Variation in LLS and leaf CC
LLS varied between 56 and 162 d in herbaceous species, and between 80 and 802 d in woody species (Fig. 3a, Table 1). Our sample did not include evergreen trees with very long-lived leaves (e.g. Pinus), which explains why the range for woody species is substantially narrower than that observed by Diemer (1998a) reviewing data on Mediterranean shrubs and trees. For herbaceous species, our data seem to be the first available to date in a Mediterranean environment. In hemicryptophytes, LLS was 109 d on average (Table 2), which is longer than for species from zones with a shorter growth period than Mediterranean France (56 d to 82 d: Sydes, 1984; Diemer et al., 1992; Craine et al., 1999; Ryser & Urbas, 2000) and shorter than for species from a-seasonal Andes (193 d: Diemer, 1998b) where growth occurs all year long. Overall, these results support the hypothesis of a positive association between LLS and length of growing season, previously formulated for regions where growth is seasonally constrained by temperature (Prock & Körner, 1996; Diemer, 1998b; Ryser & Urbas, 2000). Our data extend it to the Mediterranean climate, where growth is constrained by both temperature and water stress. This absence of differences between climates is probably the consequence of the small effect of drought on LLS: a similar range of LLS was found by Reich et al. (1999) for American sites differing in dryness, whereas 10% lower in LLS was documented for Australian species from arid places when compared with species from wetter places (Wright et al., 2002).
As recorded under other climatic conditions, LLS increased along the successional gradient (Fig. 3a). A number of potential environmental constraints on plants have been identified in successional habitats (Grime, 1979; Tilman, 1990). The following discussion will focus only on those likely to affect the traits measured. First, secondary succession may be considered to be characterized by a double gradient of decreasing disturbance frequency and intensity, and of increasing light competition resulting from the accumulation of standing biomass. To these gradients corresponds a gradient of life forms (Escarréet al., 1983; Inouye et al., 1987; Prach et al., 1997), from low to high stature plants. The increase in LLS along the succession actually corresponds to this gradient, with LLS increasing in the order: therophytes < hemicryptophytes < chamaephytes = deciduous phanerophytes < evergreen phanerophytes, the largest difference been found between the two classes of phanerophytes, even though evergreen trees with very long-lived leaves were not included in species set (Table 2, see also Kikuzawa & Ackerly, 1999). To account for this, Kikuzawa & Ackerly (1999) recently proposed a cost-benefit model in which the optimum LLS of a species depends not only on the construction costs of the leaf itself, but also on those structures (shoots in particular) supporting the leaf. A prediction of this model is thus that LLS should increase with plant height or mass. This was indeed found for the species studied here, whose height varied from 0.02 to 0.05 m (therophytes of the first stage) to more than 7 m (some phanerophytes of the latest stage) (E. Garnier et al., unpublished data): Log(LLS, d) = 0.28 * Log(plant height, m) + 5.27 (r2 = 0.50, P < 0.001). Changes in LLS along disturbance (Ryser & Urbas, 2000) or successional gradients (Shukla & Ramakrishnan, 1984; Hegarty, 1990; Reich et al., 1995; Bazzaz, 1996) have also been recorded within particular life forms, but to what extent this also relates to changes in plant stature or mass cannot be assessed with the data provided in these studies.
Second, secondary succession may involve changes in availability of soil resources either in absolute amounts or relative to light availability (discussed in, e.g. Tilman, 1990; Bazzaz, 1996). As LLS has a strong impact on nutrient conservation within the plant (Aerts, 1995; Garnier & Aronson, 1998; Eckstein et al., 1999), the increase in LLS with successional status of the species could be interpreted as a response to a decrease in nutrient availability as succession proceeds. Such a decrease could be due to the progressive development of an imbalance between nutrient supply by the soil and nutrient demand by the vegetation as a large standing biomass develops during the course of succession. Whether this actually holds in our particular situation remains to be established.
Tilman (1990) has further postulated that several of the underlying environmental constraints of successional habitats (including those discussed above) should lead to trade-offs in maximal growth rates of the species. High rates of resource capture and relative growth are indeed prominent plant traits characterizing the ‘ruderal’ syndrome associated with disturbed environments (Grime, 1979) and early stages of succession (Gleeson & Tilman, 1994). These high rates are assumed to be adaptive in such an environment, where rapid growth is both possible because of low competitive interactions with an established vegetation, and necessary to produce rapidly numerous seeds leading to a high colonization potential. These rates, together with a number of correlated traits, usually decline along successional gradients (Bazzaz, 1979; Gleeson & Tilman, 1994; Reich et al., 1995; Bazzaz, 1996) as the frequency of disturbance decreases and/or the absolute or relative levels of resources changes. This is the case for specific leaf area (SLA, the ratio of leaf area to leaf mass) in particular (Reich et al., 1995), a trait negatively related to LLS (Reich et al., 1997; Diemer, 1998b; Wright & Westoby, 2002). Such a decrease in SLA has been found for the species studied here (data from Garnier et al., 2001), and a negative association between SLA and LLS was observed as well (Log(SLA) =−1.19 * Log(LLS) + 3.51, r2 = 0.46, P < 0.001). The increase in LLS could thus be the direct consequence of the decrease in the growth potential of species along the succession.
Leaf CC were in the range found for many species (e.g. Poorter & Villar, 1997; Baruch & Goldstein, 1999; Villar & Merino, 2001), and were slightly lower in herbaceous than in woody species (−8%: Table 2), as found previously (Poorter & Villar, 1997; Baruch & Goldstein, 1999). This was confirmed comparing herbs and woody species within a family. Contrary to Villar & Merino (2001), we did not find higher leaf CC in evergreen than in deciduous species, maybe because evergreens in our sample had relatively short LLS. Differences among species were so small that changes in leaf CC along the succession were not significant (Fig. 3b). To our knowledge, the only other study where leaf CC of species differing in successional status have been assessed is that conducted on seven Piper species by Williams et al. (1989) in a tropical forest. These authors found higher leaf CC (and shorter LLS) in early successional species compared with late successional ones, but these differences were related to the 100-fold difference in the mean irradiance experienced by the seven species. These results are thus not comparable with ours, obtained on well-lit leaves for all species.
Leaf CC were found to be only weakly related lo LLS (Fig. 5a), which confirms the conclusions put forward by Sobrado (1991) and Villar & Merino (2001) for woody species. As already found by Williams et al. (1989) and Sobrado (1991) for tropical trees, leaf longevity was positively related to leaf payback time, calculated here as leaf CC : Amax (Fig. 5b). However, the relationship became hardly significant when evergreens were removed from the analysis. This weak relationship may indicate that the variation in Amax does not reflect the actual variation of daily carbon gain in the wide spectrum of species studied here as it does for tropical trees. Kikuzawa (1991) proposed that optimal LLS depends on the (b * leaf CC : Amax) ratio where b is the leaf age when photosynthesis becomes zero. This ratio could not be calculated here because b was not available for our species. Following this author, we may suggest that LLS of all species but evergreens is only controlled by b because leaf CC : Amax did not significantly differ among those species (Fig. 5b) and the decline in leaf photosynthesis with age has been shown to be faster in species with short LLS (Kitajima et al., 1997). In evergreens, both parameters could well be of importance.
Whatever the actual form and variables involved in the relationship between LLS and leaf CC, our data further confirm that the impact of differences in leaf CC among species on LLS is likely to be relatively low (Kikuzawa & Ackerly, 1999; Villar & Merino, 2001).
Influence of leaf dynamics on LLS
Mediterranean flora is characterized by a large variability in leaf phenology among life forms (Floret et al., 1989, M.-L. Navas, unpub. data), explaining why it is an appropriate model to test for relationships between leaf dynamics and leaf functioning. The only consistent pattern of variation in leaf dynamics parameters among species differing in successional status was found for the period of leaf loss (Fig. 4b): tL increased regularly, but not significantly, with successional status, in parallel with LLS (compare Figs 4b and 3a). The period of leaf production and the time lag between the end of leaf production and beginning of leaf loss show an abrupt and significant change for species of the latest stage (Fig. 4a,b). Species prominent in the first four stages display a large variability in phenology: herbs have a long period of leaf production, overlapping with a period of leaf loss, whereas most chamaephytes produce leaves in the autumn and/or spring and loose them next spring or summer. By contrast, most phanerophytes prominent in the oldest stage have a flushy spring production of leaves, separated from the period of leaf loss occurring the next autumn for deciduous species and more than 1 yr after production for evergreens. Therefore, two groups of species can be recognized along the successional gradient; these groups been also found within Rosaceae. Restricting the comparison to stages 4 and 5 where most woody species occur, this result is in line with studies showing that trees occurring early in successional environments exhibit continuous leafing and produce short-lived leaves, while trees forming mature forests tend to produce long-lived leaves in flushes (Kikuzawa, 1983; Lechowicz, 1984; Kikuzawa, 1988; Koike, 1988). These differences in leaf dynamics and structure have been related to changes in resource availability during succession (Kikuzawa, 1983, 1988), but whether this applies to our particular situation is unknown (see previous section).
LLS depended on the period of leaf loss and the time lag between the end of leaf production and the beginning of leaf loss (Fig. 5) The lack of relationship between LLS and leaf production period, already found by Kikuzawa (1988) was probably due to the negative link between t and tp. Using our framework, there were only two species, starting to lose leaves as soon as they were produced, for which the quality of fit between measured and estimated LLS was low (Fig. 7). In that case, a delay in the recording of the beginning or end of periods of leaf production or loss, due to the observation of cohorts instead of individual leaves, could have resulted in false estimations of tp and tL. These results demonstrate that our framework can be used for estimating LLS of a large range of species.
The quality of prediction given by our framework has to be compared with results from models previously proposed. LLS has previously been assumed to depend on leaf loss in species forming a single cohort (Southwood et al., 1986) or in species where the rate of leaf loss is similar to the rate of leaf production, resulting in a constant number of leaves per plant or branch (e.g. Jow et al., 1980; Ackerly, 1999; Duru & Ducrocq, 2000). In these cases, LLS can be properly estimated by dividing the number of leaves by the birth (or mortality) rate (Southwood et al., 1986; Ackerly, 1999). In terms of the model presented here, this means that LLS can be approximated by the sum of tp and t (see Appendix 2). To test whether this two-parameter model holds for the species studied here, we compared the LLS observed with that estimated by this sum. The relationship is significant when all species are included or for woody species, but not for herbaceous species or when evergreens are excluded from the analysis (Table 3). In another approach, LLS was assumed to depend both on birth and death rates of leaves, with a further hypothesis that these were also constant, but not necessary equal (Jow et al., 1980; Williams et al., 1989; Craine et al., 1999). Williams et al. (1989) calculated LLS as the difference between the number of leaves recorded at a date ti after the beginning of the experiment divided by the death rate, and the number of leaves that appeared until ti divided by the birth rate. Using our model, this means that LLS is equal to the difference between tL and tp (cf. Appendix 2). Table 3 shows that the LLS of our species are significantly predicted with this second model when all species are considered. The quality of the prediction decreases when evergreens are excluded, while accurate estimates of LLS are found for both herbaceous and woody species. The model proposed here (Eqn. 1) gave the best estimates of LLS for all sets of species (Table 3, Fig. 7). Our data also show that when species displaying a large range of phenological patterns are compared, the period of leaf loss and the time lag parameter to a lesser extent, have a stronger impact on LLS than the period of leaf production.
Table 3. r2 of relationships between observed leaf life span (LLS) and estimates provided by different models proposed by Southwood et al. (1986) (model 1), Williams et al. (1989) (model 2) and the present study (eqn. 1) (model 3). Derivations of LLS from models are (I) Model 1: LLS = tp + t; (2) Model 2: LLS = tL– tp; (3) Model 3: LLS = (1/2) * (t p + tL) – t where tp, tL and t are the periods of leaf production and loss, and time lag between the end of leaf production and the beginning of leaf loss. See Appendices 1 and 2 for more details on calculation
|Species||n||Model 1|| ||Model 2|| ||Model 3|| |
|All species but evergreens||40||0.06||Ns||0.49||***||0.77||***|
|Evergreen phanerophytes|| 5||0.13||Ns||0.00||ns||0.17||ns|