Spatial and temporal variability in positive and negative plant–bryophyte interactions along a latitudinal gradient

Authors


Correspondence author. E-mail: simondoxford@googlemail.com

Summary

  1. According to the Stress Gradient Hypothesis, facilitation and competition are considered to be important at opposite ends of an environmental gradient. However, recent research has questioned the generality of this idea. One limitation is that the small-scale and short-term nature of much research into plant interactions limits our understanding of how their strength and importance changes over large temporal and spatial scales.
  2. Here, we experimentally measured the interactions between bryophytes and a community of winter annuals in sand dunes along a c. 1500 km latitudinal gradient over two generations. We carried out bryophyte-removal experiments within permanent quadrates at eight field sites along the Atlantic coast of Western Europe, to measure the spatio-temporal variation in the strength and direction of biotic interactions.
  3. We found a striking contrast in the nature of plant–bryophyte interactions observed in consecutive years of study, using three measures of plant demographics: density, plant performance and population growth. In the first year facilitation dominated, but showed an overall decline in strength from south to north. In the following year facilitation was largely replaced by competition, which also declined in strength from south to north along the latitudinal gradient.
  4. Interaction strengths are correlated with population growth rate, so that facilitation dominated in sites and years where population growth was low, whereas when population growth was high competition appeared to be more important. This is consistent with the Stress Gradient Hypothesis, but further to previous studies we demonstrate that this mechanism operates over time and in terms of population growth rates.
  5. The effects of bryophyte removal were more strongly related to population density and population growth than to individual plant performance (i.e. reproductive output). This result suggests interactions are most important during germination and early establishment of annuals, and not in terms of inter-plant competition for resources.
  6. Synthesis: This study is one of the first to show that there may be extreme spatio-temporal variation in the strength and direction of interactions, with substantial changes in their impact over a short time period. Our results suggest that the Stress Gradient Hypothesis may operate within and along gradients, as well as inter-annually. This may therefore play a role in buffering population dynamics.

Introduction

The occurrence and performance of plants in a community is determined by a combination of environmental factors and biotic interactions, and the degree to which they co-vary (Grime 1977; Tilman 1994; Callaway & Walker 1997; Holmgren, Scheffer & Huston 1997). A gradient of environmental conditions can systematically alter the nature of the interactions occurring along it, changing between net positive and negative effects on plant performance (Goldberg et al. 1999; Brooker et al. 2008; Pugnaire, Armas & Maestre 2011). As a result of this, facilitation and competition are widely considered to be important at opposite ends of stress gradients (see the Stress Gradient Hypothesis; Bertness & Callaway 1994). However, studies have now reported both increases and decreases in positive and negative interactions along stress gradients, as well as no measurable trend at all (Tielbörger & Kadmon 2000; Callaway et al. 2002; Maestre, Valladares & Reynolds 2006). It is likely that biotic interactions will include both positive and negative elements occurring simultaneously and that their net effect will depend on particular environmental conditions (Holzapfel & Mahall 1999; Madrigal-Gonzalez, Garcia-Rodriguez & Alarcos-Izquierdo 2012).

Currently facilitation/competition research is often limited by short-term, small-scale studies (Callaway 2007). One consequence of this is that the Stress Gradient Hypothesis refers to the change in interactions observed along an environmental spatial gradient, yet it is unclear how this relationship will be altered by inter-annual variation in the environment (but see Tielbörger & Kadmon 2000; Maestre et al. 2009). Few studies have tackled this issue and there is a need to understand the effects of both spatial and temporal variation in the balance between biotic and abiotic factors (Becerra et al. 2011; Soliveres et al. 2011). Specifically, the role of plant–plant interactions in response to climate change is unclear (Tylianakis et al. 2008). Climatic factors such as drought may increase the occurrence and importance of facilitation while general warming and increased nutrient deposition may result in a greater incidence of competition (Bertness & Ewanchuk 2002; Klanderaud 2005; Soliveres et al. 2011). Plant interactions may act to mitigate the effects of climate change or they may themselves be fundamentally altered with a knock-on effect for species diversity and community structure (Brooker 2006).

Soil surface bryophyte layers are present in many habitats and are important in generating competitive and facilitative interactions as they can both positively and negatively affect local environmental conditions for vascular plants. Bryophytes can buffer soil temperatures and increase water retention (Zamfir 2000; Gornall et al. 2011) and their physical structure provides seed ‘safe-sites’ protected against predation (Jeschke & Kiehl 2008). Alternatively, deep, dense moss acts as a barrier to light reaching seeds, and also prevents seeds from penetrating the soil during germination (Donath & Eckstein 2010). The outcome is complex, with the precise effects on vascular plant performance depending on multiple abiotic and biotic mechanisms (Zamfir 2000; Sedia & Ehrlenfeld 2003; Soudzilovskaia et al. 2010).

Small-sized and short-lived species of annuals are likely to have strong associations with bryophytes. The recruitment of these species depends on favourable conditions during germination and establishment, making them highly sensitive to inter-annual variation in climate (Sher, Goldberg & Novoplansky 2004). This may in turn affect population processes such as density dependence acting during germination and early recruitment (Watkinson, Freckleton & Forrester 2000). Their recruitment may therefore benefit from climate amelioration and seed refugia when conditions are unfavourable but they may also be in competition with bryophytes in a mesic environment (Chesson & Huntly 1997; Brooker & Callaghan 1998). The effect of variability in weather patterns means the type of interaction may not be easily generalized across multiple years (van Tooren 1990). Both the presence of bryophytes and changes in plant interactions, driven by environmental conditions, may act to structure plant communities.

In this study, we examine the interaction between bryophytes and a community of winter annuals along a latitudinal gradient in European sand dunes, which encompasses a range of environmental conditions. We analyse the change in interactions at multiple sites on this gradient and across two generations. We achieve this through in situ removal of the bryophyte layer and measurement of its rate of regrowth. We then assess the impact of bryophyte removal on the population density and performance of individuals of targeted annual species. Environmental stress is quantified by variation in the performance of the plants/populations studied. Specifically we address the following questions:

  1. How do the strength and nature of plant–bryophyte interactions vary between the sites and years studied?
  2. What components of demography are affected by interactions, e.g. population density or individual plant performance?
  3. What do these results allow us to conclude about the mechanisms responsible for the interactions observed?
  4. What are the roles of climate and the potential impact of climate change on the strength and direction of interactions?

Materials and methods

Study system and species

The study was conducted on fixed sand dunes on the Atlantic coast of western Europe. A total of eight sites were studied along a latitudinal transect of c.1500 km between northwest Portugal in the south and eastern England in the north. Their locations are given in Table 1. These sites were chosen as a result of their similarity, typically sharing many of the characteristics of “grey dune” (vegetation community SD19; Rodwell 2000). This community consists of a diverse mosaic of moss and lichen species carpeting the soil surface. Common species belong to the genera Cladonia, Racomitrium and Hypnum. Vascular plants grow within this moss and lichen crust and include a high diversity of winter annuals (20–40% of the flora in UK dunes) and perennial herbs and grasses (Watkinson & Davy 1985). Mosses made up the vast majority of the ground cover in our study and as such we refer to the removal of, and interaction with “bryophytes” for simplicity, although this may include some non-bryophytic lichens.

Table 1. The location of the field sites studied with a summary of their topography, soil chemistry, climate and grazing intensity. Mean annual rainfall and temperature values are given for the time periods studied (2009–2010/2010–2011). Grazing intensities show counts of rabbit faecal pellets per m2. Mean values (± SD) for slope, moss and lichen cover and depth, and grazing intensities are based on m2 quadrat surveys (n = 145–302; M.K.J. Ooi & R.P. Freckleton, unpublished data)
Site location (site code)Lat.Long.SlopeMoss and lichenSoil nutrients (mg 100 g−1)Temp. (°C)Rainfall (mm)Grazing Intensity
% CoverDepth (cm)N P
Praia do Tocha, Aveiro, Portugal (P2)40.328.854.2 ± 3.94 ± 90.8 ± 0.62.6 ± 2.11.7 ± 1.016.2/15.595.7/76.8Very low (NA)
Praia do Amorosa, Viana do castelo, Portugal (P3)41.65−8.8210.3 ± 8.968 ± 312.6 ± 1.96.4 ± 1.93.0 ± 2.315.5/14.7135.7/123.5Very low (NA)
Monte Blanco, Galicia, Spain (S1)43.23−8.9311.5 ± 5.581 ± 191.3 ± 1.0NANA16.5/14.9163.5/112.4Low (5)
Ondres Plage, Aquitaine, France (F1)43.58−1.480.1 ± 0.969 ± 311.6 ± 0.87.5 ± 1.52.3 ± 1.214.0/14.1113.4/113.9Medium (35)
Les Conches, Vendee, France (F2)46.39−1.496.9 ± 5.367 ± 321.9 ± 1.34.1 ± 0.22.7 ± 0.0412.7/13.358.3/53.0Medium (31)
Sainte Barbe, Bretagne, France (F3)47.60−3.153.3 ± 3.488 ± 213.3 ± 1.114.1 ± 0.033.1 ± 0.1911.6/12.484.0/68.1High (86)
Hatainville, Normandie, France (F4)49.40−1.827.4 ± 5.080 ± 221.3 ± 0.66.4 ± 3.73.0 ± 0.00311.8/11.558.2/52.0High (72)
Holkham, Norfolk, UK (GB2)52.980.776.3 ± 4.987 ± 181.8 ± 1.04.0 ± 0.82.7 ± 0.0039.9/11.150.1/49.8High (85)

These plant communities are confined to water limited, highly leached areas and require rabbit grazing or other disturbance (e.g. wind) to maintain their structure and prevent succession of perennials (Provoost et al. 2004). Their soils are well-drained, often acidic, sand, which are low in nutrients and organic matter (Rhind, Stevens & Sanderson 2006). The prevalence of drought stress in dunes suggests that climate is also an important factor for their development and persistence. Details of soil chemistry, prevailing climate and grazing intensity for each site are presented in Table 1. Previous authors have suggested that the nutrient deficiency of dune soils limits plant growth (Pemadasa & Lovell 1974) and the nutrient levels we observed were consistently low and in line with previous studies (Boorman & Fuller 1982). Mean temperature at the field sites declines directly with increase in latitude, while rainfall is more variable. Grazing is primarily by rabbits and was markedly greater in northern sites. The structure and composition of the plant community differed among sites, reflecting major climatic differences between them with large perennial shrubs and few annual species in south being replaced by smaller herbaceous perennials and a greater diversity of annuals in the north. However, the habitat type and many of the annual plant species are fundamentally very similar.

Our study focuses on the interaction between bryophytes and a group of winter annuals. Overall 15 native European species were included which occurred frequently at multiple sites along the latitudinal transect, these species represent approximately 60–80% of all annual species recorded at the sites in the study. The species included were: Arenaria serpyllifolia (L.), Asterolinum stellatum (L.), Cerastium diffusum (L.), Cerastium semidecandrum (L.), Erodium cicutarium (L.), Geranium molle (L.), Myosotis ramosissima (Rochel), Phleum arenarium (L.), Polycarpon tetraphyllum (L.), Saxifraga tridactylites (L.), Senecio vulgaris (L.), Veronica arvensis (L.), Viola tricolor (L.), Viola kitaibeliana (Schultes) and Vulpia fasciculata (Forskal). Despite this large number, overall, five species, A. serpyllifolia, C. diffusum, C. semidecandrum, E. cicutarium & P. arenarium, account for 87% of the individuals recorded across the study. All species germinate in the autumn or early spring and complete their life cycle by early summer, with the exception of E. cicutarium, which is occasionally biennial.

Experimental design and set up

The experiment was set up in the spring of 2009. Each field site was visited once between March and May beginning with the most southerly site and moving northwards to track spring phenological events. Between six and nine permanent 0.2 × 0.4 m quadrat plots were established at each field site in areas where both continuous moss/lichen cover and annual plants occurred. The plots were surveyed by a scale of a grid of 32 0.05 m × 0.05 m squares. The percentage cover of moss, lichen, bare soil and perennial vegetation was recorded in each. All individual annual plants from the species listed above were recorded and a count of inflorescence number (or inflorescence size for P. arenarium) was taken.

The plots were then divided into equal halves of 0.2 × 0.2 m each; one-half received no treatment, and is termed the control, while in the other all mosses and lichens were removed leaving only annual plants and perennial vegetation. Typically, areas of continuous moss and lichen cover, where the plots were located, had minimal perennial cover (mean 6.6%). The plots were re-surveyed in the two subsequent springs. In 2010 many regrown bryophytes were counted and removed. Air temperatures for each site were recorded on data loggers every three hours throughout the study. In the case of site F3 restricted access in 2009 meant that the plots could only be established in 2010 and re-surveyed once in 2011.

Statistical methods: plant–bryophyte interactions

The data were analysed using generalized and general linear models in r (R Development Core Team 2011). Plot density (number of plants per plot) and the mean performance (number or size of inflorescences depending on species) of annuals recorded in the quadrat plots were analysed using negative binomial regression. The factors Site (eight sites along the transect), Plot (individual plots within sites), Treatment (removal of bryophytes or not) and Year (2010 or 2011) were used as predictors of plot density and inflorescence production. The interaction term Site × Treatment × Year was fitted for each model. Population growth rate (defined as log[Nt+1/Nt] where N is number of plants per plot) was modelled in a similar way, but assuming a normal rather than negative binomial error distribution and including initial population density (Initial density) as a covariate in the analysis to control for difference in population density between sites and years.

We also fitted bryophyte regrowth as a covariate in each analysis to control variability in the recovery of the bryophyte layer between removal treatments. This was calculated as change in the natural cover of bryophytes for the control and regrowth from zero for the removal treatment. The analysis of plot density, plant performance and population growth was repeated on each year's data separately to assess the effect of Treatment in individual years, the analysis was also carried out for each site in isolation to examine the variability between them (see Tables S1–S4 in Supporting Information).

Testing for changes along gradients

The effect of the removal treatment was summarized using the log of the ratio of the control and removal treatment in each plot with a mean value calculated per site to provide a measure of the strength and direction of plant–bryophyte interactions. A positive value occurred where population density, plant performance or population growth was increased by the presence of bryophytes and indicates facilitation, while a negative value is suggestive of competition. We used this measure of plant–bryophyte interactions to analyse (linear regression) changes in the effect of the bryophyte-removal treatment along the latitudinal gradient and between study years.

Linear regression was also used to test the relationship between plant–bryophyte interaction strength and the variability in population density, plant performance and population growth. Here these measures are calculated as means of both control and removal treatments. This approach avoids the issue of regression to the mean, where the ratio of two variables is plotted against one or other of them (see Figure S1 in the Supporting Information for a full example).

The Stress Gradient Hypothesis predicts that along a gradient of increasing stress there is a change from facilitation to competition. One difficulty is defining what is meant by stress in a relevant way. Here, we present our results initially in terms of the latitudinal position of the sites. This reflects average environmental conditions and spans a wide range of mean temperatures and rainfall, but ignores site-specific and annual variations. We specifically define and analyse a gradient of stress in terms of plant performance and demographics: we assume that the less stressed environments are those in which population growth, density or plant performance are higher. These measures reflect the outcome of the biotic and abiotic factors that influence the relative performance of a species in a given set of environments.

Climate effects

To quantify the relationship between the variation in climate and the direction of interactions, the effects of the removal treatment on plot density, plant performance and population growth were compared with a selection of weather variables. Measures of mean temperature were calculated for autumn and spring from temperatures recorded in the field. Rainfall data were taken from nearby weather stations (www.metoffice.gov.uk; www.meteofrance.com; records courtesy of the Universidade de Aveiro) and used to calculate total autumn rainfall and total spring rainfall. Autumn and spring correspond with key life-history events, germination/establishment and flowering/reproduction, in the annual plants studied.

The weather variables were used as predictors for the effect of the removal treatment on density change and plant performance and population growth in a linear regression. Each variable was analysed individually, and including a two-way interaction with year to test whether the effects of weather variables were consistent between years. We were particularly aiming to identify weather variables with effects that are consistent between years, i.e. that in which the main effect is significant but shows no interaction with year. Although this approach involves quite a large number of significance tests, we note here that we were unable to identify consistent predictors and therefore our results are not likely to be compromised by multiple testing.

Results

Plant–bryophyte interactions

The most striking finding of our data was the strong contrast in plant interactions observed between the 2 years studied. The analyses revealed that there were effects of bryophyte removal, but that the direction of these effects varied between both sites and years (Table 2a–c). Overall, in 2010 bryophyte removal resulted in a lower density of annuals (Fig. 1a,b), exerted a limiting effect on plant performance, relative to the control (Fig. 2a,b) and there was a greater negative population growth (Fig. 3a,b). This indicates a facilitative effect of the bryophyte layer on annuals. Facilitation occurred consistently from sites P2-F2 although the apparent strength of the interaction was variable. At sites F4 & GB2 at the northerly limit of the latitudinal gradient there was a switch to weak competitive interactions (see Tables S1–S3 for analysis of each site in isolation and Table S4 for analysis of each study year in isolation).

Table 2. The results of models for three measures of annual plant demography in response to bryophyte removal in the sites and years studied. (a, b) Chi-squared analysis of deviance tests for negative binomial regression of population density and plant performance (measured as number/size of inflorescences). (c) An F-test for regression (normally distributed error) of population growth rate
ResponsePredictord.f.DevianceP-value
(a) Population densitySite7590.6
Year144.2
Plot50250.7< 0.0001
Bryophyte regrowth14.760.029
Treatment18.67
Site × Year638.9< 0.0001
Site × Treatment714.4< 0.05
Year × Treatment114.5< 0.0001
Site × Treatment × Year616.50.011
Residual133234.3 
(b) Plant performanceSite7282.87
Year1109.18
Plot50173.91< 0.0001
Bryophyte regrowth13.300.07
Treatment10.70
Site × Year678.82< 0.0001
Site × Treatment18.250.31
Year × Treatment15.500.02
Site × Treatment × Year617.910.01
Residual133250.01 
(c) Population growth rateSite745.5
Year177.8
Plot5029.50.09
Bryophyte regrowth10.920.15
Initial density1108.0< 0.0001
Treatment11.55
Site × Year614.4< 0.0001
Site × Treatment74.680.16
Year × Treatment14.86< 0.001
Site × Treatment × Year60.500.35
Residual132  
Figure 1.

(a, c) Log population density at each site for control (bryophytes present) and removal (bryophytes removed) treatment at each site in 2010 and 2011. Facilitative (f) or competitive (c) effects of bryophyte removal are indicated. (b, d) The mean effects of bryophyte removal by site, positive values indicate facilitation while negative values indicate competition. Error bars show standard error. (e) Shows population density against the effect size of the removal treatment (F1,105 = 2.61, = 0.109). Open circles = 2010 and closed circles = 2011.

Figure 2.

(a, c) Log of plant performance at each site for control (bryophytes present) and removal (bryophytes removed) treatment in 2010 and 2011. Facilitative (f) or competitive (c) effects of bryophyte removal are indicated. (b, d) The mean effects of bryophyte removal by site, positive values indicate facilitation while negative values indicate competition. Error bars show standard error. (e) Shows plant performance against the effect size of the removal treatment (F1,105 = 0.439, = 0.509). Open circles = 2010 and closed circles = 2011.

Figure 3.

(a, c) Log ratio of change in density of annuals (population growth rate) recorded at each site for control (bryophytes present) and removal (bryophytes removed) treatment in 2010 and 2011. Facilitative (f) or competitive (c) effects of bryophyte removal are indicated. (b, d) The mean effects of bryophyte removal by site, positive values indicate facilitation while negative values indicate competition. Error bars show standard error. (e) Shows population growth against the effect size of the removal treatment (F1,105 = 13.14, = < 0.001). Open circles = 2010 and closed circles = 2011.

In the following year (2011) this pattern was reversed, with competitive interactions dominating at the vast majority of sites. Population density was greater as a result of removal (Fig. 1c,d) and plant performance increased (Fig. 2c,d). There was also a greater positive population change (Fig. 3c,d). There was an overall trend of decreasing strength of competition moving northward along the latitudinal gradient. Bryophyte regrowth showed a linear increase with latitude varying between 0% and 98% cover at the plot level and was highly consistent between years. The rate of regrowth appears to affect population density but has no impact on plant performance or population growth (Table 2).

We performed linear regression analyses of the ln(Control treatment/Removal treatment) measurement of the type/strength plant–bryophyte interactions against sites along the latitudinal gradient and between years (Figs 1–3b,d). We found a statistically significant Year × Site interaction for population growth rate (F1,103 = 4.72, = 0.032), but nonsignificant interactions for population density and plant performance. A significant difference between years but not sites was found for population density (F1,103 = 7.10, = 0.009). This result highlights differences between years but does not provide a consistent measure of environmental stress across the study.

Figure 4.

Linear regression analysis of weather variables against the effect of the bryophyte-removal treatment on population density, plant performance and population growth rate of annuals. Air temperature (oC): (a–c) mean spring temperature, (d–f) mean autumn temperature. Rainfall (mm): (g–i) total autumn rainfall, (j–l) total spring rainfall. Open circles = 2010 and closed circles = 2011.

To quantify stress we assessed the relationship between the direction/strength of the plant–bryophyte interaction and the variability in mean (across treatment) population density, plant performance and population growth rate at the plot level over the study (Figs 1–3e). There was a significant linear relationship with population growth rate (F1,105 = 13.14, = < 0.001); the strongest facilitation occurred where there was the greatest negative population change while competition occurred during greatest population growth (Fig. 3e). A similar trend was apparent for population density although it was nonsignificant (Fig. 1e). There was no clear relationship between interaction strength and plant performance (Fig. 2e).

Notably, the strength of plant–bryophyte interactions appeared consistently limited at the extreme ends of our latitudinal gradient. When repeating our analysis at the level of individual sites we find a nonsignificant Year × Treatment effect at P2 and GB2, suggesting reduced importance of bryophyte removal at these sites, while the Year × Treatment interaction was significant at all intervening sites along the gradient for at least one of the measures of annual plant growth (see Table S1–S3).

Climate effects

We observed relatively consistent mean monthly temperatures across the field sites between study years. Total monthly rainfall was also generally comparable, however, it was apparent that there was reduced rainfall in spring 2011 (54% of that in 2010 when averaged across all sites). There were clear relationships between the nature of the plant interactions, indicated by the relationship between the ln(Control/Removal) treatment effect, and the various weather variables calculated from temperature and rainfall data (Fig. 4). Plant–bryophyte interaction strength increased linearly with both increase in temperature and rainfall, however, this was true of both competition and facilitation with the result that we again observed apparently contrasting results between study years. Interaction strength was most strongly dependent on climate when measured from population growth rate. The relationships with population density and plant performance, although showing contrasts between years, are nonsignificant. Particular statistical significance and between year variation was associated with spring temperature and autumn rainfall (Table 3).

Table 3. The results of linear regression analyses (F-tests) of the effect of bryophyte removal on population density, plant performance and population growth rate against a selection of weather variables. Bold text highlights statistically significant Year × Climate Variable interactions (results are presented as: F-valued.f.(error d.f.); P-value: ** < 0.01, * < 0.05, (*) < 0.1)
ResponseClimate Variable (C)Year C Year × C R 2
Population densityAutumn temperature5.091,11*2.361,110.331,110.41
Spring temperature4.721,11(*)0.031,111.661,110.37
Autumn rainfall5.231,11*1.231,111.841,110.43
Spring rainfall5.241,11*2.821,110.251,110.43
Plant performanceAutumn temperature1.981,110.031,111.191,110.23
Spring temperature3.561,111.821,113.021,110.40
Autumn rainfall2.331,110.191,113.251,11(*)0.34
Spring rainfall2.081,110.0031,111.891,110.27
Population growth rateAutumn temperature9.541,11*0.281,112.161,110.52
Spring temperature 13.151,11** 1.29 1,11 6.231,11* 0.65
Autumn rainfall 13.841,11** 0.15 1,11 8.351,11* 0.67
Spring rainfall10.531,11**0.271,113.571,11(*)0.57

Discussion

Biotic interactions are a key driver of plant population dynamics and community composition and will be an important component of plant responses to global climate change. However, despite increased research into positive interactions in particular, there is little consensus over the relative roles of positive and negative interactions along a gradient in environmental conditions. This study is one of very few including both a spatial and temporal gradient in the environment acting on multiple populations. A striking result is that overall we observed opposite interactions occurring in consecutive years, with a switch from net positive to negative effects of bryophytes on plant growth. The effect of bryophyte removal on annuals appears to be the product of both positive and negative mechanisms acting during germination/establishment (thereby impacting population density/growth rather than plant performance) and whose net outcome is dependent on variation in the environment. Climate appears to be important to the strength of interactions observed, however, these cannot be consistently predicted by rainfall or temperature in isolation. When using population growth rate as a measure of overall environmental stress we find a strong linear relationship with interaction strength and direction (Fig. 3e), which is in agreement with the Stress Gradient Hypothesis (SGH). However, our results are novel in incorporating temporal variation and finding this relationship occurring within and between years.

Interactions along an environmental gradient

A key issue when studying environmental gradients lies in defining the effect of changes in environmental stress in a spatially varying environment. Attempting to quantify this using discrete variables such as temperature, nutrients, disturbance etc. is limited by our perception of these factors and our knowledge of species-specific responses to them. Instead, stress can be defined by measuring individual responses to the environment along a gradient with the differences in plant/population performance indicating the quality of the environment (Sears & Chesson 2007).

This approach is particularly applicable here where we observe a striking change in the nature of the plant–bryophyte interactions within the same sites between years. We find a strong linear relationship between population growth rate and interaction strength with facilitation operating during a population crash and an inter-annual switch to competition in a better year that shows population growth (Fig. 3e). A similar, but nonsignificant, trend is found when measuring stress through differences in population density while there is no apparent relationship with plant performance. These results can be contrasted with those observed when attempting to use average conditions as indicators of environmental stress or considering only spatial variation in the environment (e.g. by using latitude). This also highlights the importance of the chosen measure of environmental stress to the outcome of studies into plant interactions (Holmgren, Scheffer & Huston 1997; Maestre, Valladares & Reynolds 2005).

Although an influential concept, the SGH has recently received criticism because several empirical studies have failed to support its main prediction, namely that positive interactions should increase linearly along a gradient of environmental stress (Tielbörger & Kadmon 2000; Maestre, Valladares & Reynolds 2005). An alternative model of plant interactions suggests a breakdown of the SGH at the extreme end of environmental gradients and that facilitation may be more important under moderate conditions (Holmgren & Scheffer 2010). Site P2, at the extreme southerly limit of our study, appears to support this idea as there is little evidence of plant–bryophyte interactions occurring. This site is characterized by reduced bryophyte cover, is highly disturbed, shows the highest mean and maximum temperatures and has relatively low rainfall resulting in the prevalence of drought conditions (Table 1). Bryophytes are known to facilitate the growth of annuals by buffering soil temperatures and increasing water retention (Zamfir 2000; Gornall et al. 2011). However, it seems likely that along a gradient of environmental stress this mechanism will eventually break down as, for example, the effect of drought exceeds the bryophytes ability to retain water (Malkinson & Tielbörger 2010).

Balance between positive and negative interactions

The net result of plant interactions which can be measured through effects on population density and individual plant fitness are commonly thought to be the product of many competitive and facilitative effects occurring simultaneously (Bertness & Callaway 1994; Brooker & Callaghan 1998; Pugnaire & Luque 2001). In this study, there are a multitude of abiotic and biotic mechanisms that contribute to the interaction between bryophytes and dune annuals. Positive mechanisms include climate amelioration and physical protection against predation, particularly in seeds (Zamfir 2000; Jeschke & Kiehl 2008; Donath & Eckstein 2010; Gornall et al. 2011). Alternatively, direct resource competition can occur, or bryophytes may act as physical barriers to light and substrate access and in some cases they have been shown to secrete alleopathic chemicals (Sedia & Ehrlenfeld 2003; Soudzilovskaia et al. 2010).

We find an effect of variability in moss regrowth particularly for population density (Table 2a), however, we also observe highly consistent rates of regrowth in the bryophyte crust between years. Based on these results we conclude that the contrasting pattern in plant–bryophyte interactions is not directly a result of inter-annual variation in the regrowth of bryophytes. The dramatic switch from facilitation to competition in consecutive years suggests a balance of positive and negative interactions, potentially highly dependent on fine scale fluctuations in the environment i.e. climate variables (Fig. 4; Spitale 2009). This contrast highlights the importance of a temporal perspective on plant interactions, as their nature and strength is clearly dependent on variation between years. Bryophyte presence/absence appears to become largely unimportant when environmental conditions are either highly limiting or nonlimiting for plant growth. Under these conditions the importance of facilitative and competitive interactions with bryophytes are being overwhelmed entirely by other processes (Holzapfel & Mahall 1999).

When comparing the effect of bryophyte removal on our respective measures of annual plant population density, performance and population growth, we find a stronger association with population density and population growth than plant performance. This result indicates that the plant–bryophyte interaction is at its most important during germination and early establishment of annuals, where environmental conditions play a particularly key role (Levine, McEachern & Cowan 2008). Further growth and development may therefore be largely independent of bryophytes.

Implications of climate change

Climate change is likely to affect the balance of abiotic and biotic processes within communities, with knock-on effects for population growth and community composition (Tylianakis et al. 2008; Weedon & Facelli 2008). The prevalent role of drought stress in the fixed dune sites studied here indicates that climate is an important factor to their structure and persistence (Rhind, Stevens & Sanderson 2006). Our results show that two markedly contrasting types of habitat, with and without bryophytes, are preferable for annual plant growth in consecutive years. As these results are likely to be influenced by changes in environmental conditions this highlights the importance of habitat heterogeneity in mitigating climate change (Thomas et al. 2001).

The strength of interactions and the rates of bryophyte regrowth are related to the prevailing climate, which can vary considerably between sites and years (Table 1 and Fig. 4). However, we find both facilitation and competition increasing with increase in temperature and rainfall. Conventionally, facilitation and competition would be expected to increase at opposite ends of an environmental gradient. Similarly, rainfall is expected to be limited in high temperature environments. However at some of the sand dune sites studied here high temperature and high rainfall coincide. Under these conditions plant growth is likely to be promoted perhaps resulting in competitive interactions for limiting resources. Contrastingly high temperature and low rainfall (as a result of temporal variation) acting together will result in drought conditions where the moisture retaining properties of bryophytes will likely facilitate annuals. This could in part explain the relationships in Figure 4. The results of our climate analysis do not allow the identification of a consistent predictor of interaction strength/ net effect, although the linear relationships with rainfall and temperature are indicative of an important role in change in population density in particular.

Conclusion

This study emphasizes the results of much recent research in finding that the nature of plant interactions along an environmental gradient are highly dependent on the conditions and environment studied as well as the measured effect on plant growth. We highlight the limitations of both a restricted temporal and spatial scale by demonstrating high variability at the landscape scale and between study years. Our results are unusual in demonstrating such a strongly contrasting pattern in plant interactions with a switch from facilitative to competitive interactions along a spatial gradient in consecutive years. We find evidence that the interaction between bryophytes and a community of dune annuals appears to be overwhelmed or unimportant under highly stressful or, conversely, nonlimiting environmental conditions at the extreme ends of a latitudinal gradient. When incorporating temporal and spatial variation in the environment and using population growth rate as a measure of environmental stress we find a linear relationship with biotic interactions consistent with the Stress Gradient Hypothesis. The complex patterns of inter-annual variation we observe combined with the lack of studies incorporating temporal variation into the classic stress gradient concept shows a need for further research in this area.

Acknowledgements

We would like to thank N. Fowler for useful comments on the manuscript. During this work S.W.D. was funded by a PhD studentship and M.K.J.O as a post doctoral researcher via a Leverhulme Trust Research Leadership Award to R.P.F. R.P.F. is funded by a Royal Society University Research Fellowship.

Ancillary