Multiple mechanisms sustain a plant-animal facilitation on a coastal ecotone

Theory suggests that species distributions are expanded by positive species interactions, but the importance of facilitation in expanding species distributions at physiological range limits has not been widely recognized. We investigated the effects of the nurse shrub Tamarix chinensis on the crab Helice tientsinensis on the terrestrial borders of salt marshes, a typical coastal ecotone, where Tamarix and Helice were on their lower and upper elevational distribution edges, respectively. Crab burrows were abundant under Tamarix, but were absent in open areas between Tamarix. Removing Tamarix decreased associated crab burrows with time, while simulating Tamarix in open areas by shading, excluding predators, and adding Tamarix branches as crab food, increased crab burrows. Measurements of soil and microclimate factors showed that removing Tamarix increased abiotic stress, while simulating Tamarix by shading decreased abiotic stress. Survival of tethered crabs was high only when protected from desiccation and predation. Thus, by alleviating abiotic and biotic stresses, as well as by food provision, Tamarix expanded the upper intertidal distribution of Helice. Our study provides clear evidence for the importance of facilitation in expanding species distributions at their range limits, and suggests that facilitation is a crucial biological force maintaining the ecotones between ecosystems.

their interacting animals, few studies have disentangled the roles of food provision vs. non-trophic facilitation by plants on animals 24 .
We report two-year field experiments examining the interactions between the shrub Tamarix chinensis and the crab Helice tientsinensis on the terrestrial borders of two northern Chinese salt marshes (see Methods). These borders are a typical coastal ecotone 10 , which is flooded only during spring and storm tides, and are dry and hypersaline, leading to extreme thermal and desiccation stress for marine animals 9,16,17 . Tamarix is a nurse shrub distributed in coastal uplands without tides. Although Tamarix persists on the terrestrial borders, it is absent from lower marshes 9 . In contrast, Helice is a crab found mainly in intertidal marsh habitats, and is absent from coastal uplands 25,26 . On the terrestrial borders, both Tamarix and Helice are on their habitat edges. Helice burrows are abundant under Tamarix, but absent in open areas between Tamarix.
We conducted Tamarix removal, Tamarix simulation and crab tethering experiments to test the following hypotheses: (1) Tamarix is a critical factor in expanding Helice distributions on the physiological range limits, and (2) the facilitative effects of Tamarix on Helice are due to alleviation of both abiotic and biotic stresses (i.e. associational defense against predation).

Results
At both sites, crab burrows were generally absent in open areas, but were dense under Tamarix ( Fig. 1). At the beginning of the Tamarix removal experiment, there was no significant difference in burrow density between Tamarix and Tamarix removal plots (P . 0.05; Fig. 1). Removing Tamarix led to gradual decreases in crab burrows, and decreased by .50% after two years (Fig. 1). Tamarix removal significantly elevated solar radiation, soil temperature, and air temperature and humidity to levels similar to open plots but significantly different from Tamarix plots (Fig. 2). Soil salinity was significantly higher in open than in Tamarix plots in September, but not in May (Fig. 3). In contrast, soil moisture was significantly lower in open than in Tamarix plots in September, but not in May (Fig. 3). At the beginning of the experiment, there was no difference in soil salinity and moisture between in Tamarix and removal plots (Fig. 3). Removing Tamarix increased soil salinity and decreased soil moisture to levels similar to open plots but signficantly different from Tamarix plots (Fig. 3).
Simulation of Tamarix's shading and predation-protection using shade houses significantly increased crab burrows in open areas ( Fig. 4; F 1,20 5 90.8, P , 0.0001). Addition of Tamarix as crab food also significantly increased crab burrows in open areas ( Fig. 4; F 1,20 5 7.02, P 5 0.015). There were no significant interactive effects of shade house and food addition on crab burrows ( Fig. 4; F 1,20 5 0.04, P 5 0.84). Differences were also not found in the number of crab burrows between procedural control and control treatments (df 5 1, x 2 5 0.29, P 5 0.59). Shading open areas with shade houses significantly decreased solar radiation, soil and air temperature, air humidity and soil salinity, and increased soil moisture, all of which became similar to Tamarix plots but different from open plots (P , 0.05, Supplementary Fig. S1; also see Figs. 2 and 3).
Survivorship of crabs tethered in Tamarix plots was 60% (Fig. 5), significantly higher than in open plots (P 5 0.0046) and cage plots (P 5 0.0046), but not significantly different from shade-house plots (P

Discussion
Our results reveal that Helice is dependent on Tamarix on the terrestrial borders of salt marshes, and that this dependence is not only because Tamarix is a food resource for Helice, but also strongly due to non-trophic facilitation. Thus, in the coastal ecotone, facilitation expands the landward distribution of marsh crabs. Our work pro-vides an unambiguous demonstration for the critical role of facilitation in mediating species distributions in natural communities, and in maintaining ecotones.
Facilitation and species range expansion at physiological range limits. Our results show that thermal and desiccation stresses on the terrestrial borders of salt marshes are extreme and lethal to Helice (tethered Helice without shading all died, and Helice burrows are absent in open areas). Where Tamarix occurs, shading by its canopy retains soil moisture and decreases thermal and desiccation stress, creating microhabitats that are physically suitable to Helice. Thermal and desiccation stresses are the major abiotic factors limiting the distribution of marine animals in the upper intertidal, and alleviation of these stresses by shading has commonly been found www.nature.com/scientificreports to drive their facilitative interactions 15,16 . Although it has been argued that facilitation should collapse with extreme stress due to diminished effects of neighbors, or switch to competition for limiting resources (reviewed in Ref. 8), our results provide no support for these arguments (also see Ref. 27). The two species studied in our work are on different trophic levels and do not compete for limiting resources anyway. Although adult Tamarix persist on the terrestrial borders of salt marshes, abiotic stress such as salinity strictly limits its growth and regeneration 9 . This suggests that even in habitats where benefactor species themselves are severely limited 28 , facilitation can still function as a structuring force of communities.
Our work provides a straightforward example of how facilitation expands species' realized niche at their vertical distribution limits. Facilitation by Tamarix drives the expansion of the distribution of Helice from marshes at low elevations to the upper terrestrial border. Without Tamarix, the distribution of Helice would retract to lower elevations that are more frequently flooded. Facilitation has also been shown to expand the high intertidal borders of algae and invertebrates 15 , the low intertidal limits of marsh plants 9,29 , the arid borders of plants in dry habitats 8,30 , and the landward expansion of mangroves 31 . Our work, together with these studies, support the hypothesis that including facilitation in niche theory leads to species realized niches being larger than their fundamental niches 7,8 .
Biotic drivers of plant-animal facilitation in physically stressful habitats. Our results also show that in addition to alleviation of abiotic stress, Tamarix's associational defense against predation is a mechanism of facilitation of Helice. Many seabirds such as terns are abundant at our study sites and feed on Helice. It is also known that some seabirds prefer to feed in unvegetated habitats, possibly due to the ease of walking and attacking without vegetation (see Ref. 17). The existence of Tamarix on the largely bare terrestrial borders thus reduces Helice's risk of predation. A previous study also suggested that the facilitation effects of vegetation on fiddler crabs in hypersaline marshes in Georgia (USA) is likely to due to association defense against avian predators 17 , but had no experimental tests. Bortolus, et al. 16 also conducted a crab tethering experiment in an Argentinean high marsh, but found no evidence for plant associational defense as a mechanism of plant facilitation on crabs. This may have been due to the fact that their tethering experiments lasted only a few hours and this may not have been long enough to detect predation. Avian predators, even those resident in an area, are often not continuously present and may move around quite a bit. Since predation is potentially high in many physically stressful habitats including the high intertidal 21,32 , our work emphasizes the overlooked importance of associational defense in plant-animal facilitations in these habitats.
Highlighting the importance of facilitation in ecotones. The importance of facilitation in the coastal ecotone demonstrated in our study is consistent with other studies on other types of ecotones, such as steppe-woodland 33,34 , alpine treeline 12,13 , and open water-lake shore ecotones 35 . Ecotones are often abiotically extreme to species originated from at least one of the adjacent ecosystems 12 or both (our study). These species are able to persist on ecotones where neighbors ameliorate abiotic stress to their physiological tolerance range, while beyond the ecotones they are limited by abiotic stress. Associational defenses can also be a mechanism driving facilitation on ecotones. For example, ecotones between open water and lake shores provide refugia for fish that would be at risk of predation in open waters or desiccation stress on lake shores 35 . Other mechanisms also likely occur, such as entrapment of propagules 31 . Ecotones have been long known as biodiversity hotspots, and our study together with these previous studies suggest that facilitation is likely one of the key biological forces enhancing diversity in ecotones 36 . Ecotones are often species boundaries sensitive to environmental change, and have been widely used for monitoring the effects of climate change 10 . Future research on facilitation and community organization on species distribution borders or ecotones will be critical to understanding how environmental change affects natural communities.

Methods
Study sites. Field work was conducted on the terrestrial borders of two salt marshes in the Yellow River Delta, northern China: Huanghekou (37u439 N, 119u149 E) and Yiqian' er (38u059 N, 118u429 E). The climate is temperate monsoonal, with cool, dry springs and hot, rainy summers. The long-term annual precipitation is 537.3 mm, and the average temperature is 12.8uC 37 . The terrestrial borders at both sites were bare flats, with scattered Tamarix trees and Suaeda salsa (Linnaeus) Pallas, an annual succulent. These borders are typically dry and extremely saline (salinities . 100-200 PSU), having a layer of salt on the soil surface.
Tamarix is a shrubby recretohalophyte that inhabits coastal and riparian zones in East Asia (native) and North America (invasive 38 ). In the Yellow River Delta, Tamarix is absent from flooded marsh habitats at lower elevations, but is abundant on the borders between marshes and terrestrial uplands at upper elevations 9,39 . Tamarix also occurs in terrestrial uplands, but is limited by competition from perennial grasses 9 .
Helice is a grapsoid crab common in East Asia, and is primarily herbivorous 26,40 . The distribution of Helice along intertidal gradients in the Yellow River Delta has been previously quantified 41,42 . Helice occurs in low abundances in flooded mudflats and low marshes at low elevations where other crabs (e.g., Macrophthalmus japonicas) are dominant. At higher marshes, Helice becomes abundant, being the only common crab species there. On the terrestrial borders at upper elevations, Helice occurs only under the canopies of Tamarix trees. Helice is completely absent from terrestrial uplands beyond tidal influence. Common avian predators include Larus spp., Sterna spp. and Ardea spp. 43 , which forage on the terrestrial borders of salt marshes and nearby water bodies (Qiang He, personal observation).
Tamarix removal experiment. To test the hypothesis that removing Tamarix would reduce the density of crab burrows, we performed a removal experiment in May 2012. We marked 10 blocks on the terrestrial border at each site. Within each block, we randomly selected two Tamarix trees with 2.0-2.5 m 2 canopies (.3-5 m between trees), and cut down one of the two trees at the soil surface. We established a permanent 1.5 3 1.5 m plot centered at the base of each Tamarix tree, and a 1.5 3 1.5 m plot in open areas between trees within each block. Removal treatments were maintained monthly as necessary. We counted number of crab burrows in each plot in May, September 2012 and September 2013 at both sites, and also in May and August 2013 at Huanghekou. We estimated crab abundance by quantifying the density of crab burrows, which followed previous studies from similar habitats 16,17,25,44 . We examined differences in crab burrow density among treatments at each site and sampling date with randomized-blocked ANOVAs followed by Tukey HSD multiple comparisons at the significance level P , 0.05. All statistical analysis was done with JMP 10 (SAS Institute, NC, USA).
Edaphic and microclimate conditions were quantified following Tamarix removal. Soil cores (5.05 cm diameter 3 5 cm depth) were collected in each plot in May and September 2012 and 2013. To determine soil moisture, soil cores were weighed, ovendried at 60uC for 48 hours, and reweighed. Soil salinity was determined using the initial soil moisture content and the salinity of the water extract from the soil cores (determined using a conductivity meter; model JENCO 3010, Shanghai, China) 45 . Photosynthetically active radiation (PAR), air temperature, air humidity, and soil temperature were quantified at 11:00-13:00 on a cloudless day in early September 2012. PAR, air temperature, and air humidity were quantified 20 cm above the soil surface with quantum light meters (model 3415, Spectrum Technologies, Aurora, IL, USA) and temperature/relative humidity pens (Spectrum Technologies), and soil temperature 5 cm below the soil surface with a soil thermometer (model 6310, Spectrum Technologies) 46 . Differences in edaphic/microclimate factors between treatments on each date were analyzed with randomized-blocked ANOVAs followed by Tukey HSD multiple comparisons. Soil salinity and moisture data were log 10 (x)and (x) 3 -transformed, respectively, to increase normality when necessary.
Tamarix simulation experiment. To test the hypothesis that the dependence of Helice on Tamarix is due to facilitation by shading and predation-protecting, as well as due to food provision, we conducted a shrub simulation experiment crossing shade house and food addition treatments. In May 2012, we located 30 1.5 3 1.5 plots in open areas at Huanghekou, and randomly assigned 6 to each of the five treatments: shade house, food addition, shade house plus food addition, control, and procedural control. Plots assigned to shade house treatments were covered by shade cloth (1.5 3 1.5 3 0.7 m, l 3 w 3 h) on all sides. PAR in the shade houses ranged from 100 to 300 mmol/m 2 ?s, which was comparable to those under the trees (30-220 mmol/m 2 ?s). Shade house treatments also excluded bird access, and no bird footprints/feces were observed within these plots. To allow crab access, shade cloth was 10-15 cm above the soil surface. Procedural controls had shade cloth only on two sides. For food addition treatments, five live 30 cm long Tamarix branches were cut, tied at one end, and placed in the center of the plot, and were replaced at least weekly throughout the experiment. We counted the number of crab burrows in each plot in mid-September. The experiment was discontinued in the following year, due to potential winter ice and human damage. We examined the separate and interactive effects of shade house and food addition on number of crab burrows (sqrt-transformed) with a two-way www.nature.com/scientificreports SCIENTIFIC REPORTS | 5 : 8612 | DOI: 10.1038/srep08612 ANOVA, and differences between control and procedural control treatments with a Wilcoxon test.
To quantify the effects of shade treatments on edaphic and microclimate conditions, we determined soil salinity, soil moisture, PAR, air temperature, air humidity, and soil temperature in each plot in September, using the same methods as described above. We examined differences in each factor between treatments using nonparametric multiple comparisons for air temperature data (Dunn method for joint ranks; the air temperature data did not meet the normality assumptions of parametric tests) and Tukey HSD multiple comparisons for all others.
Crab tethering experiment. To test the hypothesis that Tamarix's alleviation of both abiotic stress and predation is critical to the fate of Helice on the terrestrial borders of salt marshes, we conducted a crab tethering experiment. In August 2012, we tethered Helice (carapace width 24-29 mm; hand collected in the field) on the terrestrial border at Huanghekou. We placed a tethered crab in the center of each of 24 open area plots, 12 Tamarix plots and the 12 shaded plots used in the shrub simulation experiments. Crabs in half of the open area plots were protected from predation with cages (25 3 25 3 20 cm) of galvanized-steel hardware cloth (7 mm mesh size). Tethers were constructed of 15 cm long fishing line, tied around the carapace of crabs, secured to the carapace with cyanoacrylic glue, and held by steel stakes pushed flush with the soil surface 16 . We deployed Helice at dusk, and examined their survivorship after 24 hours. Crabs that died from either desiccation stress (with intact body) or predation (with dismembered body and presence of bird footprints) were counted. Treatment differences in survivorship were analyzed with chi-square tests (Fisher's exact test). This work is licensed under a Creative Commons Attribution 4.0 International License. The images or other third party material in this article are included in the article's Creative Commons license, unless indicated otherwise in the credit line; if the material is not included under the Creative Commons license, users will need to obtain permission from the license holder in order to reproduce the material. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/ www.nature.com/scientificreports SCIENTIFIC REPORTS | 5 : 8612 | DOI: 10.1038/srep08612