Relationship between post-fire regeneration and leaf economics spectrum in Mediterranean woody species

Authors

  • S. Saura-Mas,

    Corresponding author
    1. Department of Animal and Plant Biology and Ecology, CREAF (Centre for Ecological Research and Forestry Applications), Unit of Ecology, Autonomous University of Barcelona, E-08193 Bellaterra, Barcelona, Spain; and
    Search for more papers by this author
  • B. Shipley,

    1. Département de Biologie, Université de Sherbrooke, Sherbrooke, Quebec, Canada J1K 2R1
    Search for more papers by this author
  • F. Lloret

    1. Department of Animal and Plant Biology and Ecology, CREAF (Centre for Ecological Research and Forestry Applications), Unit of Ecology, Autonomous University of Barcelona, E-08193 Bellaterra, Barcelona, Spain; and
    Search for more papers by this author

*Correspondence author. E-mail: s.sauramas@creaf.uab.es

Summary

  • 1Recent work has identified global-scale relationships between key leaf traits (leaf economics spectrum). However, it is important to determine whether this approach can be applied at local scale with smaller subsets of species facing similar environments. Since fire is a key process in Mediterranean shrubland dynamics we analyze whether fire-related life-history traits influence the pattern of correlation between the leaf economic spectrum and leaf moisture traits.
  • 2Using structural equation modelling and exploratory path analysis, we developed alternative models to test how interspecific leaf traits are related to the seasonal variation of water content (leaves and shoots) and to the type of post-fire regeneration of Mediterranean woody species.
  • 3This study demonstrates that for these species seasonal variation in water content and fuel moisture would be better predicted by the presence or absence of a trait describing post-fire seedling establishment than by the leaf economic spectrum traits. However, leaf dry matter content (LDMC) is influenced by both the leaf economic spectrum and the post-fire regenerative type.
  • 4Seeder species (those that recruit via seeds immediately after fire) present lower LDMC and higher relative seasonal variation of relative water content (RWCrsv) than non-seeders. We hypothesize that since seeder species mostly evolved under the Mediterranean climate, they developed a particular strategy of drought tolerance (without causing an effect to the relation between the volume occupied by cytoplasm relative to the volume occupied by cell walls), which is the cause of the observed relation between LDMC and RWCrsv.
  • 5This study suggests that the leaves of Mediterranean woody species would follow the general leaf economics spectrum (Wright et al. 2004) but that specific selective forces, such as disturbance regime, acting at regional scale also play a relevant role to explain leaf traits related to water content.

Introduction

The relationship among leaf traits has been intensely studied because of their ecological significance. Besides affecting species’ performance, some leaf traits such as leaf dry matter content (LDMC), specific leaf area (SLA) and leaf nitrogen content are also indicators of ecosystem functioning (Lavorel, Mcintyre & Grigulis et al. 1999; Garnier et al. 2001a). Since these traits describe a given functional unit (leaf), it may be expected to share common, fundamental trade-offs that can be scaled up to the ecosystem level (Wright et al. 2004; Shipley et al. 2006; Whitfield 2006). Therefore, understanding how these traits are interrelated should allow us to understand plant functioning in a more integrative way, and summarize their variation (within and between habitats) by the use of a more simplified trait axis (Wright et al. 2006).

Reich, Walters & Ellsworth (1997) demonstrated that there are globally valid relationships between such functional leaf traits, such as increase of carbon gain (photosynthesis) and loss (respiration) in similar proportion with decreasing leaf life span, increasing leaf nitrogen content, and increasing leaf surface area : mass ratio. These relationships would allow the development of more quantitative and predictive models of vegetation productivity, distribution and dynamics. Wright et al. (2004) described similar relationships using a global plant trait network (GLOPNET data set) including a larger number of species. They proposed a global spectrum of leaf economics, according to chemical, structural and physiological properties, that run from highly active, cheap, short-lived leafs to tough, little active, expensive, long-lived leaves. They concluded that this spectrum operated largely independent of growth form, plant functional type or biome.

Despite such large-scale results, it is important to determine to what extent this approach can be applied at local scale to smaller subsets of species facing similar environments. If the trait correlations defining the leaf economic spectrum are caused by basic physiological or morphological trade-offs experienced by all leaves, then the strength of the pattern should be consistent across communities in specific environments. In the present study we analyze several variables (SLA, LDMC, leaf nitrogen, phosphorus and carbon content), in a group of 28 Mediterranean woody species that coexist in a community that is periodically affected by fire. Mediterranean-type ecosystems are subjected to a strong seasonal climatic regime with hot summers, mild winters, and rain concentrated in the autumn and spring. For this reason, we also considered in our analysis the seasonal variation in relative water content (RWC) of leaves and live fine fuel moisture. Here, we test whether fire life-history traits influence the patterns of correlation describing the leaf economic spectrum and leaf moisture traits.

Shipley et al. (2006), using structural equation modelling and exploratory path analysis, studied the multivariate patterns of covariation between four key variables of the leaf spectrum (leaf mass per area, nitrogen content per mass, leaf longevity and maximum photosynthetic rate) and proposed an explanation for these patterns based on two general causes. The first cause reflected Kikuzawa's (1995) hypothesis that natural selection optimizes plant carbon fixation over the life span of a leaf such that leaf longevity should be negatively related to maximum photosynthetic rate and positively to leaf construction cost. Shipley et al. (2006) found that there was strong statistical evidence that interspecific patterns of correlation among these four leaf traits were generated by a single unmeasured variable (or group of tightly correlated variables) (i.e. a latent variable) that was a common cause of each measured variable. Shipley et al. (2006) hypothesized that this single latent variable was the ratio of cytoplasmic volume : cell wall volume (Vc/Vw). This ratio describes a necessary trade-off between the processes occurring in the cytoplasm whereby larger cells (thus, high cytoplasmic volume) with thin cell walls (thus, low cell wall volume) are expected to be more photosynthetically active, while smaller cells (thus, low cytoplasmic volume) surrounded by thick cell walls (thus, low cell wall volume) should show greater longevity under a regime of lower light availability. This fundamental trade-off is likely to scale up to leaf traits and ultimately to individual-level attributes.

Shipley et al. (2006) theorized the presence of the latent variable, but an unambiguous demonstration of its physical nature was not provided. Although this critical step remains to be done, in this paper we further explore the theoretical explanation. In particular, we hypothesize that if this latent variable corresponds to Vc/Vw then this should also explain the relationship between seasonal variation in RWC of leaves and LDMC that is observed in a group of Mediterranean woody species. The rationale for this hypothesis is that plants successfully able to reduce cell volume under dehydration conditions (Canny & Huang 2006) (that is, those with higher seasonal variation in RWC of leaves) are expected to show higher cell elasticity, which is inversely related to cell wall thickness (Marshall & Dumbroff 1999). Specifically, in our causal analysis, we include SLA, LDMC, leaf nitrogen, phosphorous and carbon contents, and then we extend the analysis to include seasonal variation in RWC of leaves and of live fine fuel moisture.

Since fire plays an important ecological role in Mediterranean ecosystems, the mode of post-fire regeneration (PFR) was also considered as a variable. Most Mediterranean woody species display effective PFR mechanisms (Trabaud 1987; Trabaud 1991; Keeley 1995; Ackerly 2004). Seeders are ‘propagule-persisters’, because a local population persists in propagule form (seed, fruit) after a fire. In these species, populations regenerate by establishing seedlings just after fire from a persistent seed bank. Indeed, the recruitment of new individuals is often enhanced by fire (Pausas et al. 2004) and seedlings established shortly after fire usually present higher survival and growth rates than those establishing in periods between fires (Lloret 1998). ‘Non-seeders’ are species in which the propagule (seed, fruit) does not persist after fire. Consequently, effective propagules arrive only by dispersal from outside the area affected by fire. Commonly, seeders are species with a relatively short life cycle and a high recruitment after a fire while non-seeders are species with longer life cycle and lower recruitment after fire (Pate et al. 1990; Verdaguer & Ojeda 2002; Lloret 2004; Knox & Clarke 2005; Schwilk & Ackerley 2005). Thus, in this study, we compare species in two groups according to their regeneration properties after fire: seeders and non-seeders. As a result, we hypothesize that both a species position in the leaf economics spectrum and its PFR type will influence leaf trade-offs and eventually the functional properties of ecosystems subjected to frequent-fire regimes.

Methods

study species and study area

The study was carried out on 28 evergreen woody plant species (Table 1) growing in the study region. This subset of plant species was selected to belong to as many different families and regenerative strategies as possible in this type of community (16 seeders and 12 non-seeders). Species were classified into two regenerative groups depending on their PFR strategies: seeders (R–S+ and R+S+sensu Pausas et al. (2004) and (Pausas & Verdú 2005)) and non-seeders (R+S–). Whereas the ability to resprout after loss of above-ground biomass is a widely distributed trait among Angiosperms and habitats, thus representing an ancestral attribute, most of the seeder species would have appeared more recently (during the Quaternary) linked to the Mediterranean climate conditions, characterized by a seasonal drought, when wildfires are likely to occur, (Herrera 1992; Verdú et al. 2003; Pausas & Verdú 2005). Therefore, we preferred this comparison instead resprouters vs. non-resprouters.

Table 1.  Study species, family, post-fire regenerative type (PFR) and life form (according to Raunkiaer classification). Chamaephyte = C, Nano-phanerophyte = NP, Macro-phanerophyte = MP, Phanerophyte-Vine = PV
Study SpeciesFamilyPFRLife form
Argyrolobium zanonii (Turra) P. W. BallLeguminosaeSeederC
Cistus albidus L.CistaceaeSeederNP
Cistus monspeliensis L.CistaceaeSeederNP
Cistus salviifolius L.CistaceaeSeederNP
Coronilla minima L.LeguminosaeNon-seederC
Daphne gnidium L.ThymelaeaceaeNon-seederNP
Dorycnium hirsutum (L.) ser. In DC.LeguminosaeSeederNP
Dorycnium pentaphyllum Scop.LeguminosaeSeederNP
Erica arborea L.EricaceaeNon-seederMP
Fumana ericoides (Caav.) Gandg.CistaceaeSeederC
Fumana thymifolia (L.) Spach.CistaceaeSeederC
Globularia alypum L.GlobulariaceaeSeederNP
Helianthemum appeninum (L.) MillerCistaceaeSeederC
Lavandula latifolia Med.LabiataeSeederC
Lonicera implexa AitonCaprifoliaceaeNon-seederPV
Olea europaea L.OleaceaeNon-seederMP
Osyris alba L.SantalaceaeNon-seederNP
Phillyrea angustifolia L.OleaceaeNon-seederNP
Pinus halepensis Mill.PinaceaeSeederMP
Pistacia lentiscus L.AnacardiaceaeNon-seederMP
Quercus coccifera L.FagaceaeNon-seederNP
Quercus ilex L.FagaceaeNon-seederMP
Rhamnus alaternus L.RhamnaceaeNon-seederMP
Rosmarinus officinalis L.LabiataeSeederNP
Smilax aspera L.LiliaceaeNon-seederPV
Staehelina dubia L.CompositaeSeederC
Teucrium polium L.LabiataeSeederC
Thymus vulgaris L.LabiataeSeederC

To characterize the species, we used both published information (Cucó 1987; Papió 1988; Lloret & Vilà 1997; Verdú 2000; Alberdi & Cavero 2003; Lloret & Vilà 2003), and direct field observations from a recently burned area in Montgrí. Species that present both strategies, seeding and resprouting, were included in the group of seeders, because in these species the recruitment of new individuals is enhanced by fire. Moreover, previous studies have showed that they have a more similar water-use strategy to seeders than to the species that can only resprout after fire (Saura-Mas & Lloret 2007).

All individuals were sampled in the same study area, sited on the Massís del Montgrí, a Mediterranean protected coastal area located in the NE of Catalonia, (north-east Iberian Peninsula, 42°16′N, 3°24′W). Vegetation grows on limestone and is mainly dominated by open pine forests and by Mediterranean shrublands with a prevalence of Quercus coccifera, Cistus albidus, Cistus monspeliensis, and Rosmarinus officinalis (Polo & Masip 1987). Sampling was conducted in mature shrubland vegetation (1–2 m high shrubs) that had been untouched by wildfire for over 10 years. The mean annual precipitation is c. 654 mm, with cool winters (mean minimum annual temperature: 4·1 °C) and warm summers (mean maximum annual temperature: c. 27 °C) (Ninyerola, Pons & Roure 2000; Ninyerola et al. 2003).

leaf traits measurements

We studied SLA (mkg−1), LDMC (mg g−1), leaf nitrogen (LNC, mg g−1), phosphorus (LPC, mg g−1) and carbon (LCC, mg g−1) contents, as well as seasonal variations of relative leaf water content (RWCrsv), and of live fine fuel moisture (LFFMrsv), that is, of branches < 0·6 cm diameter and with foliage. Mean measures of each variable were obtained from 10 individuals of each species from the same area. However, measures of all variables could not be obtained from the same individuals, because the removal of leaves and shoots would influence the estimation of other variables such as leaf nutrient contents.

To obtain SLA, fully developed and water saturated leaves (to achieve water saturation they were stored 8 h at 4 °C in dark conditions (Garnier et al. 2001b)) were digitized with a flatbed scanner and leaf area was measured with image processing software. After that, samples were oven-dried at 60 °C for at least 48 h, and weighed to obtain the dry weight of the leaf.

LDMC is the proportion of the leaf matter, related to the mass of the leaf with the maximum water content (Garnier et al. 2001b). LDMC was measured along the four seasons, and a mean across the four seasons was obtained for each species.

LNC, LPC and LCC are the proportions of nitrogen, phosphorus and carbon mass relative to the dry weight of the leaf. Measures were obtained from fully developed leaves of each species that were sampled during February, when the studied species are likely to remain at a very similar phenological state (Floret et al. 1989; Milla et al. 2006). LPC analyses were measured using ICP-OES (Optical Emission Spectroscopy with Inductively Coupled Plasma) in a Perkin Elmer, Optima 4300 (Shelton, MD) while LNC and LCC were measured with an elemental analyzer NA 2100 (Thermofisher Sicentific, Milano, Italia).

Relative Seasonal Variation of Relative Leaf Water Content (RWCrsv) is defined as the proportional change in the RWC of leaves between the winter and summer: RWCrsv = ((RWCwinter – RWCsummer)/(RWCwinter)), where RWC is the proportion of the leaf water content related to the maximum water content that can potentially be achieved by the leaf (RWC = 100 × ((Mf – Md)/(Mt – Md)), where Mf is the fresh mass, Mt is the turgid mass after re-hydrating the leaves, and Md is the dry mass after drying the leaves in an oven, 70 °C at least 48 h).

Relative Seasonal Variation of Life Fine Fuel Moisture (LFFMrsv) is defined as: LFFMrsv = ((LFFMwinter – LFFMsummer)/(LFFMwinter)), where LFFM is the water content of shoots (< 6 mm of diameter) under field conditions in relation to its dry mass (LFFM = 100 × ((Mf – Md)/Md)).

Some data of relative seasonal variations of RWC (RWCrsv), Live Fine Fuel Moisture (LFFMrsv) and LDMC have been previously published in Saura-Mas & Lloret (2007).

statistical analyses

To assure homogeneity of variance and linearity, LDMC was transformed to its log-odds ratio (log (LDMC – (1/LDMC)), and LCC, LNC and LPC were transformed to their natural logarithms. The hypothesized causal relationships were tested using structural equations modelling (Shipley 2000) using the EQS 6·1 statistics program (Multivariate Software Inc., http://www.mvsoft.com). Fit of the data to the models was judged by the overall null probability of the model, by the probabilities of each path coefficient being zero, and by the likelihood ratios (i.e. the natural logarithm of the ratio of the null probabilities) of each model relative to our most likely model; Royall (1997) suggests that a likelihood ratio greater than 3 gives better support to the alternative model, a likelihood ratio less than 1/3 gives better support to the chosen model while values in the interval (3, 1/3) indicate no preference. All probability values of the obtained structural equations were obtained with MCX2, a program to obtain probability estimates for the Maximum Likelihood χ2-statistic based on small sample sizes (Shipley 2000).

Based on the pattern of correlation between the measured variables, we proposed a series of alternative causal hypotheses and translated these into structural equation models. Our first model involves only the five variables directly related to the leaf economics spectrum (SLA, LPC, LNC, LCC and LDMC) and it is tested simply to determine if our data follow the general scheme proposed by the model of Shipley et al. (2006) (Fig. 1), although we did not include leaf longevity and maximum photosynthetic rate because these data were not available for our populations. Our second model assumes that the two measured variables related to seasonal variation in water content are simply additional manifestations of the same basic model (see Fig. S1 in Supporting information). To develop our alternative models (models 3–8) we started with two constant submodels. First, we specified that the correlations between LCC, LNC, LPC, SLA and LDMC are due solely to a common latent cause, following Shipley et al. (2006); we call this the ‘leaf economics’ submodel. Second, we allow RWCrsv and LFFMrsv to freely covary since there are no obvious reasons to hypothesize that either seasonal changes in the water content of leaves or in fine woody tissues would cause each other; we call this the ‘seasonal water variation’ submodel. Based on these two submodels we proposed a series of alternate hypotheses concerning how the two submodels are linked with respect to the PFR types (Fig. 2 and see also Figs S2–S4 in Supporting information).

Figure 1.

Model 1 (see Methods) showing the relationship between leaf traits, along with the path coefficients. Standard errors are in parenthesis; * indicates that it is a significant path coefficient. Values of r2 are also in the model. L1 and L2 are the latent variables. L1 is an unmeasured variable representing the ratio of cytoplasmic volume to cell wall volume. Fit of the data to the model is shown in Table 2.

Figure 2.

Several models relating leaf traits and post-fire regeneration type (see Methods). Figure 2a presents the model 5 (the best fitting one), 2b model 6, 2c model 8 and 2d presents model 9 relating. The models are presented along with their path coefficients. Standard errors are in parenthesis; * indicates that it is a significant path coefficient. Values of r2 are also in the model. L1 and L2 are the latent variables. L1 is an unmeasured variable representing the ratio of cytoplasmic volume to cell wall volume. L2 is the PFR (post-fire regenerative type). Fit of the data to the models is shown in Table 2.

Model 3 hypothesizes that the leaf economics and seasonal water variation submodels are independent of each other and unrelated to the PFR type. Model 4 hypothesizes that species having a large potential proportion of their leaf fresh mass being water (i.e. a low LDMC) causes them to express a larger variation in seasonal water content, again with the PFR type being independent of both submodels. Model 5 hypothesizes that LDMC and RWC are correlated due to a common selection with respect to PFR types but that this common selection is not directed to the underlying morphological driver of the leaf economics spectrum itself. Model 6 is similar to model 5 except that, in addition to the common selection due to the PFR types, we allow a second causal path linking the variation in water content directly to the potential for high leaf water content. Model 7, in contrast to model 5, assumes that the common selection due to the PFR types acts directly on the underlying morphological driver of the leaf economics spectrum, thus implying only a spurious correlation between LDMC and RWC. Model 8 is similar to model 7 except that selection for the PFR acts both on the underlying morphological driver and on LDMC. Finally, model 9 posits that each of LDMC, RWC and the PFR type are common causes of some unmeasured cause (L2) (Fig. 2d).

Results

Using data for all study species and variables (see Table S1 in Supporting information), structural equations models defining different possible causal relationships were obtained.

Model 1 was not rejected (Table 2), and it is consistent with the original model of Shipley et al. (2006); the overall fit was good and each path coefficient was significantly different from zero (Figure 1). When RWC and LFFM were added as additional indicators of the same underlying latent variable (Model 2), the path coefficients linking the latent variable to each of these two variables were weak and clearly non-significant (Fig. S1), meaning that RWC and LFFM are not simply additional variables responding to the same underlying structure (Table 2). We therefore rejected model 2. Models 3, 4 and 7 were unambiguously rejected at the 5% significance level (Table 2). Thus, the two submodels are not independent (Fig. S2), any dependence between the two submodels must involve the PFR type (Fig. S3) and the effect of the PFR type does not act directly on the underlying driver of the leaf economics spectrum, this is the unmeasured (latent) variable (Fig. S4).

Table 2.  Summary of the fit statistics of the nine models (see Figs 1, 2 and see also S1, S2, S3, S4 in Supporting information). Likelihood ratio was calculated as the P-value of each model divided by the P-value of the best fitting model (model 5, Fig. 2a) and measures the strength of the evidence favouring each model relative to model 5
Modelχ2dfPLikelihood ratio relative to model 5
12·90250·742 ± 0·0192·040
220·807140·147 ± 0·0160·403
346·332220·007 ± 0·0040·018
440·889210·017 ± 0·0060·047
523·647200·364 ± 0·0211
621·816190·407 ± 0·0211·120
735·585200·045 ± 0·0090·124
823·645190·322 ± 0·0200·884
923·512190·322 ± 0·0200·884

Models 5, 6, 8 and 9 were not rejected based on the overall model fit (Table 2, Fig. 2). However, the LDMC→RWC path in model 6 (Fig. 2b) that differentiates it from model 5 (Fig. 2a) is not significantly different from zero. Similarly, the PFR→L path in model 8 (Fig. 2c) that differentiates it from model 5 is not significantly different from zero. Thus, model 5 is the most parsimonious model relative to 6 and 8. Finally, model 9 (Fig. 2d), while providing an almost equivalent level of fit relative to model 5 (Table 2) and with all path coefficients being significantly different from zero, has no significant residual variance associated with PFR (residual variance = 0·033, SE = 0·071, Z-value = 0·468, P = 0·6736), implying that the second latent variable (L2) is essentially the same as PFR, thus leading again back to model 5 (Fig. 2a).

Discussion

Our results support the importance of the leaf economics spectrum in explaining leaf trade-offs, but additionally suggest an important role for other factors, which vary at a local scale. The most parsimonious fitting causal model (model 5) shows that water relative seasonal variation (RWCrsv) is explained by the PFR type which, at the same time, also causes an effect to LDMC. The influence of PFR strategy in some variables of the leaf economics spectrum therefore suggests that local factors determining plant properties may influence the patterns associated with this spectrum. Specifically, we have shown that seeder species have lower LDMC and higher RWCrsv than non-seeders. Model 5 states that the PFR type does not affect the latent driver variable (L1), which we interpret as the relation between cell volume and cell wall volume. However, covariation between LNC, LCC, LPC and SLA is independent of the post-fire regenerative type of the species, while LDMC would be due both to L1 and to some other cause associated with being a seeder or non-seeder (PFR).

The main reason for the patterns that we have found is probably local adaptation to climate. Species in our study region with the seeder trait have evolved under the Mediterranean climate established in the Quaternary, while the resprouter attribute exhibited by many non-seeders is an ancestral trait found in many ancient lineages and widely distributed in many ecosystems (Lloret et al. 1999; Verdú 2000; Bond & Midgley 2003; Bond & Keeley 2005). In fact, it is considered that non-seeder species evolved in the Tertiary before the establishment of the Pre-Mediterranean climate (Herrera 1992; Verdú 2000; Pausas & Verdú 2005). The seeders’ tolerance to higher water content seasonality (RWCrsv) and its lower LDMC contents may be the result of the adaptation to the Mediterranean climate which is characterized by seasonal drought periods and high between-years climatic variability. Leaves of most studied seeders are short-lived and they are renewed at the end of the summer (Floret et al. 1989; Villar & Merino 2001; Navas et al. 2003). According to the global spectrum of leaf economics, we do not expect that species with short-lived, thin leaves would have low mass-based net photosynthesis, because a low instantaneous productivity plus a short leaf duration would result in a low total productivity over the leaf life span, making it unlikely for a leaf to maintain the costs of the leaves’ economics (Reich et al. 1997; Ackerly 2004). Since seeders have short-lived thin leaves, we expect that seeders will have a high mass-based net photosynthesis rate. In fact, there is some evidence of this pattern in some species. Varone & Gratani (2007) studied eight Mediterranean maquis species and the seeder species Cistus incanus presented higher net photosynthesis rate than resprouters such as Arbutus unedo or Phyllirea latifolia. Oliveira & Peñuelas (2004) also showed evidence that the seeders Cistus albidus presents higher values of net photosynthesis rate than the resprouter Quercus ilex. Also, short leaf life span would also result on low LDMC (Haj Khaled et al. 2006).

Frequent disturbances, such as fire, and high between-years climatic variability are likely to have acted as selective pressures in favoring shorter longevity of the new species arising in the Mediterranean conditions. Accordingly, seeders often show a short life cycle with high recruitment after disturbances such as fire (Lloret & Vilà 1997; Lloret, Pausas & Vilà 2003; Pausas et al. 2004), but also after episodes of intense drought (Lloret, pers. obs.). The ability to deal with high water relative seasonal variation is likely linked to drought tolerance and less strict stomata control (Choi et al. 2005), contrary to that experienced by many non-seeder, resprouters. Thus, seeders would have higher photosynthetic activity associated to lower stomatic control in drought periods. As a result, this high mass-based net photosynthesis plus the short leaf life span and the short life cycle could be some of the intrinsic factors of being seeder that explain the low values of LDMC and the high RWCrsv for seeder species.

As expected from the model of Shipley et al. (2006), LNC, LCC, LPC, SLA, LDMC are all apparently caused by the same latent variable, (L1). The fact that we did not have information on two of the measured variables included in the model of Shipley et al. (2006) (leaf longevity and maximum photosynthetic rate), does not affect the causal relationships between our variables and the latent. Following these authors, we tentatively interpret this latent variable as the ratio between the volume of the leaf occupied by water (Vc) to the volume of the leaf occupied by the cell wall (Vw), although this physical interpretation has not yet been experimentally verified. More important to the present paper is to understand how variables related to the seasonal variation in water content of leaves and fine branches relates to the leaf economics spectrum. Variables related to the seasonal relative variation of water are very important to the functioning of Mediterranean plants (Larcher 1995; Terradas 2001), and they also seem to be narrowly related to perturbations such as wildfires. Thus, is this seasonal variation simply another manifestation of the leaf economics spectrum, is it independent of the leaf economics spectrum, or is it related through another variable?

Our analysis suggests that the seasonal relative variation in tissue water is not simply a further manifestation of the leaf economics spectrum. Although the small sample size of this study certainly reduces the statistical power to reject incorrect models, the models that we did reject are not affected by this. However, the small sample size means that we must be cautious in our acceptance of our preferred model. Following the causal explanation based on the current interpretation of the latent variable (i.e. the ratio of cell cytoplasmic volume to cell wall volume), it could be expected that higher values of this latent variable would cause lower LDMC, as it is supported by our results. In turn, Canny & Huang (2006) have shown that palisade cells of Eucalyptus pauciflora submitted to dehydration experience a reduction and deformation that implies shrinkage of the cell. This distortion implies marked changes in the elastic properties of the cell walls, which are inversely correlated to their thickness (Marshall & Dumbroff 1999). Given this, a higher ratio of cell cytoplasmic volume to cell wall volume would result in higher elasticity, which would imply higher seasonal variation in RWC. However, our analysis suggests that the variables related to seasonal change in water content were not directly caused by the latent variable (L1, Vc/Vw). Other studies have reported a reduction in cell-volumetric elasticity during tests of drought response (Fan, Blake & Blumwald 1994), suggesting that at present we only have a partial knowledge of the cellular mechanisms involved in leaf water content variation. Of course, these considerations do not question the significance of evapotranspiration control of water loss during drought periods, for example by stomata control.

In conclusion, this study suggests that the leaves of Mediterranean woody species follow the general leaf economics spectrum (Wright et al. 2004) but that specific selective forces acting at regional scale also play a relevant role to explain leaf traits related to water content. In our case, PFR type influences one variable, LDMC, which is strongly linked to the general leaf economic spectrum, but also to water seasonal variation (RWCrsv). We suggest that since seeders species evolved under the Mediterranean climate, they developed a particular strategy of drought tolerance (without causing an effect to Vc/Vw) and frequent disturbance which would be the ultimate cause of the observed relationship between LDMC and RWCrsv in Mediterranean woody species.

Acknowledgements

The Authors thank everyone who helped in the field work. This study was funded by the Department of Universities, Research and Information Society of the Generalitat de Catalunya, the European social funds, and the Spanish MCYT projects REN 2003-07198, and CGL2006-01293.

Ancillary