Exploitation of intertidal grazers as a driver of community divergence

Authors

  • Gustavo M. Martins,

    Corresponding author
    1. Secção Biologia Marinha, Departamento Biologia, Universidade dos Açores, 9501-801 Ponta Delgada, Açores, Portugal
    2. Marine Biology and Ecology Research Centre, Marine Institute, University of Plymouth, Plymouth PL4 8AA, UK
    3. Centro Interdisciplinar de Investigação Marinha e Ambiental (CIIMAR), Rua dos Bragas 289, 4050-123 Porto, Portugal
      Correspondence author. E-mail: gmartins@uac.pt
    Search for more papers by this author
  • Richard C. Thompson,

    1. Marine Biology and Ecology Research Centre, Marine Institute, University of Plymouth, Plymouth PL4 8AA, UK
    Search for more papers by this author
  • Ana I. Neto,

    1. Secção Biologia Marinha, Departamento Biologia, Universidade dos Açores, 9501-801 Ponta Delgada, Açores, Portugal
    2. Centro Interdisciplinar de Investigação Marinha e Ambiental (CIIMAR), Rua dos Bragas 289, 4050-123 Porto, Portugal
    Search for more papers by this author
  • Stephen J. Hawkins,

    1. School of Ocean Sciences, Bangor University, Menai Bridge, Anglesey LL59 5AB, UK
    2. Marine Biological Association, Citadel Hill, Plymouth PL1 2PB, UK
    Search for more papers by this author
  • Stuart R. Jenkins

    1. School of Ocean Sciences, Bangor University, Menai Bridge, Anglesey LL59 5AB, UK
    2. Marine Biological Association, Citadel Hill, Plymouth PL1 2PB, UK
    Search for more papers by this author

Correspondence author. E-mail: gmartins@uac.pt

Summary

1. The possibility that different assemblages of species may represent alternative stable states has been the subject of much theoretical and empirical work. Alternative stable states may in theory arise from a perturbation of sufficient magnitude that pushes an assemblage from one stable equilibrium point to another. Overfishing is one such disturbance that can lead to cascading community-level effects. Yet, whether these different assemblages represent alternative stable states or are the consequence of chronic disturbance from fishing is still a matter of debate. Understanding the mechanisms that drive community stability is fundamental if we are to assess the consequences of anthropogenic impacts on the structure and function of ecosystems to better inform management of disturbed habitats.

2. To investigate the extent to which present-day community state is stable versus being maintained by chronic exploitation, we manipulated the time and intensity of physical disturbance and grazing by limpets in a system where over-harvesting of limpets has led to a regional-scale shift in community structure to one in which algal turfs have replaced barnacles as the primary space occupier in the mid-intertidal.

3. After a 1-year period since disturbance was applied, assemblages in disturbed areas were significantly different from undisturbed areas, but the timing of disturbance and its intensity had little effect on the outcome of succession. Undisturbed areas were highly resistant to new colonization and persisted unchanged throughout the study period.

4. Manipulation of limpet abundance in disturbed patches showed that, where present, limpets successfully prevented the recolonization of space by algal turfs. Moreover, there was evidence that grazing by limpets at the turf/open-rock boundary effectively pushed the turfs back, extending the area of open-rock.

5.Synthesis and applications. Our findings provide evidence that in this system the dominance by algal turfs does not represent an alternative stable state but that chronic exploitation of limpets leads to the persistence of this community. Conservation strategies aimed at protecting or enhancing limpet abundances (e.g. no-take marine reserves) should allow the gradual restoration of this community to its pre-disturbed state.

Introduction

Different assemblages of species are often observed in the same physical environment. The possibility that these may represent stable alternative community states has been the subject of debate for decades (e.g. May 1977; Connell & Sousa 1983; Knowlton 1992). Recent work by Petraitis and co-workers in the Gulf of Maine on sheltered rocky shores (Petraitis & Dudgeon 1999; Petraitis & Latham 1999; Dudgeon & Petraitis 2001) has re-ignited this debate and led to controversy over the evidence required to demonstrate alternative stable states (Bertness et al. 2002). Alternative stable states may in theory arise from different starting conditions or from a perturbation of sufficient magnitude that pushes an assemblage from one stable equilibrium point to another. Such perturbations may arise from both natural (e.g. ice scour) and anthropogenic (e.g. over-harvesting) sources.

In South Africa, two distinct subtidal communities can be found in two adjacent islands separated by only 4 km (Barkai & McQuaid 1988; Castilla, Branch & Barkai 1994). At Malgas, the subtidal community is regulated by strong top-down control by a rock lobster Jasus lalandii. At Marcus, however, the community is regulated by a predatory mollusc Burnupena papyracea although rock lobsters were apparently abundant in 1960. It is not clear what has led to community divergence but these communities have remained unchanged for over 20 years and represent a clear example of stable alternative community states. However differences in community structure do not necessarily indicate a stable state. In South America over-harvesting of intertidal fauna led to significant changes in community structure, but these changes were only maintained because of chronic harvesting (e.g. Branch & Moreno 1994). This became apparent when harvesting by humans was restricted along parts of the Chilean coastline resulting in a return to the original intertidal community structure over the following five years (Branch & Moreno 1994; Castilla, Branch & Barkai 1994). Hence it appears that disturbance in some regions can shift community composition to an alternative stable state, whilst in other regions communities states are only maintained whilst disturbance persists. Understanding the mechanisms that drive community stability is fundamental if we are to assess the consequences of anthropogenic impacts on the structure and function of ecosystems (Hughes et al. 2005) and better inform management of disturbed habitats (Young et al. 2008).

The Azorean intertidal zone has been the subject of intense exploitation, chiefly for the limpets Patella candei and P. aspera (Hawkins et al. 2000). Although levels of limpet abundance prior to the start of exploitation are not known, their density has been reduced from approximately 61 to 7 individuals m−2 over a period of twenty years at some locations (Martins 2009). Amongst the Azorean islands, limpet densities range between 80 and 9 individuals m-2 as a direct consequence of exploitation intensity (Martins et al. 2008a). Empirical evidence suggests that the well-substantiated decline in patellid populations has led to a significant change in the mid-shore community structure (Martins et al. 2008a). On islands where the abundance of limpets has been reduced, turf-forming algae (e.g. Gelidium spp., Caulacanthus ustulatus) have replaced barnacles as the dominant space occupier at mid-shore levels. Over-harvested islands now support different assemblages, which differ in structure and functioning from that of islands where limpets are still abundant (being net producers vs. net consumers). The stability of this exploitation-driven community state is, however, not known although it has apparently been maintained at least since 1995 (Neto 2000).

Turfs, which bind sediment, are a dominant occupier of primary space in the rocky intertidal. In the central Mediterranean, for instance, neither the type of disturbance (removal versus abrasion), time of disturbance nor the depositional environment of sediments severely affected the development of algal turfs, which showed remarkably fast recovery rates compared to other algal morphotypes (Airoldi 1998). The ability of turfs to withstand invasion by other spatial competitors (Sousa, Schroeter & Gaines 1981) in addition to their tolerance of physical and biotic stresses (Hay 1981) and their resistance to and quick recovery from disturbance (Sousa 1980) suggest that algal turfs may dominate and persist under a wide range of conditions (see Airoldi 1998 for review).

Patellid limpets have a strong community-structuring role on shores of northwest Europe through grazing on microscopic algae and early life-stages of larger macroalgae (see Jenkins et al. 2008 for review). However, despite their ability to control macroalgal biomass experimental work suggests they may have little impact on algal turfs (Jenkins, Hawkins & Norton 1999). As in other well-studied patellid limpets (i.e. Patella vulgata) Patella candei is thought to be a microphagous grazer (Martins 2009). Hence once turfs are established, it is likely that only a physical disturbance that frees up space by removing turfs can trigger community transformation. In the Azores, over-harvesting of limpets allowed the development of well-established assemblages of algal turfs at mid-shore heights where barnacles used to be the dominant space occupier (Neto 2000, Martins et al. 2008a). Thus the Azorean intertidal presents a system in which human exploitation may have shifted the community to an alternative stable state. Alternatively, it may simply be that continued exploitation of limpets is the key driver allowing turfs to dominate. To test these alternative hypotheses we examined the effects (i) of disturbance size and timing and (ii) grazing by limpets on the dominance of algal turfs at mid-shore level. In a first experiment, we manipulated the size (a surrogate for disturbance intensity) and timing of disturbance and examined community development. Here, the abundance of limpets was left unaltered to examine the ability of the system to recover under a scenario of intense limpet exploitation. In the second experiment, we manipulated limpet density to specifically test the hypothesis that in disturbed areas, grazing by limpets at natural (unexploited) densities controls space acquisition by algal turfs and thus promotes community transformation, for example, into an assemblage dominated by barnacles.

Materials and methods

Study sites

The study was conducted at mid-shore level at two moderately exposed locations (Lagoa and Caloura) on the south coast of São Miguel Island, Azores. These locations are of volcanic origin (basalt s.l.) and consist of small rocky platforms intermingled between small cobble and boulder beaches. The rocky platforms are steep; the substratum is convoluted and presents many pits and crevices. In the eulittoral zone, conspicuous organisms include the barnacle Chthamalus stellatus Poli, the limpet Patella candei d’Orbigny, the littornid Littorina striata King and turf-forming algae (e.g. Gelidium spp.). Small patches of the fucoid Fucus spiralis L. also occur. The ephemeral algae Ulva spp. and the cyanobacteria Rivularia sp. can be seasonally abundant. The encrusting alga Nemoderma sp. is common in the damper areas, whilst Ralfsia sp. is more common on well-drained rock (see Martins et al. 2008b for further descriptions of these locations). Both locations are exploited and are hence representative of the alternative exploitation-driven community state where the mid-shore is dominated by algal turfs and barnacle-dominated areas are restricted to the high shore (Martins et al. 2008a).

Sampling design

To investigate the extent to which present-day community state is stable versus being maintained by chronic exploitation of limpets, two experiments were established to examine (i) the persistence and resilience of algal turfs to physical disturbance, and (ii) the role of grazing by limpets in areas where turfs had been disturbed to produce patches of open rock.

The role of timing and size of physical disturbance in the persistence of the algal turf

At each of the two locations, 24 patches of 25 × 25 cm were haphazardly selected and marked with screws in the mid-shore where turf-forming algae (mostly Gelidium microdon Kützing) covered more than 90% of the substratum. To examine the effects of disturbance size on the persistence/recovery of algal turfs, these patches were then randomly assigned to one of three treatments (n = 8 each): (i) unmanipulated controls, (ii) total removals, and (iii) partial removals. Unmanipulated controls were established to examine the persistence of algal turfs over the entire study period. In the total removal treatment, the entire area of each patch was scraped and wire-brushed of all biota until no macrobiota were visible. In the partial treatment, only half the area of each patch was cleared hence simulating a disturbance of lower intensity. In this treatment, clearance of biota was achieved by dividing the patch in 10 equally sized smaller areas and the biota was removed from 5 randomly selected areas. This was intended to deliberately intensify edge effects, the prediction being that edge effects increase with perimeter. The smaller cleared areas also better mimic the size of natural disturbances such as those produced by mechanical abrasion from movement of cobbles by wave action (G.M. Martins, personal observation).

Many marine invertebrates and algae have discrete temporal windows for recruitment and these could affect the outcome of competitive relationships amongst colonizing species (Hawkins 1981; Benedetti-Cecchi 2000). Thus, the influence of timing of disturbance was examined by replicating the experiment twice: the first experiment was run from February 2007 to June 2008 (16 months), and the second from October 2007 to October 2008 (12 months).

Sampling was done every month during the first 4 months and at approximately 3–4-month intervals thereafter. A 5 × 10 cm sampling quadrat with 10 intersections was used, which was haphazardly laid three times in each patch so that a total of 30 sampling points corresponding to 150 cm2 were sampled in each replicate. Sampled areas thus corresponded to approximately 25 and 50% of the scraped area in the total and partial removals treatments respectively. Note that in the partial removal treatment, only scraped areas were sampled. The cover of sessile organisms was converted to percentage cover whilst mobile animals (limpets) were counted and their abundance expressed as density.

The role of grazing by limpets in preventing re-establishment of algal turfs in areas of open rock

To test the hypothesis that limpets influence the ability of turf-forming algae to re-gain space via lateral vegetative growth we estimated the change in space (bare rock) of disturbed areas over time. At each location, 15 additional patches of 10 × 10 cm were marked within the turf matrix and an area of approximately 5 × 5 cm was scraped clean of all biota, as above, within the centre of the marked areas. The size of the clearings is within that frequently produced by natural disturbances (G.M. Martins, personal observation). Individuals of Patella candei were collected nearby and carefully transplanted into disturbed patches according to each of the three treatments (n = 5 each): (i) no limpets, (ii) one limpet, and (iii) two limpets. The latter two treatments corresponded to limpet densities of 400 and 800 individuals m-2, respectively. Such high limpet densities are far greater than the average abundance currently observed at these locations (Martins et al. 2008a) and probably greater than average abundance prior to exploitation. However it is not unusual to observe small aggregations of two to three limpets in similarly sized patches in the Azores (G.M. Martins, personal observation) and the Mediterranean (Benedetti-Cecchi et al. 2005) and hence these treatments represent realistic densities.

Transplanted limpets ranged between 10 and 15 mm shell length, the modal size at these exploited locations (Martins et al. 2008a). The experiment was initiated in January 2008 and run until July 2008. During this period, limpet abundance inside treatments was maintained, if necessary, approximately every two weeks.

The area of open rock (available for colonization) was estimated through time using a 10 × 10 cm quadrat with 100 subdivisions (1 subdivision = 1 cm2). In each subdivision, the cover of bare rock was given a score from 0 to 4 (corresponding to 0, 25, 50, 75, and 100% cover). The percentage cover of bare rock was later converted to the corresponding area. Sampling was done prior to the establishment of the experiment and every 2 months thereafter. Temporal variation in the area of bare rock (disturbed area) was analysed with reference to the area of bare rock at the start of the experiment and expressed as percentage change.

The percentage cover of colonizing organisms (those recruiting from the plankton) was sampled using a 5 × 5 cm quadrat with 49 intersections. However colonization was negligible so data are not presented. In addition, no mobile animals (e.g. littorinids) were ever observed inside the cleared patches.

Data analysis

Data were generally analysed using analysis of variance (anova). Lack of temporal independence meant that time was not considered as a factor and hence data were analysed separately for each sampling occasion. Prior to anova, data were checked for homogeneity of variances using the Cochran’s test and transformations applied where necessary (Underwood 1997). Pooling procedures were used where appropriate (α > 0·25) to improve the power of tests concerning terms of interest (see Underwood 1997 for further details). Student–Newman–Keuls (SNK) tests were used to compare means within significant terms.

Analysis of the effects of timing and size of disturbance on the persistence of algal turfs was done on 12-month old assemblages to guarantee equivalence (same duration since start) between the two starting dates.

In the second experiment, confidence intervals were used to determine if the size of disturbed areas changed (relative to the size at the start of the experiment) over time in response to limpet grazing. Inspection of standard deviations showed that variability was small and evenly distributed amongst treatments suggesting that confidence intervals were not affected by the small sample size.

Results

The role of physical disturbance in the persistence of the algal turf

Undisturbed areas remained relatively unchanged over the study period with turf-forming algae dominating the space and successfully preventing the colonization of other biota (see Appendix S1, Supporting Information). In contrast, disturbed patches were readily colonized by other biota. Ephemeral algae (mostly Ulva rigida) quickly colonized disturbed areas but were gradually replaced by turf-forming algae (which colonized space mostly via lateral vegetative growth), barnacles, encrusting algae and limpets.

Overall, 12 months after the start of the experiment, the assemblage structure differed amongst treatments (Fig. 1). Thus, the cover of algal turfs in disturbed areas was approximately half that of undisturbed areas. In contrast, the cover of encrusting algae and barnacles had increased as a consequence of disturbance to levels 14 and 34 times higher than control areas. Limpets were absent in control areas but reached a mean (±SE) density of 2·7 ± 0·6 (≈ 43 individuals m−2) in disturbed areas.

Figure 1.

 Abundance (mean ± SE, n = 8) of conspicuous taxa at the end of the experiments to examine the role of physical disturbance on the persistence and recovery of algal turfs. Date 1 and 2 correspond to the experiments initiated in February and October 2007, respectively. Different letters indicate statistically differences in the abundance of taxa amongst treatments (see Appendix S2, Supporting Information). Limpet density was standardized amongst treatments for 625 cm2. Note the different scale of y-axes.

Disturbance size (partial vs. total) had a non-significant effect on the outcome of succession with the abundance of animals and plants being generally similar between the two disturbed treatments (Fig. 1, see Appendix S2, Supporting Information for anovas). Similarly, timing of disturbance had little effect on the structure of developing assemblages except for encrusting algae, which achieved a significantly greater cover when disturbance was applied in October 2007. Timing of disturbance also seemed to influence barnacle abundance at one of the two locations examined (Fig. 1) although this was not statistically significant (Appendix S2).

A strong and positive correlation was found between the abundance of encrusting algae and barnacles and between limpets and barnacles, whilst the abundance of limpets, barnacles and encrusting algae were all negatively correlated with the abundance of algal turfs (Table 1) highlighting the suppressive effect of turfs on the remaining taxa.

Table 1.   Correlation coefficients between the abundance of taxa at the end of the experiments in the disturbed treatments (n = 32)
 Algal turfsEncrusting algaeBarnacles
  1. *P < 0·05, **P < 0·01, ***P < 0·001, otherwise n.s.

Algal turfs  
Encrusting algae−0·26* 
Barnacles−0·54***+0·29**
Limpets−0·51***+0·12+0·54***

The role of grazing by limpets in preventing re-establishment of algal turfs in disturbed areas

Significant differences in the way the amount of bare rock changed amongst treatments were already clear 2 months after disturbance and these were consistent in time and space (Fig. 2, Table 2). Where limpets were absent, the algal turfs surrounding the disturbed patches gradually colonized the available space via lateral vegetative growth so that the amount of bare rock decreased in the disturbed area. In contrast, the presence of limpets successfully prevented the re-invasion of the disturbed patches by algal turfs and there was generally an increase in the area of bare rock, which was greater at higher limpet densities (Fig. 2, Table 2). Where limpets were included, patch growth was generally more pronounced during the first two months and tended to stabilize thereafter (Fig. 2) suggesting that the effect of grazing by limpets was density-dependent. In contrast, patch shrinkage (in the absence of limpets) occurred throughout the experiment as would be expected by the continuous growth of algal turfs.

Figure 2.

 Change (mean ± SE, n = 5) in the area of bare rock in disturbed areas exposed to different grazing levels by Patella candei (0L, 1L and 2L stand for 0, 1 or 2 limpets per patch, respectively).

Table 2.   Two-way factorial anova testing for differences in relative change in the area of bare rock (%) in disturbed patches exposed to different levels of grazing by limpets (0L – no limpets, 1L – one limpet, 2L – two limpets). Location was a random factor whilst Treatment was fixed
SourcedfMarch 2008May 2008July 2008F-ratio
MSFMSFMSF
  1. *P < 0·05, **P < 0·01, ***P < 0·001, otherwise n.s.

Location = L10·30·211540·81·493·90·19Residual
Treatment = T221·0195·19**30591·71053·68***84·668·08*L × T
L × T20·20·1029·00·031·20·56Residual
Residual241·61034·92·1 
Transformation sqrt (+ 1)nonesqrt (+ 1)  
Cochran’s test = 0·39= 0·32= 0·44 
SNK tests 0L < 1L < 2L0L < 1L < 2L0L < 1L = 2L 

At the end of the experiment, the size of disturbed patches for all three treatments was significantly different from that at the start of the experiment (Table 3). That is, where limpets were present, the area of open rock available for colonization at the end of the experiment was on average 30 and 90% (in the 1L and 2L treatments, respectively) greater compared to initial conditions. Where limpets were absent, however, the area of the disturbed patch was on average 32% smaller.

Table 3.   Area of disturbed patches (as a percentage of initial area) and the upper and lower confidence limits (CL 95%) for each treatment: 0L – no limpets, 1L – one limpet, 2L – two limpets. Time refers to sampling dates
TimeLocationTreatmentMean area (%)Upper and lower CL
  1. A CI interval that does not overlap with 100 indicates a significant change in the area of the disturbed patch in relation to the initial disturbance area (indicated as bold).

March ‘08Lagoa0L86·970·5–103·3
1L125·697·5–153·7
2L170·2145·2–195·2
Caloura0L87·976·6–99·4
1L116·194·9–137·2
2L164·0117·6–210·5
May ‘08Lagoa0L84·462·9–105·8
1L140·9103·6–178·2
2L197·8158·9–195·2
Caloura0L73·962·1–85·8
1L124·5107·2–141·8
2L181·3150·3–212·4
July ‘08Lagoa0L70·449·0–91·8
1L164·9106·2–223·6
2L198·3165·7–230·8
Caloura0L64·548·9–80·1
1L124·7103·6–145·7
2L189·9159·7–220·0

The surrounding assemblage of algal turfs proved a valuable barrier to the movement of limpets that became ‘imprisoned’ inside the experimental patches. However, algal die-off during the hot summer months meant that by August, limpets in some experimental patches (irrespective of the treatment) were released of their ‘imprisonment’ and were able to disperse elsewhere and hence the experiment was terminated.

Algal recruitment from the plankton was negligible throughout the experiment. It should be noted, however, that recently recruited individuals of the barnacle Chthamalus stellatus could be identified in all the disturbed patches by July suggesting that had the experiment been maintained for a longer period then barnacles would have probably become a dominant space occupier in patches where limpets prevented the re-invasion of algae.

Discussion

The experimental manipulation of limpets showed that they successfully prevented the establishment of algal turfs in disturbed patches and, to a certain extent, mediated community divergence. Moreover, there was evidence that limpets also grazed at the turf boundary increasing the area free of algal turfs. Such an effect of limpets in well-established patches of macroalgae has also been documented elsewhere (Jenkins, Hawkins & Norton 1999; Davies, Johnson & Maggs 2007). In undisturbed patches, however, algal turfs showed remarkable persistence and resistance to invasion, in accordance with previous work (see Airoldi 1998 for review). These results indicate that a switch from the current turf-dominated community in the mid-intertidal back to one dominated by sessile filter feeders and bare space is contingent on two factors: (i) a disturbance that removes turf-forming algae allowing other organisms to recruit, and (ii) limpets recruiting into these disturbed areas before recolonization by turf-forming algae. The rocky intertidal is a highly dynamic system where small-scale disturbances are frequent and play a central role in maintaining species diversity (Sousa 1979). Creation of artificially disturbed areas within the turf matrix showed that the time required for algal turfs to fully recover from disturbance is long enough to allow the recruitment of limpets. The fact that recruitment was observed in the turf-dominated mid-intertidal emphasizes that there is a source of larvae, presumably from limpets which extend vertically up the shore above the turf boundary. Our results from experimental manipulations of limpet densities within the turf matrix also suggest that limpets on the high shore could extend their range down the shore by extending into the turf zone, provided that limpet abundance is sufficiently high. Hence, our results suggest that in this system the dominance of algal turfs at the mid-intertidal is not a stable condition and is maintained by low levels of limpet abundance in turn caused by continued exploitation.

In considering the balance between turf-forming algae and barnacles it is worth noting that barnacles are inferior competitors for space, have no negative effect on algae (in fact, dense patches of barnacles reduce algal grazing by limpets by limiting their access to algae; see Hawkins 1981; Lubchenco 1983) and attain no size refuge from algal overgrowth. Persistence of barnacles lower on the shore, at the barnacle-algal boundary, is maintained by fluctuation-dependent mechanisms (sensuChesson 2000). That is, there is a trade-off between competition and resistance to disturbance. Algae are superior competitors but have little resistance to grazing by limpets, which is relatively substantial at this shore level. Barnacles are inferior competitors but are not as negatively affected by limpets (but see Dayton 1971 for the effects of bulldozing by limpets on recently settled juvenile barnacles) and are thus able to persist provided that grazing by limpets is maintained (Underwood 1980; Hawkins & Hartnoll 1983). Therefore, in the absence of limpets, barnacles have no ability to resist competition by algae, which eventually extend higher on the shore until limits on their ability to resist desiccation prevent further extension. This is probably the situation in the most exploited islands of the Azores such as São Miguel.

Petraitis & Latham (1999) and Dudgeon & Petraitis (2001) suggested that only large-scale disturbances can trigger a shift between alternative assemblages (such as that between Ascophyllum nodosum vs. filter-feeder dominated assemblages) and that in smaller-scale disturbances strong edge effects might preclude such change. In our system, disturbance size had little qualitative or quantitative effect on the outcome of succession in disturbed patches even though the treatment corresponding to the smaller-scale disturbance (partial removal) was prepared in a way that deliberately increased the perimeter–area ratio (a surrogate for edge effects). It is possible that differences in assemblage composition between the two systems (Azores vs. Maine, USA) can explain these divergent results. For instance, whereas shores in the Gulf of Maine are sheltered from wave-action, Azorean shores experience considerable hydrodynamic forces. Moreover, the biota of the two regions is itself much different with large canopy algae (i.e. Fucus, Ascophyllum) and key space occupiers (i.e. mussels) dominating in the Gulf of Maine, but being generally absent in our system which is dominated by turf-forming algae and small chthamalid barnacles.

Due to seasonality in reproductive patterns and growth, timing of disturbance is also expected to have a strong effect on succession of disturbed patches. In some systems, the timing of disturbance, although affecting the sequence of species that colonize the free space, has little effect on the outcome of succession with the stronger competitor eventually dominating (Hawkins 1981). In other systems, however, multiple end-points are possible and the timing of disturbance can determine the outcome of succession via priority effects (Benedetti-Cecchi 2000). In our study, with the exception of encrusting algae, the timing of disturbance did not have a strong effect on the abundance of organisms. Our study, however, was not carried out for a period long enough to clearly assess whether the quantitative differences in the abundance of taxa between treatments initiated at different times could result in different end-points.

What then does determine the stability of alternative community states? Alternative states may arise from a perturbation where the dominant species is differentially suppressed allowing other species to increase in abundance. The stability of such alternative states, however, will only be maintained if the ‘new’ assemblage of species is able to successfully secure resources (e.g. space) and resist invasion by the previous dominant species. In some regions, stability is maintained because the ‘new’ dominant species is able to grow to a size that allows it to resist predation by the suppressed species (Paine, Castilla & Cancino 1985). In other regions, stability is determined by density-dependent reversals of the predator-prey roles between the two dominant species (Barkai & McQuaid 1988; Castilla, Branch & Barkai 1994). However, in many cases the cause of stability is more elusive and is probably maintained by the continued influence of some external factor (e.g. predation, exploitation, pollution or disease) (e.g. Moreno, Sutherland & Jara 1984; Hawkins et al. 2002). This suggests that stability of alternative communities states is context-dependent and may be influenced by the identity or trophic level of the dominant species that characterize the different community states. Connell & Sousa (1983) suggested that when different community states are maintained by some external factor (e.g. exploitation) they do not represent ‘true’ alternative community states. Our results are in agreement with this; when limpets were present at densities sufficiently high (simulating pre-disturbance scenarios), they successfully prevented space monopolization by algal turfs and promoted community divergence despite the fact that turf-forming algae have dominated mid-shore communities for a period far greater than the species turnover (Neto 2000). Hence, this turf-dominated community cannot be considered as an alternative stable state.

Understanding the mechanisms that drive community stability is of theoretical as well as of applied interest. For instance, experimental enhancement of the densities of the exploited lobster Jasus lalandii in South Africa could not be maintained because they were eliminated by a local reversal in the roles of predator and prey (Barkai & McQuaid 1988). In cases like this, ecosystem reversal would require the active removal of the dominant predator species. In contrast, our study provides evidence of a community state that apparently is not stable but is instead maintained by the chronic exploitation of a key species. This finding implies that ecosystem restoration could be achieved by conservation measures aimed at protecting or enhancing the stocks of limpets. No-take marine reserves have been successfully used as a way to rapidly enhance the abundance and size of exploited stocks (e.g. Halpern 2003) and would provide the ideal setting to test this hypothesis. Unfortunately, insufficient levels of enforcement and illegal harvesting within protected areas along the Azorean coastline (Martins 2009) preclude the use of the existing reserves for such a purpose.

Acknowledgements

This work was part of the requirements for the completion of a PhD and was supported by Postgraduate Grant SFRH/BD/22009/2005 awarded to G.M.M. by Fundação para a Ciência e Tecnologia (FCT, Portugal). Thanks are due to all of those from the Secção de Biologia Marinha da Universidade dos Açores who helped in the fieldwork but who are too numerous to individually acknowledge. The study complies with the laws of the country in which the work was carried out. This manuscript greatly benefited from comments by Peter Fairweather and George Branch.

Ancillary