Under pressure: how a Mediterranean high-mountain forb coordinates growth and hydraulic xylem anatomy in response to temperature and water constraints

Authors


Summary

  1. Plant growth in Mediterranean high mountains is limited by the double climatic stress of low winter temperatures and summer drought. Elevational shifts in response to climate change may be complex for species whose distribution is constrained by several climatic factors.
  2. We used herb-chronology, that is, the analysis of annual rings in the secondary root xylem of perennial forbs, to evaluate life-long factors constraining secondary growth and xylem hydraulic anatomy along an elevational gradient from the upper to the lower distribution limits of the alpine forb Silene ciliata at its southernmost distribution range.
  3. Generalized additive models (GAM) showed that annual ring width (RW) in S. ciliata was greatest at intermediate elevations and smallest at the upper and lower limits of its elevational range. In contrast, maximal vessel area (MVA) was greatest at lower elevations. RW responded to climatic conditions in early spring and late summer, suggesting the presence of a bimodal xylogenesis. Structural equation modelling (SEM) indicated a positive influence of MVA on RW in the same year; improved hydraulic efficiency seemed to promote higher secondary growth.
  4. The observed greatest secondary growth (RW) and maximal vessel area (MVA) at intermediate and intermediate-low elevations, respectively, contrasts with previous evidence of an improvement in plant reproduction and recruitment with increasing elevation for S. ciliata. However, our results are in agreement with other indicators suggesting that best conditions occur at intermediate elevations, such as better seed quality or larger genome size.
  5. This study reinforces the evidence that the response of high-mountain plants to climatic change under simultaneous temperature and drought stress is complex and that models that simply assume an increase in elevation as a response to higher temperatures may fail to predict future responses to climate change.

Introduction

World-wide rising temperatures deriving from enhanced concentrations of greenhouse gases in the atmosphere are causing shifts in species' distributions (Wilson et al. 2007). Distribution ranges of species are readjusting to these new thermal conditions, generally leading to declines at the warmer and expansions at the cooler distribution edges (Parmesan & Yohe 2003; Thomas, Franco & Hill 2006). Mountains are particularly prone to experiencing these changes, since steep temperature gradients occur over very short distances (Körner 2003), thus allowing species to rapidly colonize suitable habitats. Mathematical models predicting elevational shifts in species distribution under future climatic scenarios (Randin et al. 2009; Engler et al. 2011) are receiving strong support from empirical evidence both at local (Grabherr, Gottfried & Pauli 1994; Sanz-Elorza et al. 2003) and continental scales (Chen et al. 2011; Gottfried et al. 2012).

Range shifts in response to climatic change may be far more complex for species whose distribution is concomitantly constrained by additional environmental factors such as water availability. In fact, novel climatic conditions are already leading to elevational descents of vegetation belts following increased rainfall under a Mediterranean climate (Crimmins et al. 2011). A major characteristic of the Mediterranean climate is the so-called double stress (i.e. low winter temperature and summer drought), which has been shown to limit plant growth (Mitrakos 1980; Camarero et al. 2012). In Mediterranean high mountains, this double stress seriously constrains plant growth (Giménez-Benavides, Escudero & Iriondo 2007a; García-Cervigón et al. 2012) and leads to a highly specialized flora (Väre et al. 2003). Current climatic scenarios for the Western Mediterranean Basin project a higher than average increase in temperature and strong reductions in rainfall leading to enhanced summer drought conditions (Nogues-Bravo et al. 2008), thus rendering this area particularly sensitive to climatic change. Under these anticipated conditions, plants in Mediterranean high mountains may not be able to cope with climatic change by simply increasing their elevational limits. Accordingly, empirical models and long-term monitoring data project a decrease in Mediterranean high-mountain flora richness (Benito, Lorite & Penas 2011); a pattern that contrasts with the observed increase of species richness evidenced in other European mountains (Engler et al. 2011; Pauli et al. 2012). More importantly, species loss is particularly high among endemic species (Pauli et al. 2012), thus contributing to the global biodiversity loss.

Unveiling the factors controlling plant growth and performance in alpine forbs (von Arx, Archer & Hughes 2012; Franklin 2012) is necessary to move from an observational framework to a mechanistic explanation of responses to climatic change. The application of herb-chronology, that is, the analysis of annual rings in the secondary root xylem of perennial forbs, provides a temporal baseline to describe factors controlling secondary growth (Dietz & von Arx 2005; von Arx & Dietz 2006; Liu & Zhang 2010; Olano, Eugenio & Escudero 2011; von Arx, Archer & Hughes 2012; Franklin 2012). Moreover, annual rings also contain information on a plant's anatomical adjustments to hydraulic constraints (Fonti et al. 2010; von Arx, Archer & Hughes 2012). In particular, earlywood vessels (which are the widest) may be subject to a safety vs. efficiency trade-off (Hacke & Sperry 2001; Tyree 2003; Hacke, Jacobsen & Pratt 2009; Sperry et al. 2012): they may be vulnerable to both freezing- and drought-induced failure by cavitation. Yet, the widest vessels contribute disproportionally to overall hydraulic conductivity, because potential hydraulic conductivity increases with vessel diameter at the fourth power, according to Poiseuille's law (Tyree & Zimmermann 2002). Consequently, maximizing vessel diameter may increase hydraulic efficiency and therefore plant performance. In this sense, wider vessels have indeed been shown to occur in moister microhabitats in plants growing in alpine environments (von Arx, Archer & Hughes 2012). Such plasticity in anatomical characteristics may enhance the ability to tolerate a wider range of climatic conditions (Villar-Salvador et al. 1997; Alla & Camarero 2012; von Arx, Archer & Hughes 2012) and to cope with environmental changes such as climate warming (Matesanz, Gianoli & Valladares 2010).

In order to improve our knowledge of factors controlling plant growth and performance in Mediterranean high-mountain plants, we evaluated secondary growth and hydraulic xylem anatomy of Silene ciliata Pourr. in four populations along a 450 m gradient from its lowest to its highest elevational limits at Peñalara peak (Sierra de Guadarrama National Park, Madrid, Spain). The study sites were selected because Peñalara peak constitutes this species' southernmost limit of distribution and because abundant information on S. ciliata's demography, reproductive performance and genetic variability along this particular elevational gradient is available (Giménez-Benavides, Escudero & Iriondo 2007a,b; Giménez-Benavides et al. 2011a,b; García-Fernández et al. 2012, 2013), thus allowing a more comprehensive interpretation of the data. We hypothesized that (i) summer drought would constrain secondary growth, resulting in increasing RW with increasing elevation, as temperature and summer drought stress decrease; (ii) the widest earlywood vessels (MVA) would occur at the intermediate elevation, and MVA would decrease towards the lower and upper range limit to reduce cavitations during summer drought and late frosts, respectively and (iii) greater potential hydraulic conductivity obtained through wider vessels would result in wider root rings.

Materials and methods

Study Species

Silene ciliata Pourr. (Caryophyllaceae) is a rosette-forming long-lived nanochamaephyte inhabiting circum-Mediterranean mountains. Flowering has a strong photoperiodic control, taking place from early July to mid-September (Giménez-Benavides, Escudero & Iriondo 2007b; Giménez-Benavides et al. 2011b). Seeds are produced at the end of the growing season from August to September, and emergence is delayed until next spring (Giménez-Benavides, Escudero & Iriondo 2007a), with summer drought being the main bottleneck for plant establishment (Giménez-Benavides et al. 2011a). The species reaches its southernmost (driest) Iberian limit in the mountain ranges of central Spain, where it grows from the tree line (±1900 m a.s.l.) up to the highest summits (±2600 m a.s.l.).

Study Area

The study area was located in Sierra de Guadarrama National Park, 50 km north of Madrid in Central Spain (40°46′N, 4°19′W). Climate of the area is oromediterranean (Rivas-Martínez & Loidi 1997) with an average annual temperature of 6·4 °C. Mean annual precipitation is 1350 mm, with a pronounced drought from July to August, when <10% of total annual rainfall occurs (Fig. 1). Snowfall begins in October, and the snow-free season begins in April – June; average snow cover duration is 100–140 day year−1 (Palacios, de Andrés & Luengo 2003). Over the period 1960–2004, mean monthly temperature increased 0·4 °C per decade for winter months and 0·3 °C per decade for growing season months (Giménez-Benavides, Escudero & Iriondo 2007b). No trends are observed for total rainfall, but the snow-free period increased by 19·7 days over the same 44-year period (Giménez-Benavides, Escudero & Iriondo 2007b). Climatic records on mean monthly temperature and total monthly precipitation were obtained from Navacerrada Pass meteorological station, at 1860 m a.s.l., 5–7 km away from the study sites.

Figure 1.

Climatic diagram obtained from Navacerrada Pass meteorological station (1860 m) for the period 1965–2010. Dashed line corresponds to rainfall and continuous line to temperature.

Vegetation is open and gradually changes from sparse shrub formations dominated by Cytisus oromediterraneus Rivas Mart. & al. and Juniperus communis L. subsp. alpina (Suter) Čelak near the tree line, to discontinuous cryophilic pastures at higher elevations. Densification and elevational ascent of the shrub belt have been reported for this mountain (Sanz-Elorza et al. 2003).

The survey was conducted along an elevational gradient of 450 m around Peñalara Peak, with sample sites at four elevational levels (S1: 1986 m, S2: 2084 m, S3: 2256 m, S4: 2413 m). Three sites corresponded to areas analysed in previous studies (S1-Low, S3-Intermediate and S4-High) (Giménez-Benavides, Escudero & Iriondo 2007a,b; Giménez-Benavides et al. 2011a,b; García-Fernández et al. 2012, 2013). S1 and S2 were south-east-oriented concave areas with S. ciliata growing in clear patches between shrubs, whereas S3 and S4 were north-west-oriented open pastures, flat to slightly concave in S3 and convex in S4. Soils were poorly developed and slightly acidic with pH values ranging from 5·5 to 6·2 (Giménez-Benavides, Escudero & Iriondo 2007a). Texture showed remarkable differences among sites; particle size increased with elevation, shifting from the dominance of fine materials in S1 and S2 to that of fine gravel in S3 and of gravel in the summit region (S4) (Giménez-Benavides, Escudero & Iriondo 2007a), this fact being presumably associated with a decrease in water holding capacity along with elevation. The snow-free period starts 14–21 days earlier in the low site (S1) than at the peak (S4) (Giménez-Benavides, Escudero & Iriondo 2007b). Soil water content is depleted at all sites during summer (mid-June–August), with drought starting 14–21 days earlier at the lowest than at the highest site (Giménez-Benavides, Escudero & Iriondo 2007a). A recovery in soil water content tends to occur after late July-August storms (Giménez-Benavides, Escudero & Iriondo 2007a).

Sampling Design and Anatomical Analysis Protocol

A total of 40 individuals of S. ciliata (10 per site), representing medium-to-large-sized plants, were selected. Two perpendicular diameters of the basal rosette were measured per plant. Afterwards, plants were carefully harvested in September 2010 and kept frozen for preservation.

In the laboratory, roots were mounted in a sledge microtome (© H. Gärtner / F.H. Schweingruber, Birmensdorf, Switzerland) in order to obtain 10- to 15-μm-thick cross sections from the root collars. The histological cross sections were then stained with Alcian blue (1% water soluble) and safranin (0·125%) and permanently embedded with Eukitt® glue (O. Kindler GmbH, Freiburg, Germany). Images of the cross sections were captured with a digital camera (Nikon D-90) mounted on an optical microscope (Nikon Eclipse 50 Nikon Corp., Tokyo, Japan), using 100× magnification. Generally, three overlapping images from the bark to the pith from a sector of the cross section were taken, accounting for approximately one eighth of the total xylem surface. Images were analysed using the image analysis tool ROXAS (version 1.2.0.68, © G. von Arx, Birmensdorf, Switzerland, www.wsl.ch/roxas2; von Arx & Dietz 2005), developed for Image Pro Plus (version 6.1 for Windows XP x64; Media Cybernetics, Inc., Silverspring, MD, USA).

In a first step, annual growth rings were manually delineated starting from the cambial zone (Fig. 2) up to the point where annual ring borders could not be clearly identified due to blurry limits (which represent the first years of life of the plants). Plant age was determined by counting the number of delineated growth rings and adding the estimated number of the poorly distinguishable growth rings along the remaining tissue up to the pith. Series' cross-dating was visually performed on RW measurements. Vessel recognition was performed by ROXAS. Only vessels >45 μm2 were considered, thus comprising all earlywood vessels. The analyses provided annually resolved information on (i) ring width (RW) and (ii) maximal vessel area (MVA). MVA was selected because largest vessels contribute disproportionally to overall hydraulic conductivity, and at the same time they are under the strongest climatic control (García-González & Fonti 2006, 2008).

Figure 2.

Permanent histological cross section of the root collar of Silene ciliata. Lines indicate the annual ring borders. In yellow, all sampled vessels for year 2009. The widest earlywood vessel of each annual ring within the current view (referred to as maximal vessel area MVA, in the text) is shown in black. Scale bar represents 100 microns.

Statistical Analysis

In order to distinguish the relative effects of plant age, climate and site on RW and MVA, generalized additive models (GAM) (Hastie & Tibshirani 1986) were performed. GAMs are a flexible extension of the generalized linear models (McCullagh & Nelder 1989; Zuur et al. 2009) allowing the inclusion of linear and nonlinear complex response shapes in one model. The ability to model nonlinear responses may be particularly useful in order to model the complex pattern of ontogenetic trends occurring in medium-lived perennial plants. Thus, the fixed part of every model included cambial age at each analysed year, climatic parameters (monthly average temperature and accumulated rainfall from March to September) and site as a nominal parameter. Based on preliminary exploration, cambial age was modelled with a smoother (Zuur et al. 2009), and linear relations were assumed for the remaining parameters. It is important to note that this approach was selected because it has the potential to distinguish age-related effects from climatic or site effects. Homoscedasticity of residuals was visually checked. The optimal model was chosen by comparing all potential nested models. The best model was selected based on AIC using maximum likelihood method (ML) (Zuur et al. 2009). When several models showed similar AIC values (ΔAIC < 2, Burnham & Anderson 2002), the model with the simplest fixed component was chosen for the sake of parsimony. Once selected, the best model was refitted by means of REML to obtain estimates of factor effects and their significance. Statistical analyses were performed in R, version 2.13.1 (R Development Core Team 2011) including the packages ‘nmle’ and ‘mgcv’.

Structural equation models (SEMs) were used to model the interactions among the different xylem parameters (RW, MVA) and how they were influenced by rosette size and age. Only data corresponding to the sampling year (2010) were used, since rosette size was only available for that year. SEM evaluates how well data support a set of hypothesized relationships among variables, including the strength of both direct and indirect relationships (Grace et al. 2010). The model included a positive effect of plant age on rosette size; both parameters exerted a direct effect on annual ring characteristics: RW was expected to decline with age as a consequence of an allometric relationship between RW and basal area, whereas MVA was expected to increase. Plants having a larger rosette were expected to grow faster and have wider vessels because larger plants require a more efficient water transport system (Tulik, Marciszewska & Adamczyk 2010). Plants showing larger MVA values were expected to overall have a more efficient water uptake and to grow faster, leading to larger RW (Tyree 2003). Endogenous variables met multinormality, and estimation was based on maximum likelihood estimates. Goodness-of-fit (χ2) test was used to compare the significance of the model against a full saturated model, a nonsignificant difference in goodness-of-fit indicating that the most simple model does not differ in descriptive ability from the full saturated model (Iriondo, Albert & Escudero 2003). The validity of the model was tested by the goodness-of-fit index (GFI) and root mean square error of approximation (RMSEA). GFI is independent of estimation methods and ranges between 0 and 1, with values above 0·90 indicating good fit (Tanaka 1987). RMSEA is based on predicted vs. observed covariance and includes a correction for model complexity. RMSEA is <0·05 for very good models (close fit), <0·1 for models that fit adequately, and >0·1 for poorly fitted models. Analyses were conducted with AMOS 5.0 (Amos Development Corporation, Chicago, IL, USA).

Results

Sampled plants were aged between 10 and 45 years, with an average age of 25 ± 7 years (mean ± standard deviation). Average plant rosette diameter was 8 ± 2 cm, ranging from 3 to 15 cm. The best model explaining rosette size (R2 = 0·38; F = 5·285; = 0·002) included the additive effects of plant age (F = 10·041; = 0·003) and site (F = 4·419; = 0·01). For a given age, the largest rosettes were found in site S3, and the smallest at the lowest site (S1; Table 1).

Table 1. Elevation and Silene ciliata parameters description for the four sampling sites (mean ± standard deviation)
SiteElevation (m)Sample sizePlant age (years)Number of analysed ringsRosette size (cm)RW (μm)MVA (μm2)
  1. RW = ring width; MVA = maximal vessel area.

S119861031 ± 821 ± 78 ± 277 ± 58263 ± 127
S220841022 ± 818 ± 69 ± 4103 ± 41253 ± 117
S322561022 ± 618 ± 57 ± 194 ± 36239 ± 121
S424131024 ± 420 ± 36 ± 180 ± 32215 ± 106

A total of 768 annual rings and 30,213 vessels were measured in the 40 individuals. The average number of measured rings per plant was 19 ± 5, ranging from 8 to 30 (Table 1). Estimated plant age and number of measured rings were highly correlated (r = 0·73; < 0·001); thus, more annual rings were measured from older individuals. Secondary root growth averaged 87·6 ± 39·5 μm. Average area of largest vessel lumen (MVA) per ring was 242·2 ± 119·6 μm2, ranging from 49·2 to 779·6 μm2.

Generalized additive models model for RW (n plants = 40; n rings = 768; < 0·001) indicated that this parameter strongly depended on plant age, as well as climatic and site conditions (Table 2). Ring width decreased with plant age, with a steep decrease along the first 15–20 years of life, and a more gradual decrease thereafter (Fig. 3a). Climatic signal was determined by conditions at the beginning and at the end of the growing season. Rings were narrower in years featuring warm March and September weather and wider when April was warm and August rainy and warm. Site conditions exerted a strong effect on RW, with the widest rings occurring at S2, followed by S3, and with only a slight difference between the top and bottom sites (see Table 2).

Table 2. Results of generalized additive models (GAM) analysis for ring width (RW) and maximal vessel area (MVA) in Silene ciliata
 EstimateSE F P
RW
Intercept36·2110·853·34<0·001
March T−1·940·42−4·66<0·001
April T1·090·422·610·009
Aug T1·800·602·300·003
Aug P0·170·035·84<0·001
Sept P−0·05−0·02−2·680·008
Site 1 (1986 m)−3·92d1·96−2·000·046
Site 2 (2084 m)24·54a2·2610·88<0·001
Site 3 (2256 m)11·13b1·925·78<0·001
Site 4 (2413 m)0c   
 edf   
Age4·66 51·31<0·001
 EstimateSE F P
  1. edf = estimated degrees of freedom.

  2. Different superscript letters for site estimates indicate significant difference at P < 0·05.

MVA
Intercept272·1016·9616·05<0·001
March T−3·421·067−3·200·001
April T−6·711·067−6·29<0·001
May T−3·821·193−3·210·001
Jun T−3·171·077−2·940·003
Site 1 (1986 m)46·84b6·087·70<0·001
Site 2 (2084 m)61·87a5·8610·55<0·001
Site 3 (2256 m)27·75c5·495·06<0·001
Site 4 (2413 m)0d   
 edf   
Age1·97 224·4<0·001
Figure 3.

Predicted values from generalized additive models (GAM) of (a) root ring width (RW) and (b) maximal vessel area (MVA) for a plant established in 1980 at the four different sites. Climatic effect was modelled considering observed climatic conditions for each of the years. The figure allows the comparison of the weight of the different factors included in the additive models. Site, causing the differences between lines, plant age, causes the declining/increasing trend, and climate, responsible of year to year variability.

The GAM model for MVA (n plants = 40; n rings = 768; < 0·001) indicated that such parameter tended to increase monotonically with cambial age (Table 2; Fig. 3b). Climatic effects were restricted to a negative effect of temperature from March to June, the strongest signal occurring in April. MVA differed among sites, with a clear rank (S2 > S1 > S3 > S4), with the widest vessels in the lower study sites and the narrowest vessels at the peak (see Table 2).

Structural equation modelling model provided a good overall fit for the data set. Probability of chi-squared estimate was >0·05 (d.f. = 1, χ2 = 0·581, = 0·446), with GFI being >0·90 (0·993) and RMSEA being lower than 0·1 (<0·001). The selected model included a positive effect of plant age on rosette size with older plants being larger (Fig. 4). Plant age and rosette size had contrasting effects on RW; older plants had narrower rings, whereas plants with larger rosettes had wider rings for a given age. MVA was not related to rosette size, but showed a strong positive response to plant age. There was a remarkable positive effect of MVA on RW.

Figure 4.

Structural equation model describing the interactions among ring width (RW), maximal vessel area (MVA), rosette size and age for 2010. Panel a shows the full model and panel b the selected model after removing non-significant relationships. Model statistics: χ2 = 0·581, d.f. = 1, = 0·446; GFI = 0·993 and RMSEA <0·001. Arrows depict causal relationships. Positive effects are indicated by solid lines and negative effects by dashed lines. Arrow widths are proportional to path coefficients. The numbers in the paths indicate standardized regression weights. R2 – explained variance for each endogenous parameter. *< 0·05, **< 0·01.

Discussion

This is the first study investigating the plasticity of xylem growth and vessel traits to environmental changes along an elevational gradient for perennial forbs from Mediterranean high mountains. Our results show that, in contrast to alpine environments, greatest secondary growth was found at intermediate elevations and widest vessels occurred at low locations. This result reflects the existence of a double climatic limitation at lower (dry) and upper (cool) elevational limits. Secondary growth was constrained by conditions at the beginning and at the end of the growing season, which supports the existence of a summer arrest and the existence of a bimodal growth response as in other lowland Mediterranean species (Camarero, Olano & Parras 2010; Camarero, Palacio & Montserrat-Martí 2013).

Environmental Control of Root Xylem Characteristics

In alpine environments, snowmelt is one of the main sources of water for plant growth (Walker & Webber 1994). Accordingly, the observed negative influence of March temperature on RW suggests that lower March temperatures may contribute to the persistence of a larger snow pack in early spring. This prevents runoff of melting water before the start of the growing season, while snow cover reduces exposure to damaging low temperatures. However, the effect of temperature becomes positive 1 month later. Since xylogenesis is activated by a threshold temperature (Deslauriers et al. 2008; Rossi et al. 2008; Moser et al. 2010), the positive signal of April temperature may reflect an earlier activation of xylogenesis (late April or early May). April temperatures in the study area may be low when xylogenesis initiates, and the early timing of this signal may be a consequence of the ability of low stature plants to benefit from solar radiation, thus achieving higher cambial tissue temperature and a prolonged growing season (Körner 1998). Furthermore, it is widely recognized that ground-level temperatures may in fact be much higher than those recorded at the World Meteorological Organization standard of 1·25–2 m above-ground (Graae et al. 2012). In fact, a similar pattern has been observed for a prostrate shrub in Mediterranean mountains, for which the spring temperature signal on secondary growth occurred 1 month earlier than in coexisting trees (García-Cervigón et al. 2012). In Mediterranean environments, an earlier initiation of the growing season enables plants to take advantage of the favourable hydric conditions in spring and is usually related to higher investment in secondary growth (DeSoto et al. 2012).

The importance of temperature at the beginning of the growing season, and the absence of a climatic signal from May to July, contrasts with the dominant role of temperature during the middle of the growing season in arctic and alpine environments (Bär et al. 2008; Franklin 2012). Moreover, a strong response to conditions at the end of the growing season was also observed, with secondary growth showing a positive response to warm and wet August conditions. This may reveal the existence of a second growth pulse after the period of summer drought. Bimodal xylogenesis, a partition of secondary growth in two different temporal periods, has been described for Mediterranean environments as a response to the separation of favourable conditions into two distinct periods, before and after summer drought (de Luis et al. 2007; Camarero, Olano & Parras 2010; Camarero, Palacio & Montserrat-Martí 2013), and our results suggest that this process might also occur in Mediterranean high-mountain flora.

Maximal vessel area was mainly controlled by factors occurring prior to or at the beginning of the growing season, as previously reported (Fonti, Solomonoff & García-González 2007). The positive relationship of low early-season temperatures with MVA is probably mediated through the associated lower levels of the growth hormone auxin (Gray et al. 1998). Low auxin levels extend the time required for vascular cell differentiation, which leads to wider vessels (Aloni & Zimmermann 1983; Aloni et al. 2006). In addition, lower temperatures in the early season promote slower snowmelt, thus maintaining favourable soil moisture for a longer period. Increased water availability would promote larger vessels and would be consistent with a safety vs. efficiency trade-off (Hacke & Sperry 2001). The long period affecting MVA (March–June) may reflect temporal differences in the onset of the xylogenesis and the finalization of vessel enlargement along the elevational gradient.

Although an increase in MVA with larger plant sizes is a common phenomenon reflecting tapering or widening effect (Anfodillo et al. 2006), our results indicate that MVA increased along with plant age, but not with above-ground (rosette) size. This unexpected result may be related to the fact that most of these relationships have been described for trees, in which above-ground biomass mostly persists and increases throughout plant life. However, in forbs like S. ciliata, above-ground biomass shows a strong annual turnover and shoot height does not necessarily increase with plant age, which contrasts with the generally accumulating growth pattern of the root system (von Arx et al., unpublished data). In this sense, the observed age effect on MVA might not reflect a purely ontogenetic signal, but rather the role of plant age as a surrogate for below-ground size. This would underline the need to understand the role of roots – in addition to above-ground structures – when investigating the response of alpine forbs to climate. High conductance has been proposed as a necessary prerequisite for rapid growth rates (Tyree 2003), and higher water transport capacity has already been related to higher growth rates at an interspecific level (Ze-Xin et al. 2012). However, to the best of our knowledge, this is the first study showing that this relationship may be also valid at the intraspecific level: individuals with wider vessels at the start of the growing season would therefore produce a wider ring in the same year.

Response to Elevation

A decrease in secondary growth with elevation has been described for perennial forbs and interpreted as a response to generally harsher conditions and a shorter growing season (von Arx, Edwards & Dietz 2006). Our study species departed from this pattern, with secondary growth being greatest at intermediate elevations and smallest at both upper and lower limits, as recently shown for mountain forests in California (Trujillo et al. 2012). This may be an effect of the presence of a double climatic stress in Mediterranean mountains that constrains plant growth at low and high sites. Higher temperature at low sites allows an earlier onset of xylogenesis, but involves an earlier start of summer drought stress. The decrease in temperature and later initiation of snow melting that occur along with increasing elevation may alleviate this summer stress, enabling the persistence of water in soil for a longer period (Giménez-Benavides, Escudero & Iriondo 2007a), but at the highest elevations, this extra growing time may not be enough to compensate for the delay in the onset of xylogenesis. Changes in soil texture may affect differentially water retention capacity along the elevational gradient, although observational studies show that depletion in soil water content is slower at higher elevations (Giménez-Benavides, Escudero & Iriondo 2007a).

In contrast to RW, MVA was greatest in plants at lower sites, a pattern that may seem counterintuitive, since these plants are expected to suffer from more severe summer drought stress. However, the higher vegetation cover at lower sites may lead to a higher competition among plants and therefore a faster exploitation of soil water. In this competitive situation, wider and thus more efficient vessels in early season, when soils are still moist, may be advantageous and may in fact promote higher secondary growth (RW; see Fig. 4). Support for this explanation comes from a study by Martín et al. (2010), in which the authors interpreted larger tracheids from pines of drier provenances as an adaptation aimed to maximize water uptake during shorter growing seasons. Moreover, the assumed higher cavitation risk related to wider vessels may be counterbalanced by plant local adaptation to drought, since common garden experiments have shown that low-site plants exhibit a higher tolerance to drought (García-Fernández et al. 2013), and by the ability to refill embolised vessels (Secchi & Zwieniecki 2011; Cao et al. 2012).

Secondary growth (higher at intermediate elevations) and MVA as a surrogate for competitive ability (higher at intermediate to low elevations) suggest that environmental conditions are not optimal at the highest elevations. This result contrasts with previous studies on the same species showing that plant reproduction (Giménez-Benavides, Escudero & Iriondo 2007b) and demographic rates (Giménez-Benavides et al. 2011a,b) improve with increasing elevation. Such disparate results may be a consequence of the different timing of the factors controlling growth and reproduction in S. ciliata. Flowering and seed set occur during summer, thus reductions in temperature and delays in snow melting date associated with higher elevation alleviate summer drought and may improve plant reproduction (Giménez-Benavides, Escudero & Iriondo 2007b). Summer drought is also the most critical stage for plant recruitment for Silene ciliata (Giménez-Benavides, Escudero & Iriondo 2007a), as in other Mediterranean environments (Olano, Eugenio & Escudero 2011), thus leading to higher seedling mortality rates at lower elevations. In contrast, secondary growth and xylem anatomy are mostly controlled by environmental conditions occurring between snowmelt and summer drought. Secondary growth can avoid summer stress through growth arrest during the drought the effect of period and reactivation in late summer when climatic conditions are favourable again, but reproductive processes must withstand summer conditions. Interestingly, our results are in agreement with other indicators suggesting optimal conditions for S. ciliata at intermediate elevations, such as higher seed quality (Giménez-Benavides, Escudero & Iriondo 2007a) or larger genome size (García-Fernández et al. 2012).

The considerable variance in xylem hydraulic traits likely reflects plastic plant responses to variable water availability, which demonstrates the importance of such functional traits for plant growth and survival in water-limited habitats such as Mediterranean high mountains. The existence of divergent strategies for individuals at the extremes of the elevational gradient is congruent with the hypothesis of demographic compensation proposed by Doak & Morris (2010). This hypothesis suggests that favouring different demographic traits may allow plants to persist under very different climatic conditions, as recently shown for a different species in the study area (García-Camacho, Albert & Escudero 2012). In a similar fashion, S. ciliata plants at the dry edge of their range would favour rapid growth and tolerance to drought, whereas individuals at the cool edge would prioritize reproduction. This study reinforces the evidence that Mediterranean high mountain plants respond to climatic change in a complex way. It also shows that in areas experiencing complex gradients of limiting factors such as the Mediterranean high mountains, models based on a simple increase in elevation as a response to increased temperatures may fail to predict future plant responses to climate change.

Acknowledgements

Authors thank L. Schneider for training us on preparing histological sections. We thank the Natural Park Peñalara for permission and to Adrián Escudero, David Sánchez, Alfredo García-Fernández and Txema Iriondo for hints on species biology and field assistance. Txema Iriondo very kindly reviewed a previous version. Ana García-Cervigón collaborated with figure edition. David Brown edited the English. This work has been supported by the Spanish Ministerio de Ciencia e Innovación projects ISLAS, CGL2009-13190-C03-03 and CGL2012-34209.

Ancillary