1. We hypothesized that in Mediterranean plant communities seedling survival of different species during the drought period would be related to their ability to use below-ground resources, particularly water and nitrogen.
2. For 5 years we studied under field conditions the summer seedling survival of 11 dominant species of a Spanish Mediterranean shrubland. We related seedling survival of the different species to their seed size and root allocation estimated as the slope of the function root biomass vs log shoot biomass. We used δ13C for the estimation of water-use efficiency and δ15N and nitrogen concentration to determine the sources of nitrogen utilized. We correlated these variables with root allocation.
3. Seedling survival of the different species was positively correlated with root allocation and seed size. Root allocation was also positively correlated with seed size. δ15N and nitrogen concentration were also positively correlated with root allocation, but δ13C was not.
4. Under the relatively moist conditions occurring during summer 1996, higher root allocation was associated with the use of nitrogen from more nitrogen saturated microsites.
Root:shoot biomass ratio of adult plants in Mediterranean ecosystems tends to be higher than in more temperate ecosystems, possibly as an adaptation to the summer dry season (Hilbert & Canadell 1995). Large root:shoot ratios could therefore be expected to be advantageous for seedling survival in drought-prone and perhaps nutrient-poor environments (Gedroc, McConnaughay & Coleman 1996). However, Jurado & Westoby (1992) studied seedling growth and survival from species of arid Australia, focusing on recently emerged seedlings that were grown under glasshouse conditions and they did not find significant correlations between seedling survival and root:shoot ratio. Long-term studies considering the coexisting species in a community are rare but they are needed to check for the significance of plant features influencing seedling survival under field conditions. In spite of the great interannual rainfall variability and the soil heterogeneity occurring in Mediterranean ecosystems, most seedling mortality occurs in the first dry season of their life cycle, so the relevant value of the root:shoot ratio in relation to seedling survival should be determined just before this period.
The existence of differences among the patterns of seedling survival of coexisting species may provide insights on the main constraints operating at the community level. Differences in the development stage of seedlings and the variability of microenvironmental conditions may result in differences in the root:shoot ratios of individual plants. However, we assume that species characteristics largely override development stage or environmentally induced influences and for a given species, root and shoot sizes show characteristic allometric relationships regardless of the development stage, which makes it possible to compare the pattern of root vs shoot allocation among different species (Kohyama & Grubb 1994).
In the Eastern Iberian Peninsula early successional shrublands have become dominant because of the abandonment of agricultural fields and the increase of fire frequency in recent decades (Masalles & Vigo 1987). These shrublands are dominated by short, evergreen shrubs and the substitution by taller, late successional shrublands appears to be slow. In these communities, grazing is currently not very intense and the main constraints to the establishment of new individuals are seed and microhabitat availability (Lloret 1998).
We aimed to determine, in 11 dominant perennial species of one of these Mediterranean shrublands, first the relationship between seedling survival during the summer dry season and seedling root:shoot ratio and seed size, and second the relationship between seedling root:shoot ratio and water-use efficiency and nitrogen utilization. We hypothesized that seedling survival would be higher in species with more allocation to root growth before their first drought season and with larger seeds, and that these trends would correlate with water-use efficiency and nitrogen use, estimated by δ13 and δ15N, respectively.
FIELD SITE AND SPECIES DESCRIPTION
The study was conducted in the Garraf Natural Park, about 30 km south-east of Barcelona, Spain (latitude 41 ° 18 ′N, longitude 1 ° 54 ′E). The site is located between 250 and 350 m a.s.l. and 8 km from the coast. The climate is typically Mediterranean. Mean annual rainfall at the closest weather station is 548 mm (Vilanova i la Geltrú, 11 km to the South-east, averages over 29 years) (Allue 1990). There is a pronounced summer drought (100 mm) from June to August. Mean annual temperature is 16·7 °C. Mean maximum temperature in the hottest month (August) is 30·6 °C. In 1994 both spring and summer were very dry, while there was abundant rainfall in August 1993, 1995 and 1996 and in springs 1992, 1993 and 1996. Overall 1994 was the driest year, while 1996 was the wettest (Fig. 1).
The study site was located in field terraces which were abandoned approximately a century ago. Stony and basic soils have developed here over compact Cretaceous limestone. The current vegetation on these terraces is shrubland about 1 m high, belonging to the Rosmarino–Ericion Br. Bl. 1931 alliance (Folch 1981), that regenerated after a fire in 1982. The dominant species are Quercus coccifera L., Erica multiflora L., Globularia alypum L., Ampelodesmos mauritanica. (Poiret) T. Durand et Schinzand, and Rosmarinus officinalis L., contributing, respectively, 20%, 15%, 12%, 10% and 5% to the total ground cover. The undergrowth is dominated by the grass Brachypodium retusum (Pers.) Beauv., and several evergreen half shrubs species, that is, with non-lignified upper parts of their shoot systems (Pate, Dixon & Orshan 1984), which account for up to 40% of the ground cover. Total ground cover is c. 80%, but is not uniform, and open areas of about 1–10 m2 are common.
All the species selected for this study were common shrub or half shrub members of the community and all regenerated from seeds, although some of them are also able to resprout (E. multiflora, G. alypum, C. minima). Growth form, adult size and seed mass are shown in Table 1. Seedling emergence occurs from autumn to spring, associated with rain events. Summer storms are occasional, but the resulting replenishment of water in the upper soil horizons, where seedling root systems develop is only transient. Although the existing microhabitat heterogeneity may influence seedling dynamics, clear patterns of emergence and survival have not been found in relation to canopy cover (Lloret 1998).
Table 1. . Species name, family, growth form (GF), range of adult height (AH), seed mass (SWT), years of available data for seedling survival, date of seedling harvest used to estimate root log–1 shoot (R/logS) ratios and date in which seedlings were collected for isotope analysis. Nomenclature follows de Bolòs et al. 1990
Materials and methods
Seedling survival was recorded from 1992 to 1996. In June each year, seedlings that had emerged since the previous summer were labelled. They were recorded again in the following September. The number of labelled seedlings of each species and year varied according to their abundance, ranging from 30 to 745. Some species could not be recorded every year because of great variability in their seedling emergence. Species suspected of having incurred mortality by predation during the summer were excluded from the analyses (Table 1).
Additional sets of the above-ground and below-ground parts of 25–30 seedlings were harvested in June 1995, June 1996 and September 1996 (Table 1). Special care was taken to avoid the breaking of fine roots and only undamaged intact samples were accepted. Roots and shoots, including leaves, were separated in the field, placed in paper envelopes, dried at 80 °C for 48 h and weighed to the nearest 0·001 mg.
The aim of the study was not to follow the process of seedling development, but to investigate the status of root vs shoot sizes for each species just before onset of the drought season. Harvested seedlings were not obtained from growth sequences; instead, they were at different development stages and grew under different microenvironments. The use of single means to characterize the root vs shoot relation of one species was not recommended here because of the great variability between seedling histories, which can be better accounted by the slope of a function relating root and shoot masses. We estimated the biomass allocation to root vs the allocation to shoot as the slope of the root vs log shoot regression, named from here on as R/logS ratio. This slope was calculated from the seedlings of each species harvested in June 1995, June 1996 and September 1996, respectively. This relationship, which corresponds to the function y = a exp bx, where y = shoot biomass and x = root biomass, takes into account that the shoot is related to the size of the root in a non-linear way.
Seed mass data were obtained from seeds collected in the same locality during the spring and summer of 1996. More than 300 seeds were collected from 10 plants of each species, and a set of 30 randomly selected seeds for each species was weighed.
Samples of 3 and 5 mg of seedlings aerial parts collected in September 1996 were used to analyse δ13C and δ15N, respectively (five samples for each species and analysis) (Table 1). When the biomass of a single seedling was too small to perform the analysis, several seedlings of similar R/logS were pooled together. Carbon and nitrogen leaf concentrations were analysed with a Carlo Erba NA1500 Analyser, using the standard configuration for those determinations. The δ15N and δ13C ratios were measured on a SIRA Series II isotope ratio mass spectrometer (VG Isotech, Middlewich, UK) operated in direct inlet continuous flow mode after combustion of the samples in an elemental analyser (NA1500, Series 1, Carlo Erba Instrumentaziones, Milan, Italy). The reference CO2, calibrated against standard Pee Dee belemnite (PDB), was obtained from Oztech (Dallas, TX, USA). A system check of analysis was achieved with interspersed working standards of cellulose, atropine and urea (Sigma, St. Louis, MO, USA). Data on δ18O composition of the samples were used for correction of δ13C. The accuracy of the method was ± 0·1‰ for δ13C and 0·2–0·5‰ for δ15N.
Some of the statistical analyses such as the regression models, were conducted with species mean values, and therefore with values having known associated variance. Thus a model I regression is not strictly appropriate and a model II regression was used (Coleman, McConnaughay & Ackerly 1994). Although, there is not a consensus about the optimal technique to compute this model II regression, we used the reduced major axis method (Sokal & Rohlf 1995), which has been used in the recent literature (Shipley & Peters 1990; Jurado & Westoby 1992). Logarithmic transformations to homogenize variances were applied to species survival and seed size. These variables were correlated with the species R/logS ratios obtained in the June 1995 and June 1996 harvests. Species δ13C, δ15N and N concentration (% dry mass) were correlated with species R/logS ratios obtained in the September 1996 harvesting.
Seedling survival of all species during summer showed positive significant correlations with the root/log shoot ratio in 1995 and 1996, the only years in which data on root and shoot masses were available (Fig. 2). The slope of the regression equation was higher in 1995 than in the wetter year 1996. Seedling survival also significantly increased with species seed mass in the 5 years of our observations (Fig. 3). The slope of the regressions were higher in 1994, an extremely dry year, and decreased in years with wet summers (1995 and 1996). Species R/logS ratio also showed significant positive correlations with species seed mass in 1995 and 1996 (Fig. 4).
Species seedling δ13C, and therefore water-use efficiency did not show significant correlations with species R/logS ratio or with species seed size (Fig. 5). There was, however, a positive significant correlation between the seedlings nitrogen isotopic δ15N, and the R/logS ratio of the different species (Fig. 5). The correlation between δ15N and the seed size was not significant. The concentration of nitrogen in the above-ground parts of the seedlings was positively correlated both with the R/logS ratio of the species and with the seed size (Fig. 5).
When considering the seedlings of all species combined, δ13C and δ15N did not show significant correlations with seedling size, but nitrogen concentration did (r = 0·49, P < 0·001). Nitrogen concentration (%) in above-ground organs was not significantly correlated with δ13C, but it was correlated with δ15N (r = 0·35, P = 0·018).
Seedling survival during summer increased with allocation to below-ground organs and seed size. As far as we know, this is the first report under field conditions showing that the survival of seedlings during the summer dry season in a Mediterranean community was related to the ability of the species to produce roots rather than shoots and leaves. Larger root systems may achieve deeper soil levels in which water remains available throughout the summer (Canadell & Zedler 1995). This may not be relevant in the studied seedlings because their roots rarely are present below 10 cm. However, a greater allocation to root biomass may allow a more efficient use of superficial water (Dodd et al. 1984; Hilbert & Canadell 1995). There was no trend in δ13C among the species in relation to R/logS and the water-use efficiency was similar for all the species studied. The irregularly fluctuating soil water regime and the short length of the seedling stage could make insufficiently beneficial the selection for a particular WUE. However, the absence of a relationship may be also a consequence of the wet conditions occurring during the 1996 summer (Fig. 1). In the relationships between survival, R/logS and seed size in the 1992–1996 period, there was still a general trend of increasing slope associated with drier summers, suggesting a more important role for these variables in dry summers.
Root system development may also be related to more efficient extraction of available nutrients, such as nitrogen or phosphorous which several authors have considered to be limiting in many Mediterranean ecosystems (Specht & Moll 1983; Keeley & Keeley 1988; Lamont 1995). There were increasing shoot δ15N values and higher shoot nitrogen concentrations associated with increasing R/logS ratios (Fig. 5). These relationships suggest that species with larger root allocation (larger R/logS ratio) such as C. minima, P. rupestris and H. syriacum are able to exploit more efficiently soil systems and to use more nitrogen from more nitrogen-saturated soil sites (sensuAber et al. 1989). The legume C. minima is moreover able to fix nitrogen, which also increases its δ15N values (Shearer & Kohl 1986). Therefore, the between species differences in δ15N values reported here are probably a consequence of specific differences in the exploitation of nitrogen sources.
Seed size and seedling survival in the first 10 days after germination have been shown to be positively correlated for species of Australian semiarid environments, grown in coarse sand and under standard conditions in a glasshouse (Jurado & Westoby 1992). We have found this same pattern under field conditions for 5 years. A greater independence of seedlings from mineral soil resources in species with larger seeds has been proposed to explain this trend (Atkinson 1973; Jurado & Westoby 1992). Seed size has also been found to be positively correlated with seedling size in other poor nutrient environments such as Mediterranean woodlands (Stock et al. 1990). This hypothesis of greater independence from mineral soil resources in species with larger seeds is also supported in our study by the positive correlation between seed size and shoot nitrogen concentration in seedlings. This relationship may also be owing to the higher root allocation in large seeded species, which increase their ability to search for soil nitrogen. On the other hand, the highest values of δ15N and percentage N were found in C. minima (Fig. 5), a legume in which the pattern of nitrogen use may already be influenced by its N fixing symbionts.
We also report a positive relationship between seed size and root allocation in natural communities. Jurado & Westoby (1992) did not obtain this result in species from arid habitats in Australia, but they only considered early seedling development under standard conditions in a glasshouse. More resources from larger seeds may also enhance the ability of seedlings to minimize the effects of the limiting factors of a given environment (Armstrong & Westoby 1993). In Mediterranean ecosystems these limitations mainly lay in the soil and hence more allocation to roots is to be expected.
Thus, in conclusion, differences in seedling survival among species were partially explained by the ability to allocate resources to below-ground vs above-ground organs and this pattern was apparently associated with the ability to take up soil resources. Although the mechanisms involved remain insufficiently explored, we have also observed a positive relationship between seed size and seedling survival, probably related to the early pattern of plant development and resource use. Our results indicate that, in addition to water deficit, nutrient availability influences survival in the early stages of plant development in Mediterranean pioneer shrubs and this effect operates at the small scale at which seedlings interact with the environment.
We thank anonymous referees for suggestions to the manuscript. This work was supported by CICYT (AMB94–0881, AMB94–0199, and CLI97–0344) and CIRIT.