: the paradox of Rubisco

Rubisco is the most abundant enzyme on the Earth and has two different catalytic functions: catalysing the initial step of the carbon reduction and oxidation cycles, using CO2 and O2 as the substrates, respectively. The two cycles work against each other as the oxidation cycle (photorespiration) uses light energy to release recently assimilated C as CO2. Overall photorespiration decreases the net efficiency of photosynthesis by 20–50%, depending upon temperature (Zelitch, 1973). Compared with the other enzymes in the Calvin cycle, Rubisco is inefficient at carboxylation, with a much lower substrate affinity for CO2 than the other enzymes for their substrates. As C3 photosynthesis is not saturated at current (rising) atmospheric concentrations of CO2, if trees were C-limited, why has there not been a strong evolutionary pressure to increase the efficiency of the carboxylation function of Rubisco during the last 26 million yr, during which atmospheric CO2 concentrations have been lower than at present (Pearson & Palmer, 2000)? This paradox is discussed in relation to C limitation in trees by considering the role of Rubisco as a storage protein, the evolution of CO2 concentrating mechanisms in plants and the response of Rubisco in elevated CO2 studies.

1. The role of Rubisco as an N storage protein

The model of Farquhar et al. (1980) predicts that C3 photosynthesis is limited by the amount of Rubisco and is borne out by the majority of physiological studies which have genetically manipulated the amount of Rubisco in (mainly crop) plants (reviewed by Parry et al., 2003). However, we now know that despite its relative inefficiency at C assimilation, trees can have Rubisco concentrations in excess of that needed for photosynthesis (Warren et al., 2000, 2003; Warren & Adams, 2001; Table 1). The catalytic activity of Rubisco is under complex regulation involving an activase enzyme (Andersson & Taylor, 2003) and at any point in time there are two pools of Rubisco in a leaf, comprising activated and inactive protein. The balance between the two can be modified by N supply, with an increased leaf N status increasing the amount of Rubisco protein, but causing a decline in the Rubisco activation state (Cheng & Fuchigami, 2000; Manter et al., 2005).

Table 1.  The effect of autumn nitrogen (N) supply to Scots pine on the number of current season needles, their dry weight, nitrogen and Rubisco content, and their photosynthetic capacity
Needle parameterN supply after flushing
No NLow NHigh N
  1. Each value is the mean and standard error of five replicate trees; different means across rows (P < 0.05) are postmarked with different letters. Values within a row having the same letter are not statistically significantly different (P < 0.05).

  2. Three-year-old seedlings of Pinus sylvestris were grown in sand culture in individual pots with a N supply of 200 cm3of a nutrient solution containing 2 mol NH4NO3 m3 three times a week from 14 May until 31 July, by which time terminal resting buds had been set. From 1 August to 22 November, the trees received 200 cm3of nutrient solution three times a week but containing 0 (no N), 2 (low N) or 8 (high N) mol NH4NO3 m−3. The photosynthetic capacity of the current year's needles was measured at saturating light conditions (1500 µmol m−2 s−1) and at a leaf temperature of 18°C, and the needles were then harvested and counted and the N, Rubisco content (using the method of Catt & Millard, 1988) and dry weight measured.

Needle number (per tree)1698 ± 423 a1566 ± 324 a1493 ± 132 a
Needle dry weight (mg per needle)  10 ± 1.9 a  10 ± 2.0 a  14 ± 2.9 a
Needle N content (µg per needle)  95 ± 20 a 185 ± 32 b 350 ± 38 c
Rubisco content (mg m−2)  82 ± 29.8 a 770 ± 180 b1506 ± 184 c
Photosynthetic capacity (µmol m−2 s−1)  8.3 ± 1.23 a 12.7 ± 3.43 a  9.3 ± 2.87 a

Accumulation of an excess of Rubisco in response to a generous N supply is because Rubisco has a third function, as a N storage protein. This has been demonstrated in herbaceous plants (Millard, 1988; Quick et al., 1992) and is common in species that have monocarpic senescence, N being released from Rubisco in older (shaded) leaves as they senesce, providing N for both new leaf growth and reproduction (Millard & Catt, 1988; Suzuki et al., 2001). There is evidence that in a range of both deciduous and coniferous, evergreen trees, Rubisco can temporarily store a significant proportion of the N that is used subsequently for leaf growth (Table 2). For deciduous species in temperate ecosystems, for example, this allows a sink for sustained N uptake in summer, once leaf growth has finished, continuing into autumn while leaves senesce (Millard & Thomson, 1989). Coniferous, evergreen trees store N in their youngest class of needles once a terminal bud has been set (Millard & Proe, 1993; Millard et al., 2001), allowing for continued N accumulation (as Rubisco) in the autumn, without growth or enhancement of needle photosynthetic capacity. This was demonstrated in an experiment where young Scots pine trees were given contrasting N supplies once shoot terminal buds had been set and shoot extension growth finished for the year (Table 1). Needle number per tree, needle size and photosynthetic capacity were unaffected, whereas needle N content increased and Rubisco concentrations rose nearly 20-fold in response to a generous autumn N supply (Table 1). In coniferous, evergreen trees, N remobilization from Rubisco occurs during flushes of growth, without needle senescence (Table 2). Taking a ‘carbon-centric’ view, it makes no sense that a plant would store N in Rubisco, given the large energy costs associated with the synthesis of such a complex protein and the need to maintain the machinery for its regulation. However, using Rubisco as an N store compared with either nitrate or amino acids has the advantage of allowing large amounts of N to be stored without osmotic or toxic consequences (Millard, 1988). Also, unlike most other proteins, Rubisco has a very slow turnover rate (and therefore maintenance costs) unless the plant is stressed and leaf senescence is triggered (Suzuki et al., 2001). It is also important to stress that the ability to store N in Rubisco is not evidence of N sufficiency, as physiologically this will be more important for N-limited than N-replete plants (Millard, 1988) and so will occur without there necessarily being a surplus of Rubisco for photosynthesis.

Table 2.  Summary of the evidence for Rubisco acting as a nitrogen (N) storage protein in trees, along with those studies that found no evidence for such a role for the protein
EvidenceSpeciesLeaf habitReferences
Studies finding evidence for Rubisco as an N store
Over-investment in Rubisco for photosynthetic requirementsSclerophyllous shrubsEvergreenWarren et al. (2000)
Pinus pinasterConiferous evergreenWarren & Adams (2001)
Pinus sylvestrisConiferous evergreenWarren et al. (2003)
Pinus sylvestrisConiferous evergreenP. Millard (unpublished)
Preferential loss of Rubisco N from leaves during senescenceMalus domesticaDeciduousKang & Titus (1980)
Loss of Rubisco N during leaf senescence providing 30–48% of N used for leaf growth the following yearMalus domesticaDeciduousMillard & Thomson (1989)
Specific decrease in Rubisco protein from 1-yr-old needles during flushingPseudotsuga menziesiiConiferous evergreenCamm (1993)
Rubisco activation state decreases with increasing leafMalus domesticaDeciduousCheng & Fuchigami (2000)
N statusPseudotsuga menziesiiConiferous evergreenManter et al. (2005)
Studies finding no evidence for Rubisco as an N store
Loss of N, but not Rubisco, from old leaves to support new leaf growthEucalyptus globulusBroadleaf evergreenWendler et al. (1995)
Nothofagus fuscaBroadleaf evergreenStephens et al. (2001)
Eucalyptus globulusBroadleaf evergreenWarren (2004)

2. Evolution of Rubisco and CO2 concentrating mechanisms

Understanding the role of Rubisco as a storage protein is important, when considering if a potential C limitation to trees has resulted in an evolutionary pressure to increased its carboxylation efficiency. Isolation of Rubisco from a diverse range of species and comparison of the relative specificity of the enzyme for its two substrates, CO2 and O2 (Srel), has shown differences, suggesting that natural selection for increasing the carboxylation function has occurred. However, a higher specificity for CO2 seems to be acquired at a cost, as the catalytic efficiency of Rubisco (Kcat; mol CO2 mol Rubisco active sites s−1) decreases (Bainbridge et al., 1995). Kcat has also been shown to vary amongst C3 species in relation to temperature, with species from a cool climate having a higher Kcat (Sage, 2002), thereby potentially increasing their photosynthesis and photosynthetic N use efficiency. Sage (2004) suggested that the relatively narrow range of Srel found in C3 plants reflects a balance for selection for enhanced Srel and high Kcat. So, in the absence of further improvements to Rubisco, several examples of CO2 concentrating mechanisms have evolved in higher plants, such as C4 photosynthesis and crassulacean acid metabolism (CAM). These adaptations have allowed for an increased carboxylation efficiency of Rubisco and thus a smaller investment in the protein.

C4 photosynthesis involves fixation of CO2 by PEP carboxylase in mesophyll cells. This is followed by decarboxylation of the C4 acids in specialized bundle sheath cells, thereby elevating CO2 concentrations up to 10-fold, so that it is saturated for reassimilation by Rubisco (von Caemmerer & Furbank, 2003), thereby suppressing photorespiration. This adaptation appears to have evolved over 45 times independently within 19 families of angiosperms (Sage, 2004). It has been suggested that one of the main drivers for the evolution of C4 photosynthesis was decreasing atmospheric [CO2] (Ehleringer et al., 1997), coupled with rising O2 concentrations (Sage, 2004). CAM also involves the fixation of CO2 by PEP carboxylase (with the resultant organic acids often being stored in vacuoles) followed by decarboxylation of organic acids for assimilation by Rubisco (Lüttge, 2004). There is a great plasticity in the CAM adaptation to photosynthesis, with some species fixing atmospheric CO2 at night whilst the stomata remain closed during the day, and others fixing atmospheric CO2 throughout the 24 h period (Dodd et al., 2002).

However, in relation to trees, neither of these adaptations is widespread. C4 photosynthesis is found in some woody shrubs and small trees but no large trees, while CAM is found only in neotropical trees of the genus Clusia (Lüttge, 2006). Both C4 and CAM pathways are relatively recent adaptations of the ancestral C3 pathway. In both these cases, the amount of Rubisco needed by the plant has been decreased and these adaptations are found predominantly in ecosystems where N availability is not the major limitation to photosynthesis or growth. C4 plants are found predominantly in warm temperate, subtropical and tropical regions at higher temperatures (Ehleringer et al., 1997), where P availability is often a greater limitation than N. CAM plants are found predominantly as succulents in arid environments or as epiphytes in rainforests and probably evolved as a consequence of multiple environmental stresses, including salinity, nutrients and water (Lüttge, 2004). As a consequence, the fact that both CAM and C4 mechanisms result potentially in a smaller storage capacity for N is unlikely to affect plant growth or competition.

The answer to the paradox of the continuing Rubisco catalytic inefficiency is probably its role as a N storage protein. The first appearance of Rubisco in terrestrial plants was some 500 million yr ago, when atmospheric CO2 concentrations were much higher than at present and O2 concentrations lower. This probably explains its low affinity for CO2. Perhaps as atmospheric CO2 concentrations fell, plants were still not C-limited as, overall, there was a greater nutrient limitation. As already noted, young soils (in geological time scales) probably had very slow mineral weathering rates and had little input of N from N2 fixation, causing an overall N limitation (Chadwick et al., 1999). In young, highly leached soils, P limitation can also quickly develop (Richardson et al., 2004). As a consequence of nutrient limitations, the inefficient enzyme for carboxylation did not need to evolve further, as it was also useful as an N store. The fact that trees (unlike many other higher plants) have not utilized C4 or CAM-type adaptations to photosynthesis suggests that there was little pressure to assimilate more C. Even if photosynthesis is limited by the amount of Rubisco, it does not necessarily follow that in long-lived plants such as trees, which assimilate C over decades, growth is C-limited. Thus at the biochemical scale, considering the roles of Rubisco, it would appear that there is little evidence that the current atmospheric CO2 concentration is a major limitation to the functioning of trees. So, given the implication from the functioning of Rubisco, that plants in general are more nutrient- than C-limited, are the biochemical responses of tree leaves (and Rubisco in particular) to elevated [CO2] consistent with the hypothesis that trees are more nutrient- than C-limited?

3. Rubisco and elevated CO2

In most species studied, elevated CO2 has two primary effects. First, increased net C assimilation rates are the result of both increased carboxylation and competitive inhibition of oygenase reaction, thereby decreasing C losses through photorespiration. Second, there is often an effect upon stomatal aperture to reduce stomatal conductance, thereby increasing the water use efficiency of photosynthesis. Some studies have reported an initial stimulation in net photosynthesis slowing down or disappearing after a few years (Roberntz & Stockfors, 1998; Griffin et al., 2000). The reason for this is photosynthetic acclimation, which can be defined as ‘any adjustment in the C acquisition system that may develop over time in plants grown continuously in elevated CO2’ (Wolfe et al., 1998). This often involves down-regulation of the Rubisco protein, accompanied by a change in the balance of its active vs inactive forms, and is a particularly common response under low-N conditions (Stitt & Krapp, 1999).

In a meta-analysis of data from a range of European experiments growing trees with elevated CO2, Medlyn et al. (1999) examined photosynthetic model parameters. Using the Farquhar model of photosynthesis, they found that under elevated CO2 there was a down-regulation of potential electron transport rate (Jmax) and maximum Rubisco activity (Vc max) of the order of 10%, but found little evidence (only one study out of six) for a downwards shift in the relationships between Jmax, Vc max and leaf N concentration. They concluded that down-regulation of photosynthesis had several causes: reduced leaf N (because of a limited N availability compared with increased C availability) and a shift in the relationship between photosynthetic parameters and leaf N, probably as a result of an accumulation of nonstructural C in the leaves (Medlyn et al., 1999). Using a similar approach, Rogers & Humphries (2000) calculated the minimum Vc max capable of supporting acclimated photosynthetic rates observed under elevated CO2. They found a strong correlation between observed and predicted values, which they interpreted as Vc max and investment in Rubisco being coupled to requirements for C assimilation (Rogers & Humphries, 2000), again suggesting that as there is a loss of active Rubisco under elevated [CO2], the trees were not C-limited. The view based on findings from the free-air carbon dioxide enrichment (FACE) experiments is that there is a selective loss of Rubisco enzyme under elevated CO2, benefiting N use efficiency (Long et al., 2004), without there necessarily being much change in leaf C assimilation rate.

Thus in N-limited ecosystems, the key effect of the plants’ response to elevated CO2 is not removal of a limitation but an increase in efficiency of N use for photosynthesis (Drake et al., 1997). This view might well not be correct in the medium term if less investment in Rubisco protein compromises the ability to store and internally cycle N for growth. However, while this has never been studied experimentally, it is interesting that the acclimation of Rubisco to increased [CO2] in 1-yr-old needles of pines was not found in the current year's needles (Turnbull et al., 1998; Tissue et al., 2001; Rogers & Ellsworth, 2002). As current needles are the main site of N storage in pines (Millard et al., 2001), this suggests that acclimation does not occur until after Rubisco turnover to release N from storage from the youngest needles at the onset of spring growth. Acclimation would therefore not occur at the expense of N storage by the tree.

III. The ecophysiological scale: the profligate use of carbon

Trees are profligate in how they use and lose C. There are complex feedbacks between C inputs from trees to soil and nutrient cycling processes regulating soil fertility (Fig. 2). As a consequence, trees allocate a lot of their C in order to gain nutrients, but have evolved elaborate mechanisms for conserving, storing and internally cycling nutrients. Evidence for a C limitation for tree growth will be discussed, by considering C allocation at the whole-tree scale in relation to the persistence of soluble carbohydrate pools in perennial tissues, suggesting that much of the C is sequestered and not stored; leaf senescence and the efficient withdrawal of nutrients compared with carbohydrates; and allocation of C below-ground to roots and associated microbes to acquire nutrients.

Figure 2.

A schematic representation of carbon (C) inputs from a tree to soil and the consequent pathways for nutrient acquisition by the roots. The pathways of C input into soil are as follows: (1) via biomass turnover through leaf senescence (1α), root senescence (1β) and turnover of extraradical mycorrhizal networks (1δ); (2) C transfer from roots to mycorrhizal hyphae; (3) rhizodepostion by nonmycorrhizal (3α) and mycorrhizal (3β) roots, influencing soil biota and mycorrhizal saprophytic activities; (4) secretion of enzymes by the mantle/intraradical hyphae (4α) and extraradical hyphae (4β) which break down soil organic matter (SOM); (5) C uptake by soil biota (i.e. all soil organisms except mycorrhizal fungi) through decomposition of soil organic matter (5α), consumption of rhizodeposits (5β) and grazing of mycorrhizal hyphae and roots (5δ). The pathways for nutrient return to the roots which are fuelled by the C inputs are as follows: A, root uptake of mineral and low-molecular-weight soluble organic N/P compounds produced by soil biota; B1−4, transfer of N/P from mycorrhizal partners to tree roots, through competition with roots to take up soluble nutrients released by soil biota (B1), or utilization of complex sources of organic N/P (B2), which can then be utilized for fungal growth or transported and used across the mycorrhizal network (B3) and/or transferred to the roots (B4). In addition, there is mineral weathering (C) and subsequent uptake of nutrients by mycorrhizal fungi (C1) or directly by roots (C2). The potential influence of rhizodeposits, soil biota and/or mycorrhizal hyphae is indicated by dotted lines and could be mediated via C compounds and/or other physical and chemical factors.

1. Carbon storage or sequestration?

Storage (as defined by Millard, 1988) involves C being held in a pool, with the potential to be reused subsequently for growth or maintenance of another. In contrast, sequestration represents a metabolic dead end, with the C no longer physiologically active and so not affecting metabolism. So do trees store C? The short answer is yes, but much of the nonstructural soluble carbohydrates (NSC), such as starch, that are accumulated in woody tissues are sequestered and so probably not reusable by the tree. Seasonal fluctuations in the concentrations of NSC such as starch have also been found in many species, especially in seedlings (Gansert & Sprick, 1998). The ability to store C (and nutrients such as N and P) is important in allowing trees to recover from disturbances such as defoliation, and the relative ability to store C has been associated with seedling survival and growth in temperate (Kobe, 1997), tropical (Gleason & Ares, 2004) and boreal species (Kagawa et al., 2006a). C remobilization in deciduous trees has also been demonstrated as an important process during bud burst (Maurel et al., 2004), providing up to about 40% of the C used for new leaf growth in a boreal species with a short growing season (Kagawa et al., 2006a) and contributing to early wood formation (Kagawa et al., 2006b). C remobilization can also be important for recovery from winter embolism (Ameglio et al., 2004).

Trees certainly accumulate large pools of NSC. Hoch et al. (2003a) estimated that in a temperate forest the concentrations of NSC found in mature trees would allow the whole leaf canopy to be replaced four times. Even when tree growth is constrained to a short season by temperature (e.g. at the tree line) the build-up of NSC suggests that C availability is not a limitation to growth (Hoch & Körner, 2003). Furthermore, NSC pools in trees are never fully depleted in the same way that N storage pools are (e.g. Hoch et al., 2003a). Even girdling the phloem does not deplete NSC in roots completely (Jordan & Habib, 1996; Bhupinderpal-Singh et al., 2003), suggesting that roots contain pools of NSC which are not utilized, even when the supply of current assimilates is cut. So is the build-up of NSC in leaves, trunks and roots of trees C storage or sequestration? It seems to be partly a mechanism to avoid down-regulation of photosynthesis, sequestering C that cannot be reused by the tree. This is a mechanism comparable to arginine accumulation in conifers in response to excessive N deposition (Näsholm, 1994).

In comparison, it is well established that trees store nutrients (Millard, 1996) and that, particularly for N, remobilization of stored resources can provide the first source used for growth during flushing and the majority used for growth above ground each year (Dyckmans & Flessa, 2001; Millard et al., 2001; Carswell et al., 2003). N remobilization from storage is source-driven (Millard et al., 2001). In consequence, in contrast to C in NSC pools, all the stored N is depleted during remobilization, as has been seen in temperate species by, for example, the complete disappearance of bark storage proteins during the spring (Wetzel et al., 1989; Cooke & Weih, 2005).

2. Leaf senescence

Nutrient withdrawal from leaves during senescence is often very efficient and in deciduous trees plays an important role in their internal cycling of N and P (Millard, 1996). The remaining nutrient fraction returns to the soil as litter, with the mass and nutrient content of litter produced being the primary determinant of nutrient availability in forest soils (Prescott, 2002). The fraction of mineral nutrients withdrawn from leaves before their abscission is variable with differences associated with leaf life span, overall leaf nutrient content, soil fertility (Niinemets & Tamm, 2005) and environmental conditions such as drought (Del Arco et al., 1991). In a review comparing over 100 deciduous shrubs and trees, Aerts (1996) reported that 54 ± 16% of the N and 50 ± 20% of the P was withdrawn from a leaf, with resorption efficiencies usually lower for trees growing on sites with higher fertility. Van Heerwaarden et al. (2003) suggested that these values could be increased by a further 6–10% when accounting for changes in leaf mass and areas during senescence, the implication being that the majority of leaf nutrients were reused by the tree. In addition, trees have also evolved pigment systems (such as anthocyanins) to protect leaves from radiation damage and so facilitate nutrient recovery during senescence (Hoch et al., 2003b). One of the first events to occur during senescence is a turnover of Rubisco and export of N (as amino acids) from the leaf. This selective breakdown of Rubisco can lead to an approximately 80-fold greater rate of turnover than of other enzymes of the Calvin cycle (Crafts-Brander et al., 1990), thus emphasizing the role of Rubisco as a N storage protein.

In contrast to nutrient withdrawal from senescing leaves, much of the soluble C present at the end of the summer remains in the abscised leaf. Indeed, leaf senescence is now thought to be triggered by the accumulation of both soluble sugars and starch in the leaf (Ono et al., 2001; Pourtau et al., 2004) and it is the water-soluble C content (i.e. sugars) that may determine initial litter decomposition rate (Allison & Vitousek, 2004). In the longer term at the community scale, leaf content in secondary C compounds (such as lignin, polyphenols and tannins) is negatively correlated with litter decomposition rate. As slow-growing trees tend to produce leaves rich in such compounds, the feedback between leaf quality and decomposition rate contributes to maintaining lower levels of soil fertility and is thought to explain the dominance of such trees in their habitat (Aerts, 1999; Cornelissen et al., 1999). Hence C in leaf litter, in the form of complex secondary compounds or sugars, plays an important part in determining the rate of soil nutrient cycling. C investment in secondary C compounds is usually regarded as a means to protect the key organs in C acquisition against herbivory. However, many evergreen tree species store N in their leaves and, in their case, antiherbivory compounds may serve primarily to protect the nutrients stored within the leaves (Millard et al., 2001). Thus trees tend to protect and then reuse the nutrients held in their leaves, while allowing much of the soluble leaf C to abscise with the leaf. Taken together, these physiological and structural plant mechanisms suggest an evolutionary pressure led by nutrient rather than C limitation.

3. The importance of C allocation to roots and associated microbes

In addition to the C and nutrients returned to the soil as litter, trees allocate much of their C below ground. Giardina et al. (2005) strikingly summarized the importance of below-ground C allocation: half of the 120 Pg C fixed annually by terrestrial plants is allocated below ground, with tree-based ecosystems accounting for most of this flux, amounting to 20 times the annual release of C by combustion of fossil fuels. Below-ground C allocation has been estimated as being between 35 and 50% of net primary production (NPP) in a tropical forest (Giardina et al., 2003), and as much as 73% in Douglas fir (Fogel & Hunt, 1983) and black spruce (Ruess et al., 2003) forests. Figure 2 shows the pathways for C input to the soil: from litter (pathway 1); by transfer from roots to mycorrhizal fungi (pathway 2); directly into the soil as mycorrhizodeposits (pathway 3) or secreted enzymes (pathway 4); and through grazing by soil fauna (pathway 5). The profound impact of this C (particularly through pathway 2 in Fig. 2) on the functioning of soil microbial communities has been demonstrated in large-scale field experiments. Girdling trees has been used to stop the flow of photosynthates to roots (Högberg et al., 2001), thereby reducing soil respiration. In a tropical eucalyptus forest, girdling reduced soil respiration by 16–24% (Binkley et al., 2006), despite tree canopies remaining intact for 3 months and live fine root biomass showing no decrease for at least a further 2 months, indicating C storage capable of sustaining root maintenance and respiration. Girdling lowered soil respiration by more than 50% in forests in northern Sweden (Bhupinderpal-Singh et al., 2003) and Germany (Subke et al., 2004) and 31–44% in Colorado (Scott-Denton et al., 2006). Furthermore, there was an average decline of 32% of the soil microbial biomass within 3 months at the Swedish site, attributed by Högberg & Högberg (2002) to a loss of the extra radical ectomycorrhizal mycelium.

The formation of associations with mycorrhizal fungi is ubiquitous in the plant kingdom. Roots of forest trees are heavily colonized by ectotrophic (ectomycorrhizal, EM) and/or endotrophic (arbuscular mycorrhizal, AM) fungi. For example, Ruess et al. (2003) found 100% of first-order roots to be EM in three mature black spruce forests, whilst Adams et al. (2006) reported colonization rates by EM fungi of between 76 and 100% of root length in eucalyptus forests. Gehring & Connell (2006) reported a colonization rate of up to 61% of root length of seedlings in tropical and subtropical rain forests. A meta-analysis of data for root colonization by AM fungi found an average (across all plant species) of 36 ± 10, 23 ± 3 and 24 ± 8% root length in tropical, temperate and boreal forests, respectively (Treseder & Cross, 2006). Boreal and temperate forest trees predominantly associate with EM species, whilst the majority of forest trees growing in tropical climates have AM symbioses. There are, of course, exceptions to this crude classification, with a minority of species forming associations with EM fungi in tropical forests (Alexander, 2006) or AM fungi in temperate forests (e.g. sugar maple, sweetgum, some eucalypts and poplars). When co-occurring in the same ecosystem, EM and AM fungi have been shown to occupy different niches, with a greater abundance of EM species in organic soil horizons, and with AM species predominating in mineral horizons (Neville et al., 2002; but see Moyersoen et al., 1998). Some species also exhibit successional mycorrhizal associations, often with an increase in the abundance of EM and a decrease of AM fungi as the trees age, as reported by Adams et al. (2006) in their comparison of field-grown seedlings with mature Eucalyptus grandis trees. Similarly, Egerton-Warburton & Allen (2001) found a shift in the relative abundance of AM and EM fungi in oak trees along a 1–30 yr age sequence (presumably as tree carbohydrate storage capacity increased) and suggested that C supply to the roots partly controls succession from AM to EM, adding weight to the view that EM and AM fungi exhibit different C sink strengths (Lynch & Whipps, 1990). So how much C is actually transferred from roots to EM and AM mycorrhizal fungi?

Microcosm studies of EM conifers have shown that up to 30% of total photoassimilates are transferred to the fungal partner (reviewed by Soderström, 2002). Soil C balance calculations for a field site in a northern hardwood forest estimated that 17% of total below-ground C allocation was to mycorrhizal fungi (in a mixed AM and EM community) and exudation (Fahey et al., 2005). However, to our knowledge, there has been no direct quantification in situ of the total C flux between tree roots and their mycorrhizal partners. The main impediment to direct measurements of C allocation to mycorrhizal fungi is a lack of adequate techniques. C transfer to mycorrhizal fungi can be considered as two fluxes: C exchange at the root–fungus interface (flux 2α, Fig. 2) and C allocation between intra- and extraradical hyphae (flux 2β, Fig. 2). The quantification in situ of C transfer at the plant–fungus interface (flux 2α, Fig. 2) is virtually impossible, without either full characterization of transport systems and specific compounds involved in resource transfer, or identification and isolation of a pool of fungal/plant specific compounds with relatively rapid turnover (such as nucleic acids) which could be used as markers of C incorporation. This second approach, together with the rapid development of molecular techniques, may allow study of both function and structure of the mycorrhizal community intercepting plant C (Johnson et al., 2005a). This would be particularly relevant, given emerging evidence that the structure of the EM community is affected by treatments changing C supply below ground (Fransson et al., 2001; Parrent et al., 2006). Root-excluding mesh cores, allowing ingrowth of mycorrhizal mycelia, combined with isotopic C labelling, have been used to quantify the flux of C to extraradical mycelia of AM fungi in grassland ecosystems (Johnson et al., 2002), hence describing the flux of C transfer from intraradical to the network of extra radical hyphae (flux 2β, Fig. 2) which can extend centimetres (AM) to metres (EM) into the soil and link plant roots together (Selosse et al., 2006). Such an approach has been used in forest ecosystems to measure fungal biomass, but not for a full C budget, because C loss through fungal respiration was not measured (Godbold et al., 2006). Another major limitation to establishing an accurate C budget of the mycorrhizal community in forest ecosystems is our ability to measure mycorrhizal biomass accurately; particularly to distinguish extraradical mycorrhizal hyphae from mycelia of saprotrophic species (Wallander, 2006). Extraradical hyphae may contribute substantially to soil C stock. It has been estimated that over 60% of the accumulation of soil organic C at a poplar FACE site was derived from external mycorrhizal hyphae, twice that from leaf litter (Godbold et al., 2006). In terms of litter production (flux 1δ, Fig. 2), such large amounts of fungal biomass in soils may have important implications for rates of soil organic matter (SOM) decomposition, particularly when comparing EM- and AM-dominated forests. The two types involve fungal species with fundamentally different morphology and chemical composition, which potentially affects their palatability to soil fauna. This will have inevitable consequences on the substantial amounts of C (Johnson et al., 2005b) going from mycorrhizal fungi to the soil food web (flux 5δ, Fig. 2). Hyphal turnover rates also differ between AM and EM (Staddon et al., 2003; Treseder et al., 2004), and AM produce substances (e.g. glomalin) which can stabilize soil aggregates (Rillig & Mummey, 2006). The chemical and morphological characteristics of mycorrhizal hyphae may also affect the fine root decomposition rate. When investigating the decomposition rate of EM vs nonmycorrhizal pine roots, Langley et al. (2006) showed that EM roots lost only one-third of the C of nonmycorrhizal roots. Langley & Hungate (2003) hypothesized that mycorrhizal type may ‘substantially influence fine root decomposition and soil carbon processing rate’.

So why do trees allocate such a substantial proportion of their C below ground and specifically to mycorrhizal fungi? The common view is that through their mycorrhizal partners, trees access nutrients otherwise unavailable for direct root uptake (such as complex forms of organic P and N), or present in insufficient quantities in the vicinity of the root (e.g. orthophosphate; Hinsinger, 2001). For example, using a microcosm system, Brandes et al. (1998) calculated uptakes of 73% of tree N and 76% of P via the EM fungus Paxillus involutus colonizing the roots of Norway spruce. In addition, van Breemen et al., 2000) suggested that external hyphae play a substantial role in nutrient return to the tree, via mineral weathering (Fig. 2, pathway C). Read (1991) hypothesized that there is a link between plant biome, mycorrhizal type and overall SOM amount, soil pH and N : P status. Read's hypothesis, later reinforced by Read & Perez-Moreno (2003), explains both the broad relative distribution of both mycorrhizal types and the body of evidence showing that AM fungi contribute predominantly to their host P requirements (but see Hodge et al., 2001) through enhancement of phosphorus uptake, while EM fungi have a greater capability to mobilize N and P from SOM. In boreal forests, this could satisfy a considerable proportion of the annual nutrient requirement of the trees (Read & Perez-Moreno, 2003). Recently, understanding the functional difference between AM and EM might have been taken a step further. Lindahl et al. (2007) demonstrated that saprotrophic fungi were mainly found in the soil horizons where primary decomposition of recent leaf litter occurs, whilst EM fungal taxa dominated the underlying horizon where N was mobilized from partly decomposed litter. Their study added weight to the view that AM forests (tropical) have an extravagant nutrient cycle relying heavily on N mineralization, whereas EM forests (temperate/boreal) have a conservative nutrient cycle where trees access organic N via EM fungi, bypassing mineralization processes by free-living microbes (Chapman et al., 2006).

It may be assumed that the ‘C cost’ of nutrient acquisition through symbiotic partners is higher than that of acquiring nutrients through direct root uptake from the soil solution. Does this imply that the C demand by the symbiotic partner leads to C limitation of tree growth? The consensus view is that this is not the case. Mycorrhizal infection has been shown to cause an up-regulation of photosynthesis in young EM trees in microcosms (Loewe et al., 2000) and pots (Wright et al., 2000) and young AM trees (Lovelock et al., 1997) in pots, suggesting that when the mycorrhizal associations impose an extra C demand on the tree, it can be met by increasing C assimilation. The initial presence of the fungi may lead to a growth depression for trees under conditions of high soil fertility, as shown in field experiments in highly managed agrosystems (Graham & Eissenstat, 1998); presumably as the benefit to the plant of the increased nutrient availability conferred by the fungus diminishes. However, such levels of soil fertility are not found in forest ecosystems. To assess the C cost to the tree of nutrient acquisition via mycorrhizal fungi in situ, we need to understand the regulation of C exchange at the plant–root interface by both biological (e.g. plant C and nutrient status) and environmental factors (e.g. nutrient availability in the mycorrhizosphere, temperature). This is assuming that the amount of C passed on to its symbionts (2α, Fig. 2) is mainly controlled by the plant itself. However such a phytocentric view is now being questioned, and it is argued that control of C use and allocation (flux 2β, Fig. 2) by mycorrhizal partner(s) also need to be considered (Lindahl et al., 2002; Staddon, 2005), especially given the extent of extraradical mycorrhizal hyphae produced by EM fungi. In a meta-analysis comparing the response ratio of AM/EM plants (not exclusively trees) and mycorrhizal fungi to elevated CO2, Alberton et al. (2005) reported evidence ‘for the mycocentric view in EM, but not in AM systems’.

In addition to transferring C to their root symbionts, trees lose C from their roots as rhizodeposits (pathway 3, Fig. 2). Rhizodeposition involves a wide range of compounds (Grayston et al., 1997), including exudation of low-molecular-weight substances such as sugars, organic acids and amino acids. The availability of these C substrates is considered the factor most limiting to the growth of free-living soil microbes (Wardle, 1992), explaining the greater microbial activity in rhizosphere compared with bulk soil. There are few reliable quantitative estimates of the flux of C entering the soil as exudates. A review of whole-plant 14C-labelling studies performed in soil on a wide range of plant species suggested that exudation accounted for 5–10% of net C assimilation (Farrar et al., 2003), although Jones et al. (2004) highlighted a possible overestimation resulting from methodological bias and suggested that a true estimate of root exudation was likely to be only 2–4% of net fixed C. Under a variety of stress conditions (including nutrient or water stress) the C flux from roots as exudates is increased, mainly because of damage to membranes or disruption of normal cell metabolism (Neumann & Römheld, 2001). As noted by Jones et al. (2004), the vast majority of studies quantifying root exudates have not considered the quantitative or qualitative impact of the (ubiquitous) mycorrhizal fungi colonizing roots (Fig. 2, pathway 3β), despite numerous studies demonstrating exudation of hydrolytic enzymes by mycorrhizal mycelium (mainly EM fungi). Recently Phillips & Fahey (2006) calculated that microbial biomass, C/N mineralization rates and phosphatase activity were 25–30% higher in the rhizosphere than in bulk soil under EM tree species. Under AM trees, rhizosphere and bulk soil differed by only 10–12% for similar parameters. Tree C inputs to soil as rhizodeposits can also act as primers for the degradation of existing SOM (Fontaine et al., 2004; Hoosbeek et al., 2004) through both abiotic and biotic mechanisms (reviewed by Kuzyakov et al., 2000). However, the trade-off between C priming and nutrient availability, on the one hand, and tree growth on the other remains largely unknown. Under elevated CO2, exudation is thought to be increased on a per-plant basis via stimulation of root growth, although experimental evidence is still lacking.

Taken together, C transferred from trees to root symbionts and rhizodeposition account for a significant proportion of net C assimilation. The tree benefits from this C loss in terms of microbial activity and the consequent nutrient acquisition and cycling through the turnover of SOM. These complex interactions between the tree and the soil suggest that C availability is not the primary resource limitation for tree growth. If C were limiting tree growth, it would only be through C limitation to nutrient cycling in forest soils.

IV. The ecosystem scale: how forests respond to environmental change

The third line of evidence that tree growth is seldom limited by C comes from a consideration of the impact of global environmental change on forests. Forest growth has increased in recent decades, as shown by studies of temperate forests in North America (Turner et al., 1995), Europe (Hunter & Schuck, 2002; Solberg et al., 2004) and tropical forests in Amazonia (Baker et al., 2004). Does this mean that the increased forest growth we now see, as atmospheric CO2 concentrations rise, is because trees have been, and are, C-limited? There have been two different approaches used to try and explain these growth increments. Manipulative experiments (especially applying elevated CO2 to trees) and models have been used to try to determine the combined and interactive effects of elevated CO2, changes in climate (such as temperature and rainfall) and pollutant loadings (particularly N deposition) on forest growth and C sequestration.

1. Responses to elevated CO2 and the importance of soil processes

Hundreds of papers have now been published describing the responses of trees to elevated CO2, including many reviews (Ceulemans & Mousseau, 1994; Curtis & Wang, 1998; Saxe et al., 1998; Gielen & Ceulemans, 2001; Long et al., 2004; Nowak et al., 2004; Ainsworth & Long, 2005). While many have reported that elevated CO2 stimulates forest tree growth, some have found only a short-term growth enhancement or no stimulation of above-ground growth, but enhanced root growth, so NPP is increased. Interpreting the results from these studies is complex, as responses are confounded by the timescale of the studies, the stage of stand growth (expanding vs full canopy) and the relative availability of other resources, such as nutrients (see Körner, 2006). However, in the few studies where elevated CO2 has been given to forest systems with a near to steady-state nutrient cycle and full canopy development, initial responses in above-ground growth have declined through time (Körner, 2006).

The scale of most elevated CO2 experiments is not that of the forest, and even the longest duration study still only reflects a small fraction of the lifetime of a tree (Norby et al., 1999). The experiments that have provided elevated CO2 at a scale closest to that of the forest are the FACE experiments. There have been four of these: at Duke Forest, NC, USA, growing loblolly pine (from 1996, Hendrey et al., 1999); at Rhinelander, WI, USA, growing aspen, sugar maple and birch (from 1997, Karnosky et al., 1999); at Oak Ridge, TN, USA, growing sweetgum (from 1998, Norby et al., 2001) and at the POPFACE site in Italy, growing poplars (from 1999, Miglietta et al., 2001). These experiments have studied the response of plantations, which between them cover a wide range of NPP (Norby et al., 2005), to an addition of an extra 200 µmol mol−1 CO2 above ambient. In addition, there has been one experiment which has used a derivation of the FACE technology (called Web-FACE) to study the response of a range of 35-m-high deciduous forest trees to a 530 µmol mol−1 CO2 atmosphere over 4 yr (Körner et al., 2005).

A common response to elevated CO2 in all the tree FACE experiments has been an increase in both above-ground and below-ground production, as a result of faster photosynthetic rates (Nowak et al., 2004). Despite both the AspenFACE and POPFACE experiments being initiated on bare soil, all the FACE sites will have ‘primed’ the system with extra C, causing a disruption so the system is no longer in equilibrium. The feedback loop (B) shown in Fig. 1 between the forest soil and the tree operates at a timescale of up to decades, while the flux of C from the tree into the soil (3) is regulated by physiological processes operating over a considerably shorter timescale. However, it is not clear if any of the experiments have yet reached a new equilibrium or stable state (Körner, 2006). Therefore, it is difficult to know if the responses seen to the abrupt change in atmospheric CO2 concentrations arising in FACE experiments would be the same as those to the (comparatively) gradual increase predicted over the next 50 yr or so, where the C priming effect will be much less and forests remain closer to a steady state throughout the transition. A major question then is as follows: will nutrient supply rate keep pace with the demand created by faster tree growth under elevated CO2? This has been questioned (Zak et al., 2000) and it has been suggested that available soil N may increasingly limit growth as C and N are sequestered in woody biomass and SOM (Luo et al., 2004). However, the actual evidence for this in FACE experiments is still sparse, with recent reviews showing little evidence at the Oak Ridge site yet (Norby & Iversen, 2006) and, while there has been immobilization of N in tree biomass and soil at Duke Forest site, ecosystem C : N ratios have increased under elevated CO2 and there has been a large accumulation of ecosystem N capital (Finzi et al., 2006).

Trees have two options for maintaining faster growth rates under elevated CO2 if soil N availability does not keep pace with their increased demand for N. First, they can increase their N-use efficiency. This will happen as a result of the kinetic properties of Rubisco, as the oxygenase function is suppressed at higher [CO2]. It can also occur through photosynthetic acclimation, reducing N lost through leaf senescence or changing patterns of C/N allocation, as discussed above. Second, they can use their extra C to acquire more N from the soil through increased allocation below ground, specifically to fine roots and their mycorrhizal fungi. One mechanism for this could be a tighter coupling between litter production and direct uptake of organic N from the litter by mycorrhizal fungi (Fig. 2). So far, the potential for mycorrhizal fungi to bypass litter decomposition by saprotophs by direct nutrient transfer from litter to tree roots has been widely overlooked (Chapman et al., 2006). Yet it could play a significant role in meeting tree N demand under elevated CO2, because litter production is increased (Finzi et al., 2001).

To increase the nutrient capital available to trees in the longer term, SOM turnover rates would have to increase, and consequently soil respiration rates would also increase. Soil respiration has two components. The autotrophic respiration (Ra) is associated with the use of C recently assimilated by the tree, including growth and maintenance of roots and their mycorrhizal fungi and the rhizosphere microbial respiration tightly coupled to the supply of C from rhizodeposition (flux 6 in Fig. 1). Heterotrophic respiration (Rh) involves the wide range of organisms involved in the SOM decomposition food web (flux 5 in Fig. 1), possibly including respiration of extraradical mycorrhizal network. Thus Rh relies upon ‘historical’ C but, through the turnover of SOM, accelerates the rate of soil nutrient cycling. Forest soil respiration rates vary considerably in relation to soil moisture and temperature (Tedeschi et al., 2006). However, the biotic factors governing Rh in forest soils are not well understood. Therefore, distinguishing Ra from Rh, and particularly how the latter is regulated, is important for understanding the processes regulating SOM turnover and, ultimately, whether with rising atmospheric CO2 concentrations nutrient cycling in the soil will be affected.

It is well established that elevated CO2 increases soil respiration rate under a wide range of vegetation types, including trees (reviewed by Zak et al., 2000) and in tree FACE experiments (King et al., 2004; Bernhardt et al., 2006). But is there any evidence that Rh and N cycling increases with elevated CO2? After 6 yr of elevated CO2 at the Duke Forest site, there was evidence of a dynamic or transient increase in the C sink in the forest floor (Lichter et al., 2005), although the magnitude of the increase was limited by fast turnover times (estimated using the δ13C signature of the CO2 used to supplement the atmosphere as a tracer for the C). This was interpreted as being the result of increased fine root turnover, suggesting that Rh would be unaffected. At the Oak Ridge site, enhanced NPP led to greater fine root production (Norby et al., 2004) and a slight increase in Rh (Norby et al., 2002). However, no effect of elevated CO2 was found on the microbial biomass or N cycling processes (Sinsabaugh et al., 2003). In a study of soil N cycling across the three FACE experiments in the USA, Zak et al. (2003) found that increased amounts of root and leaf litter had no effect on microbial N cycling pools or processes. At the POPFACE site in Italy, the rate of soil respiration under the young trees was increased by elevated CO2 (King et al., 2004), but the annual production of leaf litter was not affected, while litter chemistry was, with a decrease in N concentration (Cotrufo et al., 2005). As a consequence, the leaf litter showed a slightly slower decomposition rate, although the rate of degradation of control litter was also very slow, perhaps reflecting the recent land use change at the POPFACE site affecting the capacity of the soil microbial community to degrade a ‘new’ substrate of poplar litter (Cotrufo et al., 2005). The Web-FACE experiment found a rapid transfer of C to soil and an increase in soil CO2 concentrations (interpreted as enhanced soil respiration), but no measurable stimulation in stem growth (Körner et al., 2005). A meta-analysis of 117 FACE and open-top chamber studies showed increased soil C concentrations under elevated CO2, but no increase in total soil N, net N mineralization or N2 fixation (De Graaff et al., 2006). In unfertilized ecosystems, microbial N immobilization was found to decrease any stimulation of plant growth to elevated CO2 (De Graaff et al., 2006).

So it is clear that there is a C-priming effect at the FACE sites, with increases in NPP (DeLucia et al., 2005), but there is not yet evidence that the system has reached a new steady state with respect to the interaction between soil C and N cycling in any of the experiments. While soil respiration has been increased by FACE, at some sites (with established forest) the stimulation in rate has declined through time (King et al., 2004; Bernhardt et al., 2006) and there is little evidence that Rh has been affected or soil N cycling altered. One possible reason for the contradictory results from FACE experiments of enhanced tree N uptake and even soil N depletion, without measurable changes in soil N mineralization, is that our measures of mineralization and soil N availability are not sensitive enough (Johnson, 2006), or are addressing a process of minor relevance for assessing nutrient availability to trees on infertile soils (Schimel & Bennett, 2004). However, taken together, these results suggest that, in the medium term, N cycling is unlikely to keep pace with the increases in NPP caused by C priming, resulting in an overall nutrient limitation rather than C limitation to forest growth.

2. Ecosystem modelling

Turning now to consider modelling, several studies have used large forest inventory datasets along with climate and other environmental data to determine the causes of observed growth changes in forests (Nellemann & Thomsen, 2001; Ollinger et al., 2002; Solberg et al., 2004). The advantage of such approaches is that multiple factors can be accounted for, including changing weather patterns and N inputs from atmospheric deposition. The modelling has suggested that in Europe the majority of forest growth increment can be accounted for by N deposition (Nellemann & Thomsen, 2001; Solberg et al., 2004) and very little by elevated CO2. In North America the picture is further complicated, with ozone pollution in the USA potentially offsetting much of the gain attributable to either elevated CO2 or N deposition (Ollinger et al., 2002). However, an analysis of tree ring chronologies has suggested that atmospheric CO2 increases since 1950 have increased the growth of ponderosa pine growing on water-limited sites in the Pacific Northwest (Soulé & Knapp, 2006). A similar approach, comparing tree rings in individuals growing near a natural CO2 vent with control trees at ambient CO2, demonstrated after 30 yr a 12% growth response, found mainly when the trees were young and in years that had a dry spring (Hättenschwiler et al., 1997). In boreal Canada, disturbances caused by fire or insect attack appear to have resulted in an overall reduction in forest growth in recent years (Kurz & Apps, 1999). There have been fewer studies of tropical forests, although it has recently been suggested that their response will be highly dependent upon factors such as increased rainfall patterns, which are predicted to reduce growth (Schuur, 2003) and could be interpreted as a potential nutrient limitation as a result of increased leaching.

In addition to these large-scale modelling studies, models of forest growth have been used to simulate the long-term effects of elevated CO2 on forest growth and plant–soil C dynamics (Medlyn et al., 2000; McMurtrie et al., 2001). Such studies have questioned the relevance of short-term experimentation in systems where longer-term biogeochemical feedbacks operate (Kirschbaum et al., 1994; Rastetter et al., 1997). In particular, the importance of soil processes (especially N immobilization and mineralization) in regulating long-term responses to elevated CO2 has been highlighted (Medlyn et al., 2000; McMurtrie et al., 2001). These ecosystem simulation models suggest a down-regulation of NPP in response to an increase in atmospheric CO2, when plant uptake of N exceeds the rate of replenishment via mineralization (Rastetter et al., 1997; Luo & Reynolds, 1999). With a model calibrated to an old growth forest in the north-eastern United States, Rastetter et al. (1997) identified this as happening on a timescale of several decades, which is significantly longer than the duration of any of the FACE experiments.

V. Conclusions

It should be clear from the arguments presented that, at each of the scales considered, there is little evidence for a direct C limitation of tree growth. While no one piece of evidence is conclusive, taken together the weight of evidence suggests that the physiological functioning and growth of individual trees is not limited by the availability of C. Evidence from both the biochemical (leaf) and physiological (whole tree) scales suggests that forest trees have an abundance of C compared with nutrients such as N or P. However, given what is known about the complex interactions between the C and nutrient dynamics of forest ecosystems, discussed earlier, it is tempting to think that the overall rates of SOM turnover and biogeochemical cycling of nutrients are limited by the inputs of C from vegetation.

So if tree growth is not C-limited, is their nutrient supply? Does C cycling interact with the N cycle to impose an overall limitation on tree growth? Based on current evidence, the answer to this is ‘probably not’. Where FACE experiments have provided a C priming to a forest system, there has been a faster cycling of C through the trees to the soil and back to the atmosphere because of increased soil autotrophic respiration. But the lack of evidence for an impact on N cycling processes suggest that all this extra labile C entering the soil has little impact on nutrient supply through SOM turnover. This suggests that the faster tree growth that has been observed with FACE has been supported by increased N use efficiency and N cycling, possibly through the interactions of mycorrhizal fungi and leaf litter.

Although carbon assimilation increases with increasing atmospheric CO2, it does so at a diminishing rate. One reason for this could be a feedback effect resulting from an eventual sink limitation on photosynthesis, once the rate of autotrophic respiration has reached a new steady state (determined perhaps by a nutrient limitation). Second, respiration is an exponentially increasing function of temperature. Thus, as global change proceeds, the rate of increase of CO2 assimilation by terrestrial ecosystems will slow, while rates of soil respiration will increase. One possible feedback might be that, if the turnover of SOM increases, this could increase nutrient cycling and so enhance soil fertility. In the short term, there will be a positive effect on growth and therefore on CO2 uptake, but over longer time frames (decades), the net effect is likely to be a decrease in the ability of the terrestrial biosphere to absorb CO2.


We thank Eric Paterson and David Whitehead for their constructive comments on the manuscript. PM and G-AG thank Roderick Dewar for his hospitality at INRA, Bordeaux, while writing the first draft of the paper and Joelle Salmon for her support and encouragement. The Scottish Executive Environment and Rural Affairs Department funded PM while writing this paper.



  •  Summary 11

  • I. Introduction 12
  • II. The biochemical scale: the paradox of Rubisco 13
  • III. The ecophysiological scale: the profligate use of carbon 16
  • IV. The ecosystem scale: how forests respond to environmental change 21
  • V. Conclusions 23
  •  Acknowledgements 23

  •  References 23