Our study provides strong support for the hypothesis that variation in LL is tightly coupled to variation in the cost of water transport as measured by KLeaf, mass. Across two orders of magnitude in LMA, we observed a strong negative relationship between KLeaf, mass and LMA and no correlation between LMA and KLeaf, area. Similarly, across two orders of magnitude in LL we observed a strong negative relationship between KLeaf, mass and LL and a relatively weak correlation between KLeaf, area and LL. This lack of a consistent correlation between LMA and KLeaf, area has led previous investigators to conclude that KLeaf, a pivotal plant water-relations trait, is orthogonal (i.e. statistically independent) to traits influencing LL and, by extension, lifetime carbon gain (Sack et al., 2003; Sack & Frole, 2006). However, standardizing leaf traits by leaf surface area obscures existing variation in the carbon cost of light interception, gas exchange, and water transport that are, in fact, present if the traits are standardized by dry mass. For example, A. menziesii, P. ponderosa, and U. californica all possess dense, long-lived leaves and showed higher maximum KLeaf, area than Metasequoia glyptostroboides, a species with short-lived leaves (Fig. 1a). Yet, when KLeaf was standardized by leaf dry mass, M. glyptostroboides had greater KLeaf, mass than both P. ponderosa and U. californica, and was equivalent to A. menziesii because of a lower LMA (Fig. 1c,d). Similarly, Reich et al. (1997) and Wright et al. (2004) found that variations in LL and LMA were weakly correlated with variation in photosynthetic capacity when expressed on an area basis (Aarea). Although previous research has shown that the cost of water transport, as measured by the fractional investment of lamina biomass in the midrib, is tied to variation in LL (Niinemets et al., 2007), the present study shows for first time that the cost of leaf water transport, as measured by KLeaf, mass, is tightly coupled to variation in LL.
Functional basis for the coordination between leaf lifespan, LMA and KLeaf on a mass vs area basis
Variation in LMA occurs via changes in leaf density (LD; g mm−3) and/or leaf thickness (T; mm) where . Across genera, LD is relatively independent of variation in leaf thickness (Niinemets, 1999) and a recent meta-analysis by Poorter et al. (2009) showed that variation in LMA is more closely related to changes in LD than to changes in thickness. Further, previous research has shown that LL is positively correlated with LD and often independent of variation in thickness (Kitajima & Poorter, 2010; Wright & Westoby, 2002; although see Wright & Cannon, 2001). Taken together, the positive correlation between LD and LMA (Poorter et al., 2009) and LD and LL (Wright & Westoby, 2002; Kitajima & Poorter, 2010) provides firm evidence that LD strongly influences the coordination between LMA and LL.
Variation in LD, both within and between species, occurs through changes in the relative contribution of different leaf tissues toward whole leaf dry mass. Although leaves are composed of a wide variety of cell and tissue types, the total volume of a leaf (VL) can be divided into three broad component volumes that directly influence the overall structure and function of leaves: the volume occupied by air spaces (Va), the volume occupied by cell walls and the cuticle, that is, the structural volume (Vw), and the volume bound by the plasma membrane (Vc; Shipley et al., 2006; Roderick et al., 1999a,b). In general, Vw represents the vast majority of leaf dry mass such that variation in leaf dry mass is approximately equal to the product of Vw and the specific gravity of cell walls (d ), which is relatively constant at c. 1.5 across species (Desch, 1973; Siau, 1984). As such, LD can be approximated by:
Thus, increases in LD, and by extension LMA, can occur through decreases in the ratio of Vc/Vw and/or Va/Vw. Changes in Vc/Vw are driven largely by changes in the average leaf cell wall thickness relative to the average size of leaf cells, whereas changes in Va/Vw are dictated by variation in the packing of mesophyll cells, that is, changes in the fractional airspace, Va/VL. These changes in leaf architecture are expected to directly influence liquid- and/or vapor-phase transport within and between the leaf tissues that define the leaf transpiration stream (Boyer, 1974; Cruziat et al., 1980; Tyree et al., 1981; Fricke, 2000; Sack et al., 2004; Aasamaa et al., 2005; Zwieniecki et al., 2007; Ye et al., 2008; Pieruschka et al., 2010).
Although the effect of changes in Vc/Vw and/or Va/Vw on leaf water transport has not been evaluated, previous research suggests the hydraulic linkages between the xylem and extra-xylem transport pathways and the relative contribution of different leaf tissues to the transpiration stream are influenced by changes in Vc /Vw and/or Va /Vw and therefore LMA. For example, if we assume that the rehydration kinetics of leaves are symmetrical with the transpiration stream, then data from Zwieniecki et al. (2007) suggest that for leaves with relatively large bundle sheath extensions and a dense mesophyll (i.e. mesophyll with low Va/Vw), the vast majority of extra-xylem tissues are likely to participate in the fast phase of the transpiration stream; whereas for leaves with a relatively diffuse mesophyll, the epidermis appears to be the major extra-xylary tissue involved in the transpiration stream, with much less water originating from the palisade and spongy mesophyll tissues. Therefore, according to the interpretations by Zwieniecki et al. (2007), as mesophyll density increases so does the volume of liquid-phase transport pathways involved in the fast phase of the transpiration stream. Since the hydraulic architecture of a leaf approximates a hierarchical network of parallel resistors (Sack & Holbrook, 2006; McKown et al., 2010) and resistors in parallel add in reciprocal, then, all else being equal, an increase in the number of parallel resistors (e.g. space filling through greater packing of conduits) should result in lower resistance to liquid-phase transport. However, if this were true, as LMA and the number of extra-xylem pathways increase, leaf hydraulic conductance should also increase, resulting in a positive correlation between KLeaf, area and LMA. Yet, across genera no consistent relationship is observed between KLeaf, area and LMA (Fig. 1a) or LD (see appendix D from Sack & Frole, 2006). Instead a negative correlation is seen between KLeaf, mass and LMA, and KLeaf, mass and LL (Fig. 1c,d). This lack of an area-based correlation suggests that the extra-xylem pathways for water transport in leaves with a high LD and LMA are, on a pathway-by-pathway basis, more resistant than the extra-xylem pathways for water transport in leaves with a low LD and LMA, such that summing the resistances in parallel has little to no effect on KLeaf, area. Therefore, the lack of correlation between KLeaf, area and LMA, and the negative correlation among KLeaf, mass, LMA and LL provides strong evidence for a biophysical tradeoff between resource allocation to extra-xylem tissues specialized for liquid-phase processes (i.e. high KLeaf, mass) vs structural rigidity and longevity (i.e. low KLeaf, mass).
Implications for lifetime leaf water loss and carbon gain
The data presented here suggest that lifetime leaf water use is optimized to maximize leaf carbon gain. Across two orders of magnitude of variation in LA, we observed strong positive isometric scaling between KLeaf and LA. Isometric scaling between KLeaf and LA means that, on average, any increase in LA is accompanied by a proportionally equal increase in KLeaf, and thus KLeaf, area is independent of variation in LA, as seen in previous evaluations and modeling simulations (Sack & Frole, 2006; McKown et al., 2010). Additionally, during periods of leaf water loss and CO2 assimilation, we observed a relatively constant driving gradient for water transport from stem to leaf (i.e. ΔΨStem–Leaf). Furthermore, for the species grown in a common environment, Δ13C was uncorrelated with variation in LA. According to the coupled photosynthesis–leaf water balance model shown in Eqn 6, a relatively constant ΔΨStem–Leaf along with relatively small variations in Δ13C, and by extension time-integrated ci/ca (Eqn 9), suggests KLeaf, area is directly proportional to Aarea. However, despite relatively constant Δ13C and ΔΨStem–Leaf, the actual amount of carbon gain per unit leaf water loss (i.e. ) is tightly coupled to variation in the water vapor concentration difference from leaf to air (i.e. (wi–wa); see Eqns 5, 6). It is widely recognized that (wi–wa) is tightly coupled to variation in leaf form, in particular variation in leaf area. This is because, larger leaves heat up faster in sunlight than smaller leaves, as a result of increased resistance to sensible heat loss, resulting in greater wi (Gates, 1980). Thus in order to minimize (wi–wa) and optimize leaf water loss for carbon gain, leaves should decrease in size in response to increasing radiation interception and atmospheric demand for water (i.e. lower wa). If LA varies in a way that minimizes (wi–wa) then the maximum potential lifetime carbon gain of a leaf is dependent on the covariation between KLeaf, mass and LL as shown by Eqn 7.
We observed negative isometeric scaling between KLeaf, mass and LL across the 24 species evaluated here (Fig. 1d). Negative isometric scaling between KLeaf, mass and LL means that, on average, any increase in LL is associated with an equally proportional decrease in KLeaf, mass. Therefore, according to the lifetime leaf water use model described in Eqn 8, since variation in LL results in an equally proportional change in KLeaf, mass, but in the opposite direction (i.e. negative isometric scaling), and ΔΨStem–Leaf is relatively constant, lifetime leaf water use per unit leaf dry mass () is relatively constant and independent of variation in LL. Similarly according to the lifetime leaf carbon gain model described in Eqn 7, since Δ13C and ΔΨStem–Leaf are relatively constant and KLeaf, mass scales negatively and isometrically with variation in LL, lifetime leaf carbon gain per unit dry mass () is relatively constant and independent of variation in LL. Therefore, according to Eqns 7 and 8 the data reported here suggests lifetime leaf water loss has been optimized for carbon gain such that is independent of variation in LL.
Although negative isometric scaling between KLeaf, mass and LL suggests that lifetime leaf water loss has been optimized for carbon gain (i.e. is relatively constant) it is important to note that Eqns 7 and 8 are estimates of the maximum potential carbon gain and water use, yet the actual lifetime carbon gain and water use of a leaf is much more complex (Mediavilla & Escuders, 2003; Kikuzawa et al., 2004; Kikuzawa & Lechowicz, 2006; Blonder et al., 2010). For example, LL often overestimates the physiologically active LL. This is especially true for species that inhabit a seasonal climate, when temperature and water availability regularly fall below the minimum requirement for photosynthetic metabolism on an annual basis (Kikuzawa et al., 2004). Further, changes in water availability and subsequently ΨLeaf can have negative consequences for KLeaf via water stress-induced hydraulic failure (Sack & Holbrook, 2006), although, to date, there is no evidence of a tradeoff between KLeaf and the susceptibility to water stress-induced hydraulic failure on either an area basis (KLeaf, area; Blackman et al., 2010; Sack & Holbrook, 2006; Sack et al., 2003) or a mass basis (based on data from table 1 of Blackman et al., 2010; KLeaf, mass = KLeaf, area × 1/LMA). However, a weak positive relationship has been observed between LMA and the water potential at which 50% hydraulic failure occurs. Therefore, the data presented here suggest that further research into the susceptibility of KLeaf, mass to hydraulic failure, when exposed to variation in water availability, is needed to better understand the biophysical limitations on LL and the lifetime carbon gain of leaves.