Comparison of leaf construction costs in woody species with differing leaf life-spans in contrasting ecosystems


  • Rafael Villar,

    Corresponding author
    1. Departamento de Biología Vegetal y Ecología, Universidad de Sevilla, Apdo 1095, 41080 Sevilla, Spain;
    2. present address: Area de Ecología, Universidad de Córdoba, Colonia San José n°3, 14071 Córdoba, Spain;
    • Author for correspondence: Rafael Villar Tel: +34 957 21 86 35 Fax: +34 957 21 82 33

    Search for more papers by this author
  • José Merino

    1. Departamento de Biología Vegetal y Ecología, Universidad de Sevilla, Apdo 1095, 41080 Sevilla, Spain;
    2. present address: Departamento de Ciencias Ambientales, Universidad Pablo Olavide, Carretera de Utrera Km 1, 41013 Sevilla, Spain
    Search for more papers by this author



  •  The construction costs (CC) are reported of leaves from 162 wild woody species from 14 contrasting environments (desert to rain forest) and with different leaf life-spans.
  •  Calorimetric methods were used to estimate the CC of deciduous, semideciduous and evergreen leaves.
  •  Leaf CC showed a wide range (78%) between species, and deciduous species showed a slightly lower CC (6%) than both semideciduous and evergreen species. Mean leaf CC differed between ecosystems, with the highest and lowest CC in the tundra and rain forest, respectively. Leaf CC was positively correlated with lipid concentration. Leaf size (log) and specific leaf area (SLA, leaf area per leaf dry mass) were negatively correlated with leaf CC. Leaf CC did not show differences between different leaf life-spans or ecosystems when leaf size (log) or SLA were included as covariates.
  •  The small differences in leaf CC among leaf life-span types and ecosystems (6% and 23%, respectively) suggest that SLA is more important in determining differences in the carbon balance between species than leaf CC. Leaf size is shown to be an important trait associated with other leaf characteristics.


A, ash concentration; CC, construction cost per unit dry mass; CCA, construction cost per unit area; Eg, growth efficiency; Hc, ash free heat of combustion; N, nitrogen; SLA, specific leaf area.


Species with different leaf life-spans differ in many traits related to carbon flux, such as photosynthetic and respiration rates (Mooney & Gulmon, 1979; Field & Mooney, 1986; Reich et al., 1992; Larcher, 1995; Villar et al., 1995; Reich et al., 1997). Fast growing species from favourable habitats show short leaf life-span and have high photosynthetic and respiration rates per unit mass (Poorter et al., 1990; Reich et al., 1992; Atkin et al., 1996; Reich et al., 1997). These differences could help explain the advantages of the different leaf life-spans in different habitats, and thus the distribution of species with different leaf longevity. However, to fully elucidate the advantages in terms of carbon balance of different leaf life-spans in different habitats, we also need to know the energetic costs of tissue construction (i.e. the construction cost, CC). The construction cost includes the glucose required to build carbon skeletons, and the glucose consumed in respiration to supply reductant and ATP for energy-requiring processes in the biosynthesis of the tissue constituents (Penning de Vries et al., 1974; Williams et al., 1987).

Although differences in leaf CC among species with different leaf life-span (evergreen and deciduous) have been studied for about three decades, no clear picture has emerged. Early studies suggested that evergreen leaves have higher CC than deciduous ones (Johnson & Tieszen, 1976; Orians & Solbrig, 1977; Miller & Stoner, 1979), since the former are richer in defensive compounds (such as lignin and antiherbivore compounds), which are expensive to synthesize (Mooney & Gulmon, 1979; Chabot & Hicks, 1982). Subsequent studies have supported this hypothesis (Merino, 1987; Diamantoglou et al., 1989; Gower et al., 1989; Damesin et al., 1998) but others have not (Merino et al., 1982; Chapin, 1989; Williams et al., 1989).

Several factors could explain the contradictory measures of CC reported in the scientific literature. Firstly, relative estimates of CC will depend on the units with which they are expressed. For example, Sobrado (1991) found that evergreen leaves had higher CC per unit area (g glucose m−2) than deciduous ones. However, recalculation of that data per unit dry mass (g glucose g−1) showed that there were no significant differences in CC between these leaf types. Secondly, the size of the leaf seems to affect leaf CC. For example, Merino (1987) found higher leaf CC in evergreen than in deciduous leaves, but he did not find any difference between these leaf types when species with similar leaf size were compared. Thirdly, comparisons were often made between only two or three species. Moreover, in some cases, the species compared were native to different ecosystem types (tropical, mediterranean, arctic, etc.). Since, differences in resource availability, such as light, nitrogen and phosphorus, appear to affect CC (Griffin, 1994; Poorter, 1994; Griffin et al., 1996; Poorter & Villar, 1997), we may expect ecosystem type to influence CC. Finally, because of the diversity of methods used to calculate CC, a comparison of the available data from different literature sources is unlikely to be a good method to test the hypothesis of whether CC differ among contrasting species and/or between contrasting environments. For example, comparisons of CC values obtained by different methods for the same plant material may differ by as much as 20% (Williams et al., 1987; Griffin, 1994).

There is little information available on the differences in CC among species from contrasting ecosystems. A major focus of our study was therefore to assess whether leaf CC differ among species with different leaf life-span, and/or among species from contrasting ecosystems. The aims of this study were to test (1) if species with long leaf life-span have relatively high leaf CC, and (2) if mean leaf CC differs between contrasting ecosystems. The relationships between leaf CC and other leaf traits (size, chemical composition, etc.) were also investigated.

Materials and Methods

Different leaf traits of 162 wild woody species from 14 contrasting ecosystems (Table 1) were studied. Included in this sample, were published data on seven tropical dry forest species in South America (Sobrado, 1991) and 35 species from two rain forests in Africa (Waterman et al., 1980). Both studies followed the same approach as in the present study. The species were classified according to the mean leaf life-span into the following categories: deciduous (4–8 months), semideciduous (5–12 months) and evergreen (> 12 months). Semideciduous species are those that have a leaf longevity of less than 1 yr, but in contrast to deciduous species, keep leaves throughout the year. Data on leaf life-span were taken from literature and from field observations. Nomenclature of the species agreed with the classification given in taxonomic texts for each region (Thomas, 1961; Branwell & Branwell, 1974; Porsild & Cody, 1980; Benson & Darrow, 1981; Moore, 1983; Valdés et al., 1987; Petrides, 1988).

Table 1. Type of the ecosystems studied, latitude and longitude, location and the number of deciduous (Dec), semideciduous (S-Dec) and evergreens species (Ever) considered in each ecosystem. Code of ecosystems used is as in Fig. 2. Data of ecosystem 12 are from Sobrado (1991) and data of ecosystems 13 and 14 from Waterman et al. (1980)
CodeEcosystemLatitude and longitudeLocationDecS-DecEver
 1Tundra75° N 82° WDevon Island, Canada 1 0 2
 2Desert28° N 106° WChihuahua, USA 2 1 3
 3Xeric forest28° N 17° WCanary Islands, Spain 4 4 6
 4Chaparral36° N 122° WCalifornia, USA 1 2 3
 5Xeric mediterranean forest37° N 6° WAndalucía, Spain 011 9
 6Mesic mediterranean forest36° N 122° WCalifornia, USA 6 0 4
 7Mesic mediterranean forest37° N 6° WAndalucía, Spain13 2 9
 8Temperate forest44° N 80° WToronto, Canada 4 0 0
 9Warm temperate forest35° N 80° WNorth Carolina, USA 7 0 1
10Austral forest55° S 70° WTierra del Fuego, Argentina 2 0 5
11Lauriphyll forest28° N 17° WCanary Islands, Spain 1 017
12Tropical dry forest10° N 67° WCharallave, Venezuela 4 0 3
13Rain forest 5° N 10° WDouala-Edea Forest, Cameroon 1 020
14Rain forest 0°, 32° WKibale Forest, Uganda 3 011

The sampling was done following the same protocol during summer of 1990 and 1991. For each species, several individuals were sampled, taking branches found in different positions of each individual. All the leaves present in each branch were sampled, excluding those with injuries. Leaf blade area was determined in one subsample either using an image analyser (Skye Instruments, Ltd.) or by making photocopies of leaves with paper of known specific weight and weighting the leaf images. Leaf samples were oven dried at 80°C until constant weight, ground and homogenized for subsequent analysis. Ash concentration was determined gravimetrically after combustion of the sample for 4 h at 500°C. Total organic nitrogen concentration was determined by Kjeldhal analysis. Protein concentration was estimated by multiplying nitrogen concentration by 6.25 (Merino et al., 1984). Heat of combustion was determined with an adiabatic bomb calorimeter (Phillipson Gentry Instruments, Inc., USA) with correction for ignition wire melting (Phillipson, 1964).

Lipid concentration was determined in leaves of 43 species, most of them native from xeric and mesic mediterranean forests (Spain), austral forest (Argentina), and chaparral scrub (CA, USA). We also include the data on lipid concentration of seven species from tropical dry forest in Venezuela (Sobrado, 1991). Lipid concentration was obtained gravimetrically from soluble diethylether extracts (Allen, 1974).

Leaf CC (g glucose g−1) was calculated using a formula based on the growth efficiency of the leaf tissue, heat of combustion and ash and nitrogen concentration of leaves according to Williams et al. (1987):

inline image

where Hc is the ash free heat of combustion (kJ g−1), A is the ash concentration (g g−1), k is the oxidation state of the nitrogen source (+5 for nitrate or −3 for ammonium), N is the organic nitrogen concentration (g g−1) and Eg is the growth efficiency. The value used in this study for Eg was 0.89 (Williams et al., 1987). In the calculations, we assumed that the nitrogen source was nitrate for all the species, as it is the principal source of nitrogen that is available to higher plants under most field conditions (Taiz & Zeiger, 1991). However, there is a broad consensus that in some ecosystems, for example tundra, the main nitrogen source is ammonia, although tundra species can also use nitrate (Atkin et al., 1993). So, in the case of tundra species, we also consider ammonia as the nitrogen source for calculation of leaf CC. Heat of combustion, ash, and nitrogen and lipid concentration were measured from two different samples obtained from the homogenized leaves for each species. In cases in which variation was higher than 5%, a triplicate sample was considered.

The cost of protein synthesis (g glucose spent in protein synthesis per gram of dry tissue) was calculated by multiplying the protein fraction in the tissue by the specific cost of protein synthesis [2.775 g glucose (g protein)−1] (Poorter, 1994). The percentage of CC dedicated to protein synthesis was calculated as the ratio: (cost of protein synthesis/CC) * 100.

Statistical analysis of data

Statistics were performed using Statistica (StatSoft, 1996) and SPSS (SPSS, 1999). Differences in leaf traits were analysed with a non-parametric test (Kruskal–Wallis) with leaf life-span or ecosystem as class factor. Comparison of leaf traits between leaves with different life-spans (class factor) were made in two ways: (1) pooling all species from different ecosystems and (2) independently in each ecosystem with two or more species belonging to at least two of the three different leaf life-span classes (deciduous, semideciduous or evergreen). Note that most ecosystems studied did not have species belonging to the three leaf types considered, and also that in some ecosystems the majority of species belongs to only one leaf life-span category (Table 1 and Appendix 1).

To detect differences in leaf traits between contrasting ecosystems, a non-parametric test (Kruskal–Wallis) with ecosystem type as class factor was performed. In doing so, the differences in leaf CC between contrasting ecosystems could be affected by the dominant leaf life-span of the species in each ecosystem. Therefore, to check if leaf CC were affected by ecosystem type within each leaf life-span type, we performed a Kruskal–Wallis test (ecosystem type as class factor) on two data sets separately; one for deciduous species and the other for the evergreen ones. In this analysis, only those ecosystems with at least four evergreen species or four deciduous ones were considered. Species with semideciduous leaves were not included in the analysis due to the low numbers of species of this type (i.e. there were only two ecosystems with at least four species).

A general linear model was fitted to leaf CC data with ecosystem, leaf life-span and leaf area (log) or SLA as explanatory variables using maximum likelihood methods. Ecosystem and leaf life-span were introduced as factors (14 ecosystems, two classes of leaf life-span: deciduous and evergreen) and leaf area (log) or SLA as covariates.

Although leaf CC results from the values of three independent variables (heat of combustion, nitrogen and ash concentration) (Eqn 1), the importance of each one of these in explaining the value of leaf CC was unknown. We explored the sensitivity of leaf CC to changes in each component (Hc, N or ash concentration) keeping the other two components constant (similar approach as Griffin et al., 1996). Mean values of Hc, N and ash concentration obtained from our data set were chosen as constant values, and sensitivities of estimates of CC were calculated on the basis of a change in a variable value of plus or minus two times its standard deviation. We calculated the percentage of change in leaf CC that was caused by increasing each one of the independent variables from − 2*S.D. to + 2*S.D., maintaining the other two variables constant.

All means are presented with ± standard deviation.


Leaf CC of the 162 species ranged from 1.08 g glucose g−1 (Chaetacme aristata, rain forest, Uganda) to 1.92 g glucose g−1 (Erica scoparia, mesic mediterranean forest, Andalucía, Spain) (see Appendix 1), with the mean leaf CC for all species being 1.52 ± 0.12 g glucose g−1.

Causes of variation in leaf CC

Both the sensitivity of CC to small changes in variable value and the actual variation in parameter values contribute to the relative importance of each parameter in determining variation in CC. For example, CC was shown to be sensitive to small changes in Hc. However, there was very little variation in this measure between samples (C.V. = 6%), and therefore Hc contributed less than expected to the observed variation in CC. Contrary to this, CC was not very sensitive to changes in ash concentration, but this measure showed considerable variation between samples (C.V. = 50%) and therefore it determined more of the variation in CC than expected on the basis of its sensitivity (Fig. 1a). In any case, the most important parameter was shown to be Hc. The sensitivity analysis of leaf CC showed that increasing Hc from − 2*S.D. to + 2* S.D. caused an increase in leaf CC of 30%. The increase in ash concentration determined a decrease in leaf CC of 15%, whereas the increase in N showed the lowest effect on leaf CC, increasing about 6% (Fig. 1a).

Figure 1.

(a) Sensitivity analysis of leaf construction cost to the increase or decrease in only one component (ash free heat of combustion, Hc; nitrogen, N; or ash concentration) keeping the other two components constant. Mean values of Hc, N and ash concentration of our data set were chosen as constant values and the amount of increase or decrease in the variables to detect its effect on leaf construction cost were ±2*S.D. Relationships between leaf construction cost (g glucose g−1) and (b) ash free heat of combustion (r = +0.92, P < 0.0001), (c) ash concentration (r = −0.62, P < 0.001), and (d) nitrogen concentration of leaves (g g−1) (r = −0.11, P > 0.17).

Leaf CC was positively correlated with Hc (r = +0.92, P < 0.0001; Fig. 1b) and negatively correlated with ash concentration (r = −0.62, P < 0.001; Fig. 1c). However, leaf CC was not correlated with N (r = −0.11, P > 0.17; Fig. 1d).

Since Hc is the main determinant of the differences in leaf CC, it is worth investigating the parameters related to the variation in Hc. The value of Hc is determined by the chemical composition of the tissue (Williams et al., 1987). We found a positive relationship between lipid concentration and both Hc (r = + 0.61, P < 0.0001) and leaf CC (r = + 0.54, P < 0.05, Fig. 2a). This suggests that lipid concentration could be one of the main factors responsible for the observed differences in leaf CC associated with leaf life-span and ecosystem type.

Figure 2.

Relationships between (a) leaf construction cost (g glucose g−1) and lipid concentration (g g−1) (r = +0.54, P < 0.05) and (b) protein concentration (g g−1) and ash concentration (g g−1) (r = +0.38, P < 0.00001).

Proteins, which are one of the most expensive compounds to synthesize were positively correlated to minerals (r = +0.38, P < 0.00001, Fig. 2b), which have a null construction cost. Therefore, leaves with higher protein concentration have relatively high concentrations of minerals, which tends to keep CC values close to average CC.

Leaf CC between different leaf life-span and ecosystem type

Mean leaf CC of deciduous species (1.46 ± 0.12 g glucose g−1) was significantly lower (6%; P < 0.05) than those of semideciduous and evergreen species (1.55 ± 0.10 and 1.55 ± 0.12 g glucose g−1, respectively) (Fig. 3a). There were no differences in leaf CC between semideciduous and evergreen species. Leaves of evergreen and semideciduous species showed a higher heat of combustion, a lower nitrogen and ash concentration and a smaller leaf size than those of deciduous species (P < 0.05, Table 2). Deciduous species also show a significantly higher proportion of the leaf CC dedicated to protein synthesis (26%, P < 0.001) than evergreen and semideciduous species (19.6 and 17.5%, respectively) (Table 2).

Figure 3.

(a) Mean leaf construction cost expressed per unit dry mass (g glucose g−1), and (b) per unit area (g glucose m−2), and (c) specific leaf area (SLA, m2 kg−1) in relation to the life span of the leaves (Dec, deciduous; S-Dec, semideciduous; and Ever, evergreen). Box limits correspond to ± SE and bars to ± SD. Different letters mean a significant difference (P < 0.05).

Table 2. Mean values (± SD) of ash free heat of combustion (Hc), nitrogen and ash concentration, leaf size and the proportion of construction cost dedicated to protein synthesis (CC proteins, [glucose used in protein synthesis/construction cost]*100) in leaves with different life span (deciduous, semideciduous and evergreen) from the 14 ecosystems considered (Table 1). In brackets, number of species considered. For leaf size the number of species considered were 41,19 and 56 for deciduous, semideciduous and evergreen, respectively. Different letters in one column means a significant difference (P < 0.05)
 Hc (cal g−1)Nitrogen (mg g−1)Ash (mg g−1)Leaf size (cm2)CC proteins (%)
Deciduous (n = 49)20.40 ± 1.03a22.04 ± 7.7a84.3 ± 48.9a96.9 ± 373.5a26.2 ± 9.0a
Semideciduous (n = 20)21.32 ± 1.34b15.65 ± 5.2b57.9 ± 18.5b 3.0 ± 3.3b19.6 ± 6.2b
Evergreen (n = 93)21.24 ± 1.41b17.30 ± 7.4b59.0 ± 30.3b16.6 ± 21.08c16.2 ± 6.2b

The analysis of differences in leaf CC between deciduous and evergreen species within each ecosystem shows that only in the case of xeric forest (Canary Islands, Spain) and rain forest (Uganda), were evergreen leaves more costly to construct than deciduous leaves (0.05 < P < 0.10). No differences between CC of evergreen and deciduous leaves were found in the five other ecosystems where evergreen and deciduous were present (Table 1).

We found significant differences (P < 0.0001) in leaf CC between ecosystems (Fig. 4a) that were mirrored by significant differences in Hc (P < 0.001). Leaves of tundra species showed the highest CC (1.72 ± 0.11 g g glucose g−1) whereas the lowest leaf CC correspond to species from rain forest (Uganda) 1.40 ± 0.13 g g glucose g−1. When ammonia was assumed to be the principal nitrogen source in tundra species, the mean leaf CC was still higher than in other ecosystems (1.60 ± 0.16 g g glucose g−1). The maximum difference in mean leaf CC between ecosystems was 0.32 g glucose g−1, corresponding to about a 23% difference (0.32/1.40). However, when excluding the tundra species because of their low representation (only three species harvested), the ecosystems with highest leaf CC were xeric mediterranean forest (Andalucía, Spain) (1.58 ± 0.10 g glucose g−1) and chaparral (1.58 ± 0.06 g glucose g−1), and then the difference in mean leaf CC between ecosystems was much lower (13%), but still significant. We found a near significant correlation between mean leaf CC of each ecosystem and latitude (P = 0.06, r = +0.51), but excluding the tundra species there was no significant correlation (P > 0.70).

Figure 4.

(a) Mean leaf construction cost expressed per unit dry mass (g glucose g−1), and (b) per unit area (g glucose m−2), and (c) specific leaf area (SLA, m2 kg−1) of the species from 14 contrasting ecosystems. Ecosystem code as in Table 1. Box limits correspond to ± SE and bars to ± SD. No data are available for SLA in ecosystems 13 and 14.

Deciduous leaves from different ecosystems did not show differences in CC (P > 0.3; range: 1.40–1.50 g glucose g−1). In contrast, leaf CC of evergreen species were significantly different (P < 0.001) between ecosystems, with the highest values for the species from the xeric forest (Canary Islands, Spain, 1.66 g glucose g−1) and xeric mediterranean forest (Andalucía, Spain, 1.63 g glucose g−1). In these ecosystems the evergreen leaves also showed a higher Hc (P < 0.05) than in the other ecosystems.

Leaf CC was negatively correlated with SLA (r = −0.28, P < 0.0001) and with the logarithm of leaf blade area (r =−0.42, P < 0.00001, Fig. 5a). SLA and leaf size showed a positive relationship (r = +0.41, P < 0.00001, Fig. 5d).

Figure 5.

Relationships between log leaf area (cm2), and (a) leaf construction cost (g glucose g−1) (r = −0.42, P < 0.00001), and (b) lipid concentration of leaves (g g−1) (r = −0.39, P < 0.01), and (c) ash concentration (g g−1) (r = +0.22, P < 0.05), and (d) specific leaf area (m2 kg−1, SLA) (r = 0.43, P < 0.0001).

Taking into account the effect of leaf area when comparing the CC of leaves leaf life-span or ecosystem type were shown to have no significant effect (Table 3). Similar results were obtained when using SLA as covariate.

Table 3. Results of two ways ANCOVA of leaf construction cost as dependent variable and leaf life-span (deciduous and evergreen) and ecosystem as main factors. In the analysis, log leaf area is included as covariate
Log leaf area 10.000
Leaf life-span 10.626
Leaf life-span * Ecosystem 80.094

The logarithm of leaf size was negatively correlated with lipid concentration (r = −0.39, P < 0.01, Fig. 5b) and positively correlated with ash concentration (r = +0.22, P < 0.05, Fig. 5c).

Leaf CC per unit area and specific leaf area

Leaves with different life-span showed much larger differences in CC when expressed per unit area (CCA) (130, 319, 237 g glucose m−2 for deciduous, semideciduous and evergreen species, respectively, P < 0.00001), than when CC was expressed per unit dry mass basis (Fig. 3a,b). These large differences in leaf CCA (calculated as CC/SLA) were due more to the high differences in SLA (13.8, 5.5, 7.2 m2 kg−1 for deciduous, semideciduous and evergreen species, respectively) (Fig. 3c) than to the differences in leaf CC per unit dry mass (Fig. 3a).

Similarly, mean leaf CCA were significantly different (P < 0.0001) between ecosystems with values ranging from 86 g glucose m−2 (warm temperate forest, North Carolina, USA) to 321 g glucose m−2 in the chaparral (California, USA) (Fig. 4b). The differences in leaf CCA between ecosystems were also mainly due to the differences in SLA, which ranged from 20.9 m2 kg−1 in warm temperate forest species (North Carolina, USA) to 5.1–5.6 m2 kg−1 in xeric mediterranean forest species (Andalucía, Spain) and chaparral (California, USA), respectively (Fig. 4c).


Leaf CC between different species

Leaf CC of the species studied was in the range published for leaves of woody species from different ecosystems (Miller & Stoner, 1979; Merino et al., 1982; Merino, 1987; Chapin, 1989; Sobrado, 1991). Studying young leaves, some authors (Merino et al., 1984; Sobrado, 1994) obtained higher leaf CC than in our study (about 2.2 g glucose g−1). The data in the present study refer to samples representative of all the leaf ages present in the plant (young, medium-aged and mature leaves). Thus, the results obtained represent a mean value of CC of the leaves of all age classes for each species.

Leaf CC showed a wide range between different species from 1.08 to 1.92 g glucose g−1, which represents a 78% difference. Assuming similar differences in roots and stems, it could be significant for the carbon balance of a species, as individuals could grow expending 78% less energy than others; which could result, all things being equal, in higher relative growth rates or more energy allocated to defence and/or reproduction (Poorter & Villar, 1997).

Higher leaf CC results from higher Hc and lower ash concentrations (Fig. 1b,c). However, leaf CC is not positively correlated with N (Fig. 1d), perhaps due to the positive correlation between N (or protein concentration) and ash concentration (Fig. 2b). Because ash has a null direct cost (Penning de Vries et al., 1974), a higher ash concentration is related to a lower CC. This explains the negative relationship of CC and ash concentration found in a previous study in tomato cultivars (Gary et al., 1998) and in our study. Hc was positively correlated with lipid concentration, which explains nearly 40% of the variation in Hc. Similarly, Pantis et al. (1987) and Peng et al. (1993) found a positive correlation between lipid concentration and either Hc or CC. Lipids are one of the components with the highest energy content per unit mass and are one of the most expensive compounds to synthesize (3.030 g glucose g−1, Penning de Vries et al., 1974). In contrast, Poorter & Bergkotte (1992) did not find any relationship between lipid concentration and leaf CC in herbaceous species from central Europe, which could result from the low lipid concentration in these species (Poorter & Villar, 1997). Other compounds with high specific cost, such as lignin or phenols, could also explain the higher CC in some species.

Leaf CC between species with different leaf life-span

In contrast to the large differences in leaf CC between species (78%), the difference in mean leaf CC between leaf life-span types was small (6%, Fig. 3a). Differences in mean relative growth rate between deciduous and evergreen are much higher (98 and 15 mg g −1 d−1, respectively) (Reich, 1998). The small difference observed in leaf CC between species with different leaf life-spans is therefore probably unimportant in determining differences in carbon balance between these groups.

Higher leaf CC in evergreen and semideciduous species result from their higher Hc and lower ash concentration (Table 2). The higher values of Hc in evergreens and semideciduous are caused by their higher lipid concentrations and possibly other compounds such as lignin or phenols (Poorter & Villar, 1997). Evergreen and semideciduous species from mediterranean ecosystems usually have a thick cuticle (Lillis, 1992), which is also found in species with low SLA (evergreens and semideciduous) from other ecosystems (Turner, 1994). The cuticle is rich in lipid compounds, which may be involved in reducing water loss (Larcher, 1995).

However, the differences between deciduous and evergreen leaves are not straightforward. First, the size of the leaves seems to play some role affecting leaf CC as small leaves have higher CC than large leaves (Fig. 5a). Because evergreen and semideciduous leaves considered in the present study have a smaller size than deciduous leaves (Table 2), leaf area could confound the effect of leaf life-span on CC. In fact, when leaf size (log) was included as a covariate, there was no significant difference in CC between leaf life-span types (Table 3). This result agrees with those of Merino (1987), who did not find any difference between leaf CC in deciduous and evergreen species when leaves of similar size were compared. In the present study, smaller leaves showed a low SLA (Fig. 5d) and have a higher lipid concentration (Fig. 5b). Smaller leaves are better suited in water-limited habitats since they have lower water requirements for temperature control (Gates, 1976). In addition, thicker leaves and cuticles reduce water expenditures, the latter generating higher CC (waxes and lipids are expensive to synthesize). Larger leaves require more structural support (i.e. more veins) (Niinemets & Kull, 1999), so they need to allocate more biomass to structural compounds, such as hemicellulose and cellulose, which have low specific costs. Large leaves also showed a high ash concentration (Fig. 5c), which could in part contribute to their lower leaf CC. Therefore; the higher leaf CC of evergreen and semideciduous species could be a consequence of their smaller size (or low SLA) ‘per se’, independently of their leaf life-span.

Secondly, the relationship between leaf CC and leaf life-span types is even less robust when we include semideciduous leaves in the comparison, as they show a life span similar to deciduous leaves (5–12 and 4–8 months for semideciduous and deciduous, respectively) and a CC similar to evergreen ones.

Finally, when we compare leaf CC of deciduous species with that of evergreen species within each ecosystem, we only found significant differences in two of the seven ecosystems compared.

Considering our results, there seems to be no clear relationship between leaf CC and leaf life-span.

Leaf CC between different ecosystems

The difference in leaf CC between ecosystems (about 23%) was higher than the differences in leaf CC between different leaf life-spans (about 6%). The higher leaf CC in different ecosystems was also related to higher Hc, which may be due to higher lipid concentrations. Lipids can play an important role in energy storage in plants from cold climates such as tundra (Bliss, 1962) or alpine ecosystems (Pantis et al., 1987), which may explain the high leaf CC we observed in tundra species (Fig. 4a). Other differences in chemical composition, such as lignin or phenols, may also affect leaf CC.

Different hypotheses have been postulated regarding the effect of habitat on CC or Hc. Golley (1961) hypothesized that Hc (and thus leaf CC) should increase with latitude. The results of some studies agree with this hypothesis (Bliss, 1962; Adamandiadou et al., 1978; Siafaca et al., 1980). In our study, we also found a positive trend between leaf CC and latitude, but this pattern was lost when tundra species were excluded. Other authors (Penning de Vries et al., 1974; Amthor, 1989) have suggested that the CC should be higher in stressful habitats. In contrast, Pitelka (1978) hypothesized that in stressful habitats, the energetic investment in the construction of one organ should be minimized to enhance fitness, resulting in lower CC. Our results do not support any of these hypotheses: species in stressful habitats (tundra and desert) showed high and intermediate leaf CC, respectively. Similarly, Merino (1987) did not find any pattern related to leaf CC in a set of 16 species growing in habitats differing in the degree of water and nutrient availability. One explanation is that in stressful habitats the individual plants could compensate either for higher, or lower, leaf CC by changes in their leaf biomass. Also, different types of stress could have different effects on leaf CC. Thus, a decrease in nitrogen availability appears to decrease leaf CC (Laffite & Loomis, 1988; Griffin et al., 1993), while a decrease in phosphorus availability (Peng et al., 1993), or an excess in light availability, appears to increase leaf lipid concentration (Loveless, 1961) and corresponding values for leaf CC (Williams et al., 1989; Niinemets, 1999). The contrasting effects of different environmental factors on CC as mentioned above, could be one of the reasons for the lack of a clear pattern in leaf CC between ecosystems.

Also, it is important to consider that the pattern in leaf CC between ecosystems could be affected by the leaf life-span dominant in each particular ecosystem (i.e. in some ecosystems only one type of leaf life-span was found). Thus, the best way to compare ecosystems is to focus on one particular type of leaf life-span. It is remarkable that deciduous leaves showed similar CC in different ecosystems around the world, while evergreen leaves from two of the ecosystems considered (xeric forest of Canary Islands and xeric mediterranean forest in Spain) have higher CC than the evergreen leaves from all the other ecosystems. The former could be due to the fact that the environmental conditions in habitats where deciduous species are present, are quite similar (at least during the time of the year when the deciduous leaves are present).

Importance of leaf CC, chemical composition and leaf structure

The similarity of mean leaf CC between leaves with different life-spans and from different ecosystems is surprising. This similarity of mean CC cannot be due to the inability to measure differences in CC, because a wide range of leaf CC was found between species.

The positive correlation between components with high and low cost of synthesis, for example, that between proteins and minerals (Chapin, 1989; Poorter, 1994; Poorter & Villar, 1997, Fig. 2b), or the negative correlation between expensive compounds, for example, protein and lignin, lignin and tannin (Chapin, 1989), or protein and wax (Martínez et al., unpublished), could explain the observed similarity in mean leaf CC in species with different leaf life-span or from different ecosystems. This means that a very different chemical composition might result in a similar leaf CC making leaf CC insensitive to differences in ecosystem or leaf life-span. For this reason, it may more adequate to consider how the energy is allocated in the leaf, more than the leaf CC value itself. A plant can allocate the energy to construct a leaf, either synthesizying different compounds, thus having a different chemical composition, and/or having a different leaf structure [i.e. changing the ratio leaf area (leaf dry mass)1, SLA]. Species with different leaf life-spans differ markedly in both aspects. Species with short leaf life-spans (deciduous) show a higher nitrogen concentration than evergreens (Cornelissen et al., 1997, Table 2) suggesting that a higher fraction of the energy allocated to leaf constrution is used for protein synthesis (Table 2). Deciduous leaves are also characterized by a high SLA (Fig. 3c). Both variables, leaf nitrogen concentration (or protein concentration) and SLA are positively correlated to leaf photosynthesis (Field & Mooney, 1986; Reich, 1993; Reich et al., 1997).

Leaf life-span has been found to be positively related to the ratio CC/daily C gain (Williams et al., 1989) or CC/Amax (Sobrado, 1991), being Amax the maximum photosynthetic rate. This ratio (CC : C gain) is called payback time, which is defined as the time required for a leaf to fix the carbon necessary to equal the cost of leaf construction (leaf CC). Since Williams et al. (1989) and Sobrado (1991) did not find any difference in leaf CC between deciduous and evergreens, we can assume that the positive relationship between leaf life-span and payback time was mainly due to differences in C gain between the two types of leaves, more than to any relationship with CC. In fact, a general negative relationship of net photosynthesis and leaf life-span has been found for a wide range of species from different ecosystems (Kikuzawa, 1991; Reich, 1993; Larcher, 1995; Reich et al., 1997). This suggests that leaf life-span seems to be more related to carbon gain than to energy spent in the construction of the leaf. The higher carbon gain of species with leaves of short life-span is then related to high leaf N and high SLA. As leaf N and SLA are highly correlated (Reich et al., 1992; Reich et al., 1997, our data set), and the range in SLA is larger between contrasting species than leaf N range, it follows that SLA comes up as the main single variable associated with differences in leaf longevity. In fact, SLA has been considered a key variable to explain differences in leaf life-span and relative growth rate in wild species (Lambers & Poorter, 1992; Atkin et al., 1996; Cornelissen et al., 1996; Poorter & van der Werf, 1998; Reich, 1998).

In summary, we conclude that leaf CC per unit dry mass showed large differences between species; lipid concentration (and probably, other chemical compounds) being the main factor responsible for these differences in leaf CC. However, functional groups (evergreen vs deciduous) and ecosystems showed small differences in leaf CC. These differences are very much higher when leaf CC is expressed in area basis, due to the large difference in SLA among species, functional groups and ecosytems. This suggests that SLA is a more critical variable associated with differences in leaf life-span and carbon balance between species, functional groups, and ecosystem types, than leaf CC.


We dedicate this work to our friend, Diego García, a hard-working ecologist, who unexpectedly died when finishing his PhD. We will remember him for his warm and friendly personality, his goodness and strength of spirit. We thank Diego García, Francisco Muñoz for their help with the chemical analysis of the leaves, Antonio Gallardo for his help with sample collection and to Owen Atkin and Adrian Seymour for their appreciable criticism on the manuscript. Eric Garnier and two anonymous referees substantially improved the manuscript. This research was partially financed by projects AMB95/0443 and PB98/1031 (Comisión Interministerial de Ciencia y Tecnología, Spain).


Table Appendix1. Species studied, leaf construction cost (CC), specific leaf area (SLA) and leaf life-span (deciduous: D; semideciduous: S-D and evergreen: E). In the case of tundra species, the CC value calculated considering ammonia as nitrogen source is shown in brackets
SpeciesCC (g glu g−1)SLA (m2 kg−1)Leaf life-span
TUNDRA (Devon Island, Canada)
Cassiope tetragona  1.600 9.05E
Dryas integrifolia  1.837 9.42E
Salix arctica  1.73713.27D
DESERT (Chihuahua, USA)
Artemisia tridentata  1.467 7.63S-D
Chilopsis linearis  1.494 4.30D
Chrysothamnus nauseosus  1.525 8.18E
Flourensia cernua  1.538 8.63D
Larrea tridentata  1.677 8.24E
Simmondsia chinensis  1.462 5.08E
XERIC FOREST (Canary Island, Spain)
Adenocarpus foliolosus  1.77510.02E
Argyranthemum spp. 1.374 7.21D
Bupleurum salicifolium  1.561 7.12S-D
Chamaecytisus proliferus  1.740 6.59S-D
Cistus monspeliensis  1.459 4.63S-D
Cistus simphytifolius  1.475 6.05S-D
Dafne gnidium  1.652 6.28E
Dorycnium spectabile  1.51512.42D
Hypericum reflexum  1.506 9.13D
Jasminum odouratissimum  1.715 6.47E
Juniperus phoenicea  1.609 n.a.E
Olea europaea  1.683 5.09E
Pinus canariensis  1.515 n.a.E
Rumex lunaria  1.24410.71D
CHAPARRAL (California, USA)
Adenostoma fasciculatum  1.561 4.06E
Arctostaphylos crustacea  1.609 5.29E
Ceanotus cuneatus  1.490 3.60E
Diplacus aurantiacus  1.638 8.33S-D
Eriodictyon californicum  1.657 3.67S-D
Lepechinia calycina  1.548 8.85D
Arctostaphylos uva-ursi  1.687 3.14E
Cistus albidus  1.452 5.40S-D
Cistus ladanifer  1.555 3.50S-D
Cistus libanotis  1.582 3.69S-D
Cistus monspeliensis  1.548 4.18S-D
Cistus salvifolius  1.40610.50S-D
Halimium commutatum  1.463 3.40S-D
Halimium halimifolium  1.467 6.07S-D
Halimium umbellatum  1.570 3.56S-D
Juniperus communis  1.629 6.66E
Juniperus oophora  1.701 n.a.E
Lavandula stoechas  1.735 6.30S-D
Olea europaea  1.653 5.04E
Phillyrea angustifolia  1.739 4.26E
Phlomis purpurea  1.489 n.a. S-D
Pinus pinaster  1.666 n.a. E
Pistacia lentiscus  1.588 4.62E
Quercus coccifera  1.532 6.04E
Quercus rotundifolia  1.470 5.93E
Rosmarinus officinalis  1.736 4.64S-D
Aesculus californica  1.33913.92D
Alnus rhombifolia 1.46310.33D
Arbutus menziesii  1.480 6.97E
Heteromeles arbutifolia  1.542 5.19E
Quercus agrifolia  1.395 5.56E
Quercus douglasii  1.450 7.91D
Quercus keloggii  1.416 8.49D
Quercus lobata  1.462 7.63D
Rhamnus californica  1.43710.68E
Salix laevigata  1.596 8.25D
Arbutus unedo  1.536 6.48E
Calluna vulgaris  1.736 7.11E
Ceratonia siliqua  1.464 5.01E
Cistus laurifolius  1.570 3.39S-D
Cistus populifolius  1.446 4.85S-D
Crataegus monogyna  1.50510.63D
Erica scoparia  1.917 6.18E
Ficus carica  1.223 8.70D
Frangula alnus  1.51214.11D
Fraxinus angustifolia  1.50113.36D
Fraxinus ornus  1.563 6.39D
Mirtus communis  1.45310.39E
Nerium oleander  1.683 5.51E
Populus alba  1.37511.72D
Pyrus bourgeana  1.57211.81D
Quercus faginea  1.528 9.17D
Quercus lusitanica  1.537 6.76D
Quercus pyrenaica  1.44012.90D
Quercus suber  1.541 7.07E
Rubus ulmifolius  1.49811.91E
Salix sp. 1.583 9.36D
Smilax aspera  1.605 8.48E
Vitis vinifera  1.49615.78D
TEMPERATE FOREST (Toronto, Canada)
Acer rubrum  1.48822.56D
Betula papyrifera  1.49711.83D
Populus tremuloides  1.52612.30D
Quercus rubra  1.51912.10D
Carya tomentosa  1.34927.89D
Chionanthus virginicus  1.44328.95D
Cornus florida  1.34823.16D
Ilex opaca  1.532 7.36E
Liquidambar styraciflua  1.39714.00D
Liriodendron tulipifera  1.48128.13D
Platanus occidentalis  1.57823.28D
Populus heterophylla  1.43514.34D
AUSTRAL FOREST (Tierra del Fuego, Argentina)
Berberis ilicifolia  1.421 4.87E
Drimys winteri  1.479 6.85E
Embothrium coccineum  1.518 7.82E
Maytenus magellanica  1.329 7.32E
Nothofagus antarctica  1.51010.86D
Nothofagus betuloides  1.531 6.65E
Nothofagus pumilio  1.34818.33D
LAURIPHYLL FOREST (Canary Islands, Spain)
Apollonias barbujana  1.493 7.07E
Arbutus canariensis  1.490 6.97E
Erica arborea  1.819 7.60E
Erica scoparia  1.695 7.77E
Heberdenia bahamensis  1.356 8.84E
Ilex canariensis  1.567 6.34E
Ilex platyphylla  1.535 7.13E
Laurus azorica  1.563 7.76E
Maytenus canariensis  1.184 8.47E
Myrica faya  1.47810.55E
Ocotea foetens  1.643 9.10E
Persea indica  1.574 8.41E
Picconia excelsa  1.584 5.73E
Prunus lusitanica  1.631 7.72E
Rhamnus glandulosa  1.390 8.09E
Salix canariensis  1.62316.56D
Viburnum rigidum  1.582 7.02E
Visnea mocanera  1.413 7.19E
TROPICAL DRY FOREST (Charallave, Venezuela)
Beureria cumanensis  1.49613.72D
Capparis aristiguetae  1.473 8.90E
Coursetia arborea  1.42627.62D
Curatella americana  1.57411.49E
Lonchocarpus dipteroneurus  1.55023.92D
Morisonia americana  1.589 8.80E
Pithecellobium dulce  1.47316.89D
RAIN FOREST (Douala-Edea forest, Cameroon)
Anthonotha gracilliflora  1.596 n.a. E
Anthonotha macrophylla  1.611 n.a. E
Barteria fistulosa  1.616 n.a. E
Berlinia auriculata  1.545 n.a. E
Cissus producta  1.429 n.a. E
Coula edulis  1.669 n.a. E
Deidamia clematoides  1.487 n.a. E
Dichostemna caloneura  1.534 n.a. E
Dichostemna glaucens  1.523 n.a. E
Diospyros dendo  1.653 n.a. E
Diospyros hoyleana  1.594 n.a. E
Garcinia mannii  1.483 n.a. E
Garcinia ovalifolia  1.696 n.a. E
Leptaulus daphnoides  1.551 n.a. E
Librevillea klainei  1.645 n.a. E
Lophira alata  1.626 n.a. D
Mammea africana  1.599 n.a. E
Protomegabaria tapfiana  1.367 n.a. E
Rauvolfia vomitoria  1.371 n.a. E
Trichoscypha patens  1.528 n.a. E
Uapaca staudtii  1.418 n.a. E
RAIN FOREST (Kibale forest, Uganda)
Bosquia phoberos  1.372 n.a. E
Cassipourea ruwenzoriensis  1.494 n.a. E
Celtis africana  1.249 n.a. D
Celtis durandii  1.250 n.a. D
Chaetacme aristata  1.082 n.a. D
Dombeya mukole  1.443 n.a. E
Erythrina excelsa  1.392 n.a. E
Funtumia latifolia  1.564 n.a. E
Markhamia platycalyx  1.472 n.a. E
Milletia dura  1.559 n.a. E
Pancovia turbinata  1.402 n.a. E
Parinari excelsa  1.455 n.a. E
Strombosia scheffleri  1.392 n.a. E
Teclea nobilis  1.462 n.a. E