Study area

The experiment was performed in the riparian forests of the Natural Reserve of la Platière, located in the Middle Rhône floodplain, near Sablons, France (45°19′N, 4°46′E). Mean annual temperature (1970–2005) is 12.3°C and mean annual precipitation (1960–2005) is 799 mm (Meteo-France, Sablons record station). The riparian forests are located on modern alluvial deposits of a braided river. On this section, the river flow has been regulated since the 19th century by a dyke upstream from the study site (Girardon rack) and more recently by the deviation of water to a hydroelectric channel. The soils are gray alluvial soils with a sandy to a loam-sandy texture and are poor in humus (Pont, 1999). Three forest community types occur in the study area, with increasing distance to the river flow. Salix–Populus stands (hereafter called Salix stands) occur on old Girardon racks zones close to the riverbank and are yearly flooded. Acer stands replace the Salix stands on older terraces along the riparian profile, are frequently flooded and characterized by the presence of some tall and old Populus individuals. The oldest terraces, rarely flooded, are dominated by Fraxinus stands.

Studied species

We used transplants of the four dominant species of these riparian forests, two early-successional species, P. alba L. and S. alba L., the late-successional species F. angustifolia (Vahl) (Rameau et al., 1997) and the invasive A. negundo L. We used 1-yr-old seedling individuals because we wanted to use transplantable individuals and to avoid germination problems, and because the seedling stage is determinant for the structure of floodplain forests (Richardson et al., 2007). More precisely, we used 1-yr-old bare-rooted seedlings for both Fraxinus and Acer and cuttings from 1-yr-old seedlings for Populus and Salix in order to standardize the initial size of target individuals and to integrate the importance of the vegetative reproduction strategy for these species (Karrenberg et al., 2003). Seedlings were grown for 1 yr in local nurseries and were stemmed from seeds harvested in neighboring forests. At the end of the experiment, all Acer seedlings were uprooted to avoid any propagule dispersion. Regular visual inspection during the following years had confirmed that no Acer transplant had resprouted and survived.

Experimental design

The natural reserve authorities gave permission for the experiment set-up (Fig. 1) because our transplantations were established within the natural range of propagule dispersal (the invasive Acer and other native species are already present at the mature stage in the landscape). We used a completely randomized split-split-split-plot design (Underwood, 1997), with community as main effect, tree canopy as second effect, herb layer (including grasses, forbs and small woody plants from the low shrub layer) as third effect and target species (species treatment) as fourth effect. The community treatment included the three successional forest types already described, that is, the Salix, Acer and Fraxinus communities, with four replicates per community type. The area of each of the 12 sites (three communities × four replicates) was 1 ha and the sites were randomly located throughout the reserve. The tree canopy treatment was applied in each site, by delimiting a 20 m × 20 m plot in an experimental gap and another in the adjacent unmanipulated forest. The experimental gaps were created in late November 2003 to late January 2004 by removing all trees in a 40 × 40 m area. We removed tree regrowth during the whole experiment. Each of the 24 plots (three communities × four replicates × two plots (one gap + one forest)) was divided into eight subplots of at least 3 m × 3 m. The herb layer treatment consisted of removing all grasses, forbs and small woody plants monthly from four of the eight subplots in each plot by hand-pulling up. As a result, each of the four biotic interactions combinations (with and without tree canopy and with and without herb layer) was repeated four times in each of the 12 sites and represented in each site by a group of four subplots randomly separated in space. Finally, in late March 2004, the species treatment was applied by the random transplanting of three individuals (seedlings or cuttings) of each target species in the center of each of the 192 subplots with at least 50 cm-distance between individuals.

Resource measurements

In late May 2005, at least 3 d after the last precipitation event, two soil samples were randomly collected 3–10 cm deep in each biotic interaction combination within each site. The two soil samples were pooled and immediately weighed. Both NO₃⁻ and NH₄⁺ were extracted as described in Wheatley et al. (1989). The NO₃⁻ was determined by ionic chromatography (4500i; Dionex, Sunnyvale, CA, USA) and the NH₄⁺ by the blue indophenol method (Dorich & Nelson, 1983). The available soil nitrogen (N) content was calculated as the sum of NH₄-N and NO₃-N, and expressed in N mass per unit of dry soil mass (g g⁻¹). Soil samples were oven-dried at 105°C for 2 d and soil moisture was expressed in percentage of fresh soil mass. In July 2005, the fraction of photosynthetically active radiation (PAR; 400–700 nm) reaching the ground was measured at noon during a sunny day, using a portable quantum sensor (Li-188b; Li-Cor, Lincoln, NE, USA). Three PAR measurements were collected per subplot and averaged. Light availability was expressed in percentage of the mean incident PAR occurring in the four subplots of the gap without herb layer of each site.

Transplant measurements

Survival of transplants was measured in November 2005, at the end of the second growing season, and specific survival rates were calculated as the percentage of transplants alive at the end of the experiment per biotic interaction combination and per site.

We measured the basal stem diameter and the total height of each transplant at the beginning, the middle (November 2004), and the end of the experiment. For Salix and Populus cuttings, we measured the basal diameter and total height of the greatest epicormic shoot of the first growing season and the diameter of cuttings was also recorded. In November 2005, for all four species, one of the three transplants per subplot was randomly chosen to meticulously uproot and measure total biomass. However, we did not collect from subplots with null specific survival rates and this affected sample size. After oven-drying at 105°C for 3 d, we measured the total dry biomass of each individual. After different statistical tests, we found that the best allometric relation between biomass and field measurements (basal diameter, total height and cutting diameter for Salix and Populus) of each species was represented by the power function:

(B, total dry weight of the individual (g); X, basal diameter of the individual (mm); a and b are estimated parameters). We linearized the model using a log–log transformation. Biomass and diameter are both subject to natural variation and measurement error, thus we used Reduced Major Axis regression (RMA, a model II regression) to estimate the parameters, using the s.slopes function written for implementation in R (http://eeb37.biosci.arizona.edu/~brian/splus.html; see Isobe et al., 1990). We used these four specific allometric equations (data not shown) to convert our field measures into biomass values. We used the relative biomass increment during the second year as indicator of transplant growth to avoid bias caused by the type of individuals (seedlings or cuttings), calculated as:

(RBI, relative biomass increment of each individual; B 2004 and B 2005, biomass in November 2004 and in November 2005, respectively).

Biotic interactions

Tree canopy and herb layer effects on transplant performances (survival and growth) were displayed using the Relative Interaction Index (RII; Armas et al., 2004):

(X, mean value of the transplant performance in each biotic interaction combination with (F) and without (f) tree canopy and with (H) and without (h) herb layer). The high mortality rate of Populus and Salix with tree canopy and/or herb layer did not allow us to analyse the effects of biotic interactions on their growth and impeded the calculation of some RIIs.

The RII values are symmetrical around zero and included between −1 and +1, with positive values indicating facilitation and negative values indicating competition. For the indirect effects of the tree canopy, we applied the calculation method of Pagès et al. (2003) to RII. Following Pagès et al. (2003), positive RII_{indirect tree} values indicate indirect facilitation by trees, but only in the absence of direct facilitation from the tree canopy, and negative RII_{indirect tree} values indicate additional competition.

Herb layer biomass and composition

In July 2005, two 1 × 1 m samples of the herbaceous aboveground biomass were randomly collected per plot. For each sample, the different species were separated, oven-dried at 105°C for 3 d and weighed. We then gathered the species in two different functional groups (see complete list in Table 1): an ‘exploitative’ functional group for species with high relative growth rate, size and competitive ability (e.g. Urtica dioica or Renoutria japonica) as examples of the C strategy of Grime (1979) and a ‘conservative’ functional group for woody species (e.g. trees seedlings and shrubs such as Crataegus laevigata) and short shade-tolerant herbaceous species (e.g. Hedera helix), which are slow-growers and bad competitors, as examples of the S strategy of Grime (1979).

Statistical analyses

Differences in species survival and growth rates were analysed using several ANOVA models on raw data (see the Supporting Information, Table S1). Overall differences in species survival rates were analysed using a split-split-split-plot model, with community as main plot effect, tree canopy as subplot effect, herb layer as sub-subplot effect and species as sub-sub-subplot effect. The effects of tree canopy and herb layer treatments on survival and growth rates were analysed per community and per species using a split-plot on randomized block model, with tree canopy as main plot effect and herb layer as subplot effect. We used the same split-plot on randomized block model, with tree canopy as main plot effect and functional group (exploitative vs conservative) as subplot effect to test differences in herb layer biomass. Moreover, we used one-way ANOVA model to test the effect of tree canopy on herb layer biomass within each community. Differences in resource levels (light availability, soil moisture and soil N content) were analysed using one-way ANOVA model with the four biotic interaction combinations (with/without tree canopy × with/without herb layer) as four modalities of the treatment. Finally, One-sample t-tests were used to test significant deviations from zero of RII values and ANOVAs were followed by Tukey’s HSD tests when necessary. Data sets were checked for normality and homoscedasticity and all analyses were carried out with jmp 5.0.1 (SAS Institute, Cary, NC, USA).

Survival responses

Overall, results of the global split-split-split-plot ANOVA on species survival rates showed highly significant species effects, as well as single factor and in interaction with other treatments (Table 2). Thus, we used the global ANOVA to analyse changes in species responses among species and communities and within-species-and-community analyses to more precisely explore the effects of the tree canopy and of the herb layer. Overall, transplants had higher survival rates in the Fraxinus community (including gap and forest conditions) than in the Acer and Salix ones (Tukey’s HSD tests for the three effect types: Fraxinus stands, a; Acer and Salix stands, b; Table 2). Along the successional gradient, the effect of tree canopy shifted from overall negative in the Fraxinus stands to overall positive in the Salix stands (Fig. 2a, Table 2). More precisely, Salix and Populus transplants had negative responses to the tree canopy in the Fraxinus and Acer stands, whereas the survival of the invasive Acer, and to a lesser extent that of Fraxinus, were facilitated by the tree canopy of Acer and Salix stands (Fig. 2a, Table 3). Herb layer had an overall negative effect on transplants survival rates (Table 2) but this effect decreased from the Salix stands to the Fraxinus ones (Community × Herbaceous in Table 2; Fig. 2c,d). Herb layer of the three communities significantly reduced the survival rates of Populus transplants (Table 3), in particular in the gaps (Fig. 2c). The same trend was observed for Salix survival rates (Fig. 2c), except in the Fraxinus stands (Table 3). By contrast, Acer and Fraxinus seedlings had negative responses to the herb layer only under the Salix canopy (Fig. 2d). This negative effect of the herb layer almost counterbalanced the direct positive effect of the tree canopy for these two species, as suggested by the only marginally significant indirect positive effect observed for Fraxinus seedlings in the Salix stands (Fig. 2b). Finally, there was a strong additional competition for the two early-successional species in the Fraxinus and Acer stands (Fig. 2b).

Table 3.   Results of the split-plot on randomized complete block ANOVAs for the effects of tree canopy, herb layer and their interaction on the transplants’ survival rates per species within each community

Growth responses

Because Salix and Populus had high mortality rates in the plots with tree canopy and/or with herb layer, the analysis of the effects of the tree canopy and of the herb layer on the growth rates of these species were only limited to one-sample t-tests on available RII. The direct effects of both the tree canopy and the herb layer were overall negative for the growth of the four transplanted species (Fig. 3a,c,d). More precisely, the growth of Fraxinus seedlings was only marginally altered by the tree canopy of the Fraxinus and Acer stands, whereas that of Acer seedlings was significantly negatively affected by the tree canopy of the Acer and Salix stands (Table 4, Fig. 3a). In the gaps the effect of the herb layer was strongly negative for all species, and in particular for Acer seedlings in Acer stands where this effect was highly significant (Table 4, Fig. 3c). Within the forests, the negative effects of the herb layer decreased and were still significant only for Acer in the Acer stands and for Fraxinus in the Salix stands (Fig. 3d). The only significant positive indirect tree canopy effect was observed in the Acer stands for their own seedlings, likely because the herb layer competition strongly decreased for this species from the gaps to under the forest (Fig. 3b).

Table 4.   Results of the split-plot on randomized complete block ANOVAs for the effects of tree canopy, herb layer and their interaction on the Fraxinus and Acer transplants’ growth rates per species within each community

Treatment effects on resources

In the three communities the lowest light availability was observed in the plots including both the tree canopy and the herb layer (Fig. 4). However, in the Acer stands, light availability was less reduced by the herb layer in the gaps than by the tree canopy, in contrast to the two other communities (within-community Tukey’s HSD tests, Fig. 4a). Moreover, light availability was higher under the herb layer of the gaps of the Acer stands than under those of the two other communities (one-way ANOVA and Tukey’s HSD test not shown). Differences in the amounts of soil resources (N content and moisture) compared among biotic interaction combinations were significant only in the Acer community where the herb layer of the gaps induced a decrease in soil resource compared with other biotic interaction combinations (Fig. 4b,c).

Changes in herb layer composition and biomass

There were no differences in total herb layer biomass among communities (Table 5) but there was a strong decrease in dominance of exploitative species from the Salix community to the Fraxinus community (Table 5, Fig. 5). In the Acer and Salix communities, where exploitative species were dominant, the tree canopy induced a strong decrease in their biomass, whereas there were no significant changes for conservative species (Table 5, Fig. 5). There was a significant Community × Tree canopy × Functional composition interaction (Table 5) because changes in understory functional composition from gaps to forest plots were different among communities. In the Salix community, there was an exclusive dominance of exploitative species and their biomass was two times greater in gaps than in forests (Fig. 5). In the Fraxinus community, exploitative and conservative species codominated the herb layer and there was no difference between gap and forest conditions. By contrast, in the Acer community, the functional composition of the gaps was very similar to that of the Salix community (strong dominance of exploitative species), whereas the functional composition of the forest plots was rather similar to that of the Fraxinus community (codominance of both functional groups, Fig. 5).

Acer seedlings’ responses to native communities

The most significant result of our experiment was the strong increase in survival observed for Acer seedlings below the Salix canopy. Across all target species there was a dramatic shift in direct canopy–seedling interactions occurring for survival along the successional gradient, from positive tree canopy effects in the pioneer Salix community to negative effects in the late successional Fraxinus community. Such shifts in interactions during primary succession have been described in many systems (see for example Chapin et al., 1994 for glacier chronosequences) and constitute the conceptual framework of the autogenic successional model of Clements (1916), as well as that of current facilitation models (Bertness & Callaway, 1994). Facilitation is thought to prevail at the early stages of succession where environmental severity (physical disturbance and stress, sensuGrime, 1979) is the highest. Acer was strongly facilitated in the pioneer Salix community and to a lesser extent in its own community, but not in the late-successional Fraxinus community, which is consistent with general facilitation theory; in severe environments resident species may facilitate rather than compete with invaders (Bruno et al., 2003; Badano et al., 2007). Our environmental measurements did not show any significant variation among communities in the tree canopy effects on both available N and soil moisture likely to suggest a mitigation of stressful conditions. In our study, as in marine ecosystems (Ceccherelli & Cinelli, 1999; Ruesink, 2007), the driving mechanism of facilitation might have been the mitigation of physical disturbance, and in particular the reduction of flooding disturbance by the web of interlocking stems in the Salix community (field observations). However, other mechanisms such as the mitigation of flooding stress (waterlogging in spring) might also be involved, as argued by Von Holle (2005) for the invasion of an Appalachian floodplain community.

Only Acer and marginally Fraxinus were facilitated by the Salix canopy and only Salix and Populus were negatively affected by the Fraxinus canopy. Species responses to the effects of neighbors are known to be highly dependent on their functional strategies (Michalet et al., 2006). Liancourt et al. (2005) have shown that the facilitative response of a target species is positively related to its competitive response and negatively related to its stress-tolerance ability. They also demonstrated that for the same species, facilitative responses were the most frequent in the most stressed part of its ecological niche and competitive responses in relatively benign environmental conditions (see also Saccone et al., 2009; Forey et al., 2010). Following these conclusions, the two pioneer and shade-intolerant species Salix and Populus (Niinemets & Valladares, 2006) were the poorest candidates for facilitative responses because of their very low competitive abilities, as demonstrated by their strong negative responses to the tree canopy in the highly competitive environment of the shady Fraxinus community. The absence of a significant facilitative response for Fraxinus even in the Salix community is not as easy to understand because this shade-tolerant species (Niinemets & Valladares, 2006; Saccone et al., 2009) had a higher competitive response than Acer in the most shady Fraxinus community. If the cost of shade was similar for both species, we suggest that facilitation was observed only for the latter because the benefit of the adult Salix neighborhood was certainly higher than for the former. Our results for seedling survival rates in the Salix plots without tree canopies and without the herb layer showed that Acer was more negatively affected by flooding disturbance than Fraxinus (0.08 ± 0.1 and 0.33 ± 0.1 for Acer and Fraxinus, respectively). Consistent with Liancourt et al. (2005), because the abiotic environment of the Salix community was certainly harsher for Acer than for Fraxinus, the benefit of the Salix neighborhood was higher for the former, which may explain why facilitation was higher for this species.

In opposition to the tree canopy, the effect of the herb layer on seedling’s survival rates ranged from null in the Fraxinus community to negative and affecting Acer and Fraxinus seedlings in the Salix community and the gap of Acer stands. This result is consistent with the concept of biotic containment developed by Levine et al. (2004). The competitive effect of the herb layer seems to have counterbalanced the facilitative effect of the tree canopy, thus impeding the colonization of the Salix understory by Acer seedlings. We suggest that there was rather a biotic containment than a real biotic resistance, because the outcome of the opposite effects of the trees and the herb layer was positive (Fig. 2b), although not significant, likely because of the high spatial variability of the cover of herb layer in the Salix stands (P. Saccone, pers. obs.).

Tree seedlings had also globally negative growth responses to the tree canopy and the herb layer. The mortality rates of Salix and Populus seedlings did not allow us to analyse their growth responses and we could only compare Acer and Fraxinus seedling responses. The growth of the native late-successional Fraxinus was overall negatively affected by the tree and herb layer neighborhood, but less than that of the invasive Acer. Acer seedlings were particularly affected in the gaps of the Acer stands where the effect of the herb layer was stronger than in other plots.

Effect of adult Acer on native communities

The direct effects of adult Acer on survival and growth of target tree seedlings were intermediate between those of Salix and Fraxinus. By contrast, Reinhart et al. (2005) and Gomez-Aparicio et al. (2008) have shown that Acer platanoides, a European maple invasive in North America, suppressed native species and directly facilitated its own seedlings through understory and soil modifications, respectively. Such direct negative effects on native species were not observed in our experiment.

However, we found a significant indirect facilitation of the adult Acer tree canopy for the growth of its own seedlings. Although this effect was weaker than direct canopy effects, because it occurs only for this species, it is worthwhile addressing its potential role in the establishment success of Acer. Furthermore, weak indirect interactions might have strong community and ecosystem consequences (Berlow, 1999), and their role has been strongly neglected in invasion ecology (White et al., 2006). To our knowledge, there is one study which has demonstrated that an invasive tree species, Sapium sebiferum, indirectly facilitates its own seedlings to the detriment of competitive prairie grasses (Siemann & Rogers, 2003).

The occurrence of indirect facilitation only for the growth of Acer and only in its own community may be explained by the results of previous indirect facilitation studies (Levine, 1999; Pagès & Michalet, 2003; Pagès et al., 2003; Siemann & Rogers, 2003). The indirect facilitation model assumes that the direct negative effect of adult trees or shrubs on herbaceous competitors provides an indirect positive effect on tree seedlings which overrides the direct negative effect of the trees or shrubs on the seedlings (Pagès et al., 2003). Levine (1999) argued that this may occur only if the different pairs of competitors involved in the system compete for different resources or have different mechanisms to acquire resources. Indeed, Siemann & Rogers (2003) showed that the seedlings of the invasive S. sebiferum were indirectly facilitated by the canopy of adult S. sebiferum through decreasing root competition between prairie grasses and the target seedlings in the shade. In addition, Pagès et al. (2003) stressed the importance of the functional strategy of the target seedlings; in a subalpine coniferous forest of the French Alps, they demonstrated that the shade-tolerant and nutrient-demanding species Acer pseudoplatanus was a much better candidate for indirect facilitation when competing with forest weeds than shade-intolerant and nutrient conservative conifers.

In the gaps of the Acer stands the herb layer (dominated by two well-known exploitative competitors, Parietaria officinale and Urtica dioica, Grime, 1974) had strong negative effects on tree seedlings and in particular those of Acer. This negative herb effect was certainly caused by root competition, as suggested by our environmental measurements (significant decrease in N and water but not in light in these plots). In the Acer stands only, the tree canopy induced a shift in the herb layer functional composition with a strong decrease in exploitative species, which may have reduced root competition with tree seedlings. Thus, by reducing the belowground competitive effect of the herb layer, adult Acer indirectly facilitated the growth of its own seedlings, likely because this effect was not overwhelmed by the direct negative effect on light. This occurred for Acer seedlings but not for Fraxinus ones likely because the former were more sensitive to the benefit of this competitive release than the latter, as suggested by their differences in responses to the competitive effect of the herbaceous layer in the gaps of the Acer stands. In other words, among the four transplanted species Acer was the best candidate for indirect facilitation because of its high shade-tolerance associated with an exploitative strategy, consistent with the conclusion of Pagès et al. (2003) for A. pseudoplatanus.

The biotic drivers of invasion through succession

Our results stressed the importance of positive interactions for the invasion of A. negundo in the Middle Rhône floodplain, with a shift from direct to indirect facilitation through succession. We showed that direct facilitation by the tree canopy of the native pioneer Salix is the main biotic process allowing the colonization of the invasive Acer in the floodplain. Later in the invasion process, indirect facilitation by adult conspecifics is certainly the other biotic driver of invasion allowing the establishment of a population. These positive responses and effects were observed for the invasive but not for the natives, likely because this species had a particular combination of traits that do not have natives. In contrast to a number of other invasives, the persistence through time of Acer cannot be explained by higher direct competitive effects than native late-successional species. Our study adds new evidence for the importance of positive direct and indirect interactions in invasion ecology and how their roles may change along a successional gradient.