Oecologia

, Volume 158, Issue 3, pp 473–483

Switching from negative to positive density-dependence among populations of a cobble beach plant

Authors

    • Department of Ecology and Evolutionary BiologyBrown University
    • Department of BiologyCalifornia State University
  • Andrew D. Irving
    • Department of Ecology and Evolutionary BiologyBrown University
    • Scienze AmbientaliUniversità di Bologna
  • Mark D. Bertness
    • Department of Ecology and Evolutionary BiologyBrown University
Population Ecology - Original Paper

DOI: 10.1007/s00442-008-1157-0

Cite this article as:
Goldenheim, W.M., Irving, A.D. & Bertness, M.D. Oecologia (2008) 158: 473. doi:10.1007/s00442-008-1157-0

Abstract

Interactions among species occur across a continuum from negative to positive and can have a critical role in shaping population and community dynamics. Growing evidence suggests that inter- and intra-specific interactions can vary in strength or even switch direction (i.e., negative to positive) depending on environmental conditions, consumer pressure, and also among life-history stages. We tested the hypothesis that seedlings and adults of the intertidal annual forb Suaeda linearis growing on New England shores exhibit positive density-dependence under physically stressful conditions high on the shore (i.e., greater temperatures, evaporative stress), but negative density-dependence under physically milder conditions low on the shore. Among experimental treatments of plant density (dense versus sparse) at each shore height, plant biomass, length, and number of leaves/branches were greater in dense stands high on the shore (positive density-dependence), but greater in sparse stands low on the shore (negative density-dependence). Such responses were consistent among life-history stages and generally consistent between sites. As a more direct measure of fitness, per capita seed production was also positively density-dependent high on the shore, but negatively density-dependent low on the shore. These results support the current theory predicting an increase in the frequency of positive interactions with increasing environmental stress and further emphasize the previously understated role of positive interactions in shaping and maintaining populations and communities.

Keywords

FacilitationPhysical stressSpartina alternifloraStress gradient hypothesisSuaeda linearis

Introduction

Interactions among species can be key determinants of population dynamics and community structure (Krebs 1994; Begon et al. 1996). Traditionally, community ecologists have emphasized the study of interactions, such as competition and predation, resulting in a body of ecological theory that largely presents nature as a showcase of negative interactions (Tilman 1982; Menge and Sutherland 1987; Keddy 2001). While this approach has provided unquestionable insight, recent experimental research of positive species interactions, such as facilitation and mutualisms, has encouraged critical evaluation of their seemingly understated role (Bertness and Callaway 1994; Stachowicz 2001; Bruno et al. 2003). Formally considered a century ago (Phillips 1909; Clements 1916), and emphasized shortly after through description of the Allee effect (Allee 1938), examples of positive interactions are easily recognized (e.g., coral-zooxanthellae associations, origins of eukaryotic life; Goreau et al. 1979; Gupta and Golding 1996). However, incorporation of the growing quantity of empirical evidence into contemporary ecological theory of population and community dynamics is often lacking.

Positive interactions are thought to be common and pervasive in natural communities, having important effects on individual growth, fitness, recruitment, and succession, as well as the maintenance and expansion of species abundance, diversity, and distribution (see reviews by Callaway 1995, 2007; Bruno and Bertness 2001; Stachowicz 2001). Obvious and widespread examples occur where habitat-forming organisms (e.g., trees, salt marsh plants, seagrasses, mussels, etc.) facilitate other species by providing primary living space and/or modifying environmental conditions (Jones et al. 1997; Bruno and Bertness 2001). Indeed, theory accounting for the occurrence of positive interactions in nature predicts them to be most frequent in physically and/or biologically harsh environments where some organisms can ameliorate potentially limiting stressors (e.g., temperature, consumers) to create more favorable habitat for other species (i.e., the stress gradient hypothesis; Bertness and Callaway 1994; Callaway and Walker 1997; Brooker and Callaghan 1998; Bruno and Bertness 2001; Graff et al. 2007; Brooker et al. 2008; Crain 2008). For example, desert shrubs can enhance the survival, individual fitness, and seed production of associated annuals by reducing temperatures to more tolerable levels (Holzapfel and Mahall 1999), while seaweeds can enhance the recruitment, growth, and survival of marine organisms by reducing temperature and desiccation stress during low tides (Bertness et al. 1999) and providing associational defenses against grazers (Hay 1986).

Most evidence for positive interactions is inter-specific (i.e., species A facilitates/benefits species B), yet similar outcomes and mechanisms can occur among members of the same species (Allee 1938; Silander 1978; Wied and Galen 1998). Population regulation is often thought to be a function of negative density-dependence at one or more life-history stages, because higher densities can increase mortality through greater competition for resources and easier detection by consumers (Hairston et al. 1960; Connell 1972; Stachowicz 2001). In some cases, however, higher densities can also enhance the degree of habitat amelioration, which may confer benefits such as increased individual fitness, growth and survival, and, ultimately, the presence of a species within a community (i.e., positive density-dependence). Such benefits have been noted among numerous types of terrestrial plants (Wilson and Agnew 1992), as well as in populations of common intertidal invertebrates and macroalgae (Hay 1981; Garrity 1984; Bertness and Leonard 1997). Where positive density-dependence may occur, it is likely that individuals in dense aggregations are still competing with each other for resources (or otherwise interacting negatively), but that such costs do not outweigh the benefits of group living (i.e., net positive density-dependence; Callaway and Walker 1997; Bruno and Bertness 2001).

The outcome of species interactions can vary depending on environmental conditions (Bronstein 1994; Callaway et al. 2003). Even slight variations in the physical environment can have disproportionately large effects on species interactions and resulting patterns of distribution and abundance (Sanford 1999). While habitat amelioration appears decisive in physically stressful environments, it follows that such effects become less important as conditions become more benign. In such cases, negative species interactions, particularly competition, are predicted to dominate population and community dynamics (Bertness and Callaway 1994; Callaway and Walker 1997). Switches from positive to neutral/negative interactions with declining physical stress have been described in different habitats (e.g., alpine plants, Choler et al. 2001; Callaway et al. 2002; Kikvidze et al. 2005; grasslands, Greenlee and Callaway 1996; oak forests, Frost and McDougald 1989; salt marshes, Bertness and Yeh 1994; Crain 2008; rocky intertidal, Bertness and Leonard 1997; Bertness et al. 1999). Similar switches have been observed among life-history stages; for example, during the robust adult stage, there is less reliance on protection from neighbors than at the vulnerable juvenile stages (Callaway and Walker 1997). Such variation highlights the need to understand ‘conditionalities’ (sensu Bronstein 1994) of positive interactions if their roles in nature are to be properly understood.

The purpose of this study was to test how the direction of density-dependent effects (positive versus negative) among populations of a single species of intertidal plant varies with environmental stress and among life-history stages. We used populations of an annual halophytic forb growing on cobble beaches, a habitat known for steep gradients in physical stress (e.g., cobble instability) over short distances (Bruno 2000; Kennedy and Bruno 2000). This research proceeded in two stages. First, natural patterns in the density of the forb Suaeda linearis (Family Chenopodiaceae, also known as sea-blite) and the per capita biomass associated with different densities and heights on shore were quantified to suggest the conditions under which density-dependence may occur in this system. Second, we experimentally tested the hypothesis that populations of S. linearis growing high on the shore (stressful environment) exhibit positive density-dependence and thereby benefit from group-living, while such benefits are negated low on the shore (benign environment) where populations exhibit negative density-dependence instead. Experiments were replicated at two sites and life-history stages to indicate the spatial and temporal generality of responses. Multiple physical variables were sampled across the shore to identify gradients of physical stress that were then related to tests of density-dependence.

Materials and methods

Study sites and habitats

All research was done on intertidal cobble beaches in Narragansett Bay (Rhode Island, USA), a well-mixed estuary with semi-diurnal tides ranging from 0.8 to 2.0 m (see Bruno 2000 for a more detailed description of the bay environment). Cobble beaches are common on moderately protected shorelines of New England, with the seaward edge often fringed by beds of the cordgrass Spartina alterniflora that becomes almost fully submerged at high tide. When large enough (>40 m long; Bruno and Kennedy 2000), beds of S. alterniflora are present, they modify the shoreline environment by buffering against wave disturbance, thereby stabilizing cobbles behind the bed. This stabilization facilitates the establishment and persistence of a forb community comprising up to 12 annual and perennial species that are otherwise exceedingly rare (usually nonexistent) in the absence of S. alterniflora beds (Bruno 2000; Bruno and Kennedy 2000). Mechanical disturbance (cobble instability) and competition with S. alterniflora appear to control the upper and lower distribution of forbs, respectively (Kennedy and Bruno 2000; van de Koppel et al. 2006), with herbivory having little, if any, influence across the shore (Bruno 2000; Kennedy and Bruno 2000). Even so, many species within the community exhibit substantial variation in their density across the shore (authors’ personal observation). While the patterns and mechanisms driving this ‘whole-community’ facilitation have been well-described by Bruno and colleagues, ecological knowledge of the forb community itself is limited.

The experiments were replicated at two sites: Brown University’s Haffenreffer Reserve (hereafter called Haffenreffer; 41°41′N, 71°14′W) and Providence Point on Prudence Island (41°40′N, 71°20′W), part of the Narragansett Bay National Estuarine Research Reserve. These two sites were chosen for their extensive beds of S. alterniflora, as well as abundant populations of the forb S. linearis, which is a relatively hardy species that is amenable to experimental manipulation (Bruno 2000; Bruno and Kennedy 2000; Kennedy and Bruno 2000; van de Koppel et al. 2006). All experiments were done between April and August 2006, within the annual growing season of S. alterniflora and forbs in this region. Further information on the forb communities and general environmental conditions at these two sites can be found in Bruno (2000, 2002).

Natural patterns

Abundant populations of S. linearis occurred behind beds of S. alterniflora at both sites throughout the growing season. Observation of the shoreline early in the growing season (early May) revealed that S. linearis seedlings grew in a continuum of sparse (1–10 individuals per 0.1 × 0.1 m) to dense stands (>50 individuals per 0.1 × 0.1 m) across the shoreline, although the latter occurred more frequently higher on the shore and the former lower on the shore (authors’ unpublished data). Similar patterns, although generally lower numbers of individuals per area, were observed for adult plants in June.

Natural differences in the per capita biomass of S. linearis were quantified between dense and sparse stands both high and low on the shore. High shore habitat was defined as being within 1 m seaward of the high tide mark, while low shore habitat was defined as being within 1 m landward of the upper border of the S. alterniflora bed. This latter definition is important since competition with S. alterniflora can limit forb distribution (van de Koppel et al. 2006), and we avoided such biases on our results by restricting our research to the area above the S. alterniflora border. Individual plants were harvested from each group (termed high-dense, high-sparse, low-dense, and low-sparse) in mid-May for seedlings (n = 36 for both sites) and mid-June for adults (= 14 for both sites). Below-ground biomass was removed and individuals were then oven dried to a constant weight at 70°C for 48 h before being weighed. Tests for differences in the per capita biomass among treatments were done using two-way ANOVA, treating ‘shore height’ and ‘density’ as fixed and orthogonal. Observed differences were then used to derive experimental hypotheses about the occurrence and direction of density-dependent effects within populations of S. linearis.

Experimental design

Experimental tests of density-dependence were done at both sites for seedling and adult stages of S. linearis. For all experiments, stands of S. linearis growing high and low on the shore were manipulated to create four treatment groups based on the natural density patterns: dense stands high on the shore (control), sparse stands high on the shore (manipulated), dense stands low on the shore (manipulated), and sparse stands low on the shore (control). High-dense treatments comprised natural stands of more than 50 individuals per 0.1 × 0.1 m for seedlings, and more than 20 individuals per 0.1 × 0.1 m for adults. High-sparse treatments were created by thinning dense stands down to a single individual, matching the density of low-sparse treatments, and within the range of observed natural densities. Finally, low-dense treatments were created by translocating dense stands from high to low on the shore. Translocated stands of S. linearis were watered with fresh water every day for 3–5 days post-disturbance to minimize transplant shock. For seedlings, 7 replicates of each treatment were established at Providence Point and 10 at Haffenreffer; while, for adults, 13 were established at Providence Point and 11 at Haffenreffer.

Despite efforts to minimize transplant shock, the process of translocating S. linearis is a disturbance that may cause effects unrelated to, but indistinguishable from, our manipulations low on the shore. Therefore, two procedural controls were established high on the shore whereby dense stands of S. linearis were either lifted vertically before being re-planted in their original location (vertical disturbance; first procedural control) or lifted and moved a similar distance as experimental stands translocated to low on the shore, but were instead translocated to new positions high on the shore (vertical plus horizontal disturbance; second procedural control). Differences between these procedural controls and undisturbed dense stands high on the shore (experimental control) would suggest artifacts of translocation that could confound interpretation of experimental responses.

Experiments on seedlings were established in mid-May and sampled in mid-June, immediately before S. linearis began a period of rapid growth and populations appeared to undergo natural thinning along the shore (authors’ personal observation). Experiments on adults were established in mid-June and sampled in late August/early September when seeds were present in glumes and early signs of senescence were visible (e.g., withering of leaves). A single individual was sampled from each replicate, with above-ground biomass (oven dried to constant weight at 70°C for >48 h) recorded for each individual. To indicate which parts of the plant may contribute to differences in total biomass, plant length and the number of leaves (seedlings) or branches (adults) were recorded immediately before drying. For adult plants at Haffenreffer, seed production per individual was estimated by counting the number of seeds present on a representative branch and multiplying this number by the number of branches bearing seeds.

Analyses proceeded in two steps. First, artifacts of translocation were tested using ANOVA (naturally dense stands high on the shore versus procedural controls). In the absence of artifacts, analysis of experimental treatments was done using two-way ANOVA, treating both factors as fixed and orthogonal. Sites were analyzed separately due to uneven replication, but also because of some key environmental differences among sites that may have increased residual variance and obscured experimental effects. In particular, facilitation of forbs is patch-size dependent (Bruno and Kennedy 2000), and Providence Point supports one of the longest continuous S. alterniflora beds in Narragansett Bay (J. Bruno, unpublished data), while Haffenreffer is comprised of smaller, fragmented beds, not all of them large enough to facilitate forb communities (authors’ personal observations). Furthermore, Providence Point is a north-facing site bordered by salt marsh habitat, while Haffenreffer is an east-facing site bordered by a terrestrial pine and oak-hickory forest.

Physical conditions

Gradients of physical stress across cobble beaches have been demonstrated in relation to cobble instability (Kennedy and Bruno 2000), although other environmental factors may also differ among shore heights. In addition to cobble movement, we quantified soil temperature, soil moisture, and evaporative stress during July, approximately halfway through the growing season. We tested for (1) differences between ambient high- and low-shore conditions (i.e., in the absence of any vegetation, ~1–2 m away from stands of S. linearis, but still at the same heights on shore), and (2) modification of ambient conditions by dense and sparse stands of S. linearis (typically separated by ~1–2 m within each shore height).

Soil temperature was measured using a portable Omega temperature probe by inserting the sensor ~1 cm into the soil beneath loose cobbles (n = 15 per treatment). Soil moisture was quantified by collecting soil cores from each treatment (n = 5 per treatment) and calculating the difference in mass between wet and oven-dried samples (70°C for >48 h). To quantify evaporative stress, the loss of mass from moist sponges placed among treatments was measured. Sponges were placed on Petri dishes during the early afternoon and exposed to evaporation for at least 2 h over a low tide period, with the rate of water loss per hour calculated. Cobble movement was measured following the methods of Bruno and Kennedy (2000), whereby individually numbered cobbles were positioned along a transect during low tide, exposed to two high tides, and relocated the following day. The linear distance that each cobble had moved provided a measure of cobble instability among treatments. Soil temperature, soil moisture, and evaporative stress were all sampled over mid-day low tides since this time presented conditions under which the largest differences in physical stress were expected to occur. Furthermore, any possible bias of temporal autocorrelation among samples was minimized through synchronized random sampling among treatments.

Results

Seedlings

Natural differences in the per capita biomass of seedlings between dense and sparse stands were dependent on shore height at both sites (Fig. 1a; Table 1a: height × density interaction). High on the shore, seedlings in dense stands had a greater per capita biomass than seedlings in sparse stands, but this pattern was reversed low on the shore.
https://static-content.springer.com/image/art%3A10.1007%2Fs00442-008-1157-0/MediaObjects/442_2008_1157_Fig1_HTML.gif
Fig. 1

Natural patterns and experimental responses of S. linearis seedlings. Mean (±SE) per capita biomass of S. linearis high and low on the shore in naturally dense and sparse stands (a) and experimental dense and sparse stands (b). Mean (±SE) length of individuals (c) and number of leaves per individual (d) among experimental treatments. Sites are plotted separately; natural patterns (a) were sampled earlier than experimental treatment (b), which explains differences in plant size and scale of the Y axes between these two graphs

Table 1

Results of two-way ANOVAs testing the effects of shore height (high versus low) and density of stand (dense versus sparse) on natural patterns in the biomass of S. linearis seedlings (a), and the biomass (b), length (c), and number of leaves (d) of experimental S. linearis seedlings. Results from each site are shown separately

Site

Source

Natural patterns

Experimental effects

 

(a) Biomass

 

(b) Biomass

(c) Length

(d) No. of leaves

df

MS

F

df

MS

F

MS

F

MS

F

Providence Point

Height

1

0.00

1.02NS

1

0.00

0.49NS

2.17

1.55NS

0.14

0.13NS

Density

1

0.00

1.82NS

1

0.00

0.19NS

1.56

1.11NS

1.29

1.14NS

Height × density

1

0.01

141.59***

1

0.03

25.48***

41.77

29.71***

41.29

36.51***

Residual

140

0.00

 

24

0.00

 

1.41

 

1.13

 

Haffenreffer

Height

1

0.00

0.05NS

1

0.00

0.02NS

38.78

20.14***

0.90

0.92NS

Density

1

0.00

0.82NS

1

0.00

0.41NS

5.11

2.96NS

12.10

12.37**

Height × density

1

0.01

78.95***

1

0.03

87.18***

86.14

49.88***

108.90

111.38***

Residual

140

0.00

 

36

0.00

 

1.73

 

0.98

 

Cochran’s C test of homogeneity of variances: P > 0.05 for experimental effects. *** P < 0.001, NS P > 0.05

Cochran’s C test of homogeneity of variances: P < 0.05 for natural patterns, and so significance was judged at the more conservative α = 0.01. In such cases, *** P < 0.0001, NS P > 0.01

Tests for translocation artifacts revealed no effect of vertical or horizontal disturbance on seedling biomass, length, or number of leaves at Providence Point (P > 0.30 for all variables). At Haffenreffer, artifacts were not detected for seedling biomass (P > 0.08), but vertical disturbance (lifting control) appeared to reduce seedling length and number of leaves (P = 0.0113 and 0.0007, respectively). However, the second procedural control, which incorporated both vertical and horizontal disturbances and is, therefore, most similar to experimental translocations, did not differ from controls, suggesting translocation artifacts were minor.

Experiments produced results that closely matched natural patterns. High on the shore, per capita seedling biomass was greater in dense stands than sparse stands, but this was reversed low on the shore (Fig. 1b; Table 1b: height × density interaction, Student-Neuman–Keuls tests: high-dense > high-sparse, low-dense < low-sparse). Consistent with this result, sampled attributes of biomass (plant length and number of leaves) were both greater in dense stands than sparse stands high on the shore, and greater in sparse stands than dense low on the shore (Fig. 1c, d; Table 1c, d). All results were consistent between sites.

Adults

Natural differences among adult plants were similar to those observed for seedlings. Plants at Haffenreffer were of greater biomass in dense stands than sparse stands high on the shore, but this pattern was reversed low on the shore (Fig. 2a). At Providence Point, significant differences between densities were only detected low on the shore (Fig. 2a: dense < sparse; Table 2a: height × density interaction). Even so, the mean biomass of plants in dense stands ranked greater than sparse stands high on the shore (0.14 g versus 0.12 g respectively), suggesting a trend similar to that observed at Haffenreffer.
https://static-content.springer.com/image/art%3A10.1007%2Fs00442-008-1157-0/MediaObjects/442_2008_1157_Fig2_HTML.gif
Fig. 2

Natural patterns and experimental responses of S. linearis adults. Mean (±SE) per capita biomass of S. linearis high and low on the shore in naturally dense and sparse stands (a) and experimental dense and sparse stands (b). Mean (±SE) length of individuals (c) and number of branches per individual (d) among experimental treatments. Sites are plotted separately; natural patterns (a) were sampled earlier than experimental treatment (b), which explains differences in plant size and scale of the Y axes between these two graphs

No artifacts of translocation were detected at either site (P > 0.07 for all tests). Experimental manipulations of adult density produced effects similar to those observed for seedlings, although greater variability among replicates was common. At Providence Point, dense plants high on the shore had greater per capita biomass than sparse plants, but this pattern was again reversed low on the shore (Fig. 2b; Table 2b: height × density interaction). A similar result was observed at Haffenreffer (Fig. 2b; Table 2b: height × density interaction), although significant differences among densities were only detected low on the shore (dense < sparse). High on the shore, plants in dense stands were longer than those in sparse stands, while no differences in the number of branches were detected (Fig. 2c, d; Table 2c, d). Low on the shore, plants in sparse stands had more branches than those in dense stands at both sites (Fig. 2d; Table 2d), while plant length only differed at Haffenreffer (dense < sparse, Fig. 2c; Table 2c). Differences between densities in per capita seed production were not detected high on the shore (but dense ranked greater than sparse), while production was greater in sparse stands than dense stands low on the shore (Fig. 3, ANOVA: height × density interaction, F1,40 = 13.42, P = 0.0007).
Table 2

Results of two-way ANOVAs testing the effects of shore height (high versus low) and density of stand (dense versus sparse) on natural patterns in the biomass of adult S. linearis (a), and the biomass (b), length (c), and number of leaves (d) of experimental adult S. linearis. Results from each site are shown separately

Site

Source

Natural patterns

Experimental effects

 

(a) Biomass

 

(b) Biomass

(c) Length

(d) No. of branches

df

MS

F

df

MS

F

MS

F

MS

F

Providence Point

Height

1

0.01

4.27NS

1

24.99

4.77NS

16.51

0.23NS

325.00

4.28NS

Density

1

0.02

12.31**

1

0.41

0.08NS

551.20

7.60**

177.23

2.33NS

Height × density

1

0.04

23.94***

1

85.80

16.38**

1179.91

16.26***

982.23

12.92**

Residual

52

0.00

 

48

5.24

 

72.57

 

76.01

 

Haffenreffer

Height

1

0.02

6.11*

1

0.07

0.03NS

224.55

6.49NS

327.27

7.55**

Density

1

0.00

0.08NS

1

17.92

7.35*

33.16

0.96NS

245.82

5.67*

Height × density

1

0.05

12.66***

1

37.02

15.19**

518.20

14.98**

227.27

5.24*

Residual

52

0.01

 

40

2.44

 

34.59

 

43.36

 

Cochran’s C test of homogeneity of variances: P > 0.05 for experimental length at Providence Point, as well as natural patterns and experimental No. of branches at Haffenreffer. * P < 0.05; ** P < 0.01; *** P < 0.001; NS P > 0.05

Cochran’s C test of homogeneity of variances: P < 0.05 for natural patterns, experimental biomass and No. of branches at Providence Point, as well as experimental biomass and length at Haffenreffer, and so significance was judged at the more conservative α = 0.01. In such cases, * P < 0.01; ** P < 0.001; *** P < 0.0001; NS P > 0.01

https://static-content.springer.com/image/art%3A10.1007%2Fs00442-008-1157-0/MediaObjects/442_2008_1157_Fig3_HTML.gif
Fig. 3

Mean (±SE) per capita production of seeds by S. linearis among experimental treatments at Haffenreffer (seed production was not sampled at Providence Point)

Physical conditions

The ambient physical environment (i.e., without S. linearis) differed substantially between heights on shore. At both sites, high shore environments were characterized by ~4–6°C greater soil temperatures and 2–3 times greater rates of evaporative stress than low shore environments (Fig. 4a, b: Ambient; Table 3a, b). The soil was also drier in high shore environments at both sites (Fig. 4c: Ambient), although significance was only detected at Providence Point (Table 3c). Cobble movement was negligible at both sites (Fig. 4d: Ambient; Table 3d), which is not surprising given the strong stabilizing effect of S. alterniflora on cobbles (Bruno 2000).
https://static-content.springer.com/image/art%3A10.1007%2Fs00442-008-1157-0/MediaObjects/442_2008_1157_Fig4_HTML.gif
Fig. 4

Ambient environmental conditions and their modification by dense and sparse stands of S. linearis high and low on the shore. (Mean ± SE) (a) soil temperature, (b) evaporative stress, (c) soil moisture and (d) cobble movement (00 denotes no movement)

Table 3

Results of two-way ANOVAs testing for differences in the physical environmental conditions with shore height (high versus low) and density of S. linearis (dense versus sparse versus none/ambient). Results from each site are shown separately

Site

Source

(a) Temperature

(b) Evaporation

(c) Soil moisture

(d) Cobble movement

df

MS

F

df

MS

F

df

MS

F

MS

F

Providence Point

Height

1

36.04

15.24**

1

12.71

15.50**

1

4.00

5.58*

0.00

F = 0

Density

2

77.34

32.70***

2

1.32

1.61NS

2

1.94

2.71NS

0.13

5.71*

Height × density

2

18.73

7.92*

2

5.18

6.31*

2

4.07

5.68**

0.00

F = 0

Residual

54

2.37

 

48

0.82

 

24

0.72

 

0.02

 

Haffenreffer

Height

1

260.42

154.50***

1

14.61

76.48***

1

1.72

2.00NS

0.00

0.00NS

Density

2

24.48

14.52***

2

1.19

6.24*

2

2.48

2.89NS

17.63

13.66**

Height × density

2

9.22

5.47**

2

1.32

6.93*

2

0.14

0.16NS

0.00

0.00NS

Residual

54

1.69

 

48

0.19

 

24

0.86

 

1.29

 

Cochran’s C test of homogeneity of variances: P > 0.05 for soil moisture at Providence Point, and temperature at Haffenreffer. * P < 0.05, ** P < 0.01, *** P < 0.001, NS P > 0.05

Cochran’s C test of homogeneity of variances: P < 0.05 for temperature, evaporation, and cobble movement at Providence Point, as well as evaporation, soil moisture, and cobble movement at Haffenreffer, and so significance was judged at the more conservative α = 0.01. In such cases, * P < 0.01; ** P < 0.001; *** P < 0.0001; NS P > 0.01

Dense stands of S. linearis ameliorated ambient conditions high on the shore, where soil temperature was reduced by ~3–6°C (becoming similar to ambient low-shore temperatures), while evaporative stress was also reduced (Fig. 4a, b; Table 3a, b). Soil moisture was generally unaffected by dense stands high on the shore (Fig. 4c). In general, sparse S. linearis did little to ameliorate ambient physical stress high on the shore (Fig. 4). In relatively benign low-shore environments, S. linearis had variable effects on ambient physical conditions, obscuring general trends. For example, low-shore dense stands reduced temperature at Providence Point but not at Haffenreffer, evaporative stress was greater in sparse stands at Haffenreffer but not Providence Point, while soil moisture was reduced by S. linearis at Providence Point but not at Haffenreffer (Fig. 4a–c; Table 3).

Discussion

Our results show that intra-specific density-dependence can have a substantial role in determining the individual productivity and fitness of S. linearis growing on intertidal cobble beaches in New England. A key result was that the direction of density-dependence switched between shore heights—plants exhibited positive density-dependence high on the shore; however, not more than 5 m away, plants exhibited negative density-dependence low on the shore. Furthermore, this switch was observed at multiple sites and life-history stages (seedlings and adults), demonstrating spatial and temporal generality of results. Close correspondence between natural patterns and experimental effects (compare graphs a and b in Figs. 1, 2) suggests density-dependence is a key process regulating productivity of S. linearis and possibly also its abundance and distribution by enhancing seed production and, therefore, potential population size the following season (discussed in further detail below).

The observed switch in the direction of density-dependence occurred predictably along a physical stress gradient, supporting contemporary theory that suggests greater strength and frequency of positive interactions with increasing ambient stress (Bertness and Callaway 1994; Callaway and Walker 1997; Brooker and Callaghan 1998; Bruno and Bertness 2001). Similar outcomes have been observed in other intertidal systems (e.g., macroalgal canopies, Bertness et al. 1999; salt marshes, Bertness and Yeh 1994), as well as in terrestrial habitats (e.g., grasslands, Greenlee and Callaway 1996; oak forests, Frost and McDougald 1989), suggesting such switches along gradients of physical stress are a general feature of many natural communities. It is worth noting, however, that most examples describe inter-specific interactions, while our data show that such predictions also hold for intra-specific interactions. Bertness and Yeh (1994) also reported positive intra-specific interactions among populations of the marsh elder Iva frutescens, but such results contrast with the simulation model proposed by Malkinson and Jeltsch (2007), who concluded that intra-specific facilitation is a negligible process structuring shrub populations along a rainfall gradient. Nonetheless, the strength of density-dependent interactions among S. linearis in our study appears similar whether it is positive or negative. For example, differences in plant biomass between density treatments were reversed between shore heights but were of similar magnitude (Fig. 2b). Collectively, our data add to a growing body of evidence demonstrating a prominent increase in the frequency of positive interactions in physically harsh environments, highlighting their important roles in natural communities.

The precise mechanism(s) driving observed density-dependent effects were not tested in the current study, yet our sampling of physical conditions across the shore provides possible explanations. High on the shore, dense stands of S. linearis reduced thermal and evaporative stress (i.e., desiccation) to levels similar to ambient conditions low on the shore (i.e., relatively benign habitat). Such amelioration likely had substantial physiological benefits that enhanced plant growth (Callaway and Walker 1997), possibly including improved water use efficiency (with fresh water already being a scarce resource in this system) and maintaining tissue temperatures below critical thresholds. In comparison, evidence for low-shore habitat modification by S. linearis was scarce in our sampling, making benefits from habitat amelioration unlikely. Instead, individuals in dense stands were small (low biomass, short, and few leaves/branches) while plants released from neighbors (i.e., sparse stands) often grew to be among the largest sampled from all treatments, suggesting the per capita availability of resources for growth in dense stands low on the shore was limited by competition. Indeed, competition for light within dense stands may be crucial since S. linearis (and other forb species) appears sensitive to light limitation (Ellison 1987; van de Koppel et al. 2006), and ~21–72% less photosynthetically active radiation was available to plants within dense stands than in sparse because of self-shading (authors’ unpublished data).

Regardless of the mechanisms involved, the co-occurrence of positive and negative density-dependence in this system raises the interesting prediction that positive density-dependence can only occur where benefits of group living outweigh costs of coexistence (i.e., a net positive result from numerous co-occurring positive and negative interactions; Callaway and Walker 1997; Bruno and Bertness 2001; Stachowicz 2001). It is very likely that plants growing in dense stands high on the shore were competing for resources such as light, water, and nutrients (i.e., interacting negatively). However, the prevalence of positive density-dependence under these conditions would suggest that the benefits of group living (probably through amelioration of physical stressors as described above) outweighed the effects of such negative interactions, resulting in a positive outcome on balance. Under more benign conditions low on the shore, benefits of ameliorating physical stress are likely to be negligible, resulting in a net negative outcome within dense stands. The utility of predicting when and where the outcome of species interactions can change direction has previously been emphasized (Bronstein 1994; Bruno and Bertness 2001). Further work in this system may therefore be usefully directed at identifying threshold environmental conditions at which density-dependence switches from positive to negative.

Similar effects of density-dependence were observed between seedling and adults. However, effects were most obvious for seedlings and became more variable among adults, suggesting a weakening of interactions at later life-history stages. Variation in the outcome of species interactions among life-history stages is common (e.g., juvenile beneficiaries of ‘nurse plants’ can become competitors as adults: Callaway and Walker 1997), and may be caused by factors such as temporal and environmental variation, changing physiological requirements as organisms grow larger, overcoming thresholds of size-dependent sources of mortality, or increased robustness of adults to the physical environment (Miriti 2006; Schiffers and Tielbörger 2006; Sthultz et al. 2007). On cobble beaches, adult S. linearis became lignified and developed larger and deeper root systems over the course of the growing season, probably making them more robust to the environment and, therefore, less dependent on the buffering effect of neighbors to achieve adequate growth (i.e., weakening density-dependence). While experiments were only done at two sites, the high degree of consistency among results gives confidence that such density-dependent effects are a widespread phenomenon for S. linearis facilitated by S. alterniflora on cobble beaches. An obvious question not answered by this study, however, is whether such effects are general to the entire forb community, or whether variation exists among species or life-history strategies (i.e., annual versus perennial). Given the dependence of the entire forb community on the habitat-ameliorating effects of S. alterniflora beds, we predict that responses similar to those observed for S. linearis will occur for other forb species.

A critical step for the persistence of populations of annual plants is the production of seeds either for germination in the following growing season or the maintenance of a seed bank. We observed that density-dependence affected seed production in S. linearis by increasing production in dense stands high on the shore, and in sparse stands low on the shore (Fig. 3), with such effects generally corresponding with greater adult biomass in these stands. Collectively, therefore, it appears that density-dependence affects S. linearis at three life-history stages (seeds–seedlings–adults). Interestingly, seed production per unit mass of plant was independent of density at both shore heights (authors’ unpublished data), indicating that density-dependence likely affected seed production indirectly by influencing adult biomass. Even so, positive density-dependence high on the shore may establish a positive feedback mechanism between S. linearis and its environment (also see Bertness and Callaway 1994), whereby substantial seed production in dense stands creates abundant seed banks for the following growing season, which then benefit from growing in dense stands upon germination to ultimately produce large numbers of seeds themselves. In time, this feedback may expand the distribution and abundance of S. linearis until all suitable habitat behind a S. alterniflora bed is occupied.

In conclusion, the push to embrace positive interactions in ecological theory has encouraged considerable debate over the relative importance of positive versus negative interactions in ecology (Shouse 2003). Critically, evidence for positive interactions does not and should not weaken arguments for the importance of negative interactions, but instead demonstrates that both outcomes can occur along a continuum that may be contingent on environmental conditions (Bronstein 1994). It is hoped that understanding positive interactions ultimately fosters more complete theories of the key processes regulating populations and communities (Bruno et al. 2003). The major challenges are to identify the circumstances under which positive interactions occur, their strength relative to negative interactions, and the conditions under which they may switch direction. The physical environment can clearly have a large influence on the outcome of inter- and intra-specific interactions—often altering interaction strength (Bronstein 1994; Sanford 1999) and, as observed in this study, also switching interaction direction. Environmental changes and associated losses of biodiversity being experienced globally accentuate the need for understanding environment-driven variation in positive interactions, particularly where important biogenic habitat providers are threatened (e.g., forests, salt marshes, etc.). If ecologists are to provide information that leads to the successful mitigation and management of environmental problems (Padilla and Pugnaire 2006; Halpern et al. 2007), enhancing our empirical and theoretical understanding of positive interactions appears critical, if not overdue.

Acknowledgments

We thank K. Bromberg, T. Elsdon, C. Holdredge, G. Macgill, and N. Sala for assistance in the field. We are grateful to J. Bruno and two anonymous reviewers for providing insightful and constructive suggestions on a previous draft. The ongoing support of the Narragansett Bay National Estuarine Research Reserve and Brown University’s Haffenreffer Estate—both of which provided access to study sites—is greatly appreciated. Funding support to M.D.B. was provided by a Rhode Island Sea Grant and the National Science Foundation. All field and laboratory procedures complied with the current laws of the United States of America and the State of Rhode Island and Providence Plantations.

Copyright information

© Springer-Verlag 2008