Study area and species

The study area is situated on the Causse du Larzac, a 1000-km² area of limestone plateau in the south of the Massif Central, France. The altitude of the plateau varies between 560 and 920 m a.s.l. and mean annual rainfall at the two study sites (site 1 is located at 43°55′ N, 3°16′ E and site 2 at 43°57′ N, 3°13′ E) is between 1000 and 1200 mm. Rain falls mainly in autumn and winter. Winter is also characterized by heavy frosts (Marsteau & Agrech 1995).

The landscape consists of calcareous grasslands, mostly grazed by sheep, with scattered shrubs (mainly the evergreen and unpalatable Buxus sempervirens and Juniperus communis) and woods of Pinus sylvestris and Quercus humilis (plant nomenclature follows Kerguélen 1993). Woodland area is currently increasing (IFN 1995). Quercus humilis is a deciduous tree with a maximum height of 20 m and is widespread in the countries around the northern Mediterranean basin (Timbal 1975). It was a dominant species on the Causse du Larzac before forests started to be cleared about 7000 bp (Vernet 1972). At present it is recolonizing, thanks to effective dispersal by jays (Garrulus glandarius) and rodents (mainly Apodemus sylvaticus) (Rousset & Lepart 1999). The acorns reach maturity in autumn (late October or early November) and germinate in the following spring. In contrast to the distribution of older trees, that of first-year seedlings is independent of the type of plant cover (Rousset & Lepart 1999).

An experimental study on the germination and establishment of oak seedlings was conducted at site 1 (a fenced 21-ha plot, grazed every 2 years by sheep at a stocking rate of 200 ewe days ha⁻¹). The chalk grassland here has a 20% shrub cover (Buxus and Juniperus). The growth of established Q. humilis saplings was analysed at site 2, which has been ungrazed for more than 10 years and consists of a series of large clearings with a 21% cover of Buxus. Both sites are situated on the same pedological unit consisting of a fersiallitic soil overlying crystalline dolomite (Cadillon 1970).

Germination, growth and mortality of seedlings

In 1996, acorns which were about to fall were collected and stored in sand for 3 weeks in a cold room (+ 4 °C). When 200 acorns, chosen at random, were then placed onto wet sand (at a constant temperature of 24 °C), percentage germination was 99%. At the same time, 1890 acorns were planted in the field at site 1 in holes 4 cm deep, covered with fine earth and protected against rodent predation by a grill (13 × 13 mm mesh) fixed at soil level. The acorns were either planted near shrub canopy or under shrub canopy (experiments A and B), or in the centre of large areas of grassland (experiment C).

Growth of naturally established individuals

A total of 25 pairs of oak saplings were selected from regular surveys at site 2, which divided individuals into 20 cm height classes (from 10 to 210 cm; 2–3 pairs per class). Each pair consisted of two individuals from the same height class, one rooted under a Buxus shrub and the other from the immediate vicinity of the same shrub individual (less than 2 m from the canopy margin). The saplings rooted under shrub were divided into those which were shorter than the canopy (n = 8, termed ‘under Buxus’) and those which overtopped the canopy (n = 17, called ‘above Buxus’).

For each individual, the angle between the trunk and the soil, the trunk diameter at soil level and the height of the stem was measured. Five twigs (ramet of the year) were collected at random and the length, diameter at the base, number of leaves and leaf area were measured for each of them. The stems and leaves were weighed after drying in an oven at 60 °C for 48 h. A section of stem was collected from the base of each plant, and its vessels exposed. The vessels of the start of the growing season are clearly distinguishable from those at the end by their larger diameter, so that the annual rings could be clearly differentiated. The width of the annual rings were measured (to within 0.01 mm) using an Eklund machine (Elektronlund, Malmo, Sweden) connected to a recording computer. The measurements were made along a radial row of cells and were repeated in three directions at 60° to one another. To make the measurements as reliable as possible the standard cross-dating method (Fritts 1976) was used, even though false rings are not known in Q. humilis. For the ‘above Buxus’ individuals, a fragment was also collected from the main stem at the internode emerging from the shrub canopy. The number of rings was determined on this fragment to estimate the year in which the sapling emerged from the canopy.

Data on above-ground organs, other than the width of annual rings, were analysed by manova using proc anova of SAS. Two analyses were conducted: one for individuals situated under shrubs (‘under Buxus’) and for neighbouring plants on the grassland, and the other for plants that overtopped the canopy (‘above Buxus’) and their neighbours. In accordance with Scheiner (1993), we tested the normality of each variable independently. Log transformation was used to improve the approximation to normality when needed. For the variable ‘angle’, this transformation still did not result in approximation to normality, so this was analysed by the nonparametric Kruskal–Wallis test.

As there was a relation between the width of the rings and the age of plants (Cook 1990), the age-effect was corrected using the biological age of trees (method described by Cook et al. 1990). The data on ringwidths were then analysed by analysis of variance on measurements repeated with time (anovar) using the GLM procedure of SAS. Only the last 6 years, the period when all the individuals classified as ‘above Buxus’ had emerged from the canopy, were included in the analysis. The sphericity of the variance-covariance matrix was tested using Mauchley’s criterion (Winer et al. 1991). As this condition was not met, the number of degrees of freedom had to be adjusted using the Greenhouse-Geisser and Huynh-Feldt procedure before calculating F-values.

Germination and seedling growth

Experiment A: effects of shrubs

The percentage germination of Quercus humilis was significantly higher under shrub cover than outside the canopy in the vicinity of shrubs (Fig. 1 and Table 1). The north orientation improved oak germination only under shrubs (Fig. 1), resulting in a significant interaction between canopy and orientation (Table 1).

Table 1.  Results of anova for the effects of shrub species, orientation and canopy on the germination and, at date 2, on seedling height (H2), leaf dry mass (LDM2) and leaf mass per area (LMA2) of Quercus humilis. Acorns were sown in the vicinity of two shrubs (Buxus and Juniperus; factor species), in two directions (north and south; factor orientation) and near or under the canopy of the shrub (factor canopy)

Germination			H2			LDM2			LMA2
Effect	d.f.	F	P	d.f.	F	P	d.f.	F	P	d.f.	F	P
Species (SPE)	1	0.0	0.9752	1	2.5	0.1284	1	7.1	0.0134	1	1.2	0.2859
Orientation (ORI)	1	0.3	0.5857	1	0.7	0.4131	1	0.0	0.8685	1	0.1	0.7091
Canopy (CAN)	1	36.6	0.0001	1	10.0	0.0039	1	0.7	0.4189	1	7.0	0.0140
SPE × ORI	1	4.5	0.0419	1	1.1	0.2993	1	1.0	0.3216	1	1.6	0.2158
SPE × CAN	1	2.7	0.1093	1	0.1	0.8078	1	0.0	0.9186	1	0.4	0.5356
ORI × CAN	1	6.1	0.0194	1	1.0	0.3275	1	0.7	0.4247	1	0.0	0.9974
SPE × ORI × CAN	1	0.1	0.7866	1	0.5	0.4953	1	1.0	0.3338	1	0.0	0.9315
Error	32			27			25			25

At the end of the second year of growth (date 2), seedlings were taller under shrubs and the leaf mass per area was lower (Fig. 2 and Table 1). These effects were already significant at the end of the first year’s growth (anova of canopy effect: P = 0.02 for the height and P = 0.0002 for the leaf mass per area). At date 2, the leaf dry mass of seedlings associated with (i.e. either under or near) Juniperus was higher than those associated with Buxus (Fig. 2 and Table 1) but at date 1 this difference was only recorded under the shrub canopy (anova of species × canopy effect: P = 0.02, and anova of species effect: P = 0.39).

Experiment C: effects of sheep grazing and competition from herbs

Livestock grazed the plot before the end of the germination period and this life stage was therefore only analysed for the sub-blocks within exclosures. There was no significant difference in percentage germination between the competition (61.9%) and no-competition treatments (58.9%) (anova of competition effect: P = 0.73).

Outside exclosures, all the seedlings were grazed by sheep but most of them (56%) resprouted. Grazing had a negative effect on height and leaf dry mass at date 1 (anova of grazing effect: P = 0.0002 for the height and P = 0.003 for the leaf dry mass) and the effects remained significant by date 2 (Table 3). The presence of a herb layer significantly increased height and significantly decreased the leaf mass per area at date 2 (Table 3 and Fig. 3), however at date 1 only height was significantly increased (anova of competition effect: P = 0.003 for the height and P = 0.08 for the leaf mass per area). The significant grazing–competition interaction (Table 3) resulted from the fact that the total leaf mass of seedlings was only lower in the herb layer in ungrazed areas (Fig. 3).

Table 3.  Results of anova for the effects of block, grazing and grass removal on the germination, seedling height (H2), leaf dry mass (LDM2) and leaf mass per area (LMA2) of Quercus humilis at date 2. Seedlings are protected or not from sheep (factor grazing). Herbaceous plants near seedlings have been removed or not (factor competition). Bold type indicates that the effect is significant (P < 0.05)

H2			LDM2			LMA2
Effect	d.f.	F	P	d.f.	F	P	d.f.	F	P
Main plots
Block (BLO)	8	3.8	0.0491	8	0.9	0.5480	8	1.6	0.2763
Grazing (GRA)	1	52.0	0.0002	1	7.8	0.0267	1	0.1	0.7729
Main plot error (BLO × GRA)	7	1.7	0.2280	7	4.7	0.0140	7	2.2	0.1223
Sub-plots
Competition (COMP)	1	17.4	0.0019	1	6.1	0.0332	1	5.4	0.0421
GRA × COMP	1	2.9	0.1204	1	7.5	0.0210	1	0.3	0.5879
Sub-plot error	10			10			10

Seedling mortality

By the end of the experiment, mortality of Q. humilis seedlings under shrubs (where they were without artificial protection) was 45% and was exclusively due to herbivory (experiment A). Mortality was especially high in the summer of the first year. Only 4% of seedlings under shrubs disappeared while sheep were present, suggesting that small grazers are the major herbivores in such position. In the vicinity of shrubs, where seedlings were protected against rodents, rabbits and sheep, mortality was 17.1% and was related to drought stress. Mortality was twice as high to the south than to the north of shrubs.

In experiment B, where no seedlings were subjected to predation, mortality where the northern half of the shrubs had been removed was significantly lower than for seedlings near the northern edge of the shrub canopy (2.2% compared to 8.6%) (anova of soil effect: P = 0.022).

In experiment C, seedling mortality was significantly higher in grazed areas (anova of grazing effect: P = 0.002), where 44% of seedlings died as a result of sheep grazing (Fig. 4) compared to only 1% (during the same period) in exclosure cages. Mortality attributed to small grazing animals was 17% in exclosure cages, compared to 13% in unprotected areas (of the total seedlings, i.e. 23% of the seedlings resprouting after the passage of sheep). The significant grazing–competition interaction (P = 0.035) reflected a higher mortality of oaks subject to competition in exclosure cages (Fig. 4; the factor competition was not significant, P = 0.64). Many of the deaths of exclosed plants were due to drought stress especially when they were subject to competition (Fig. 4).

Growth of individuals established under buxus

The trunks of Q. humilis plants rooted under Buxus had a much lower angle to the soil than those of individuals growing in the surrounding grassland (Table 4; Kruskal–Wallis test, P < 0.01) and are pointed towards the nearest canopy margin. The above-ground organs (stems and leaves) of individuals that remained below the shrub had a slower growth (branch length, diameter and weight, and number of leaves) but a larger leaf area than plants corresponding in the grassland (Table 4; manova, Pillai’s Trace test, P = 0.03). There was no significant difference for any of the variables (except trunk angle) between individuals that had emerged above the Buxus canopy and corresponding ones in the grassland (Table 4; manova, Pillai’s Trace test, P = 0.13).

The comparison of the width of annual rings also showed that individuals under Buxus had the slowest growth (Table 5 and Fig. 5). Individuals now above Buxus had annual growth rates before 1970 which were similar to those currently under shrubs (Fig. 5). All of them were under canopy at this time, but as soon as individuals emerged above the canopy, their growth rate increased to become similar to plants on the grassland. In recent years when all such individuals had overtopped the canopy their mean relative growth rate was similar to individuals in the grassland (Fig. 5).

Effects of shrubs at different life stages of quercus humilis

Colonization by Quercus humilis on the Causse du Larzac depends on interactions with dispersing animals (Rousset & Lepart 1999), as well as with herbaceous and woody plants and with herbivores. The presence of Buxus and Juniperus has both positive and negative effects on the establishment of this tree. The effects of shrubs vary in relation to the shrub species and the life stage of Q. humilis.

Germination

Buxus and Juniperus have a positive effect on the germination of Q. humilis that is unaffected by soil conditions (litter or soil structure), as there was no significant difference between acorns sown where half the shrub canopy had been removed and those sown near to but not under shrubs. Bacilieri et al. (1993) also found similar germination rates for Q. humilis in soil samples collected from various plant formations. The facilitation mechanism during the germination stage thus appears to be related to changes in microclimate under the canopy. As germination was higher to the north than to the south of the canopy, soil moisture may be altered by the presence of a canopy (see Franco & Nobel 1988) and may be exerting an important influence at this stage (Vuillemin 1982).

Mortality

After the germination stage, the overall effect of shrubs on survival of Q. humilis seedlings remains positive, even though some negative effects could be identified. The main positive effect of shrubs is indirect. The branches of the unpalatable Buxus and Juniperus prevent sheep from gaining access to the seedlings. In open areas, sheep grazed all the seedlings and caused 44% mortality during a single passage of the flock. Even though grazing took place very shortly after the emergence of the first leaves, the remaining 56% of seedlings were able to resprout from below. Their growth was, however, severely affected by the end of the first year, leaving them little chance of surviving regular grazing, given that older plants are also grazed (Di Pasquale & Garfi 1998). The slight mortality (4%) recorded under shrubs during the sheep grazing period could be attributed to small grazing mammals, such as rodents, which grazed 1% of seedlings in cages that excluded both sheep and rabbits. Higher mortality under shrub cover could be related to a higher abundance of small mammals where they are protected from predators (Treussier 1975)

After 2 years, 45% of seedlings under shrubs had died from the effects of small grazing animals. In the grassland, mortality due to herbivores other than sheep was lower than under shrubs and was similar inside and outside cages inaccessible to rabbits but open to rodents (17% vs. 23%). Rodents are therefore probably the second most important predator of oak seedlings after sheep, while rabbits seem to play a negligible role.

In addition to protection from sheep grazing, the second positive effect of the canopy on establishment of Q. humilis was the lower mortality of seedlings during summer drought. No mortality of this type was recorded under shrubs, in contrast to all other situations. The decreased drought stress under shrubs is certainly related to a direct reduction of solar radiation and temperature by the canopy (Parker & Muller 1982), but there may also be an indirect effect related to the lower density of competitive herbs under shrubs. Two results seem to confirm the importance of the herb cover in the establishment of Q. humilis. First, seedling mortality was higher near the canopy than where half of the canopy had been removed leaving the herb cover low or absent. Secondly, in the grassland areas protected from sheep grazing, mortality was higher where there was competition from herbs.

Growth

The leaf dry mass of Q. humilis seedlings was higher under Juniperus than under Buxus and than outside. This high above-ground biomass of seedlings under Juniperus may be related to the fact that full sunlight is not the optimal condition for growth or that resources are allocated mainly to above-ground parts of the plant, to the detriment of the roots (see Jarvis 1964 for Q. petraea). The high leaf dry mass in the absence of herbs (in the grassland no-competition treatment and in experiment C when canopy cover was removed) suggests that the decrease in light intensity under Juniperus canopy may indirectly improve Q. humilis growth by preventing competition with herbs. The difference in growth under the two species of shrub was not related to soil conditions, as there was no significant difference in leaf weight or plant height between plants growing under the removed halves of Buxus and Juniperus. The lower mean daylight under Buxus than Juniperus (9.8 compared to 15.2%) did, however, have strong adverse effects on growth. Jarvis (1964) found that the compensation point for Q. petraea was 5.9% of relative light intensity. The negative effect of light shortage under Buxus lasts until the seedling emerges above the canopy, as growth (measured by annual ring widths) and the biomass of above-ground organs of saplings were much lower under shrubs than on the grassland. Light did not, however, seem to be so limited as to threaten the survival of Q. humilis, as no dead plants (others than those whose mortality was attributable to rodents) were found under shrubs. Furthermore, the growth (in term of ring width) of oaks under shrubs was similar to the initial growth of individuals that had survived and subsequently emerged from the canopy.

The establishment of Q. humilis under Buxus takes place, therefore, under microclimatic conditions that are on the whole favourable in terms of germination and survival, but unfavourable in terms of growth. Oaks remained under the canopy of the shrubs for an average of 11 years. During this period, avoidance of grazing by sheep led to the overall effect of Buxus being positive. As Q. humilis is only associated with Buxus individuals that have reached a minimum size (Rousset & Lepart 1999), the protective effect of this shrub is not seen until it reaches about 20 years of age. The appearance of shrubs of this age constitutes a shift in the balance of the ecosystem toward oak facilitation. The protection provided to oaks against sheep grazing improves until the shrubs reach a height of 1.5 m. Beyond this stage there is no further improvement and the negative effects that cause seedlings to take a longer time to emerge above the canopy come into play. The strength of facilitative effects do not, therefore, increase linearly with the size of Buxus, in contrast to the results obtained for other benefactor plants (Kellman & Kading 1992; Pugnaire et al. 1996). Once the shrub canopy has been overtopped, the balance shifts again and shrubs no longer have any effect on Q. humilis: the growth of oak individuals above the shrub canopy is similar to those in open grassland and their survival no longer seems to be threatened by grazing.

This study shows long-term variations in positive and negative effects of a nurse plant on a beneficiary plant. It also shows that the degree of facilitation depends on the shrub species. Unlike Buxus, and at least for the first 2 years, growth (as well as survival) is favoured under Juniperus and reaching the shrub canopy-top should not, therefore, pose any problem for oaks growing under the latter species.

What becomes of the shrubs? effects of q. humilis on buxus and juniperus

Juniperus communis is a pioneer species that invades open land with poor soils (Vedel 1961; Ward 1973). It is not shade-tolerant (Grubb et al. 1996) and only survives in woods with a very open canopy (Clifton et al. 1997). It is therefore highly probable that as Q. humilis woodlands develop they would have a negative effect on the growth and survival of Juniperus. Such a negative effect of a beneficiary plant has been observed in the studies that have investigated their eventual impact on the nurse plant (McAuliffe 1984, 1986; Callaway 1992; Flores-Martinez et al. 1994).

The situation is different with Buxus, which appears to benefit from the continued presence of Q. humilis. The reproduction, growth and establishment of Buxus have been studied in a Q. humilis coppice (50 years since last cutting) and compared to adjacent grassland areas (Rousset 1999). The growth of adults was found to be similar in both habitat types. However, although investment in reproduction was higher on grassland than in forest habitats, Buxus seedling density was significantly lower in grassland (Rousset 1999). After Buxus has had a facilitatory effect on the establishment and survival of Q. humilis, adult oaks may then favour the establishment of Buxus. Such reciprocal facilitation effects could explain why the phytosociological association Querceto × buxetum (oak woodland with an understorey of Buxus) is a frequent climax vegetation type of successions on plateaux in southern France (Braun-Blanquet 1970).

Facilitation and succession

As all seedlings are subject to grazing on the grasslands of limestone plateaux in southern France, facilitation appears to be an important mechanism for the regeneration of Q. humilis. Such importance of positive effects during succession has rarely been demonstrated (see Chapin et al. 1994; Callaway 1995; Fastie 1995), but this study shows that facilitation mechanism has a considerable effect on the pathway and speed of succession. In habitats where grazing pressure remains strong, Q. humilis can establish in shrub-covered areas at rates that depend not only on the density of shrubs but also on which shrub species dominates. Establishment is rapid under Juniperus but slower under Buxus and the development of oak forest on the Causse du Larzac will, therefore, be influenced by the current status of the shrub vegetation.

A model, developed by Brooker & Callaghan (1998) predicted that, after a disturbance, the balance between facilitation and competition progressively shifts towards negative interactions. They deduced that in successions, positive effects dominate in the early stages (i.e., when the environmental conditions are harshest) and then become less important, while negative effects increase in the later stages. In our study sites, and in contrast to the prediction of Brooker & Callaghan’s (1998) model, facilitation remains an important mechanism late in succession when a dominant tree species of climax vegetation is establishing. This difference is due to the importance of the effect of protection against livestock. Spatial association of plants could therefore be important in determining the vegetation dynamics in the many grasslands that are grazed by livestock. In Europe, many such ecosystems are rapidly being invaded by woody plants and populations of wild mammals associated with woodland, such as red deer, are increasing. Browsing by deer is known to prevent Quercus regeneration (Griffin 1971) and, consequently, facilitation by the spatial association of plants is perhaps a mechanism that would act to limit damage to Quercus by wild herbivores.

Received 17 February 1999revisionaccepted 23 September 1999