Introduction

The mechanical design of a plant is fundamental for survival and related functional traits can be plastic and directly influenced by the environment (Reich et al. 2003). There is much value in investigating variation in plant functional traits, especially under water limitation (Ackerly 2004). This type of information can provide crucial insights into the inter-specific variation that exists in communities that might experience changing environmental conditions, and whether differential, species-specific responses may confer a competitive advantage on some species. This is especially relevant in understanding the possible future persistence and distribution of biological invasions, especially with changing water availability. Wood anatomy, for example, is likely to be under strong selective pressure, thus the study of xylem anatomical functional traits that may determine plant responses to dry conditions in space and time has become an increasingly important research topic (Read and Stokes 2006).

Wood density, a key anatomical trait in tree species, is indicative of mechanical and physiological life history strategies and therefore a critical functional trait for understanding the evolution and ecology of whole plant performance in different environments (Swenson and Enquist 2007). This is mainly due to wood density being responsive to environmental factors such as water availability and temperature (Hacke et al. 2001; Thomas et al. 2007; Martínez-Cabrera et al. 2009). Using landscape slope as a water availability proxy (steeper gradient = less water available), Barij et al. (2007) suggested that xylem structure was influenced by in situ hydrological properties—the less water available, the denser the wood (see also Preston et al. 2006). Mean annual precipitation (MAP) as a measure of water availability was also shown to have a significant negative correlation with wood density, as shown for Eucalyptus grandis in warm temperate regions of South Africa (Naidoo et al. 2006). However, wood density can also be positively correlated with MAP (Wiemann and Williamson 2002); Chave et al. (2006) hypothesised that dense wood in wet Neotropical forests is a safety mechanism against invertebrate attack. Still, as safe hydraulic transport is a crucial trait in plants, there is generally an important (positive) link between higher wood density and high resistance against drought-induced cavitation of xylem under water-stressed conditions (Hacke and Sperry 2001; Hacke et al. 2001; Meinzer 2003; Wheeler et al. 2005; Jacobsen et al. 2007a, b; Pratt et al. 2007; Stiller 2009).

Wood density can consequently be a valuable screening tool for identifying species that may be more vulnerable to variation in water availability in a specific community (Jacobsen et al. 2007a). Fynbos riparian zones in the south-western Cape of South Africa are predicted to endure severe future drought conditions (New 2002; Steynor et al. 2009). The invasive alien plant (IAP), Acacia mearnsii, an important forestry species in South Africa and known invader of especially riparian areas, aggravates this problem as it is a profligate water user (Dye et al. 2001; Dye and Jarmain 2004). An important question has been raised concerning the future invasion success of alien plants under global climate change—would IAPs become more invasive, and thus be even more competitive, or would their invasion success be reduced (Dukes and Mooney 1999). Using vulnerability to drought-induced cavitation, an important measure of drought tolerance in woody plants, Crous et al. (2012) showed the physiological advantage that xylem of A. mearnsii has in resisting hydraulic failure compared to key indigenous riparian tree species under proxies of decreasing water availability. This accentuates the invasive potential of A. mearnsii, however, a suite of complementary physiological and anatomical traits might better explain how a species adapt against hydraulic failure under water-stressed conditions (McCulloh et al. 2012).

Of interest, therefore, is the extent to which wood anatomy varies for native and invasive alien plants across a water availability gradient (Valladares et al. 2007; Funk 2008). A lack of anatomical variation (little or no plasticity) would suggest that species might endure significant stress under reduced water availability. IAPs in particular have displayed high trait plasticity in low resource environments, therefore differences in phenotypic plasticity between native and invasive species may provide insight into how co-occurring species could respond anatomically to changing environmental conditions (Funk 2008) and future geographic distributions (Hulme 2008). It would therefore be instructive to see how species, especially IAPs, can vary in structural adaptive strategy given different water availability proxies.

Wood density per se only represents the bulk measure of the multiple functions of xylem, i.e. it is an important broad indicator of xylem “construction cost”. Therefore, wood density needs to be examined in relation to its functional and structural correlates (e.g. vessel traits), as together they can explain the variation that exists within communities (Preston et al. 2006). Important vessel traits are vessel lumen diameter, vessel wall thickness and then combined as wall-to-lumen ratios (t/b)2 (vessel mechanical resistance against implosion, Fig. 1). Density correlates significantly and positively with wall-to-lumen ratios (t/b)2 (Hacke et al. 2001; Hacke and Sperry 2001). So, plants that grow in places with lower water potentials require higher wall thickness (stronger walls) to resist implosion and drought-induced cavitation (Sperry et al. 2006). Larger lumens, although hydraulically more efficient, are seen as indicative of less dense wood (Wagner et al. 1998; Tyree and Zimmerman 2002). It is therefore helpful to seek ecological correlations between vessel lumen diameter, vessel wall thickness, (t/b)2 (among others) and wood density, as a method of investigating the link between wood density and mechanical strength against drought stress.

Fig. 1
figure 1

Diagram defining wall thickness (t) and lumen diameter (b), showing b as the width of a hypothetical, minimum-area end wall between a full and an embolized vessel, and t as the thickness of their common wall. Bending stress due to negative water pressures can result in wall bending. This bending stress may cause walls to implode, subsequently letting air in (adapted and simplified from Hacke et al. 2001)

Full size image

Another valuable correlation with wood density is minimum water potential (Ψmin). Minimum water potential (Ψmin) is a good indicator of different plants’ access to soil water (Ackerly 2004; Bhaskar and Ackerly 2006). Therefore, if density values are correlated with the water potential where the plant is most stressed, one will find that plants that experience the highest amount of stress, due to very low water potentials, are likely to have denser wood (Ackerly 2004; Jacobsen et al. 2007a).

The rationale of our study was therefore to investigate whether plastic responses in wood anatomical traits occur in the invasive alien species A. mearnsii (a major threat to functioning of riparian zones in the south-western Cape, South Africa), across proxies of water availability, and how co-occurring natives compare with that of the IAP. This was done at a catchment scale (a narrow geographic range with a similar Mediterranean-type climatic regime), where we investigated wood density and its components (t/b)2, lumen width and wall thickness among three river systems, with streamflow volume as the surrogate for water availability (sensu Swift et al. 2008). Additionally we investigated whether A. mearnsii showed variation in wood density across a biome scale (a wide geographic range), with mean annual precipitation as the surrogate for water availability. Ultimately the following key questions were investigated:

  1. 1.

    Experiment 1: At a catchment scale, how do the woody invasive tree A. mearnsii and two key native riparian tree species respond with regards to wood density across three sites that differ in streamflow? Which anatomical traits contribute to this variation, and what could this mean for species under changing environmental conditions?

  2. 2.

    Experiment 2: At a biome scale, how plastic is A. mearnsii across sites with different precipitation, and within sites, how does its wood density compare to that of a co-occurring native species?

Methods

Description of species and study sites

We studied the evergreen tree Acacia mearnsii De Wild. (Fabaceae). This Australian species is highly invasive and a threat to riparian zones in the south-western Cape, South Africa (Enright 2000; Dye and Jarmain 2004).

For the catchment scale study (Experiment 1), the anatomy of A. mearnsii was contrasted with that of native riparian species Brabejum stellatifolium (L.) (Proteaceae) and Metrosideros angustifolia (L.) (Myrtaceae), key species in south-western Cape Mediterranean-type riparian systems, and critical for replanting and restoration in degraded riparian ecotones (Prins et al. 2004; Galatowitsch and Richardson 2005). The study sites were the Eerste River at Jonkershoek (33º57.621′S, 18º55.037′E; 200 m a.s.l.), Wit River at Bainskloof (33º34.217′S, 19º08.452′E; 279 m a.s.l.) and Molenaars River at DuToitskloof (33º41.778′S, 19º13.263′E; 297 m a.s.l.). Sites were selected according to the presence of the target species, and streamflow volumes—i.e. ranging from high to low flow. Jonkershoek represents the low flow site, Bainskloof the intermediate flow site and DuToitskloof the high flow site (data shown in Crous et al. 2012). The three sites occupy slightly different relative positions in the catchment: at Bainskloof the site is situated in the mountain stream transitional zone, while the two other sites are in the upper foothills, which have a lower gradient. The sampled trees at the different sites also occupy different elevations above and distances away from the active channel (during low flow conditions). Jonkershoek catchment has been considerably modified over the last century, in contrast to the other two catchments, where streamflows likely reflect conditions closer to the natural regime. This may impact long-lived vegetation at Jonkershoek.

For the biome scale study (Experiment 2), we compared the wood density of A. mearnsii along six riparian systems that fell within regions that differed in mean annual precipitation (MAP) (Table 1). Using the MAP as the surrogate for water availability gradient (e.g. Naidoo et al. 2006), the sites were divided into two mesic, two xeric, and two intermediate rainfall sites (Table 1). Additionally, within each site, we compared wood density of the invasive woody tree A. mearnsii against one native co-occurring species, as listed in Table 1. The biogeographic differences between sites necessitated the use of different native species at the different sites. Native species selection within sites was therefore based on the most important species (aerial cover) co-occurring along with the IAP, even though none of the species are obligate riparian species, similar to the invader. This means that only A. mearnsii was tested across the water availability gradient, whilst the native species were used to study differences in wood density within each site.

Table 1 Description of the biome scale study sites, indicating the gradient in mean annual precipitation and species with which Acacia mearnsii was compared with regard to xylem density
Full size table

Wood anatomical properties

Wood density (D t ) (Experiments 1 and 2) was determined using the method described in Hacke et al. (2000) and Pratt and Black (2006). Stem segments were cut according to a standardized canopy position, compass direction and position on branch, to reduce possible within tree variation (data not shown). Five to six stem segments from five replicate trees of each species were cut and trimmed to around 2–3 cm. Stems were then dissected longitudinally so that the pith and bark could be removed. Segments were then immersed in water and degassed for 12–24 h, or until no more air bubbles were observed. Saturated volume (xylem saturated with water after degassing) was determined using the water displacement method (Archimedes’ principle) (Hacke et al. 2000). Dry weight was determined from measuring segments after dehydration in an oven at 75 °C for 48 h. Density values were calculated by the ratio of dry weight to saturated volume (g/cm3). Density data (at both scales) were statistically compared using a factorial ANOVA followed by a Fisher LSD post hoc test (Statistica Release 8, StatSoft Inc.) to identify any between site and species differences. At the biome scale (Experiment 2), a one-way ANOVA was also used to compare A. mearnsii across sites. At the catchment scale (Experiment 1), multiple correlations were made between density and vessel traits.

Vessel traits were measured using methods described in Hacke et al. (2001) and Jacobsen et al. (2007a). Transverse sections were shaved from one stem segment per individual, and mounted on a slide. Several images were taken of each transverse section with a digital camera (Leica DFC425, Leica Microsystems, Solms, Germany) attached to a light microscope (Leica DM1000, Leica Microsystems, Solms, Germany). Images were analysed using the Leica Application Suite (LAS) software. In total, 100 measurements of vessel lumen diameter (d) and wall thickness (t) were made per individual stem segment, rendering 500 measurements per species per site. As our main interest was wood specific density and not whole stem conductivity, we used the unweighted average vessel lumen diameter as opposed to the hydraulically weighted average diameter (Preston et al. 2006). The (t/b)2, a function of two adjacent xylem vessels, where t is the thickness of the common wall and b the length of the largest lumen, was obtained by measuring at least 20 adjacent vessels per individual, giving a total of 100 (t/b)2 measurements per species per site. A factorial ANOVA, followed by a Fisher LSD post hoc test (Statistica Release 8, StatSoft Inc.) was used to determine significant differences between species across different streamflow sites (Experiment 1). To investigate any possible pattern in wood anatomical traits for tree species in the Mediterranean-climate south-western Cape riparian zones [e.g. if high wood density would also indicate high (t/b)2, and how each species contributes to this relationship] (Iovi et al. 2009), average lumen diameter and wall thickness were separately correlated with wood density as d and t, and combined as (t/b)2, and examined using regression analysis. Data were transformed as necessary to adhere to statistical models.

Xylem water potentials in the field

For the catchment scale study (Experiment 1), stem xylem water potential was measured using a pressure chamber (Pockman et al. 1995) (PMS instruments, Model 1000, Oregon, USA). Measurements were made in summer (December–February, 2008–2009), when plants experience drought stress. Midday (1200–1400 hours) measurements were made to obtain minimum water potentials (Ψmin), and were carried out on the same individuals (n = 5) used for wood density measurements. Three healthy stems (with leaves) from each individual were cut and measured immediately on site (one treatment). Three replicates of each treatment, measured over three different days, were made for each species across all sites. Data were calculated as mean ± SD (n = 9), and the relationships between Ψmin and both wood density and vessel traits were examined using regression analysis.

Results

Catchment scale study

Twig wood density values from the three species investigated ranged from 0.484 to 0.641 g cm−3 (Table 2), within the range of average wood density values according to Chave et al. (2009). Wood density differed significantly across species (P < 0.001, Table 3). Specifically, B. stellatifolium and M. angustifolia did not differ in wood density across all sites whereas A. mearnsii presented variable results (Table 2). A. mearnsii consistently had significantly higher wood density compared to both native species at the intermediate streamflow (Bainskloof) and high streamflow (DuToitskloof) sites. Within sites, the intermediate streamflow site (Bainskloof) was the only site where all the species differed significantly in wood density.

Table 2 Mean values for wood anatomical traits measured across species and sites (Experiment 1)
Full size table
Table 3 Factorial ANOVA results for measured wood anatomical traits at the catchment scale study (Experiment 1)
Full size table

Lumen diameters differed significantly across species and sites (P < 0.01, Table 3). B. stellatifolium and M. angustifolia showed a consistent trend in lumen diameter, with both species having significantly smaller diameters at the low streamflow site (Jonkershoek) as opposed to the other sites (Table 2). Within sites, lumen diameters did not differ between the native species. In contrast, A. mearnsii consistently had significantly larger lumen diameters than both native species across and within all sites. In addition, A. mearnsii did not differ in vessel lumen diameter between sites.

Vessel wall thickness showed significant differences within and between sites (P < 0.001, Table 3). All species had their greatest wall thickness at the low streamflow site, Jonkershoek. Wall thickness of A. mearnsii differed significantly across all sites. A. mearnsii was the only species to react in this way (Table 2). M. angustifolia had on average the smallest wall thickness (1.13 μm) and B. stellatifolium on average the greatest (2.89 μm).

The ability of vessels to resist implosion (t/b)2 varied significantly within and between sites (P < 0.01, Table 3). B. stellatifolium had significantly lower (t/b)2 ratios than the other species (0.0143–0.0176 μm; Table 2). A. mearnsii in turn had the highest (t/b)2 ratios (ranging from 0.0291 to 0.0372 μm), but only significantly higher than both natives at Jonkershoek and DuToitskloof. M. angustifolia was the only species that differed significantly in (t/b)2 across all sites; the other two species, B. stellatifolium and A. mearnsii, did not vary across sites. Within sites, species had their highest (t/b)2 values at either the intermediate or high streamflow sites.

Measurements of wood density across all species and sites showed a significant and positive relationship with the ability of vessels to resist implosion or (t/b)2 (R 2 = 0.77; P < 0.005) and lumen diameter (R 2 = 0.47; P < 0.05) (Fig. 2a, b). However, no relationship existed between wood density and wall thickness (t) (Fig. 2c). The strongest correlation between wood density and its component anatomical traits was with (t/b)2. Measurements of minimum water potential (Ψmin) as an indicator of plant water stress were significantly and negatively correlated with (t/b)2 and wood density (Fig. 3a, b). The Ψmin explained 80 % of the variation in (t/b)2, while it only explained 67 % of the variation in wood density. All correlations showed species occupying different parts of the curve, indicating species-specific trends in the relationship between wood anatomical characteristics.

Fig. 2
figure 2

Plots of a vessel resistance to implosion or (t/b)2, b vessel lumen diameter, and c vessel wall thickness as functions of wood density across all species and sites. In each graph the solid line shape represents Acacia mearnsii, the dotted line shape Brabejum stellatifolium, and open is Metrosideros angustifolia. The low streamflow site (Jonkershoek) is represented by circles, the intermediate streamflow site (Bainskloof) by triangles, and the high streamflow site (DuToitskloof) by diamonds

Full size image
Fig. 3
figure 3

Plots of a wood density and b vessel resistance to implosion or (t/b)2 as a function of minimum water potential (Ψmin) across all species and sites. In each graph the solid line shape represents Acacia mearnsii, the dotted line shape Brabejum stellatifolium, and open is Metrosideros angustifolia. The low streamflow site (Jonkershoek) is represented by circles, the intermediate streamflow site (Bainskloof) by triangles, and the high streamflow site (DuToitskloof) by diamonds

Full size image

Biome scale study

A one-way ANOVA of wood density of A. mearnsii vs. precipitation (across sites) showed significant intra-specific variation (P < 0.001, Fig. 4). Wood density of A. mearnsii was significantly the lowest at the two mesic sites (ranging from 0.487 to 0.511 g cm−3; P < 0.05), and on average the highest at the xeric sites, as well as one coastal site, Buffeljagsriver (ranging between 0.572 and 0.593 g cm−3). Significant differences were observed between wood density of A. mearnsii and the selected co-occurring native species within sites (P < 0.05; Table 4). Densities ranged between 0.451 and 0.564 g cm−3 for the native suite of species, and between 0.487 and 0.593 g cm−3 for the IAP (Table 4). A comparison between A. mearnsii and the chosen native species within each site showed that A. mearnsii had denser wood at both xeric sites, and at one mesic site (Stellenbosch).

Fig. 4
figure 4

Mean wood density with 95 % confidence intervals for Acacia mearnsii as a function of water availability along an environmental gradient (as described in Table 1). A One-way ANOVA was used to compare density between sites, with means separated using a Fisher LSD test. Means with different superscripts differ significantly (P < 0.05). n = 5

Full size image
Table 4 Mean values for wood density of Acacia mearnsii across a precipitation gradient and between the selected native species (Experiment 2, as described in Table 1)
Full size table

Discussion

At the catchment scale, we examined how key fynbos riparian tree species co-occurring with the alien invasive tree A. mearnsii vary (if at all) in twig wood density across three sites where streamflow quantity was the proxy for water availability. Wood anatomical traits, such as vessel mechanical resistance to prevent implosion (t/b)2, vessel lumen diameter and wall thickness were also assessed at this scale as they represent important correlates of wood density and thus have ecological relevance in explaining variation in wood density associated with water availability, both intra- and inter-specifically (Preston et al. 2006).

Inter-specific differences in wood density were significant, with the invasive species A. mearnsii showing significantly denser branch wood at the two higher streamflow sites (Bainskloof and DuToitskloof). This adds to other supportive evidence that considerable phylogenetic variation in wood density exists (Chave et al. 2006; Read and Stokes 2006). Variation observed within this trait is mainly attributed to the distribution of species in relation to water availability gradients (Ackerly 2004; Preston et al. 2006; Poorter et al. 2009; Stiller 2009), though there is evidence that temperature could also contribute (Thomas et al. 2004, 2007). In view of this, Meinzer (2003) suggested that functional convergence towards higher wood density in drier areas is a safety measure against drought-induced cavitation. Therefore, as denser wood is seen as a broad indicator of adaptation to water-stressed conditions, it is clear that for these riparian systems, A. mearnsii, at least at two sites, might functionally be more responsive towards limiting drought stress than co-occurring native species. The significant positive correlation between wood density and (t/b)2 found in our study supports this argument, as species with high resistance against vessel implosion (less vulnerability to drought-induced cavitation) are likely to have high (t/b)2 values (Hacke et al. 2001; Pratt et al. 2007). This result is consistent with the findings of Jacobsen et al. (2007a).

In fact, Ψmin was also significantly correlated with both (t/b)2 and wood density. There is a tight negative correlation between wood density, its structural correlates and minimum water potentials for species in the south-western Cape riparian zones. The influence of the environment at the plant tissue level is thus linked to changes that promote drought tolerance at the cellular level (Jacobsen et al. 2007b; Pratt et al. 2007). Importantly, it is also clear that this is a species-specific response because species clumped on different parts of the curves that explained the relationship between different functional traits (see also Iovi et al. 2009). It is therefore possible to determine different plant responses to water availability within a community, and in so doing be able to broadly predict drought tolerance of species under climate change scenarios (A. mearnsii in this case) as argued by Jacobsen et al. (2007a).

Interestingly, A. mearnsii also exhibited significantly larger lumen diameters and thinner walls compared to the native species, an indication that it might anatomically select for hydraulic efficiency rather than hydraulic safety through being drought tolerant (Hargrave et al. 1994; Kolb and Sperry 1999). Preston et al. (2006) found an inverse relationship between vessel lumen area and wood density, which they attributed to vessel lumen conductivity which increases with the fourth power of lumen diameter (see also Tyree and Zimmerman 2002; Sperry et al. 2006). The result is less dense wood, as the percentage of lumen area per transverse section of xylem is increased, ultimately limiting mechanical strength (Wagner et al. 1998). However, departure from this trend does exist, where vessel traits such as lumen diameters were found to vary independent of xylem wood density (Martínez-Cabrera et al. 2009).

In support of this indistinct relationship between vessel traits and wood density, we found that lumen diameters were positively correlated with wood density. This suggests that species with denser wood would be less adapted to drought-induced cavitation, and better equipped at transporting water, a counter-intuitive response to the hydraulic safety versus efficiency relationship (e.g. Kolb and Sperry 1999). In angiosperms, however, one cannot readily correlate conduit diameter per se against water-stress vulnerability as this was found to be a weak relationship (Hacke et al. 2000; Hacke and Sperry 2001; Jacobsen et al. 2005). This is because vulnerability to drought-induced cavitation is primarily influenced by the porosity of inter-conduit pit membranes (Hacke et al. 2000; Tyree and Zimmerman 2002; Woodruff et al. 2008; Choat and Pittermann 2009). Also, Woodrum et al. (2003) found that for Acer spp. in mesic environments (similar to our catchment scale study), there were no direct trade-offs between hydraulic conductivity and mechanical properties, as water transport at their study sites might not have been limited, and therefore had no selective influence on anatomy (Sperry et al. 2006).

This suggests a non-causal relationship between specific traits such as measured wood density and vulnerability to drought-induced cavitation (Larjavaara and Muller-Landau 2010). The observed variation in vessel diameter and wood density at the catchment scale, thus adds to our understanding of plant strategies in mesic areas, and provides insight on how plants balance water availability, leaf evaporative demand and ultimately wood anatomical traits (Shume et al. 2004).

In contrast to expectations, wood density of the native species did not differ intra-specifically between sites differing in streamflow, despite considerable inter-specific variation. Clearly streamflow per se cannot be solely used to describe differences in wood density between riparian species. More evidence to strengthen this argument comes from the relationship between wood density and Ψmin. Minimum water potential (Ψmin) was significantly negatively correlated with wood density across species, similar to Jacobsen et al.’s (2007a) findings for evergreen shrubs in the Mediterranean-type climate region of South Africa. They found that plants differed significantly in their soil water use with those experiencing a greater seasonal water stress developing a higher wood density. In the Mediterranean-type climate of California, Preston et al. (2006) also argued that the best possible ecological correlate for wood density is soil water content. It would therefore be instructive to incorporate Ψmin when investigating a possible driver of wood density, as plants’ relative access to soil water (as measured by Ψmin) (Ackerly 2004; Bhaskar and Ackerly 2006) is a more accurate measure than streamflow quantity (see e.g. Dawson and Ehleringer 1991), although the two might be interdependent. Also, not only are measurements like density and minimum water potentials relatively easy to obtain (Jacobsen et al. 2007a), but also, when applied to the South African fynbos, which is predicted to undergo influential climate change, they could provide vital information on different drought adaptive abilities of species and thus insights on possible range changes in response to reductions in water availability.

To further investigate whether we can broadly predict drought tolerance of species using wood density of species, we asked whether geographically derived differences in wood density (using MAP as a proxy for plant water availability) existed within A. mearnsii; also, how the wood density of A. mearnsii compares to that of a co-occurring native species, to see if functional trait convergence exists inter-specifically. There was clear evidence that there were significant intra-specific differences in wood density for A. mearnsii, with mesic sites having less dense wood than xeric sites, indicative of higher plasticity for the IAP. As less dense wood is correlated with increased hydraulic conductance (Wagner et al. 1998; Stratton et al. 2000), this species thus show functional trait adaptive responses across a water availability gradient. However, Swenson and Enquist (2007) suggested that the evolution of wood density is mostly correlated with variation in temperature (see also Thomas et al. 2004, 2007), and not necessarily MAP. Minimum water potential might therefore have a more prominent role in determining wood density, as suggested by the catchment scale study and literature. It is evident that the specific environmental factor that determines wood density is variable, and could be influenced by a combination of MAP, temperature and Ψmin, depending on in situ microsite conditions.

Additionally, A. mearnsii had significantly denser wood compared to the native species at both xeric sites, an indication of inter-specific variation. Thus, in sites where plants have less access to soil moisture, this species should have an advantage in resisting xylem failure associated with drought conditions compared to the studied native species. This is consistent with what was found at the catchment scale study. The observed plastic response of A. mearnsii to variable water availability is very likely to have a positive influence on how this species responds to changes in hydrological and climatological conditions (Funk 2008) projected for the fynbos region.

Conclusion

At the catchment scale, wood density showed considerable inter-specific variation, but to a lesser extent intra-specific variation, indicating a higher degree of plasticity as opposed to functional convergence in this parameter of wood anatomy between taxonomically different species across these three riparian zones. In fact it was the invasive A. mearnsii which showed denser wood than the native species, particularly at the higher streamflow sites. As wood density also positively correlated with vessel resistance against implosion (t/b)2, it seems that under certain site-specific conditions, A. mearnsii is apparently the most cavitation-resistant species amongst the three. This agrees with the physiological findings from Crous et al. (2012), emphasising the ability of this species to optimise structure and function under water-stressed conditions. These findings were also evident at the biome scale which again accentuated the invasive potential of this species in a changing environment. At this scale, there were distinct differences in wood density between species (A. mearnsii versus the native species within sites), and between sites (plastic response of A. mearnsii across an environmental gradient). These results support the argument that water availability has a strong selective effect on wood anatomy (Read and Stokes 2006; Sperry et al. 2006). To summarize, based on our results, it seems that this invasive species shows generally greater plasticity in anatomical traits compared to the studied native species, a finding of which relatively little information exists in literature. Further, for A. mearnsii at least, wood anatomical traits can be an easily obtainable screening tool of possible adaptation under different environments.