Competition and facilitation structure plant communities under nurse tree canopies in extremely stressful environments

Abstract Nurse plant facilitation in stressful environments can produce an environment with relatively low stress under its canopy. These nurse plants may produce the conditions promoting intense competition between coexisting species under the canopy, and canopies may establish stress gradients, where stress increases toward the edge of the canopy. Competition and facilitation on these stress gradients may control species distributions in the communities under canopies. We tested the following predictions: (1) interactions between understory species shift from competition to facilitation in habitats experiencing increasing stress from the center to the edge of canopy of a nurse plant, and (2) species distributions in understory communities are controlled by competitive interactions at the center of canopy, and facilitation at the edge of the canopy. We tested these predictions using a neighbor removal experiment under nurse trees growing in arid environments. Established individuals of each of four of the most common herbaceous species in the understory were used in the experiment. Two species were more frequent in the center of the canopy, and two species were more frequent at the edge of the canopy. Established individuals of each species were subjected to neighbor removal or control treatments in both canopy center and edge habitats. We found a shift from competitive to facilitative interactions from the center to the edge of the canopy. The shift in the effect of neighbors on the target species can help to explain species distributions in these canopies. Canopy‐dominant species only perform well in the presence of neighbors in the edge microhabitat. Competition from canopy‐dominant species can also limit the performance of edge‐dominant species in the canopy microhabitat. The shift from competition to facilitation under nurse plant canopies can structure the understory communities in extremely stressful environments.

Interactions between individuals can also be positive (i.e., facilitation) through plants enhancing the establishment, growth, and survival of other plants (Callaway, 2007), and improving the microclimatic conditions and soil physical and chemical properties (Bonanomi, Incerti, & Mazzoleni, 2011;Brooker et al., 2008;Maestre et al., 2010). The interplay between positive and negative interactions between plants should be important in understanding species distributions, community structure, and ecosystem function (Goldberg, Rajaniemi, & Stewart-Oaten, 1999;Kikvidze et al., 2011).
The prediction that interactions should shift from negative to positive across gradients of increasing stress has been formalized as the stress gradient hypothesis-SGH (Bertness & Callaway, 1994).
Arid and semi-arid lands are unproductive and stressful for the growth of plants. Nurse plants play an important role in ameliorating the environmental conditions and creating more benign microhabitats for the understory species in arid habitats (Abdallah & Chaieb, 2012;Anthelme & Michalet, 2009;Cortina et al., 2011). Positive interactions, or the nurse plant syndrome, tend to increase with increasing aridity (Flores & Jurado, 2003). The existence of nurse trees in arid environments creates a complex network of direct and indirect interactions between the nurse plant and the understory species around the nurse plant (see Cuesta, Villar-Salvador, Puértolas, Rey Benayas, & Michalet, 2010;Michalet, Brooker, Lortie, Maalouf, & Pugnaire, 2015;Schöb, Armas, & Pugnaire, 2013). These interactions could occur between guilds (between the nurse tree and the understory species) or could occur within guilds (between the understory species; Weedon & Facelli, 2008). The extreme environment around the canopies of nurse plants in arid environments presents heterogeneity at the microhabitat scale (Gomez-Aparicio et al. 2005). In this situation, interactions between species in communities under the nurse tree canopies may shift from facilitation at the stressful edge of the canopy to competition at the most benign center of the canopy. Stress gradients controlling biotic interactions in the small microhabitats under the canopy of nurse plants in arid environments may be fundamentally important in determining community composition and species distributions. As the nature of interactions between plants potentially changes across remarkably short environmental gradients, detecting competition or facilitation in these habitats may be highly dependent on where these communities are sampled. This effect may help to explain the inconsistent results of studies testing the stress gradient hypothesis.
The predictions of the net outcome of plant-plant interactions are usually based studies of pairwise species interactions (see Gómez-Aparicio et al., 2004;He, Bertness, & Altieri, 2013;Maestre & Cortina, 2005). However, in natural communities, species interactions are not pairwise. Rather, species interact through complex multiple species interactions (i.e., neighborhood competition-see Keddy, 2001).

Negative interactions between beneficiary plants under nurse plants
have been observed in only a few studies (Schöb et al., 2013), and experimental evidence of competition between beneficiary plants is rare (e.g., Aguiar & Sala, 1994;Michalet et al., 2015). No studies have examined the potential for small-scale stress gradients to allow plantplant interactions to structure small plant communities in different microhabitats in extremely stressful landscapes.
We examined facilitation and competition under nurse trees in western Saudi Arabia. High daily temperatures, low soil moisture, and nutrient-impoverished soil make much of this region extremely stressful and unproductive. However, between-guild facilitation that occurs between the nurse tree (Acacia gerrardii) and understory vegetation creates productive microhabitats for understory vegetation. In a related study, we demonstrated that the under canopy microhabitats have high soil nutrients and water availability, and low light intensity and UV radiation relative to the surrounding areas (Al Namazi and Bonser, unpublished). In environments with A. gerrardii nurse trees, there are microhabitats differing in abiotic stress: relatively low-stress habitats under canopies and relatively high-stress habitats at the edge of canopies. We emphasize that these environments are quite stressful relative to mesic or temperate ecosystems. The distribution of herbaceous species in these arid environments tends to be habitat dependent. Some species are present at relatively high frequency under the canopy while others are present at relatively high frequency in the more stressful edge, and some even occur primarily in the open habitats (Al Namazi & Bonser, unpublished). These observations suggest that the distribution of species under the canopy of nurse plants could be controlled by increasing within-guild competitive interactions on a gradient of decreasing abiotic stress toward the center of the canopy. Intriguingly, competition could play a major role in structuring communities in these extremely stressful habitats.
We conducted a neighbor removal experiment in the low-stress canopy center microhabitats and high-stress canopy edge microhabitats produced by nurse trees on four common perennial herbaceous species with different distributions under the canopies. We tested the following predictions: (1) interactions between understory species will shift from competition to facilitation in habitats of increasing stress from the center to the edge of A. gerrardii canopies, and (2) under canopies, species distributions will be limited by within-guild competitive interactions at the center of the canopy, and facilitation at the edge of the canopy.

| Study site
The study was conducted in Sederah Natural reserve in the National However, during the experiment period (March, April, and May), the reserve received rainfall of 12.5, 21.1, and 11 mm in each month, respectively, and plant growth was sustained throughout the spring season despite the short rainy period.

| Microclimate data
Soil temperature was 37°C under the canopy compared to 47°C immediately outside the canopy during the hours around midday.
Mean photosynthetically active radiation (PAR) was 1930 ± 30 μmol/ m 2 s −1 outside the canopy (and 1700 μmol/m 2 s −1 at the canopy edge) compared with 211 ± 14 μmol/m 2 s −1 under the canopy (Al-Namazi and Bonser, in review). We also sampled the soil from canopy center and edge microhabitats. Two soil samples from the canopy center and edge microhabitats under each of the ten trees were collected from the upper 10 cm of soil and combined. Soil samples were airdried, homogenized, and sieved. Chemical composition of the soil was analysed by the department of soil science at King Saud University, Riyadh, Saudi Arabia.

| Study species
Trees in the genus Acacia are diverse and important in arid and semiarid ecosystems of the world (Ross 1981). Acacia trees contribute to increasing the productivity and diversity of understory plants that grow under their canopies (Abdallah & Chaieb, 2010Belsky 1994;Ludwig et al. 2003). Acacia gerrardii (Benth.) is one of the most common trees in the arid and semi-arid environments in Saudi Arabia.
We measured the mean abundance (the number of individuals per species in 1-m 2 quadrates) for each species in the understory, and the community density (the total number of individuals in 1-m 2 quadrates) under A. gerrardii canopies. Abundance and density were measured at the canopy center and edge microhabitats under each of the ten A. gerrardii canopies. Individuals of each species are found at both the center of the canopy and at the canopy edge, but these species are distributed differently under the nurse plant canopies (see Figure 1). S. aegyptiaca is an under canopy specialist (it is the dominant species under the canopy), F. indica is moderate canopy specialist (found more frequently under the canopy than at the canopy edge), F. aegeptia is a moderate edge specialist (it is found more frequently at the edge of the canopy than at the center of the canopy), and I. spinosa is an edge specialist (it is the dominant species at the edge of the canopy). I. spinosa is also a nitrogen fixer (Sprent & Sprent, 1990). Species were either absent outside the canopy or were present in small numbers.

| Experimental design
We conducted a neighbor removal experiment to test our predictions leaving the roots of the target plant intact. This is a common approach for neighbor removal experiments (see Aarssen & Epp, 1990). Further, the remaining root tissue would not likely significantly decompose and increase soil nutrients over the time frame of the experiment (see Lamb, Kembel, & Cahill, 2009). Competition treatments were replicated 10 times at each microhabitat for each of the four experimental species.
Target plants were harvested, and the experiment was completed at the end of May 2013. Our experiment ran over the primary growing season for the year in these habitats, as most species have only limited growth in the summer season, and individuals either remain dormant or die during the harsh dry season after the short rainy period.

| Data collection
We examined the impact of neighbors on the growth of target spe-

| Data analysis
To compare competition and facilitation effects across the stress gradient under the canopy, we used the index of relative neighbor effect (RNE, Markham & Chanway, 1996) where Dry Mass t1 is the dry mass at the beginning of the experiment, and Dry Mass t2 is the dry mass at the end of the experiment. We used mixed-model analysis of variance to examine variation in target plant growth due to species, microhabitat, and competition treatment. The species effect (and interaction terms including species) was included as a random effect while microhabitat and competition were fixed effects. Significant differences in mean growth values were assessed using Tukey's HSD. Significant differences between target species' abundance and soil characteristics were assessed using two-sample general linear models. Data were analysed with SPSS 16.0.

| RESULTS
Plants in the canopy edge microhabitat experience significantly higher abiotic stress than plants in the canopy center microhabitat. The edge microhabitat habitat had higher daytime temperatures and light intensity (see Methods) than the canopy center microhabitat. Further, the soil at the edge microhabitat had significantly lower magnesium, potassium, and sulfate than the canopy center microhabitat (Table 1).
Community density was higher at the canopy center (10.4 ± 1.01 individuals per m 2 ) than at the canopy edge (8.3 ± 0.08 individuals per m 2 ).
We found a significant effect of neighbor removal on the growth of target plants. However, the effect of neighbor removal was highly species and habitat dependent (i.e., a significant three-way interaction between these effects, Table 2). In addition, we found significant species and habitat effects on growth, and interactions between the main effects were also significant, with the exception of species × habitat and species × neighbor removal ( Table 2)  T A B L E 2 Mixed-model analysis of variance results on the impact of species, canopy position, and neighbor removal on plant growth. Species (and interaction terms including species) were included as random effects  (Table 2). For example, growth in F. indica and S. aegyptiaca (the two canopy specialist species) was significantly higher in the neighbor removal treatment at the center of the canopy. At the edge of the canopy, the growth of these two species was significantly higher in the presence of neighbors than that in the absence of neighbors (Figure 3). In contrast, the growth of F. aegyptia and I. spinosa (the two canopy edge species) remains higher in the absence of neighbors in the two microhabitats although the growth of these two species was not consistent between two microhabitats. The growth of F. aegyptia decreased between center and edge habitats in both neighbor removal treatments, and was significantly greater at the center of the canopy than at the canopy edge in neighbor removal treatments. The growth of I. spinosa increased slightly but not significantly between center and edge habitats in the neighbors intact treatment, but not in the neighbor removal treatment (Figure 3).

Source of variation
While F. indica and S. aegyptiaca were more abundant in the canopy center microhabitat, the growth of each target species was reduced by neighbors in this microhabitat, and this effect was relatively consistent across species (Figures 2, 3).

| DISCUSSION
We found complex interactions among beneficiary species that grow under a nurse plant (Acacia gerrardii). These within-guild interactions shift from competition to facilitation with increasing stress from the center to the edge of nurse tree canopies. Although our finding was observed over very small stress gradients under nurse trees, this observation supports the stress gradient hypothesis frequently observed across broad environmental gradients (Bertness & Callaway, 1994). However, the stress gradients determining the nature of species interactions occur at remarkably fine scales within these stressful arid environments, rather than at landscapes scales where the stress gradient hypothesis is typically tested (Sthultz, Gehring, & Whitham, 2007;Tewksbury & Lloyd, 2001).
In arid regions of the Arabian peninsula, Acacia trees represent "islands of fertility" with higher soil moisture and nutrient contents under the canopies than the surrounding harsh environments (Abdallah & Chaieb, 2010;Robinson, 2004). We found that the stress gradient generated by A. gerrardii canopies plays a significant role in structuring the understory herbaceous communities. On average, the intensity of competition experienced by the four target species under canopy (the less stressful microhabitat) was greater than that of their counterparts growing at the edge of the canopy (the more stressful microhabitat) of A. gerrardii. These results are consistent with the predictions strategy theory that the intensity of competition increases with decreasing the stress of habitat (Grime, 1977).
However, decreasing competitive interactions with increasing stress may not simply be due to less intense competition between plants in more stressful environments. Rather, our results suggest that the strong positive effects of neighbors in these stressful environments can outweigh the negative effects of ongoing resource competition. In the year we conducted our experiment, many plants had not started reproducing by the onset of the dry season (the wet season was short compared to other years, and reproduction may have been lower than average years). Alternately, persistence across the dry seasons could also increase species abundance in these habitats. Future studies will need to investigate differences in reproduction under competition, and in the capacity for understory herbaceous species to survive the long dry seasons in these regions to better understand the dominance of some species in the low-stress microhabitats.
At the canopy edge, abiotic stress increases its importance in controlling species abundance, but a primary impact of abiotic stress appears to be in modifying biotic interactions. For example, species more abundant in the canopy center microhabitat grew very poorly in the absence of neighboring plants in the edge microhabitat. These species appear to have low stress tolerance, and facilitation from neighboring herbaceous vegetation can promote their persistence in the higher stress canopy edge microhabitat. In contrast, the two edge specialist species (I. spinosa and F. aegyptia) did not rely on facilitation in the more stressful microhabitat. In particular, growth of I. spinosa did not differ across neighbor removal treatments or canopy position, which is consistent with a stress-tolerant strategy (Grime, 1977 The presence of a strong facilitator or nurse species in an extremely stressful environment establishes environmental conditions that promote competitive interactions and limit the impact and importance of facilitation, even at the most stressful ends of broad geographic stress gradients. The modest and sometimes equivocal support of the stress gradient hypothesis is potentially due to either a breakdown of facilitation in the highest stress environments or a switch from facilitation to competition when resources are most limited (Michalet et al., 2014).
Our results suggest that facilitation is central to community structure in high-stress environments. Further, we did not find evidence that competition will increase in importance as resources become critically limited-although the understory habitats in our study may not have crossed a threshold for resource limitation required to induce competition. Alternately, the lack of general support for the SGH could be due to neglecting the variation in life-history and ecological strategies of interacting species and the nature of the stress (Maestre et al., 2006(Maestre et al., , 2009. We demonstrate that detecting competition and In conclusion, our results show that the distribution of herbaceous species under the canopy of nurse plant in the arid environment is controlled by a complex interplay between the abiotic stresses established by the nurse trees, and the interactions between coexisting species. The nature of interacting species (i.e., competitive ability and stress tolerance) likely controls the outcome of these interactions. In the low-stress microhabitat, the dominant species should have high competitive ability. However, facilitation by stress-tolerant species and the capacity to tolerate environmental stress control the persistence of species in the more stressful canopy edge habitats. Overall, competition and facilitation play key roles in the distribution of species and the assembly of communities under nurse trees in these extremely stressful habitats. Our results are important in understanding how competition and facilitation control community assembly on stress gradients.

ACKNOWLEDGMENTS
We thank Mr. Mohammad Basharat for the help in the field experiment, and J. Cahill, S. Solivares and J. Facelli for and several anonymous reviewers for comments on previous versions of this manuscript. We thank also King Abdulaziz City for Science and Technology