1We tested the hypothesis that the more frequent occurrence of tree seedlings below the adult trees than in canopy openings might be explained by indirect facilitation. In a temperate hardwood forest, we compared the performance of five target tree seedlings (Picea abies, Abies alba, Fagus sylvatica, Acer pseudoplatanus and Quercus petraea), transplanted with or without a herbaceous competitor (Molinia caerulea), either within the forest or into experimentally created gaps.
2We quantified changes in understorey biomass, light penetration and available forms of soil nitrogen during three growing seasons.
3Photosynthetic photon flux density and total biomass of Molinia were significantly higher in the gap treatment than within the forest. Total available nitrogen was higher in the gaps in the absence of Molinia, but higher in the forest in the presence of Molinia.
4Quercus survival was very low within the forest because of fungal infection, whereas survival was very high for the four other tree species in all combinations of the two treatments.
5Although the competitive effect of Molinia on the growth of the tree seedlings was much greater in the gap treatment, seedling growth was lower within the forest. We conclude that the tree canopy imposed strong light competition, and that this direct negative influence was much greater than any indirect positive effect of increased availability of nutrients to tree seedlings, due to reduced nutrient uptake by Molinia.
6Target species responses to treatments were similar, despite strong differences in nitrogen requirements between species. This may be due to the overwhelming negative influence of the tree canopy in our experiment.
Direct interactions between plant species may be altered by indirect interactions involving additional species, i.e. when the strength or direction of the interactions between a pair of species can be modified by other species (Connell 1990). Understanding such indirect interactions may help us to understand the structure and diversity of plant communities (Miller 1994; Callaway & Walker 1997), and a number of authors have explored the consequences of indirect facilitation through modelling (Lawlor 1979; Levine 1976; Vandermeer 1990; Stone & Roberts 1991), although few experimental studies have examined indirect interactions in the field (Callaway et al. 2001; Pennings & Callaway 1996; Levine 1999; Tielbörger & Kadmon 2000). Miller (1994) found that a strong competitor, Ambrosia artemisiifolia, had indirect positive effects on two poor competitor species, Trifolium repens and Chenopodium album, by suppressing two moderately competitive species, Agropyron repens and Plantago lanceolata. However, the net effect of A. artemisiifolia on T. repens and C. album was neutral because the indirect positive effects were balanced by its direct negative effects. In a Californian salt marsh, the competitive dominant species Salicornia virginica was suppressed by a parasitic plant, Cuscuta salina, which thus indirectly facilitated Limonium californicum and Frankenia salina (Pennings & Callaway 1996).
The mechanisms involved in the model of indirect facilitation have been clarified by Levine (1999). He argued that indirect facilitation might occur in a three-part system when different pairs species compete for different resources or have different mechanisms for acquiring these resources. For example, in forest communities, if tall herbaceous species compete with tree seedlings for nutrients, whereas adult trees compete with herbaceous species and tree seedlings mostly for light, adult trees may indirectly facilitate tree seedlings. The indirect facilitation model assumes that, for the tree seedlings, the negative effect of shading by the canopy of adult trees is smaller than the positive effect of increasing nutrient availability due to lower nutrient uptake by the herbaceous species (Levine 1999). Conversely, if the negative effect of the addition of the third component is higher than its indirect positive effect, there will be a net competitive effect, i.e. this component will impose additional competition (Pages et al. 2003). Indirect facilitation may explain the coexistence of shade-tolerant tree seedlings and understorey species within closed forest communities.
We tested the model of indirect facilitation in a mature deciduous forest by comparing the response of seedlings of five tree species to tree canopy removal in the presence and absence of a dominant herbaceous neighbour, purple moor-grass (Molinia caerulea (L.) Moench). In the northern French Alps, M. caerulea ssp. altissima (Link) Domin, rapidly develops a dense cover in sunny conditions (Dobremez 1970; Taylor et al. 2001) on soils with a fluctuating water-table (El-Kalhoum et al. 2000). This compactly tufted perennial grass has been shown to exclude heather species after nutrient fertilization (Aerts et al. 1991; Bobbink et al. 1998) and it has been suggested that this species may hinder the regeneration of tree species after forest clear-cutting (Becker 1969; Dobremez 1970; Von Lüpke 1998). Preliminary recording of tree seedling distributions indicated their more frequent occurrence within the forest than in clear-cut areas (Pages 2002, and observations of local foresters), suggesting that this is a good system for testing the indirect facilitation model.
We hypothesized that the direct negative effect of adult trees on the shade-intolerant M. caerulea may induce an indirect positive effect on tree seedlings and, if this indirect effect predominates, we will observe indirect facilitation. Two evergreen coniferous species (Picea abies (L.) Karsten and Abies alba Miller) and three broad-leaved deciduous angiosperms (Acer pseudoplatanus L., Fagus sylvatica L. and Quercus petraea (Mattuschka) Liebling) were used as target tree seedlings. Leaf nitrogen content and nutrient requirements of woody and herbaceous species are negatively correlated with leaf life span (Monk 1966; Chabot & Hicks 1982; Lambers & Poorters 1992; Huante et al. 1995; Reich et al. 1998). Given the large differences in leaf life span between our five target species (two evergreen vs. three deciduous), we hypothesized that differences in their nutrient requirements may determine their ability to respond to indirect interactions. Specifically, we hypothesized that the most nutrient-demanding tree species should be the best candidate for indirect facilitation, as it is likely to benefit most from the higher nutrient availability when herbaceous species are inhibited.
Materials and methods
The experimental site is located at 650 m a.s.l. in the northern French Alps, 80 km NW of Grenoble. The site is part of the national forest of Chambaran, which is exploited for firewood. This is a mature deciduous forest codominated by sessile oak (Quercus petraea), sweet chestnut (Castanea sativa Mill.) and European beech (Fagus sylvatica). The climate is characterized by mesic conditions with annual rainfall of 1000 mm and an average annual temperature of 10.3 °C. The study site is flat with a deep silty soil that is subjected to waterlogging in winter and at the beginning of spring (redoxisol, RPE 1995). The form of humus is an oligomull (RPE 1995) with a C : N ratio of 15 and a pH in water of 4.8 (Joud 1997). The herb layer in the understorey is dominated by Molinia caerulea ssp. altissima, which forms 50% ground cover vs. approximately 100% in the openings (Joud 1997); wavy hair-grass (Deschampsia flexuosa (L.) Trin.) and honeysuckle (Lonicera periclymenum L.) are the secondary dominant species, each with 5% cover.
experimental design and layout
A 2 × 2 × 5 factorial field experiment involving creation of canopy gaps vs. natural forest understorey, removal of the herbaceous neighbours vs. neighbours left intact, and transplantation of seedlings of five tree species was conducted over three growing seasons (2000–02) to test the hypothesis that the intensity of competition between M. caerulea and tree seedlings could be modified by the removal of the tree canopy, and that the responses will be species-specific. We randomly selected four sites (150 × 75 m) within the homogeneous forest community as replicates for a block effect; blocks were at least 100 metres apart. The four sites were surrounded with wire netting 1.5 m high to limit herbivory by roe deer. In November 1999, half of each block (75 × 75 m) was clear-cut, taking care to avoid disturbance of the soil and the understorey species. In the centre of each of these eight main plots (four forested and four clear-cut), we delimited eight 4 × 4 m subplots, separated by 3-m buffer zones. To minimize edge effects, each subplot was at least 25 m from the edge of the main plot. The competition treatment (removal of herbaceous neighbours) was applied to alternate subplots (Hurlbert 1984) in each main plot. Molinia and other neighbouring species were removed chemically by treating at ground level with glyphosate (Round Up, 360 mg L−1) at the beginning of the experiment and during the first growing season. Todd et al. (2000) have shown that glyphosate has significant negative effects on Molinia. During the second and third growing seasons, the regrowth of Molinia was clipped periodically and the dead biomass removed. The target seedlings were protected from spraying during each chemical treatment by enclosing them beneath upturned plastic pots.
All five target tree species are common in the understorey of the study site, but only Fagus sylvatica and Quercus petraea occur as dominants in the canopy layer at this elevation. Light and nutrient requirements of European tree species have rarely been studied experimentally; light requirements are, however, commonly inferred from distributions in the field: Abies and Fagus are thus considered the most shade-tolerant, Quercus is intermediate, and Picea and Acer are the least shade-tolerant species (Rameau et al. 1993; Brzeziecki & Kienast 1994; Kazda et al. 1998). The obvious differences in leaf life span between the evergreen conifers and the deciduous angiosperms suggest that they differ greatly in nutrient acquisition and conservation (Chabot & Hicks 1982; Monk 1966).
Four seedlings of each species were planted at random positions in a 50 × 50 cm grid in the centre of each subplot (four seedlings × eight subplots × two plots × four blocks = 256 seedlings per species). The tree seedlings, grown from seed from the northern French Alps, were obtained from a local nursery (Robin Pépinières, Saint Laurent de Cros, France). Picea and Abies were grown for 3 years in the nursery and 1 year in pots, but because of their higher growth rate Fagus, Quercus and Acer were grown for only 1 year in the nursery and 1 year in pots.
data collection and compilation
Species growth and survival
We measured stem basal diameter and total height of all seedlings at the beginning of the experiment, in early January 2000. Survivorship was determined at the end of the experiment, in late October 2002, and all seedlings were harvested for total biomass estimation. Because we could not measure initial biomass, we analysed the relationship between biomass and size for 61–65 extra seedlings of each species in January 2000. For each of the five species there was a highly significant correlation between total biomass and basal stem diameter (Pages 2002), but total biomass was not significantly correlated with seedling height. The regression formulae derived were used to transform initial basal stem diameters to total biomass. Growth was calculated for each individual as the proportional change in biomass during the course of the experiment: [(Biomass in October 2002) – (Biomass in January 2000)]× (Biomass in January 2000)−1. Proportional change values were averaged per species per subplot, prior to statistical analysis. Survival at the end of the experiment was expressed as a percentage per block, for each treatment (gap × competition combination) and each species.
Direct and indirect biotic effects
To analyse the competitive effect of Molinia on the five species we calculated the log response ratio (LRR) (Hedges et al. 1999): LRRMolinia= log [(performancewithout neighbours) × (performancewith neighbours)−1]. For each canopy treatment and each species, LRRMolinia for survival was calculated for each block, and LRRMolinia for growth was calculated for each pair of subplots (n= 4 per plot for each block). We also used LRR to determine the competitive effect of the adult tree canopy on the seedlings, comparing plots without herbaceous neighbours with or without canopy (LRRForest). To measure the indirect effect of the canopy on the tree seedlings (i.e. the alteration, due to canopy presence, of the competitive effect of the herbaceous neighbours on the tree seedlings), we calculated a log indirect response ratio (LIRR) for survival and growth: LIRR = log [(performance with neighbours in the gaps) × (performance with neighbours below the tree canopy)−1]. When the performance of the species competing with herbaceous neighbours is decreased by the presence of the tree canopy, LIRR is positive, indicating that the trees impose additional competition. LIRR is negative when the performance of the species competing with herbaceous neighbours is increased by the presence of the tree canopy, i.e. indirect facilitation, or in the case of direct facilitation, but only indirect facilitation will also have a positive LRRForest.
To analyse environmental differences between treatments during the course of the experiment, we collected soil samples for measurements of available forms of nitrogen, and measured light penetration. To estimate variations in understorey biomass, we measured above- and below-ground biomass of Molinia. In each combination of gap × competition treatment, eight replicates of soil samples (two per block) were collected at the beginning of the growing season in June 2002, and at the end of the growing season in October 2002. Soil samples were collected from a depth of 5–10 cm and then stored at 4 °C. Extractions of and were performed immediately and then soils were dried at 105 °C for 4 days, before reweighing to allow determination of water content.and, respectively, were extracted in water and in 1 m KCl (Wheatley et al. 1989).was determined by ionic chromatography (Dionex 4500i) andwith the indophenol blue method (Dorich & Nelson 1983). On a sunny day in September 2000, we measured photosynthetic photon flux density (PPFD in µmol photon m−2 s−1) at ground level within 1 h of solar noon (1400 French Standard Time). Sixteen replicate measurements were made per gap × competition treatment with a LI-COR (LI-188b) radiometer (LI-COR, Lincoln, Nebraska, USA). In September 2002 we measured above- and below-ground biomass of Molinia in 0.5 × 0.5 m quadrats located randomly within each canopy treatment of each block (four replicates). The below-ground biomass of Molinia was obtained by separating the roots from soil, by hand. Above-ground biomass was harvested at the same time. All plant material was dried for 7 days at 70 °C and weighed.
We measured total carbon and nitrogen content in two leaves or needles of 10 seedlings of each species with a CHONS micro analyser (Carlo Erba 1500).
For survival, we tested gap and species effects and their interactions using chi-square likelihood ratios and two separate logistic regression models (SAS Institute, Cary, North Carolina, USA), one for the subplots without neighbours, and one for the subplots with neighbours. Using means of proportional changes in biomass per subplot as dependent variables, we conducted two separate three-way anovas (one for the subplots without neighbours, and one for the subplots with neighbours), with species and gap treatment as main effects, and block as a random factor (Sokal & Rohlf 1995). Differences in LRRMolinia for survival and growth were investigated using three-way anovas, with species and gap treatment as main factors and block as a random factor. Differences in LRRForest and LIRR for survival and growth were analysed with two-way anovas, with species as a main factor and block as a random factor. Shapiro-Wilk W-tests prior to analyses showed that proportional changes in biomass met the assumptions of normality. One-way anovas were used to analyse changes in above- and below-ground biomass of Molinia due to the gap treatment and differences between species in C : N ratio. Environmental variables, including light penetration and available forms of nitrogen, were also analysed with two-way anovas with gap and competition treatments as main factors. All anovas were conducted with SPSS (1997) and posthoc multiple means comparisons were performed using Tukey's honestly significant difference tests (HSD).
treatment effects and species traits
Tree canopy alone reduced PPFD by 97%, whereas Molinia reduced it by 84% in the gaps (Fig. 1a). Although the effect of Molinia was much lower within the forest than in the gaps (Fgap × competition = 569.0, d.f. = 1, P < 0.001), the lowest light level was found with herbaceous neighbours within the forest. Total biomass of Molinia was 1800% greater (Fgap = 111.7, d.f. = 1, P < 0.0001) in gaps than under the tree canopy, and this effect was stronger for above-ground biomass than for below-ground biomass (Fig. 1b). Coniferous species had higher C : N ratio in needles than angiosperms had in their leaves (Fig. 2). Differences in C : N ratio between angiosperms were significant, with Acer having the lowest ratio and Fagus the highest, whereas C : N ratio did not differ between conifers (Fig. 2). There was less total available soil nitrogen within the forests when neighbours were removed, but less in the gaps in the presence of neighbours (Fig. 3). This pattern was observed in June 2002 and October 2002 but the gap–competition interaction was only significant in October 2002 (Fgap × competition = 142.8, d.f. = 1, P < 0.0001; Fig. 3).
seedling survival and growth
There was a high rate of survival for all seedlings in all combinations of the two treatments (Table 1), except for Quercus, which survived poorly within the forest (12.3 ± 4.1% without neighbours, = 312.4, d.f. = 1, P < 0.0001; 12.6 ± 4.5% with neighbours, = 240.8, d.f. = 1, P < 0.0001). The gap–species interaction was significant both without neighbours (= 73.1, d.f. = 4, P < 0.0001) and with neighbours (= 87.2, d.f. = 4, P < 0.0001). Quercus mortality was caused by a fungal infection (Oïdium sp.), which occurred only within the forest. Adult trees were also infected.
Table 1. Mean (± 1 SE, n= four blocks) percentage survival of the five species for each combination of the two treatments (gap × competition). Different superscripts indicate significantly different means within each column (Tukey post hoc comparisons, P < 0.05)
98.5a ± 1.5
98.5a ± 1.5
100.0a ± 0.0
98.5a ± 1.5
97.1a ± 1.7
95.5a ± 2.9
95.5a ± 1.5
97.1a ± 1.7
93.8a ± 6.2
98.5a ± 1.5
97.1a ± 1.7
100.0a ± 0.0
12.5b ± 4.4
12.5b ± 4.4
97.1a ± 1.7
95.5a ± 2.9
90.8a ± 4.1
94.0a ± 2.5
98.5a ± 1.5
95.3a ± 4.75
Quercus was excluded from growth analyses because of its low survivorship within the forest. Without herbaceous neighbours, there was no difference in proportional change in biomass between the four other tree species within the forest (Fig. 4), but proportional changes in biomass were far higher in gaps, and there were significant differences in species responses (Fgap × species = 15.1, d.f. = 3, P < 0.0001; Fig. 4), with the highest growth for Acer and Picea and the lowest for Abies. When herbaceous neighbours were present differences within the forest were again very small, and the gap treatment increased the differences (Fgap × species = 40.2, d.f. = 3, P < 0.0001; Fig. 4) but in this treatment Picea and Fagus showed greater proportional changes in biomass.
direct and indirect intersactions
For growth, the competitive effect of Molinia was close to zero within the forest, with no significant differences between species (Fig. 5a). In contrast, there was a direct competitive effect of Molinia on the growth of each of the four species in the gaps (Fgap = 66.0, d.f. = 1, P < 0.0001; Fig. 5a), in particular for Acer, the species with the highest leaf nitrogen content (Fgap × species = 5.2, d.f. = 3, P < 0.01). Without herbaceous neighbours, direct competition between the seedlings and adult trees (Fig. 5b) was approximately twice as high as direct competition with herbaceous neighbours in the gaps (Fig. 5a); Abies and Fagus had significantly higher LRRForest than Picea and Acer (Fspecies = 9.9, d.f. = 3, P < 0.0001).
For survival, relative additional competition intensity was high only for Quercus (LIRR = 0.96 ± 0.15), whereas the other species had LIRR values close to zero (Fspecies = 37.4, d.f. = 4, P < 0.0001). For growth, LIRR was positive for all four species, which indicates that, with herbaceous neighbours, growth was much higher in the gaps than in the forest. In other words, the net effect of competition from both adult trees and herbaceous neighbours was clearly negative for the seedlings of all five species. There were significant differences in LIRR between species (Fspecies = 35.9, d.f. = 3, P < 0.0001; Fig. 6), with the highest values for Abies and Fagus, and the lowest for Acer.
The indirect facilitation model (Levine 1976; Levine 1999) assumes that the direct negative effect of adult trees on understorey competitors induces an indirect positive effect on tree seedlings that exceeds the direct negative effect of adult trees on the seedlings. To test this model in a temperate hardwood forest, we compared the performance of seedlings of five tree species planted within the forest and in experimental gaps, with and without the herbaceous competitor Molinia caerulea. Competition between Molinia and tree seedlings was very weak within the forest and our gap treatment produced a strong increase in biomass for Molinia, which resulted in higher competition for the target species. However, despite the presence of Molinia, the performance of our target seedlings was higher in the gaps than within the forest, because the adult trees had a greater direct negative effect on tree seedling growth than did Molinia. Thus, adult trees increased competition for tree seedlings, because the direct negative effect of trees on the seedlings was larger than the indirect positive effect on seedlings due to the release of competition from Molinia.
Levine (1999) tested the indirect facilitation model within the canopy of sedge in a Californian riparian community; he manipulated the canopy of the sedge and compared the performance of three transplanted target species, with and without the competitor Mimulus guttatus. He found evidence of indirect facilitation for a liverwort species, but not for a moss and a forb. He argued that indirect facilitation occurred for the liverwort because the sedge inhibited Mimulus guttatus by shading, preventing it from interfering physically with the shade-tolerant liverwort, as a result of species competing for different resources, or having different mechanisms for acquisition of these resources.
In our experiment the removal of Molinia in the gaps was associated with more available nitrogen, an effect not observed within the forest. This probably indicates that, in the gaps, the increased biomass of Molinia, a fast-growing shade-intolerant species (Berendse & Aerts 1984; El-Kalhoum et al. 2000; Taylor et al. 2001), dramatically reduced nitrogen availability and thus increased competition between Molinia and tree seedlings. Although in the presence of Molinia, nitrogen availability was significantly higher within the forest than in the gaps, light was far less available for the tree seedlings within the forest than in the gaps. We can conclude that, for the tree seedlings beneath the canopy, the indirect positive effect of increasing nutrient availability associated with the lower nutrient uptake of the herbaceous species did not counterbalance the direct negative effect of shade induced by the canopy of the adult trees. This may explain why we observed strong additional competition instead of indirect facilitation or even a net zero effect of competition or facilitation.
The importance of additional competition found in our study may be due to the particular conditions of our experiment as, in more open forest communities, a greater light transmission may allow indirect positive effects to exceed direct negative effects. Li & Wilson (1998) have shown that contiguous distributions of woody species enhanced woody plant establishment in the mixed-grass prairie of North America & Peltzer & Köchy (2001) demonstrated that juvenile shrubs in the same system were less affected by adult shrubs than by grasses. As in our study, Li & Wilson (1998) recorded higher nitrogen availability below adult trees than in open prairie, and other studies conducted in open savanna forests have also shown that soils under trees tend to have more available nitrogen than soils under adjacent grassland (Callaway et al. 1991; Belsky 1994; Weltzin & McPherson 1999). However, light penetration beneath shrubs was reduced by only about 70% in the Peltzer & Köchy (2001) study compared with 97% within the forest in our experiment. More experimental studies are needed to understand the optimal conditions under which indirect facilitation may occur in forest communities. Additional studies may refine the parameters of the model, by defining the threshold of light transmission by canopy trees and the shade-tolerance levels of herbaceous competitors that will permit indirect facilitation. Another plausible explanation for the intense additional competition observed in our experiment could be an underestimation of the competitive effect of Molinia in the gaps; the 2–4-year-old tree seedlings were transplanted in winter, 3 months after the clear-cut, and although the gap treatment had a strong effect on the growth of Molinia from the first growing season, our experiment imitated only the advanced natural regeneration in gaps. We might have increased the asymmetry of competition (Weiner 1990), and thus the competitive effect of Molinia in the gaps, if we had transplanted 1-year-old seedlings (Berendse & Aerts 1984). We did not observe indirect facilitation in our experiment, but the greater abundance of tree seedlings within the forest than in the gaps in our site (Pages 2002, and observations of local foresters) suggests that this mechanism is nevertheless likely to occur in this site for the tree seedlings recruited after gap creation. Hubbell et al. (1999) have emphasized the importance of historical factors in determining tree species composition in forest communities that might also influence the outcome of competition experiments.
Our second hypothesis was that the most nutrient-demanding tree species, that with the highest nitrogen content in its leaves (Jones 1944; Brzeziecki & Kienast 1994; and see Monk 1966; Chabot & Hicks 1982; Reich et al. 1998), should be the best candidate for indirect facilitation because it would have the highest sensitivity to the indirect positive effect of greater nutrient availability below the canopy of adult trees. This hypothesis was confirmed, because the tree species that had the lowest LIRR values was Acer pseudoplatanus. However, differences in LIRR between species were weak and Picea had a lower LIRR value than Fagus, despite the lower nitrogen content of its needles. The overwhelming negative influence of the tree canopy on target seedlings may explain why we did not observe a clear species effect in this study.
We are very grateful to Didier Joud (Centre Régional de la Propriété Forestière) for his advice during the site selection. We thank the Office National des Forêts for permission to conduct this experiment and in particular Stéphane Nouguier and Jean-Paul Souchon who organized the gap treatment. This study was funded by the Région Rhône-Alpes (France), through forest typology studies. We thank Rob Brooker, Steve Jordan, Marie-Laure Mougey-Pagès, Irène Till and Patrick Saccone for their helpful comments on the manuscript, Geneviève Girard, Claire Pages, Benoit Betton, Caline Elsass and Julie Siegel for field and laboratory assistance.