We still have a poor understanding of what limits tree growth under many conditions. In general, tree growth can be limited by the availability of carbon within the plant (carbon limitation) or by the tree’s ability to use the available carbon it has because of nutrient shortage, environmental conditions, or developmental constraints (sink limitation).
A growing body of literature suggests that current tree growth is sink limited under most conditions, with much of the evidence inferred from trees’ high levels of nonstructural carbon (NSC). These inferences rely on the assumption that carbon storage is passive and that NSC does not accumulate when growth is carbon limited (Bansal & Germino, 2008). However, storage may be an active process, occurring at the expense of growth (Chapin et al., 1990; Lacointe et al., 2004; Genet et al., 2010).
We argue that carbon starvation – even if it does not occur every year or even every decade in the life of a tree – is crucial to understanding carbon storage and carbon limitation in trees. We define carbon starvation, modified after McDowell (2011), as the outcome of insufficient carbon supply from current photosynthesis and available storage to sustain the metabolism, regeneration, or defense necessary for immediate survival (Fig. 1). Episodes of carbon starvation (past and present) may have led to the evolution of conservative allocation strategies favoring greater reserve formation at the expense of potential growth. Therefore, an increase in carbon storage and decrease in growth could be a plastic or evolutionary response to carbon-limiting conditions – not a result of sink limitation. To illustrate this point, we reexamine two phenomena relevant to the active debates regarding carbon limitation: tree growth at alpine treeline and under water stress.
Storage and growth limitation
Carbon storage, defined here as formation of nonstructural carbon compounds that ‘can be mobilized in the future to support biosynthesis’, can be either passive or active (Chapin et al., 1990). Passive storage forms an excess pool when carbon supply is greater than carbon demand of other processes, such as growth (Fig. 1). Active storage, or reserve formation, on the other hand, occurs when carbon is directed into storage at the expense of growth or other processes. Reserve formation allows a tree to prepare for (1) predictable asynchrony of supply and demand for carbon (e.g. early spring growth for deciduous trees) and (2) unpredictable emergencies such as disturbances that affect carbon gain or need (Chapin et al., 1990).
The potential for active reserves is often ignored when inferring the nature of growth limitation. When comparing trees in different sites or treatments, sink limitation is incorrectly inferred from one of two main patterns: (1) higher NSC content together with lower growth rates (Hoch & Körner, 2003; Würth et al., 2005; Palacio et al., 2008; Sala & Hoch, 2009), or (2) relatively stable NSC levels throughout the year that do not decline to zero (Hoch et al., 2002, 2003; Shi et al., 2008). The rationale for inferring sink limitation from the first pattern is that NSC pool size is the difference between carbon gain (source) and carbon demand (sink) within the plant (Körner, 2003). This assumption is true if all storage is passive, but not if it is active (Lacointe et al., 2004; Bansal & Germino, 2008; Genet et al., 2010). Reduced growth and increased storage is also consistent with carbon limitation if (a) populations vary genetically in the amount of active reserve formation or (b) reserve formation responds plastically to environmental or internal cues. The rationale for inference (2) is that if growth is annually carbon limited during periods when demand exceeds supply, NSC should approach zero during these periods, because growth could only be carbon limited if all storage is depleted (Hoch et al., 2003; Körner, 2003). But this complete, yearly drawdown will only occur if all reserves are stored because of annual asynchrony of supply and demand (i.e. earmarked for a particular time of year or frequent event). Reserves, though, can be formed to avoid running out of carbon during emergencies, which may be unpredictable and infrequent, and trees may actively maintain a baseline level of NSC, growing only after storage requirements are met.
Increased storage increases survival but reduces growth
Growth in trees with higher NSC levels may be more carbon limited than in those with lower NSC levels. Increased storage is often associated with survival (Canham et al., 1999; Gleason & Ares, 2004; Myers & Kitajima, 2007; although see Piper et al., 2009), recovery from disturbance (Breda et al., 2006), or resistance to biotic attack (Dunn et al., 1987). If storage can be active and increase the chance of survival, then increasing storage could be an adaptive response to an increased threat of carbon starvation. Populations that experience severe episodes of carbon starvation may experience stronger selection for increased reserves, leading to more conservative allocation strategies. However, these conservative populations, while less prone to carbon starvation, are more likely to experience carbon limitation to growth because less carbon is available after reserves are formed. Therefore higher NSC levels may indicate, and cause, greater carbon limitation.
It has been argued that large reserves cannot be beneficial given both the large opportunity costs of not growing, and also that maintaining high photosynthesis rates necessary to develop such reserves is costly, in terms of nutrients, and risky, in terms of cavitation (Sala & Hoch, 2009). But maintaining substantial reserves may have large payoffs. If a rare drawdown of storage occurs (due to a severe disturbance), a tree that has not saved enough to survive will be at a huge fitness disadvantage compared to surviving trees that continue reproducing for potentially hundreds of years. Selection for allocation patterns that minimize the risk of carbon starvation could easily lead to the emerging pattern of large carbon reserves that are never depleted within a season (as reported in Hoch et al., 2003; Körner, 2003).
Additionally, if carbon starvation is infrequent but relatively predictable by a reliable cue, plasticity of reserve allocation may be selected for. For example, when carbon uptake is low, a tree will take longer to accumulate enough carbon to recover from disturbance or defend against biotic attack unless it has sufficient reserves; therefore storing more when carbon uptake declines could help a plant survive or recover faster. Non-carbon status cues (i.e. temperature or water potential) could also be reliable indicators of increasing threat of starvation or damage, against which an increase in storage could defend. In Arabidopsis thaliana, there is accumulating evidence that plants actively shift carbon allocation between storage and growth, favoring storage when carbon uptake declines (Smith & Stitt, 2007; Gibon et al., 2009). Growth slows under salt stress and cold temperatures in part due to higher levels of DELLA proteins, which down-regulate growth (Achard et al., 2006, 2008). Mutants without these proteins grow faster and have lower survival than wildtype plants under salt stress, indicating that growth is actively, and potentially adaptively, suppressed. Suppression may allow plants to switch to a more conservative strategy of carbon use in response to environmental cues to mitigate the effects of harsh conditions (Smith & Stitt, 2007).
Re-examining purported sink limiting cases
NSC levels are often used to support the growth-limitation hypothesis, which predicts that the absence of trees above treeline is due to direct inhibition of growth by cold temperatures (Körner, 1998). The main rationale for this hypothesis is that cell division slows faster than photosynthesis as temperatures decline (Körner, 1999). If trees cannot build new tissues fast enough to keep pace with photosynthesis, carbon substrates will build up in the plant; alternatively, if carbon supply were limiting, NSC concentrations would be lower at treeline (Hoch et al., 2002). Treeline trees often grow slower and have higher NSC concentrations than trees at lower altitudes, presumably supporting the growth limitation hypothesis (Hoch et al., 2002; Hoch & Körner, 2003; Shi et al., 2008).
However, because increased storage may be adaptive and actively controlled (Sveinbjörnsson, 2000; Bansal & Germino, 2008), observed patterns in growth and storage are also consistent with avoidance of carbon starvation. Active storage can account for decreased growth and increased NSC at treeline in two ways. First, increased storage may represent a local adaptation to colder and harsher conditions. Second, allocation to storage may respond plastically to temperature, independent of sink capacity for growth. If the second is true, colder temperatures may trigger changes in gene expression that prioritize storage over growth. Such a response could be selected for if cold temperatures are indicative of periods of poor carbon gain and/or greater risk of carbon loss, such as tissue loss or damage (Grace et al., 2002). In both cases, increasing storage could protect against starvation but exacerbate carbon limitation to growth.
Several lines of evidence support the view that treeline trees actively maintain storage and suppress growth. In defoliated treeline trees, NSC stores were partially restored by the end of the growing season even though growth was reduced (Li et al., 2002). This situation does not support direct sink limitation to growth if NSC is excess carbon: growth should be unaffected since NSC was nonzero and temperature was unchanged. Instead it seems that storage was given priority over growth. Similarly, branch elongation and leaf length increased at the treeline when trees were debudded or pruned (Li et al., 2002; Susiluoto et al., 2007) and when deciduous trees were given increased CO2 (Handa et al., 2005). In the same CO2 enrichment experiment, the evergreen species did not increase growth under higher CO2 availability alone, but shoot elongation was greater when defoliated under elevated CO2. These results suggest that temperature cannot be directly limiting growth if trees can increase growth when temperature does not change, and that even for the evergreen species, growth in response to disturbance or damage seems to be carbon limited. Finally, maximum photosynthetic rates at treeline and lower altitudes or latitudes may not differ (Vowinckel et al., 1975; Sveinbjörnsson, 1983; Körner, 1999). If the higher NSC levels at treeline were passively accumulating, photosynthetic down-regulation should occur, reducing assimilation rates (Bansal & Germino, 2008).
Experiments investigating treeline growth limitation and tree height offer contrasting support for sink limitation. A recent study found that experimentally heating apical buds induced a significant increase in vertical and branch growth (Petit et al., 2011). By contrast, Susiluoto et al. (2010) found that fertilization caused more than a two-fold increase in height growth. Are these cases reconcilable? Conditions among years or treeline sites could be so variable that the mechanism of limitation also varies. However, both cases can be explained by carbon limitation and active growth suppression to maintain sufficient storage. In the first case, the experimental warming released the trees from a conservative, cold-induced storage pattern and the consequently released reserves became available for growth. In the case of fertilization, the increased leaf N that resulted (Susiluoto et al., 2010) could increase photosynthesis, and potentially increase growth if more carbon was then available after the reserve needs were met.
The debate between carbon and sink limitation also applies to the effects of water stress on tree growth. While reduced growth during the dry season, during moderate drought, and with increasing hydraulic limitation due to increasing tree height may be due to reduced carbon uptake (Yoder et al., 1994; Ryan & Yoder, 1997), it may also be the result of direct sink inhibition (Körner, 2003; Ryan et al., 2006).
The evidence supporting sink limitation under water stress is often very similar to that used to support sink limitation at the treeline. Because of turgor-limited cell expansion, low water availability may directly affect growth more than photosynthesis, leading to a passive buildup of excess NSC within the tree (Körner, 2003). Consistent with this hypothesis, tropical tree species during the dry season have lower growth rates and higher NSC content than during the wet season (Würth et al., 2005), droughted seedlings decrease growth while either showing no change in NSC (Sanz-Pérez et al., 2009) or an increase (Galvez et al., 2011), and tall trees often have lower branch elongation rates and higher NSC concentrations than shorter trees (Sala & Hoch, 2009; Woodruff & Meinzer, 2011).
While these observations are consistent with turgor-induced sink limitation, they are also consistent with active growth suppression to avoid the increasing threat of carbon starvation (McDowell, 2011). Because photosynthesis declines with increasing water stress, carbon starvation may be a major cause of death during severe drought (Breda et al., 2006; McDowell et al., 2008). Reduced carbon uptake and storage during a drought may also lead to greater canopy dieback the following season (Breda et al., 2006), reducing carbon uptake and storage even more, leading eventually to starvation (Breda et al., 2006; Galiano et al., 2011). Additionally, declining photosynthesis could reduce defense allocation, reducing a tree’s ability to avoid or recover from biotic attack (McDowell, 2011). Increasing reserves at the onset of water stress may allow a plant to avoid these fates, but at the cost of reducing growth.
Even if carbon starvation via NSC depletion is rare, increasing storage could still be an adaptation to drought. For example, carbon starvation could occur via phloem malfunction, before all reserves are depleted (Sala et al., 2010). But increased storage in all plant organs could potentially lengthen the time a plant could survive on ‘local’ carbon without long-distance transport. In which case, mortality by phloem malfunction could still select for increased storage under water stress. Furthermore, even if carbon starvation isn’t a current cause of mortality, increased storage may still be adaptive. If during a population’s evolution, starvation occurred, trees may have evolved to avoid this type of mortality, leaving other ways for trees to die (e.g. hydraulic failure). The adaptive nature of increased storage is supported by the divergent response of NSC content to drought between more and less drought-tolerant tree species: the more drought tolerant increase NSC content while the less tolerant show the opposite response (Regier et al., 2009; Piper, 2011).
We believe that increasing carbon storage and decreasing growth is an active response, either plastic or evolutionary, to reduce the risk of starvation—not a result of sink limitation. If true, we are left with the counter-intuitive conclusion that tree populations with the highest NSC concentrations may be the most carbon limited in terms of growth; this puts into question many recent conclusions about what limits tree growth. In fact, the rejection of the carbon storage pool as purely passive leads us to conclude that we may have greatly underestimated the importance and frequency of carbon limitation to growth in trees.
Our knowledge of NSC dynamics in trees remains limited, however, and much more work is needed to determine the nature of storage and whether some species or populations favor storage more than others. To determine if storage is active, we may need first to determine where and when growth is carbon limited. Increasing carbon uptake experimentally or relying on natural variation may be the surest way of determining limitation; if growth increases with greater carbon uptake, all else being equal, the tree was carbon limited – and any storage that occurred was active. To test if certain populations store more than others (e.g. cold, fire, or pest tolerant vs intolerant, late vs early-successional), common garden experiments could be used to compare NSC concentrations or pool sizes relative to growth rates. Furthermore, as genetic work with Arabidopsis has demonstrated the importance of negative growth regulators, measuring DELLA protein expression may be useful in understanding when growth is constrained by environmental conditions or actively reduced by the plant.
We thank Brenda Casper and three anonymous referees for their very helpful comments.