Notice: Wiley Online Library will be unavailable on Saturday 27th February from 09:00-14:00 GMT / 04:00-09:00 EST / 17:00-22:00 SGT for essential maintenance. Apologies for the inconvenience.
1Facilitation between neighbouring plants can promote species survival and regulate community composition. However, the role of facilitation varies along environmental severity gradients. It is important to understand the shape of this relationship to improve our ability to predict the impact of a changing environment on biodiversity.
2We used Scots pine saplings growing within heather to examine the shape of the relationship between facilitative interactions (protection from browsing) and the severity of the environment (deer browsing intensity). We also investigated whether protection from browsing translated into a biomass response of saplings.
3In the first winter following planting heather had a facilitative effect on saplings by reducing the probability of browsing. This effect was strongest at intermediate deer browsing intensities, thus producing a hump-backed relationship between facilitative effects and the severity of the environment.
4Protection from browsing did not lead to longer-term biomass gains for the saplings. The competitive effects of heather on sapling growth therefore outweighed the beneficial effects of protection from browsing.
5These results provide much-needed information on the shape of the severity–interactions relationship with respect to a key natural disturbance phenomenon (herbivory), and demonstrate that an observable interaction relationship does not necessarily translate into a biomass response.
6This illustrates the complex and potentially transient nature of plant–plant interactions, and the potential difficulty that would be associated with using shelter effects of heather as a management tool to promote Scots pine regeneration.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
Interactions play a key role in translating the physiological tolerances of individuals into a realized distribution of the species, and can have profound effects on ecosystem diversity and function. Competition may constrict distributions, excluding species from a community (Booth et al. 1988; Davis et al. 1998; Vetaas 2002), whilst positive interactions may enhance an individual's or species’ ability to survive and thereby promote diversity (see reviews by Bertness & Callaway 1994; Brooker & Callaghan 1998; Bruno et al. 2003). Therefore, in order to understand the regulation of biodiversity within ecosystems, as well as the response of biodiversity to a changing environment, we need to understand the role of interactions and, in particular, the nature of the relationship between interactions and environmental gradients (Bruno et al. 2003).
In addition, such studies often do not attempt to elucidate the principal drivers involved in the relationship, using instead the generic term ‘environmental severity’. Severity can be a combination of stress and disturbance (sensuGrime 1977; Brooker & Callaghan 1998). However, because we are often interested in the impact of particular drivers rather than a generic severity effect, it would be fruitful to understand in more detail the relationships between interactions and the separate components of severity in order to improve our predictive capabilities. Some studies have already examined what are effectively stress gradients (e.g. gradients of water availability; Maestre & Cortina 2004), but to our knowledge no study has examined in detail the shape of the relationship between plant interactions and the level of disturbance. This is important when we consider that disturbance factors such as herbivory can be major drivers of vegetation dynamics and critical components of land use impacts over vast areas of the globe (Asner et al. 2004).
In conducting these studies, we must also consider how the multiple effects of neighbours combine to determine the net outcome of interactions. The net interaction between neighbours is the result of the multiple interactions that occur between them during their lifetimes (Brooker & Callaghan 1998). Measurement of interaction effects at a single point in time will not necessarily provide an adequate understanding of this net interaction, and may miss dominant components that determine the overall interaction dynamic.
In this study we used a simple model system to investigate the shape of the relationship between interactions and herbivory (a key (disturbance) component of environmental severity) and to determine whether facilitative interactions recorded at a given point in time translate into longer-term biomass effects. Our model system was saplings of Scots pine (Pinus sylvestris L., the target phytometer species) growing within a heather (Calluna vulgaris (L.) Hull) matrix on heather-dominated hillsides in the Scottish highlands.
Scots pine is browsed by a number of native herbivores, including mountain hare (Lepus timidus L.), red deer (Cervus elaphus L.) and roe deer (Capreolus capreolus L.). Browsing can destroy saplings and reduce growth of survivors (Miller & Cummins 1982; Fenton 1985; Gill 1992; Palmer & Truscott 2003), but ground-layer vegetation, including heather, can have a beneficial effect on small saplings by protecting them from browsing (Miller & Cummins 1982; Palmer & Truscott 2003; Rao et al. 2003). However, browsing pressure varies spatially due to local management regimes and the ranging habits of the species involved (Mitchell et al. 1977; Osborne 1984), and at high browsing pressure the capacity of heather to protect Scots pine may be reduced. Therefore, we have a gradient of environmental severity along which we might expect the impact of facilitation to vary, and which we can use to examine the shape of the interactions–severity relationship. However, heather neighbours can also have negative effects on growth through competition (Miles & Kinnaird 1979; Scott et al. 2000; Norberg et al. 2001), which might outweigh the positive effect of protection from herbivores.
Using this model system we addressed the following questions:
1What is the shape of the relationship between facilitative plant interactions (protection from browsing) and environmental severity (herbivore browsing pressure)?
2Are facilitative effects, e.g. the prevention of herbivore damage, ultimately translated into biomass gains?
Materials and methods
Five field sites were selected within the Cairngorms region of the Grampian Mountains, Scotland. Field sites were selected to provide accessible areas of similar heather vegetation, i.e. in either a ‘building’ or ‘mature’ phase (Gimingham 1960), without areas of recent burn. Each site comprised a single hillside with heather moorland vegetation and no existing woodland (Table 1).
Table 1. The name and location, altitudinal range and aspect of the five heather moorland areas used as field sites
Site name and location
Altitudinal range (m)
Culblean Hill (Dinnet, Deeside)
Creag Bhalg (Mar Lodge Estate, Deeside)
Creag nan Gall (Abernethy Estate, Speyside)
Meal a Bhuachaille (Abernethy Estate, Speyside)
Creag Follais (Allt á Mharcaidh catchment, Speyside)
We used saplings of Scots pine 1 + 2 Altyre provenance (Christie Elite Ltd, Forres, Moray, Scotland), a provenance typical of north-east Scotland and known to be capable of growth in upland areas (D. Macdonald, Christie Elite, personal communication). Saplings were excavated from nursery beds in two batches, one at the start and one in the middle of the planting period (September–October 2001). Following excavation the saplings were kept in large plastic sacks at 4 °C prior to planting. The maximum time between excavation and replanting was 3 weeks.
Before being planted, excess soil was removed from the roots of all saplings. We then labelled each sapling with a unique number on a plastic tag and recorded weight, the length of the leading shoot and the length of the sapling (from the base of the highest root to the tip of the leader). In order to calculate sapling mass increments during the course of the experiment we estimated the pre-planting dry mass of saplings by selecting a random subset of 50 saplings, with a size range encompassing that of the saplings that were to be planted. These were oven dried at 70 °C for 3 days prior to weighing.
planting and treatment applications
At each site, four evenly spaced, parallel rows were set out across the slope, each row following a constant altitude. All rows were between 180 and 300 m long (mean length of c. 220 m). Five evenly spaced blocks of tagged saplings were planted along each row, each block containing 10 saplings, each of which received a different combination of canopy and root manipulations. This gave a total of 20 blocks of saplings at each site, and 100 replicates of each treatment combination across the whole experiment. Saplings within a block were arranged in two parallel lines, the lines being 5 m apart and with 5 m spacing between the five saplings within a line. Sapling treatments were randomly assigned within blocks and individual saplings were randomly assigned to treatments. In this paper we consider two treatments: control and neighbour removal. Control plants were slot-planted into un-manipulated heather. Prior to the planting of neighbour-removal plants, a 20 cm radius circle of heather was clipped down to ground level and the sapling was planted in the middle of this clipped patch. We checked clipped circles for signs of heather regrowth throughout the course of the experiment. However, regrowth was extremely limited and re-clipping was not necessary.
The occurrence of first-winter herbivory was recorded during March–April 2002, 6 months after planting, as the presence/absence of signs of herbivore damage to the saplings, including, in some cases, uprooting. Uprooting decreased markedly after the first winter, probably because the saplings started to develop root systems that helped to anchor them into the ground (Miller & Cummins 1982).
We used the density of herbivore dung, recorded during March–April 2002, as an index of herbivore presence (Putman 1984). The number of deer pellet groups (defined as a single, uniform faecal accumulation with a minimum number of 10 pellets) and individual hare pellets were counted within a 2 m wide band along each row.
Heather height was recorded 1 m upslope, down slope and to each side of each sapling. The mean of these four measurements was used as an index of heather height around the sapling.
The saplings were harvested at the end of August 2003. They were dug up, care being taken to excavate as much of the root system as possible, placed in plastic bags and stored at 4 °C prior to being washed, and measured in an identical manner to the pre-planting measurements. Signs of herbivory were also recorded. This included browsing that had occurred at any time during the course of the experiment, including during the first winter. The saplings were then dried at 70 °C for 3 days prior to weighing.
The possibility that neighbour removal by clipping could lead to an increased availability of soil nutrients, from decaying below-ground plant material, was tested by sampling soil from plots of a third treatment type at both the start and end of the experiment. This treatment involved canopy removal as described for the neighbour removal treatment, but with root cutting at the edge of the 20 cm radius clipped patch at the start of the experiment. Roots were cut using a sharp-knife with a 30-cm serrated blade. Soil samples were collected prior to planting and at the time of sapling harvest from the first block in all rows at all sites (providing a total of 20 samples at both the start and end of the experiment). The sampled treatment differed from the neighbour removal treatment only in the cutting of roots at the edge of the clipped circle. Soil samples were stored in grip-seal plastic bags at 4 °C for a maximum of 3 weeks prior to being put through a 5-mm sieve. Soils were dried at 50 °C and analysed for extractable content using an automated Dumas combustion procedure (NA1500 Elemental Analyser, Carlo Erba, Milan, Italy).
All data were analysed using SAS version 8.0. Throughout the analyses the Satterthwaite option was used to estimate denominator degrees of freedom for fixed effects (Littell et al. 1996) and data distribution assumptions were checked and, where necessary, appropriate transformations applied.
Changes in soil extractable concentrations during the course of the experiment were analysed using a linear mixed model, with altitude and sampling date and their interaction specified as fixed effects, and row (nested within site) specified as random effects.
The impact of environmental factors, including the presence or absence of heather neighbours, on the probability of sapling browsing in the winter immediately following planting was analysed using a generalized linear mixed model, with the occurrence of browsing regressed against the density of red deer pellet groups, heather height and neighbour removal treatment (removal or control) as fixed effects, and site, row nested within site and block nested within row specified as random effects. This final model was developed following stepwise selection of explanatory variables at the row and block level. The model assumed a binomial error distribution and employed a logit link function.
The impact of browsing and neighbours on the growth of those saplings that survived until the final harvest was assessed by analysing sapling mass increments. Sapling mass increment during the course of the experiment was calculated as:
(sapling dry mass at harvest)/(predicted sapling dry mass pre-planting)
Pre-planting sapling dry mass was estimated using a regression of fresh mass against dry mass calculated for the subset of 50 saplings dried and weighed prior to planting, and was conducted using the GLM procedure. There was a very close relationship between fresh and dry sapling mass (log10 dry mass = 1.023 × log10 fresh mass – 0.505, R2 = 0.97). Sapling mass increment gives an index of sapling growth that is independent of initial plant size. Mass increment was analysed using the MIXED procedure, with an identical random effect structure as above but assuming a normal error distribution. Signs of herbivory at the time of final harvest (which included any browsing that had occurred during the course of the experiment, and not simply during the first winter) and the presence or absence of heather neighbours were included in the model as fixed effects.
impact of canopy removal on soil nutrients
There was no significant effect of altitude on soil extractable concentrations (F1,17 = 2.07, NS), and there was no significant difference between concentrations at the start and the end of the experiment (F1,18 = 0.33, NS). Therefore, during the course of the experiment, there was no detectable change in nutrient availability associated with neighbour removal.
probability of browsing
Variables rejected from the regression analysis included altitude, roe deer pellet group and hare pellet density (row level) and tree height, shoot number, leader length and initial sapling fresh mass (block level). The final regression model therefore included red deer pellet group density, the height of neighbouring heather and the neighbour removal treatment as explanatory variables.
The probability of a sapling being browsed during the winter immediately following planting was significantly influenced by red deer use (pellet group density) and the presence or absence of neighbouring heather (Table 2). Increasing red deer use and removal of heather both increased the probability of browsing (Fig. 1). There was a significant interactive effect of these two variables: the rate of increase of browsing probability with respect to increasing deer use was greater for saplings without heather neighbours. However, following fitting of neighbour presence and red deer use, sapling browsing probability was not significantly affected by the height of neighbouring heather.
Table 2. Results of regression analysis of the probability of sapling browsing in the first winter, against red deer pellet group density, heather height and heather neighbour removal. In all cases numerator d.f. = 1. ** P < 0.01, *** P < 0.001
In the context of the range of red deer use encountered (as indexed by pellet group density), the impact of the presence of heather neighbours on the probability of a sapling being browsed was highest at low to intermediate levels of use by deer. At low red deer use the apparent reduction in facilitative effect may be partly because overall browsing levels are lower, and so the maximum potential impact of neighbours is reduced (Fig. 2a). However, the proportional impact of neighbours can be calculated using an index based on the relative neighbour effect (RNE) index of neighbour impact (Markham & Chanway 1996). RNEbrowsing is calculated as:
(Premoval − Pcontrol)/max(Pcontrol, Premoval)
where P is the predicted probability of being browsed for either control or neighbour removal saplings at a given red deer density. This differs slightly from the index used by Callaway et al. (2002), which was used to look at growth responses, in order to show facilitative interactions with a positive sign.
An RNEbrowsing value of 1 indicates a 100% reduction in the probability of browsing as a consequence of neighbours being present. RNEbrowsing was strongly reduced in areas of high deer use but was still relatively high at low levels of deer use (Fig. 2b).
Browsing during the first winter appears to reduce the probability of saplings surviving through to final harvest. Of the saplings that survived first winter browsing (i.e. that were not uprooted) 20% and 15%, respectively, of unbrowsed saplings in the neighbour removal and control treatments died before the final harvest, whereas the mortality levels for browsed saplings were 70% and 50%, respectively. This mortality may be due to additional browsing (browsed trees may be more conspicuous), greater competition or factors such as snow mould damage.
There was a highly significant interactive effect of neighbour removal and browsing on sapling growth (F1,107 = 6.75, P = 0.01). The removal of neighbouring heather led to increased growth of unbrowsed saplings, but not of browsed saplings (Fig. 3). The mean growth of unbrowsed saplings with heather neighbours did not differ significantly from the growth of browsed saplings, either with or without heather neighbours. Therefore the facilitative effect of heather in protecting saplings from browsing was not translated into a positive effect on sapling mass increment.
It should be noted that sapling mass increments were commonly less than 1, indicating mass loss during the course of the experiment (Fig. 3). This may be a consequence of genuine mass loss (for example due to die-back or the production of smaller needles in the field) or an artefact due to tissue loss during harvesting (although every attempt was made to minimize this). Irrespective, such effects are not likely to influence the different treatment groups other than in a manner that reflects genuine differences in growth conditions. Furthermore, we are not interested in absolute levels of sapling growth but in comparing relative increments between treatments, which we are able to do despite negative mass increments.
During the first winter of this experiment, the presence of heather neighbours clearly protected the saplings from browsing, thus supporting the findings of Rao et al. (2003) and Palmer & Truscott (2003).
Similar facilitative effects, where one plant protects a neighbouring plant from browsing by large herbivores, have been commonly recorded and tend to fall into a few main categories. Unpalatability of a facilitator can deter herbivores from a facilitated plant if the facilitator is mechanically defended, for example with thorns (e.g. Gómez et al. 2001; Kuiters & Slim 2003), or because of the chemical composition of its leaf tissue (e.g. Rousset & Lepart 2000; Oesterheld & Oyarzabal 2004). Facilitators can also reduce the apparency of the facilitated plant, with the presence of neighbours making an individual plant less conspicuous to grazers (e.g. Miller & Cummins 1982; Bellingham & Coomes 2003; Rao et al. 2003). Heather, although potentially very dense in places, is unlikely to present a mechanical barrier to browsing red deer and can be a major component of red deer diet, especially during winter (Mitchell et al. 1977; Latham et al. 1999), making reduced apparency the more likely mechanism here.
In our study, overall mean sapling height was 26.4 cm (± 4.6 cm SD) at the start of the experiment, whilst heather height was 30.9 cm (± 10.2 cm SD). Saplings were therefore commonly hidden by heather during the first winter. Removal of neighbours within 20 cm of the saplings was sufficient to make them more apparent and liable to be browsed by deer. Although by the end of the experiment, average sapling height had increased to 38.6 cm (± 8.4 cm), we cannot categorize browsing impacts into periods before and after the saplings emerged above the heather, and so cannot assess how this affected their probability of being browsed, although it is likely that it would increase. Rao et al. (2003) found that mountain hares tended to browse taller downy birch (Betula pubsescens) saplings and trees planted in shorter heather, and Palmer & Truscott (2003) found that the probability of winter browsing of Scots pine by red deer increased with sapling height and was greater if the sapling was taller than the surrounding vegetation. In our study, the short stature of the planted saplings relative to the heather, as well as the limited variation in sapling and heather heights, probably restricted the potential for finding a significant effect of either factor during the first winter.
Changes in soil nutrient availability, due to neighbour clipping and the decomposition of below-ground material, are not likely to have been the cause of sapling responses to neighbour removal. There was no significant difference in soil extractable in clipped areas between the start and end of the experiment. Although it is possible that the flush of nutrients might have peaked and declined prior to the second soil sampling date, this is unlikely because decomposition is comparatively slow in the acid, highly organic soils of heather moorland, and because heather stems provide a recalcitrant substrate for decomposers (Heal et al. 1978; Anderson & Hetherington 1999).
As predicted, the facilitative effect of heather was dependent upon the level of red deer use of an area, being greatest at intermediate levels of deer use. The reduction in facilitative effect in areas with high deer use is not surprising: the net foraging effort by deer is probably sufficiently high that effectively all saplings are found irrespective of the presence of neighbours. At low levels of deer use, the absolute level of browsing is reduced (only a small proportion of pines are browsed), and therefore the absolute reduction in browsing probability is necessarily also reduced. However, the proportional reduction in browsing probability due to the presence of heather (RNEbrowsing) remains comparatively high. This highlights the need to make a distinction between the intensity and importance of an effect (Welden & Slauson 1986). RNE provides a measure of the intensity of an interaction rather than its importance (Brooker et al. 2005). Although RNEbrowsing is high at low levels of deer use, we can see that absolute levels of browsing are extremely low (Fig. 1), suggesting that this herbivore-mediated beneficial impact of heather on pine is relatively unimportant (at the population level) at low deer densities. This supports the proposition of Welden & Slauson (1986) that the intensity and importance of interactions ‘need not be correlated’.
The shape of the relationship between the facilitative effect and the severity of the environment is similar to that found in the few other studies where sufficient data are available to allow regression analysis. For example, Callaway et al. (2002) found an asymptotic relationship between summer temperatures and the mean RNE in alpine plant communities. If the gradient of severity in the Callaway et al. study was extended into harsher conditions, the RNE value would probably return to zero as interactions break down altogether. This is akin to high deer density in our study; in these circumstances the potential for heather to hide the pine saplings is reduced simply because the net search effort by deer is high in the more ‘extreme’ environment.
Although we have shown a relationship between environmental severity and facilitation, a result of importance in terms of understanding the role of interactions along environmental gradients (Bruno et al. 2003), it is also important to demonstrate that an interaction has ecological consequences (Goldberg & Barton 1992). To influence community composition and biodiversity, interactions must translate into population-level responses (McPeek & Peckarsky 1998; Goldberg et al. 1999). There is sufficient evidence to suggest that biomass responses can be expected to translate into population level responses in the case of Scots pine. In general, browsing reduces the growth rate of saplings (Gill 1992), as well as the probability of survival of saplings of a range of tree species (Edenius et al. 1995; Van Hees et al. 1996). Therefore, the facilitative effect of heather on Scots pine, if translated into a biomass response, has the potential to be ‘ecologically meaningful’ at the population and community level.
If facilitation by heather were the dominant component of the net pine–heather interaction, we would expect saplings with heather neighbours to have greater biomass increments than those without neighbours. However, we find that unbrowsed saplings with heather neighbours grow no better than browsed saplings, and therefore that the facilitative effect of heather does not translate into a biomass response. It is likely that the negative, competitive effects of heather are suppressing the growth of the saplings and outweighing the beneficial effects of protection from herbivory. This result provides an example of a potentially beneficial effect of neighbours being offset by a second, dominant and negative effect of the same individuals, supporting previous studies that demonstrated temporal variability in plant–plant interactions (e.g. Goldberg & Novoplansky 1997; Bellingham & Coomes 2003; Schiffers & Tielbörger 2006). The assumption that interactions measured at a particular point in time will necessarily translate into biomass (and ultimately population-level) responses may be incorrect because of the multiple, temporally variable interactions that can occur between neighbours.
What is more difficult to explain is the response of the browsed trees with heather neighbours (Fig. 3). If the effects of browsing and competition were additive we would expect this treatment to have the lowest mass increment. However, it has a similar mass increment to both the browsed trees without neighbours and the unbrowsed trees with neighbours (leading to a significant browsing–neighbour interaction effect). There is only a very limited potential for compensatory growth in Scots pine (Edenius et al. 1993; Palmer & Truscott 2003; although the observation was made principally on substantially larger saplings than we used), whilst browsing commonly removed very little tissue mass (i.e. the tip of the leader; R. Brooker personal observation). If both browsing and competition cause a check on growth it might explain this pattern, i.e. there has been no potential for growth through which the differences between the treatments can become manifest.
Finally, we should consider whether this research has conservation implications. Considerable effort has been devoted to promoting regeneration of native pinewood in the Scottish uplands. Fenced exclosures have been widely used but have a number of associated problems (Staines & Balharry 2002), including detrimental impacts on woodland grouse (Catt et al. 1994). Consequently, there has been increased effort to reduce red deer densities in some areas to levels compatible with regeneration of pine (Beaumont et al. 1995; Scott et al. 2000). Based on our study it might be perceived that facilitation by heather could also be employed as a management tool to promote the regeneration of Scots pine. However, this is unlikely to be the case. We have only demonstrated the facilitative effect for saplings that are, on average, less than heather height. When the saplings emerge above the heather their increased apparency will probably make them susceptible to browsing. Moreover, and perhaps most importantly, although we have shown a beneficial effect of heather on saplings in protecting them from browsing, we have also shown that the presence of neighbouring heather reduces sapling growth, such that there is ultimately no growth benefit for the saplings from the presence of heather neighbours. These experimental results have been mirrored in a decline in pine recruitment in areas with reduced deer densities, most likely due to increased growth and cover of heather resulting from reduced grazing and trampling (Hancock et al. 2005). This illustrates the complex and potentially transient nature of plant–plant interactions, and the difficulties associated with using shelter effects of heather as a management tool to promote Scots pine regeneration.
The authors would like to acknowledge the help of RSPB Abernethy Estate, National Trust for Scotland Mar Lodge Estate, Scottish Natural Heritage and Dinnet Estate for permission to conduct the fieldwork. We are also very grateful to all of the staff from these organizations that helped us throughout this project. The authors would also like to thank Richard Michalet for comments on earlier versions of the manuscript, David Elston for statistical advice, and Linda Turner, Maddie Thurlow and Claire Brooker for help with laboratory analysis of the saplings. This study was funded by the Natural Environment Research Council (grant NER/M/S/2001/00080).